the synchrony and diachrony of time deictic adverbials

1 downloads 0 Views 3MB Size Report
been called “temporal deictic expressions” by Kurzon (2008), “temporal distance markers” by. Haspelmath (1997) ...... In Jeff Good (ed.), Linguistic Universals ...
THE SYNCHRONY AND DIACHRONY OF TIME DEICTIC ADVERBIALS: SPANISH HACER+TIME IN A CROSS-LINGUISTIC PERSPECTIVE

Author: Borja Herce Calleja Director: Carlos García Castillero Department: Classical Studies Master in Linguistics 2014-15 University of the Basque Country 1

CONTENTS Abbreviations...........................................................................................................page 4 1. Summary.............................................................. .................................................page 5 2. Background..........................................................................................................page 7 2.1. Conceptualization of time................................................................ .................page 7 2.2. The object of study in context...................................... .....................................page 9 2.3. Asymmetry.......................................... .............................................. ...............page 11 2.4. Grammaticalization.........................................................................................page 13 2.5. Word categorization........................................................................................page 16 3. Time deictic expressions......................... ............................................................page 20 3.1. Previous analyses.................................................... ......................................... page 20 3.2. Towards a typological overview of TDPs and TDMs.......................... ..............page 23 3.3. Past Deictic Markers and their phrases cross-linguistically............. ................page 27 3.4. Future Deictic Markers and their phrases cross-linguistically........ .................page 32 3.5. Conclusion............................................................ ...........................................page 36 4. The synchrony of hacer in Spanish time expressions........................ .................page 39 4.1. Introduction to hacer....................................................................................... page 39 4.2. Introduction to temporal hacer.......................................................................page 40 4.3. A competence-based approach to the synchronic properties of the adverbial construction with hacer...................................................................page 41 4.4. A usage-based approach to the synchronic properties of hacer+time...........page 45 4.5. The relationship between the adverbial and clausal constructions ................page 49 5. The diachrony of hacer/haber+time..................................................................page 56 5.1. Introduction....................................................................................................page 56 5.2. About the durative and the puntual meanings...............................................page 57 5.3. About the clausal and the adverbial constructions........ .................................page 58 5.4. Time biclausal constructions: Sources.............................................. ...............page 59 5.5. Time biclausal constructions: Outcomes.................................... .....................page 61 5.6. The emergence of the adverbial construction: Two proposals...... .................page 62 5.6.1. Preliminary observations........................................ ......................................page 62 5.6.2. Morphological haplology + reanalysis........................................................... page 63 5.6.3. Reanalysis of surface structure + syntactic back-formation .........................page 66 5.6.4. Discussion............................................................................ .........................page 69 5.7. A corpus-based quantitative analysis of the diachronic changes in time constructions with haber and hacer...................................................page 71 5.7.1. Methodology....................................................................... .........................page 71 5.7.2. Results.......................................................................................... ................page 73 5.7.2.1. The replacement of haber by hacer..........................................................page 73 5.7.2.2. Durative and punctual meanings..............................................................page 74 2

5.7.2.3. Clausal and adverbial constructions.........................................................page 75 5.7.2.4. Word order of the past deictic phrase......................................................page 77 5.7.2.5. Time adjuncts............................................................................................page 78 5.7.2.6. Adpositional uses......................................................................................page 79 5.7.2.7. Negation............................................. .......................................................page 79 5.7.2.8. Preverbal vs postverbal adverbial construction........................................page 80 5.7.3. Conclusion....................................................................................................page 82 6. Conclusion..........................................................................................................page 82 Annex 1. TDMs' word order, their polysemy vs. monosemy and the preferred word order of adpositions for the sample languages....................page 85 Annex 2. Examples of PDPs and FDPs.................................. ...................................page 87 Annex 3. Results of the native speaker competence questionnaire......................page 89 Annex 4. Synchronic usage data for hacer+time. From CREA, oral, Spain.............page 90 Annex 5. Diachronic usage data for hacer/haber+time. From CORDE, Spain........page 90 References................................................................. ............................................. page 94

3

ABBREVIATIONS

1: First person 2: Second person 3: Third person ABL: Ablative ACC: Acusative AUX: Auxiliary verb CoCl: Complement clause COMP: Complementizer DAT: Dative EKC: Euskal Klasikoen Corpusa ERG: Ergative EVID: Evidential FDG: Functional Discourse Grammar FDM: Future deictic marker FDP: Future deictic phrase F: Feminine gender FUT: Future GEN: Genitive IE: Indo-European IMP: Imperative IPF: Imperfect IPFV: Imperfective INESS: Inessive LOC: Locative ME: Middle English N: Noun

NEG: Negation NOM: Nominative NP: Noun phrase OV: Object-verb word order PART: Partitive PST: Past tense PDM: Past deictic marker PDP: Past deictic phrase PL: Plural PLUP: Pluperfect PP: Prepositional phrase PRS: Present tense PREST: Presentative RAE: Real Academia Española REFL: Reflexive RES: Resultative SG: Singular SOV: Subject-object-verb word order SBJV: Subjunctive SV: Subject-verb word order SVO: Subject-verb-object word order TAM: Tense, aspect, mood TDM: Time deictic marker TDP: Time deictic phrase VO: Verb-object word order VP: Verbal phrase

4

THE DIACHRONY OF TIME DEICTIC ADVERBIALS: SPANISH HACER+TIME IN A CROSS-LINGUISTIC PERSPECTIVE 1. SUMMARY Time is one of the most salient aspects of human experience and cognition. The coding of time in language is, therefore, a central component of grammar and has received considerable attention from the linguistic community. However, most of those efforts have been centred on the verbal domain (tense or aspect) while others, such as time adverbials, have been paid much less attention. This is, as Klein (2009:41) mentions, unfortunate. On the one hand because many languages lack tense and aspect inflectional morphology altogether. On the other hand, because, even in those languages which do have it, tense or aspect play only a secondary role in the expression of complex time relations when compared to adverbials. The present study will attempt to increase our knowledge of this comparatively more neglected domain. Expressions such as ago in English or hace in Spanish are used for marking past time adverbials like two years ago or hace dos años respectively. They constitute the head of their phrase, which also includes a time expression which specifies the time elapsed from some moment in the past until the present. The phrase as a whole is used to specify location at a specific point in the past: (1) John came [two years [ago]] Words like Eng. ago or Sp. hace have a transparent lexical origin, in the verbs go and hacer (make) respectively, whereas synchronically, their categorial status is more controversial. They tend to be remarkably idiosyncratic expressions within their languages. Ago, for instance, is arguably the only postposition of English and its Italian equivalent fa shares this same uniqueness in its language. These expressions contrast prominently with their mirror images for the future, English in and Spanish dentro de, for which a lexical/semantic origin is not recoverable and whose synchronic categorial status is much less problematic. They are most usually adpositions and tend to 'keep a low profile' in their respective languages. This pronounced asymmetry between past and future time adverbials can be explained by the cognitive and experiential difference between past and future and by different sources in the diachronic grammaticalization processes leading to those expressions. The first part of the paper is typologically oriented. Given the difficulty of trying to determine the structural characteristics or word category of the expressions in languages to which the author has only second-hand access only the readily accessible factor of word order has been the focus of the present typological investigation. With the help of a genetically and geographically varied sample of sixty languages I have verified and quantified on a cross-linguistic basis the a priori noticed asymmetry between time adverbials for the past and for the future. In the case of the more 5

remarkable past time adverbials, those expressions which are exclusively dedicated to that semantic function are shown to be the ones which actually 'stand out' in their respective languages whereas the others are actually quite unremarkable. This is hypothesized to be the synchronic result of two different grammaticalization paths: a primary grammaticalization path leading to the emergence of the expressions under study from temporal biclausal constructions and a secondary grammaticalization path having as a source adverbial expressions for other time relations. Having provided a general typological introduction to the expressions and to the grammaticalization processes whereby they arise it is time to obtain a more exhaustive understanding of the synchrony and diachrony of these elements by analysing one expression in depth. The case of Spanish hace is surveyed here. The expression is an attractive research target on the one hand synchronically. The temporal use of hace appears in two different constructions. We distinguish a so-called (Real Academia Española, 2009) clausal construction (2), where hace constitutes the main verb of the sentence and cannot be freely omitted, from an adverbial construction (3), where hace heads a syntactically optional phrase: (2) Hace dos días que vino makes two days that (he) came

'It has been two days since he came'

(3) Vino hace dos días (he) came makes two days

'He came two days ago'

Contrary to most previous analysis (e.g. García Fernández, 1999; Real Academia Española, 2009), the acceptability of the adverbial construction with non-present variants of hacer, with time adjuncts or with negation is far from clear. Some uses of the expression, such as the possibility of the adverbial construction to describe a situation extending until the present, are also subject to great variation. For this reason, a questionnaire has been distributed among native speakers to assess objectively the synchronic status of those alleged properties. That investigation of competence is complemented by a small corpus research on oral synchronic usage of the expressions. The synchronic properties of hacer+time are found to be quite different from what most previous descriptions of the language present and are subject to a great variability between different speakers and age groups. After verifying the many grammatical differences between the two constructions, I have reached the conclusion that, at least synchronically, it is more parsimonious to regard them as independent from one another, even if still admitting that the semantic and formal similarities warrant analogical connections between the two. On the other hand, hace is also diachronically an interesting expression. It is, chronologically speaking, quite a recent strategy, since it was not until the 19 th century that hacer ousted haber as the preferred, default option for locating a situation in the past. This change of strategy as well as the diachronic developments of the previous construction with haber have been investigated in-depth through a corpus study in CORDE. The characteristics of relevant expressions with haber and hacer have been analyzed for five different periods and the quantitatively observable diachronic developments are explained by establishing links to grammaticalization theory and to other related diachronic developments. After investigating the relevant parameters of the constructions, the results point towards a grammaticalization of the adverbial construction with ha and hace and towards a divergence of the adverbial and the clausal constructions, phenomena which might be still ongoing. 6

2. BACKGROUND In the previous section I have been necessarily crude and brief in order to provide a short, comprehensive survey of the present study. I have presented the particular class of expressions which will be the focus of the present study without commenting on their properties or placing them in their context next to other adverbial expressions. I have advocated for cognitive differences as an explanation for the data without providing at least a general picture of how time is conceptualised in language. I have been talking about asymmetry in grammar or about grammaticalization without explicitly defining how they are to be integrated in the present study or how the terms will be used in further discussion. A few preliminary digressions are, therefore, in order and will help clarify subsequent argumentation.

2.1. Conceptualization of time There are enormous cultural differences between human societies. The everyday experience and preoccupations of some of them have at times so little in common that it is surprising to see that their conceptualization of time is so similar, especially with time being such an abstract and elusive concept. Leaving superficial differences aside (such as that found in the conception of time as linear or cyclical), all human cultures and languages are concerned with time to begin with (i.e. they allow people to locate events in time in complex ways). They all think of time as segmentable and conceptualise it in deictic terms1, that is, they all have some sort of zero-point in time which coincides with the moment of present experience or the utterance time. This deictic centre, which is usually called origo since Bühler, is logically constantly moving in time and is quite difficult to delimit theoretically but has the crucial role in language of separating the past from the future. This fundamental division which allows speakers to specify whether an event has taken place already, is taking place now or has not taken place yet is, arguably, indispensable in the everyday life of every individual, which I take as the raison d'être of the universal character of this origo in time. Events may be located, however, not only with respect to the origo but also with respect to other events; they may happen before, later or simultaneously, they may be specified for duration etc. Another universal property of time in language that has been the object of investigation of many volumes (Tenbrink, 2007; Haspelmath, 1997 among others) is its connection to space. As explained, for example, by Haspelmath (1997:1) space and time are amongst the most important domains of human cognition. They are independent in that they are not part of a higher, more basic domain and in that none 1 As showed in Lenz (2003:vii) deixis is crucially involved in language, not only in time, but also in our conceptualizations of space and person. As he states “demonstratives, personal pronouns, tenses, certain place and time adverbials, some verbs such as come and go cannot be described without recourse to aspects of the communication act”. Similarly, Dixon (2010:114) mentions that “[a] speech act involves participants (speaker and addressee) in a place, at a time. All languages have sets of 'shifters' whose reference shifts when the role of the participants change, when the place changes or when the time changes”. Thus, for instance 'today', 'yesterday' or 'five days ago' could all refer to the same day and to an infinite number of different days depending on the day when they are uttered. These expressions therefore have a deictic component in their semantics.

7

of them can be reduced to the other. Much in the same way as other still more concrete dimensions such as body parts or landmarks are frequent metaphorical sources of spatial vocabulary (see e.g. Heine, 1997:58), the spatial domain acts as a very useful metaphorical source for the expression of the more abstract dimension of time. As Langacker (1987:148) puts it: “The experience of time clearly suggests itself as a primitive dimension of cognitive representation. The fact that we often conceive and speak of time in spatial terms only shows the utility of such metaphor for higher-level conceptualization. It does not imply that the experience of time is reducible to a purely spatial one” Space, however, is 3-dimensional whereas time is one-dimensional, which means that, when time is presented in spatial terms, one of the three axes must be chosen. For this purpose, the front-back axis is preferred in the vast majority of cases 2 because the spatial metaphor for time involves not only location, but also movement through space, since, as acknowledged by Moore (2006), there is a clear experiential connection between time and motion since the latter involves the former. In addition, motion through space is most often done with the goal (which one will reach in the future) at the front and the source (where one was in the past) at the back. We are biologically endowed in such a way that our motion is naturally oriented in the same direction as our sight. So entrenched is this habit of having the goal in front that, even when it is unnecessary, most still prefer to look in this direction. Note that we even build our means of transport (the direction of the seats) with that premise in mind:

Figure 1 It is therefore not surprising that we find expressions like Sp. el mes que viene (‘next month’ lit: the month that comes) or we are headed for summer (Moore, 2006:199) which exemplify what are commonly recognised as the ‘moving time’ and the ‘moving ego’ metaphors respectively. Regardless of these space-based metaphors, and as presented in Klein (2009) (I will be closely following Klein’s basic time structure in this section) time relations hold between two time spans which he calls temporal relata. The most prominent of them is the theme and refers to the time of the event whose location in the temporal axis is specified. The relatum is the background or reference time for the location of the theme. Let me provide an example: (4) Peter will pass his exam

-Time of Peter’s passing his exam: theme -Origo/utterance time: relatum

The theme is here located after the relatum; that is, the time of Peter’s passing his exam follows the utterance time. The relatum may be deictic (derived from the origo or utterance time) but also anaphoric (mentioned in the preceding context) or calendaric (derived from some event in history). Let’s look at another, somewhat more 2 A famous exception is, as mentioned by Haspelmath (1997:22) the use in Chinese of ‘up’ and ‘down’ for temporal ‘last’ and ‘next’ respectively. E.g. shàng (up), shàngnián (last year)

8

complex example which will illustrate the time relations holding in the expressions which will be analysed in the present work: (5) I visited New York several years ago Here, the theme or event time (E) (i.e. my visiting New York) is located before the reference time or relatum (R). The reference time here coincides with the utterance or speech time (S). The two time spans, theme and relatum are separated here by an intervening time distance (D) of “several years”. This can be represented schematically as follows:

Figure 2 With only two time spans involved, relations are easy to systematize, but it is necessary to mention that matters are not always so simple. Take, for example, the following sentence: (6) Mike lost his passport after missing his flight three hours earlier It seems that the time of Mike’s losing his passport is the one which is being located in time here (i.e. it is the theme), with respect to both utterance time (relatum) and the time of his missing his flight (another relatum?). It is located before the first but after the second. However, the time of Mike’s missing his flight is also being located in time (another theme?) with respect to both utterance time (relatum to two themes?) and with respect to the time of Mike’s losing his passport (another relatum?) which seemed to be the theme before. The analysis of complex time relations holding between several events exceeds the scope of the present study. Sentence (6) is only meant to remind us that, for theoretical purposes, we are here only dealing with a convenient, simplified model for temporal relations.

2.2. The object of study in context The temporal relations described in the previous section are linguistically encoded in many different ways, both in the verb and elsewhere in the clause. Within the verb we have tense (which locates the state of affairs in the time axis in relation to the origo or utterance time), grammatical aspect (which presents the event from a certain viewpoint; as ongoing, completed…) and lexical aspect (which refers to the different temporal properties of different verbs by virtue of their describing different situations; activities, achievements…). Tense and aspect have been extensively studied in the linguistic literature and, even if, of course, a lot remains to be done, it seems safe to say that a substantial degree of knowledge has been reached concerning the timerelated aspects of the verb. Time is, however, also encoded (one could even say “mainly encoded”) outside of the verb, were it has puzzlingly received much less attention. This is done most extensively and universally by means of time adverbials, but also by means of temporal particles or discourse strategies (basically the iconic disposition of clausal or sentential constituents). In addition, all the strategies mentioned so far are ways of presenting 9

the temporal information of the main event (i.e. they are means of presenting time as an accompaniment to a main course, so to speak). Time, however, can also be the central element of an utterance (it has dedicated verbs, can occupy argument positions…) as illustrated by sentences like “he spent two years in prison” or “time goes by so slowly”, a fact which is frequently forgotten3. Temporal adverbials are, as has been mentioned, the most important way to encode time in language. They are generally characterized (Van der Auwera, 1998) as syntactically optional elements which function mainly as modifiers of non-nominal constituents. They comprise extremely varied expressions in both their morphosyntactic composition and semantics. Following again Klein’s (2009:65) insights, temporal expressions may, according to their composition, be classified into morphologically simple (Eng. now, often…), morphologically complex (Eng. currently, later…) and syntactically complex expressions which are, in turn, very diverse. These may be not only adverbial phrases, but also adpositional phrases (e.g. in a week) or bare noun phrases (e.g. summer 1944), which should be included in this list, according to Nichols (1992) by virtue of their semantic and syntactic (optionality, position…) similarities with adverbs. Regarding their semantics, time adverbials can be of duration, of frequency, of contrast or of position, which provide information concerning the relation between the two time spans (theme and relatum) which were introduced in the preceding section. In the introductory summary, I already presented expressions such as English ago or Spanish hace. These elements, along with the phrases they introduce, will be the object of study of the present paper. The phrases headed by these expressions can now be easily described and classified with recourse to the theoretical background presented in this section and the previous one. They are a subclass of temporal adverbials. They are syntactically complex expressions (may be adverbial, adpositional phrases etc.) which provide the location in time of the main event or situation (theme) relative to the origo or utterance time (relatum). They do so by specifying, in the form of a time NP, the distance between the present and the time of the event. This may be either in the past (Eng. ago, Sp. hace, Basq. duela, Germ. vor…) or in the future (Eng. in, Sp. dentro de, Basq. barru, Germ. nach...). In order to delimit the object of study, I have included both semantic and formal considerations (the time distance to the utterance time must be specified by an NP). This is the same approach followed by Haspelmath (1997:6). As he mentions, a completely notional, semantic definition of our object of study would be impossible and thus formal restrictions are unavoidable. Given that these expressions are used to talk about time and given that they are deictic (time is counted from the origo/utterance time), I will refer to these expressions in general as Time Deictic Markers (TDM);4 to expressions such as Eng. ago or Sp. hace as Past Deictic Markers (PDM) and to Eng. in or Sp. dentro de as Future Deictic Markers (FDM) independently 3 Klein (2009) for example, however exhaustive in other respects, does not mention it among the different ways in which time information can be presented in language. 4 These expressions have been labeled in different ways in previous literature; among others, they have been called “temporal deictic expressions” by Kurzon (2008), “temporal distance markers” by Haspelmath (1997), either for distance-past or for distance-future, and the semantic contribution of the markers has been labeled “deictic scalar localization” by Bourdin (2011)

10

of their categorial status. The phrases these expressions introduce will be referred to as Time Deictic Phrases (TDP) or, when more specificity is required, as Past Deictic Phrase (PDP) or as Future Deictic Phrase (FDP).

2.3. Asymmetry The asymmetries between PDM and FDM are a central focus of the present paper. Despite being a frequently used term in Linguistics, however, and even the topic of whole volumes (Di Sciullo, 2003), the term ‘asymmetry’ is hardly ever explicitly defined. Moreover, there exist different interpretations of the term, which is seldom noticed in the literature. In Haspelmath (2008:185) for example, asymmetry is defined as a difference in grammatical coding that does not express a difference in meaning. He goes on to mention the example of the contrast between 'the book' and '(*the) my book' where, as Haspelmath mentions, the definite article is impossible although it would be expected from the inherently definite character of the involved NP. This is, however, not the kind of asymmetry I refer to here. I am here concerned with asymmetries that hold between different meanings. I propose the following definition of asymmetry in grammar: Asymmetry: The different treatment (regarding syntax, morphology, diachronic origin or any other respect) of two dimensions which are complementary in the relevant domain or mirror images of one another. Positive vs. negative polarity is a clear example of two such dimensions. In most cases, this opposition is handled symmetrically in grammar; that is, the negative polarity version of a verb differs from the corresponding positive form in a systematic way, most usually by the insertion in the negative of some linguistic item which is not present in the positive form. These examples from English illustrate the point: (Positive polarity)

(Negative polarity)

(7) I come ________________________ I don’t come

(Present)

(8) Do your homework ____ Don’t do your homework

(Imperative)

Other languages, however, don’t always behave in the same way. Spanish, for example, which otherwise negates a verb simply by including a negative adverb no (9) requires to use a subjunctive to form the negative version of the imperative (10), which constitutes then an example of a morphological asymmetry between the positive and negative forms of the imperative (see e.g. Miestamo, 2005:52): (Positive polarity)

(Negative polarity)

(9) Vengo _______________________ No vengo (Present) come.1SG.PRS

NEG come.1SG.PRS

‘I come’ (10) Haz

‘I don’t come’ los deberes _______ No

Do.IMP.2SG the homework

‘Do your homework’

hagas

los deberes (Imperative)

NEG do.SBJV.2SG the homework

‘Don’t do your homework’

Polarity, being already a black-or-white parameter, constitutes a dimension were symmetry or asymmetry are readily observable. Other more complex (non11

parametrical) dimensions, however, are also frequently coded in grammar along a number of binary oppositions, which allows for symmetrical or asymmetrical relations between both. Space, for instance, is a non-discrete dimension, but is conceptualised in terms of binary oppositions, as shown by spatial prepositions or adverbs’ arrangement into couples (up and down, in front of and behind, in and out, here and there…). Proximity vs. distance to the speaker, for example, as coded in English demonstratives appears to be quite a symmetrical division. Both demonstratives appear in the same syntactic position, have a plural form that adjectives and articles lack and even have comparable phonological material in the singular and in the plural: (11) This cat ____ These cats

(Proximity)

(12) That cat ____ Those cats

(Distance) Figure 3

The one-dimensional axis distance-to-speaker is in English arbitrarily divided into two complementary dimensions which are handled by English grammar in the same way (i.e. symmetrically). Time has properties very similar to distance. As mentioned by Langacker (1987:150) time is also a one-dimensional, unbounded, configurational 5 domain. The axis of time is divided, not in a language-specific, arbitrary manner as we saw was the case of the distance-to-speaker axis, but in a more substantial and experience-grounded way. Our conception of the world involves a conception of time divided into two complementary dimensions (past and future), separated by the present (the deictic origo we introduced in section 2.1.). Figure 4 Despite being arguably mirror-images of one another, past and future are seldom treated in grammar in the same way6. Only to mention a few grammatical asymmetries between past and future, the future tends to be expressed in the verb analytically more often than the past7: (13) I started yesterday

(inflectional past)

(14) I will start tomorrow (analytical future) In addition, as mentioned by Hengeveld & Mackenzie (2008:24) or by Comrie (1985:44), in many languages with a 3-way tense distinction, unmarked, present forms may often be used with a future but not with a past meaning:

5 In Langacker (1987:153) this is defined as the quality of a certain domain whereby they include “an indefinite number of separate points at any of which essentially the same range of experiences can occur”. 6 I disagree here with Haspelmath (1997:24) who states that “there are often asymmetries between past and future in linguistic expressions, but on the whole it is surprising to what a high degree past and future temporal expressions are symmetrical”. 7 According to WALS, only 39,6% of the languages in their sample lacked an inflectional past whereas more than 50% of them lacked an inflectional future.

12

(15) *Empiezo ayer start.1SG.PRS yesterday

‘*I start yesterday’

(16) Empiezo mañana start.1SG.PRS tomorrow

‘I start tomorrow’

Mentioned as well by Hengeveld & Mackenzie (2008:164) and by Comrie (1985:44), when languages have a two-way tense distinction, this is more often past vs. nonpast rather than future vs. nonfuture. Note, however, that both divisions are already asymmetrical to some extent, since in the deictic sense a present vs. nonpresent division would be the most symmetrical but is, in reality, very infrequent, for obvious practical reasons. As Comrie (1985:15) mentions, lexical items used to refer to any time excluding the present are frequent in languages (like English then); however, the grammaticalization of not-now as a single tense appears not no occur, and this despite the frequent grammaticalization of now as a marker for the present tense. The purported differences between the diachronic and synchronic behaviour of PDMs and FDMs would also constitute a relevant tense-based asymmetry at this point. I leave the in-depth discussion of the phenomenon to section 3.

2.4. Grammaticalization: The diachronic phenomenon of grammaticalization has received so much attention and the term for it has been in use for so long now, that it hardly seems necessary to mention here its coiner Antoine Meillet or the classical definition by Jerzy Kuryłowicz. Grammaticalization has since been well studied, systematised and delimited theoretically (Lehmann, 1995) and a great number of grammaticalization paths has been discovered and analysed (Heine and Kuteva, 2002). More concrete problems such as the question of whether grammaticalization is always unidirectional or whether it can be reduced to other phenomena is beyond the scope of this study. I do find it, however, necessary to provide here a definition which allows us to bridge over possible different uses of the term which could obscure further argumentation. A standard notion of grammaticalization would be the following: Grammaticalization: The process whereby lexical items or constructions become grammatical (i.e. a part of the grammar) or whereby a grammatical item becomes more grammatical. (Based on Lehmann (1995:11)) Sometimes 'primary grammaticalization' is used to denote exclusively the initial grammaticalization of an erstwhile lexical element whereas the term ‘secondary grammaticalization’ is used to refer to the further grammaticalization of an already grammatical element or its spread to other grammatical meanings. I will be using these terms as well whenever I need to make those distinctions. Having a concise working definition may be necessary; however, it cannot replace a more detailed insight. Grammaticalization is here a diachronic phenomenon. It is, under the previous definition, a change in the degree to which a certain linguistic item belongs in the grammar, specifically a change from less to more grammatical. Every change is, however, a transition between two states, so grammaticalization also has a 13

synchronic dimension. It may refer to the degree to which a certain linguistic item is integrated in the grammar (i.e. is more or less autonomous or dependent). In order to be able to discuss both diachronic and synchronic grammaticalization rigorously, we need to have a way of measuring the degree of grammatical character of a certain linguistic sign. This will have to be a measure of its autonomy. I will here follow Lehmann’s (2004:447) parameters of synchronic grammaticalization: paradigmatic syntagmatic

Weight Cohesion integrity paradigmaticity structural scope bondedness Table 1

Variability paradigmatic variab. syntagmatic variab.

Integrity: semantic and phonological substance of the sign. Paradigmaticity: cohesion of a sign in a paradigm. Paradigmatic variability: degree to which a sign is obligatory. Structural scope: the size of the construction relevant to the linguistic sign. Bondedness: cohesion of the sign with others (i.e. segmentability). Syntagmatic variability: freedom of the sign regarding its position. As a linguistic sign is grammaticalized, it loses in integrity (phonological attrition, desemanticization), increases in paradigmaticity (paradigmaticization), loses paradigmatic variability (obligatorification), decreases in structural scope 8 (condensation), increases in bondedness (coalescence) and its syntagmatic variability is reduced (fixation). A systematization of the different parameters that determine the degree of grammaticalization of an expression is a welcome step towards measuring it objectively. However, most of us, at hearing the word 'measure' think of numbers, so Lehmann's classic theoretical insights fall short, to some extent, of our expectations. Having a formula which would allow us to reach a concrete number would enable us to automatically compare the degree of grammaticalization of items in different languages. Indeed, some attempts have been already made in that direction albeit in restricted, relatively uniform domains. Bybee (1994:106) for example, follows a quantitative approach to grammaticalization in which she arrives at precise numbers for the phonetic reduction or loss of autonomy of a morpheme. This system, however, increases objectivity and the possibilities of comparison only apparently, since the criteria used to arrive to the number in question (i.e. relevant features and their relative weight) are inevitably largely arbitrary. Bybee's system cannot be used either to measure the grammaticalization of constructions or elements much above the morpheme. Lehmann's classic view of grammaticalization has received other more substantial criticisms. Boye and Harder (2012) for example, mention that bondedness, phonological and semantic reduction, closed class membership... which were taken to be parameters of grammaticalization by Lehmann, do frequently go hand-in-hand with 8 This has been challenged by some authors. Tabor and Traugott (1998) for example suggest that, quite on the contrary, it may be an increase in structural scope that might be involved in grammaticalization.

14

it. However, none of them is decisive, as unquestionable cases of grammaticalization do occur which spare each of them. They may be therefore epiphenomenal and thus cannot be used to define what grammaticalization actually is. Boye & Harder (2012:2) note that “[g]rammar is constituted by expressions that by linguistic convention are ancillary and as such discursively secondary in relation to other expressions”. They go on (page 12) therefore to propose the following definition for grammaticalization: “Grammaticalization is the diachronic change that gives rise to linguistic expressions that are by convention ancillary and as such discursively secondary” It becomes immediately apparent that, whereas the classic definition and theoretical delimitation of Lehmann implied a continuum between the lexical and the grammatical, quite the opposite is the case under Boye and Harder's definition of the phenomenon. A given linguistic item can be either discursively secondary by convention or not. It has to be noted that this sharp distinction between the lexical and the grammatical need not extend to other areas of language. Some theoretical models such as Functional Discourse Grammar (Hengeveld & Mackenzie, 2008:9) which otherwise argue in favour of a gradiency between the different grammatical categories, propose a synchronically sharp distinction between lexicon and grammar. In this vein, Boye and Harder (2012) propose that, given its pragmatic function, focus9 can be used as a test to diagnose whether a certain linguistic expression is grammaticalized (i.e. is part of the grammar) or not. Just as an example of Boye and Harder's test, there is in Spanish an infinitival construction like (17) below which can be used to focalize verbs. When the same construction is applied to grammatical (i.e. auxiliary) verbs (18) and (19), it results in an ungrammatical sentence, which shows that are grammatical, not lexical elements: (17) Talar

talará

el árbol, pero le va a costar horas

chop-down chop-down.FUT.3SG the tree, but to.him will.cost

hours

'He will eventually chop down the tree, but it will take him ages' (18) *Haber ha comprado algo, pero nada bueno have has

bought something but nothing good

'He has bought something but nothing good' (19) *Ir

voy

a marcharme a las cuatro, pero después de comer no contéis conmigo

go I.am.going to

leave

at

four

but

after

of

eat don't count

with.me

'I am going to leave at four o'clock but after lunch count me out' Boye and Harder also propose addressability as another test for determining the status of a certain linguistic expression since, according to them, grammatical expressions cannot be addressed in subsequent discourse. In the present work I will follow Boye and Harder's definition of grammaticalization in synchrony but without losing sight of Lehmann's classic parameters as an instrument to analyse diachronic changes in more detail. 9 Some kinds of foci such as so-called 'contrastive focus' may actually apply to grammatical elements. The linguist has to be aware of this in order to avoid them if focus is to be successfully used diagnostically like Boye & Harder (2012) propose. The distinction between different kinds of foci exceeds the scope of the present work and will not be discussed here.

15

2.5. Word categorization The process of grammaticalization has traditionally been thought of as a gradual phenomenon; as a progressive migration from lexical, open word classes to functional, closed categories. This is, however, quite problematic under some ways of understanding word categorization. As argued by Taylor (2003), Aristotelian principles of word categorization advocate for categories with clear-cut boundaries, with only two degrees of membership (either member or non-member). This position is endorsed nowadays for example within the generative theoretical tradition (whether explicitly like in Baker (2003) or implicitly). Aristotelian word categorization, however, makes it difficult to understand grammaticalization as a gradual process or to explain the existence of diachronic phenomena such as transcategorizations (i.e. the shift in the categorial membership of a linguistic item without any surface marking). Other traditions and authors (such as Cognitive Grammar or Taylor, 2003) advocate for a categorization by prototype or at least for 'fuzzy' grammatical categories (e.g. Functional Grammar; Haspelmath, 2000...). These notions imply more diffuse borders among the different word classes and, in the case of categorization by prototype, a membership based on the number of features shared with the most prototypical elements of the category. This helps, on the one hand, explain the idiosyncratic character of many linguistic items (which under this hypothesis would be peripheral members of its category) and on the other hand it would help account for the gradual implementation of grammaticalization or transcategorization processes. There are yet other ways of conceiving word categorization, however. For example, one could think of the syntactic combinatorial properties of especially closed class words as analogous to the patterns of morphological variation of verbs across different conjugations or nouns across different declensions. It is clear that speakers of a language could not afford to have a different declension for each noun. It seems analogously logical that grammatical, closed-class words cannot have each its own, unique syntactic behaviour. Such a feature is not captured by the prototype model, which rather on the contrary, leads to posit an almost infinite variation. Much in the same way as nouns in some inflectional languages are declined according to different declensions, grammatical words may also be hypothesised to form smaller, internally homogeneous groups. This does not of course preclude the possibility of having words which are also unique in their syntactic behaviour, much in the same way as inflecting languages sometimes allow nouns to have an irregular, unique declension of their own (e.g. Germ. Herz 'heart' or Rus. put’ 'path'). If this can happen with sparsely used words such as these, it should not be surprising to find closed-class, much more frequent words with a unique syntactic behaviour (e.g. English ago, Spanish hace, Italian fa…) which would strictly qualify as constituting word classes with a single member. As an example of the kind of categorization I am advocating for here (i.e. by prototype but with different subclasses)10 I present below an analysis of some linguistic expressions in Spanish based on the number of properties typical of adpositions which they display.

10 This way to conceive categorization is not very different from that proposed e.g. by Bybee (2010:18).

16

Some characteristics typical of prepositions in Spanish are the following: a) Phonetically unstressed b) Govern oblique form of pronouns (mi, ti...) c) Can take NPs as their complement d) Cannot take a bare, finite verb clause as a complement e) Invariable, they don't show inflection f) Must have an overt complement g) The constituent they head is always optional11 h) They are morphologically simple Some expressions and their properties: de: a, b, c, d, e, f, g, h (8) sin: a, b, c, d, e, f, g, h (8) para: a, b, c, d, e, f, g, h (8) en: a, b, c, d, e, f, g, h (8) a: a, b, c, d, e, f, h (7) tras: a, c, d, e, f, g, h (7) bajo: a, c, d, e, f, g, h (7) durante: a, c, d, e, f, g, h (7) salvo: a, c, d, e, f, g, h (7) ante: a, c, d, e, f, g, h (7) dentro de: b, c, d, e, f, g (6) encima de: b, c, d, e, f, g (6) como: a, c, e, f, g, h (6) detrás de: b, c, d, e, f, g (6) incluso: a, c, e, f, g, h (6) excepto: a, c, e, f, g, h (6) hace12: c, d, f, g (4) según: c, e, g, h (4) tiene: c, d, f (3) ayer: e, g, h (3) As can be seen, there are four prototypical prepositions showing all the relevant properties. As much as twelve others are lacking only one or two of them, but, unlike it could be expected if variation was random, they all pattern into just four groups regarding the characteristics that have been considered here. Three of the characteristics (c, e, f) are common to all of them and suggest themselves as more prototypical prepositional characteristics than the others:

Figure 5 11 I exclude here idiomatic uses where a complement with a specific preposition is required by a verb. 12 In the temporal adverbial construction.

17

The inclusion of ayer (yesterday) and tiene (has) is just aimed at showing how, even uncontroversial members of other grammatical categories share with adpositions some of their more typical characteristics. Adpositions are one of the most interesting grammatical categories because of their intermediate position in this respect. This is noted for example by Hagège (2010:333) who also, unsurprisingly, included expressions like French il y a in his monograph about adpositions despite it having many verbal characteristics as well. Adpositions share properties with other grammatical categories (e.g. verbs and adverbs) and occupy an intermediate position between lexical and grammatical word classes. It is thus not difficult to imagine how verbs and adverbs can easily become more adposition-like (or eventually even prototypical adpositions) by progressively acquiring more of their properties or losing some of the properties characteristic of their former word classes. A gradience between verbs and adpositions, for instance, has long been recognized. Haspelmath (1998:330) for example, shows a scale with clearly deverbal adpositionlike elements like Eng. according to, during, or past. These have all in common that, being impersonal verb forms, they are, from the onset, free of important verbal features like TAM morphology or the requirement of a subject. This is why these forms (i.e. participles, gerunds...) become adpositional quite frequently, but even finite verb forms can become more adpositional. This is what has happened with a verb like the Spanish hacer of time phrases we will be concerned with in this paper. As an impersonal verb it has lost its ability to take a subject, which is probably the single most important property defining verbs. In addition, it is the head of an optional constituent, which is also characteristic of adpositions and its phrase may be the complement of a preposition (desde hace días, de hace tiempo...) which is very unusual for a finite verb. It is also in the process of getting rid of TAM morphology, negation and other verbal features, which will be discussed in depth later on in section 4. What is interesting at this point is that, as we can see graphically in the diagram below, both verbs and adverbs (and one could even say all major word classes) may progressively develop into adpositions in small steps:

Figure 6 We have seen (Figure 5) that expressions having the same number of adpositional properties tend to cluster to form internally homogeneous groups, but connections can also be found between groups having a different number of adpositional properties. We find, for example, that the properties of hace are a subset of the ones 18

found in the group with dentro de and encima de. Similarly, the adpositional properties of según also constitute a subset of the ones of como, incluso, excepto, which in turn constitute a subset of those of the group with tras, durante, ante, salvo. This makes it still easier to see how more peripheral members could be well integrated into bigger, more prototypical groups when they advance further along the grammaticalization cline. All this internal structure we have been presenting is incorporated in the prototype representation below:

Figure 7 To close this section I would like to point out that which characteristics are typical of adpositions is subject to discussion and to cross-linguistic variation so categorization by prototype is not without its problems. Attempting to classify all closed-class words under the traditional labels ‘preposition’, 'conjunction', ‘adverb’ etc. could thus be both fruitless and arbitrary to a great extent. Some authors (Croft, 2001:45) go as far as to claim that “there are no atomic grammatical primitives”. In his opinion, crosslinguistic comparison does not allow us to identify the right distributional criteria which should be used to determine membership to a word class. A similar thought is expressed by many other authors. Dixon (2010:102) for example mentions that “word classes must be recognized for each language on grammatical criteria internal to that language”. If necessary and sufficient conditions for membership are established ad hoc, the whole argument around membership of a linguistic item to a word class could be circular.

19

3. TIME DEICTIC EXPRESSIONS 3.1. Previous analyses As was presented in 2.2, Time Deictic Phrases (TDP) are a subclass of time adverbials which provide the location in time of an event (theme) relative to the utterance time (relatum) by specifying, in the form of an NP, the temporal interval that separates both. Although these expressions have not been an exceedingly popular topic for linguistic research, they have not passed totally unnoticed either. They constitute a subset of the NP-based time adverbials studied for example by Haspelmath (1997). The synchronically peculiar properties of many Past Deictic Markers (PDM) have recently attracted scholarly attention as well (Kurzon, 2008; Bourdin 2011). Problems posed by PDMs in specific languages, both diachronic (Franco, 2012) and synchronic (Rigau, 1999; Fábregas, to appear) have also been addressed. Haspelmath (1997) and Franco (2013) also notice the frequent formal identity of TDMs and other time expressions and of TDMs and spatial expressions. This will be elaborated later on. The asymmetry between PDMs and FDMs I presented briefly in the introductory section has also been hinted at before. When considering specifically some of the sources of TDMs, Haspelmath (1997:86) already noticed a “surprising lack of symmetry” between those for the past and the future. Haspelmath presents some of the sources he identified for TDMs. For PDMs he mentions verbs like pass or exist and adverbs like back. For FDMs common sources involve spatial inclusion (within) or movement (across) or the adverb yet. The diachronic, as well as the synchronic differences frequently found between past and future deictic markers and between the phrases they introduce, however, have besides this received little attention in the literature. The fact that those expressions have chased away more than attracted linguistic research until quite recently demands certain attention. On the one hand, the neglect of these expressions has to be placed within the more general neglect of adverbials. As has been previously mentioned, adverbials and adverbs continue to be, if their crucial role in language is taken into account, a very under-studied domain in language. This can probably be explained among others by the sheer complexity of adverbials on the first place. Like we briefly put forth in 2.2, adverbials are very diverse expressions regarding their composition (both morphological and syntactic), their word categorization (may be adpositional phrases, adverbial phrases, subordinate clauses, maybe even verbal phrases…), their syntactic position (which affects the semantic scope of the adverbial itself)… Adpositions and adverbs, the word categories which most frequently head adverbials, are by themselves controversial. The former, because of its intermediate position between what have been traditionally called lexical and grammatical categories. The latter, because of the difficulty of positing it as a category in the first place, since it has frequently behaved more as a catch-all category for hard-to-classify words. On the other hand, and talking now specifically about PDMs, the fact that they are such idiosyncratic expressions makes them a difficult research target. Many of these expressions, for example, are hard to classify on a categorial basis 13. Many share 13 e.g. As mentioned by Kurzon (2008) ago for example has been analysed in previous literature variously as an adverb, as a postposition and as a preposition.

20

features of adverbs and adpositions, others of adpositions and verbs… The synchronic properties of some of these expressions are so complex and variation among speakers and dialects so substantial at times, that a comprehensive picture of PDMs has so far eluded us. What is more, the synchronic properties of PDMs (let alone their diachronic development) are at times so idiosyncratic, that when a typological approach has been pursued, appreciation errors abound. Haspelmath (1997:87) for example, presents the Basque PDM duela as the main verb in a sentence like the following: (20) Du-ela

bi ordu hemen zen

Have.3SG-COMP two hours here was.3SG

‘Two hours ago he was here’

Its lack of morphological TAM variability, as well as the presence of the subordinating suffix -ela on the PDM, speak against this analysis, however, as does the fact that no marking of subordination appears in the other verb zen. In addition, the phrase introduced by duela, i.e. duela bi ordu, can be omitted freely without affecting the grammaticality of the sentence, which could have hardly been the case if it had been the main clause. Kurzon (2008) in turn, in his attempt to classify all the PDMs of the languages in his sample into the adverbial or adpositional classes, also falls victim to some errors in his analysis of several expressions. In page 224 for example, he classifies Italian fa as a preposition despite its postpositional status: (21) Gianni è partito due anni fa

(Franco, 2012:67)

Gianni is departed two years ago

‘Gianni left two years ago’

In a brief section dealing with the diachrony of ago, Kurzon presents us with the following example, which he interprets as an instance of the distance-past meaning: (22) O, he’s drunk, Sir Toby, an houre agone (1601,Twelfth Night) (Kurzon, 2008:213) He does not notice that its meaning is different here, were it has, as pointed out by Bourdin (2011:58), an up-to-now interpretation. These and some other shortcomings are the result of an inevitable lack of knowledge of the languages in question. Finnish sitten, for example, is claimed by Kurzon (2008:219) to be a postposition “which governs the partitive case (…) when the noun in the temporal NP is in the plural (…) while in cases when the singular is referred to, the nominative is used”. This would indeed be something unique in the language, since most postpositions govern partitive but, logically, both in the singular and in the plural (Karlsson, 1991:241). What seems to be happening in Kurzon’s examples is that it is the numeral which is governing the partitive case in (23) but not in (24). Note that after numerals other than one, the partitive singular is used In Finnish (Karlsson, 1991:110): (23) Kaksi päivää sitten Two day.PART.SG ago

‘Two days ago’ (24) Vuosi

sitten

Year.NOM.SG ago

‘A year ago’

21

In the same page, Kurzon also classifies Turkish PDM önce as a postposition, rather than as an adverb preceded by an extent phrase, which is Haspelmath's (1997:82) analysis of the expression. The problem in this case seems to be that Kurzon is mistaking here the location-in-the-past (marked in English by ago) with the anterior function (indicated in English by before), which in Turkish is also coded by the same expression önce, but with a different syntactic structure. The Turkish example he provides (25) does not actually mean 'one year ago' but rather 'before one year', so it should have remained outside his object of study. What he should have included is (26) below: (25) Bir yil-dan önce One year-ABL before

'Before one year' (26) Bir

yil

önce

One year.NOM ago

'One year ago' More important still than all these errors in his analysis of particular expressions is that, from the very beginning, he seems determined to classify every single PDM in his sample as either adpositional or adverbial. To do so he has to disregard “inconvenient” properties of the expressions in question. In page 216, for example, he mentions, as an argument in support of the synchronically prepositional status of the PDMs of Romance languages like Spanish hace or French il y a, that they are “invariable in form”. This, however, is not the situation in those two languages, since non-present forms of the expressions do appear, and this not only in the main sentence, clausal construction, but also in adjunct position: (27) No le

veía

hacía

not him see.IPF.1SG make.IPF

una semana a

week

‘I hadn’t seen him for a week’ (28) Il est monté avec moi à la cabane il y a eu huit jours dimanche dernier 14 he is climbed with me to the hut

ago.PST eight days

Sunday

last

‘Last Sunday is was eight days since he climbed with me to the hut’ In page 223, similarly, he notes briefly that the relevant expression in Bislama (English creole, Oceania) might be “even verbal” (everything actually points in that direction) but he goes on merely to classify the PDM as an adverb because, apparently, from the very beginning, he had only two drawers to make his taxonomy of PDMs. As we saw in 2.5, trying to determine a rigid grammatical category for every single word may not always be the most fruitful activity. Trying to do so with such idiosyncratic expressions such as PDMs and this in languages one is little acquainted with is bound to be a failure. However, since I consider necessary to provide in this paper a typological overview of TDMs, I have chosen to focus my typological research exclusively on the word order of the TDM with respect to the accompanying time NP. Analyzing exclusively the easily accessible and unmisinterpretable factor of surface word order in relation with other parameters will allow me to avoid errors like the 14 (Rasmussen, 1981:27) 22

ones noted before while, hopefully, still providing relevant information about the properties of the expressions cross-linguistically. As Haspelmath (2000:250-251) puts it, “[a] standard argument against broad cross-linguistic studies with their necessarily superficial treatment of individual languages is that in the absence of in-depth studies of these languages, one cannot be sure that the comparison is valid because similar surface structures might turn out to be quite different at a higher level of abstraction (…) But this view is naive. For one thing, forty years of Chomskyan linguistics should have taught us that it is unlikely that we will ever arrive at a consensus of what the correct formal analysis of a structure is”.

3.2. Towards a typological overview of TDPs and TDMs In the introductory section I presented briefly some of the characteristics which I found most interesting in these expressions. The differences between PDMs and FDMs I have only sketched up to this point seem to constitute an interesting past/future asymmetry. It involves syntactic differences in word order and morphological complexity such as the ones exemplified below: PDPs

FDPs

(29) I came (two years ago)

I will come (in two years)

(30) (Du-ela bi urte) etorri nintzen

(Bi urte barru) etorriko naiz

(Basque)

(31) Sono venuto (due anni fa)

Verrò (tra due anni)

(Italian)

(32) Ya prishel (dva goda nazad)

Ya pridu (cherez dva goda)

(Russian)

(33) Je suis venu (il y a deux ans)

Je vais venir (dans deux ans) (French)

The asymmetry also involves other differences. For example, while many PDMs are synchronically seen to be based upon a lexical, open class word (most usually a verb), FDMs are based upon grammatical, semantically unanalyzable elements: a-go < prefix-go du-ela < have.3SG-COMP fa < make.3SG na-zad < to-beginning il y a < he there have.3SG

in < in barru < inside tra < behind cherez < across dans < inside

All in all, as you can see, FDMs appear to be more grammaticalized than their pasttime equivalents and tend to be quite unremarkable within their respective languages. PDMs, on the contrary, appear to be often badly aligned with regards to the dominant word order of their respective languages and tend to exhibit more interesting properties, which is probably the reason why they have attracted more scientific interest than FDMs. It has to be stressed at this point that the asymmetries that I have presented for TDMs are not paralleled by other very similar time-related but non-deictic expressions. Thus, for example, we find no such asymmetry between markers of anteriority and posteriority: 23

(34) 2 years before

2 años antes (Spanish) ‘2 years before’

(Anteriority)

(35) 2 years later

2 años después (Spanish) ‘2 years later’

(Posteriority)

These are all adverbs specified by measure phrases and heading adverbial phrases in a little surprising syntactic position. This seems to suggest that it is the opposition of past and future meaning which is responsible for the earlier differences. I believe these asymmetries found in TDMs may have a diachronic origin and may ultimately be explained by cognitive factors. I very much agree here with Haspelmath (1997:24) and Comrie (1985:43-44) who comment that there exists a huge experiential and conceptual difference between the past and the future. In Comrie's words: “there is a sense in which the future is clearly different from the past. The past (…) is immutable, beyond the control of our present actions. The future, however, is necessarily more speculative, in that any prediction we make about the future might be changed by intervening actions, including our own conscious intervention. Thus, in a very real sense the past is more definite than the future”. As a result, the strategies that emerge in speech to talk about the more abstract future also make use of more abstract, more grammaticalized resources or rely more heavily in space-based metaphors whereas the past is more often expressed by lexical means. It will be interesting to observe whether the expressions which have time deictic, as well as non-deictic uses15 tend to pattern like deictics or like non-deictics. If the asymmetry holds only for the markers devoted exclusively to past or future uses, it would be interesting as well to compare past or future time devoted TDMs to other time-dedicated morphemes (such as the verbal morphology for tense) in search of similarities16. With the aim of seeing whether this asymmetry between PDMs and FDMs is a cross-linguistically valid generalization and to further analyze the affinities and properties of the expressions that concern us here, the following sample of 60 genetically and geographically diverse languages has been chosen: English, German, Swedish / French, Spanish, Italian, Romanian, Latin / Russian, Polish, Serbo-Croatian, Bulgarian / Hindi, Punjabi / Persian / Armenian / Albanian / Welsh, Irish / Lithuanian, Latvian / Modern Greek, Ancient Greek / Finnish, Estonian / Hungarian / Udmurt / Arabic, Maltese / Hebrew / Hausa / Igbo / Swahili / Georgian / Lezgian / Chechen / Hunzib / Abkhaz / Turkish, Azeri / Japanese / Korean / Lao / Thai / Kannada, Malayalam, Tamil / Maori / Tagalog, Indonesian / Basque / Chinese / Nanai / Slave / Haitian / Quechua / Eskimo / Hopi / Evenki / Abui The geographical distribution of the languages in the sample is represented below: 15 As discussed by Haspelmath (1997:80-90), many languages lack a distinction between deictic and sequential markers; that is, they would have a single expression for a PDM like ago and a prospective like before. 16 A unified formal structural analysis of time adverbials as a whole and tense/aspect morphology has indeed been already proposed for example by Demirdache and Uribe-Etxebarría (2004).

24

Figure 8 The availability of the relevant information was the most decisive criterion in the election of the individual languages in the sample17. As a consequence, the sample is strongly biased towards Eurasian and especially towards European and Indo-European languages. With the purpose of compensating the effects of genetic concentration, when two or more languages of the same family 18 agree in every relevant parameter at issue, only one of them will be considered. For example, the Germanic languages are all predominantly prepositional and show a preposed FDM (Eng. in two days, Ger. in zwei Tagen, Swed. om två dagar) so when investigating if the order of FDM and its time NP correlates with the order of the adposition and its complement, the testimony of the Germanic languages will only count as one. Before I start to present the data I would like to note that the linguistic structure we are concerned with here (i.e. a syntactically complex but optional phrase containing a time NP specifying the time elapsed from some event in the past up to the utterance time or from the speech time to some point in the future) is widespread in the major European and Asian languages and we might take its presence for granted. However, the precise measurement, record and segmentation of time is a cultural feature which is especially prominent in Western culture but absent from other less technological societies. Thus, it must be stressed first of all that not all languages match the degree of precision with which an event can be placed in time in most European languages. Some languages, therefore, simply do not have the linguistic structures (TDMs and TDPs) we are analyzing here. Dixon (1972:115) for example, mentions for Dyirbal lexicalized, unanalyzable expressions as the only adverbial resource to locate an event in the past with a certain degree of precision. We find, among others, buluru 'very many years ago', bandagay 'many years ago', gubila 'some time ago' among others. 17 I am aware that strictly speaking this lack of randomness disqualifies this as a sample (the term 'convenience sample' is sometimes used) but the term will be kept for practical reasons to designate this group of languages I have selected for the present study. 18 Note that what I have taken to be languages of the same family (not of the same phylum) have their names separated by commas. The distinctions between family and higher order filiations are quite clear and widely agreed upon in the case of IE languages. In other 'messier' cases a time depth of 2000-2500 years has been arbitrarily chosen in this sample as the dividing line between closely related languages (belonging to the same family) and not closely related languages.

25

Something very similar describes Evans (1995:229) for Kayardild, where we find yuujbanda 'in the old days', kurdiwirdi 'some time ago' or dilaya 'a few days ago'. These expressions are outside the scope of the present study and do not lend themselves to the kind of analysis we are pursuing here. The absence of TDPs, however, is not a feature limited to Australian languages. In other linguistic areas languages seem to lack these constructions frequently as well. Everett and Kern (1997:139) mention that time adjunction as a whole is absent from Wari' (Chapacuran, Brazil). According to them, time information can only be included by the use of verbal modifiers which combine with the verb root to produce a compound. More remarkable still to western eyes is the situation in Jarawara (Arawan, Brazil). According to Dixon and Vogel (2004:409), even a specific word for 'when' is absent from the language. Even to inquire about time one must get along with circumlocutions like (36): (36) Hika bahi itara

(Jarawara)

where sun sit

'What time is it?' (lit. Where does the sun sit) There is, however, no need to go to such probably extreme cases. Many languages, when expressing precisely the location in time of some event do so exclusively with a biclausal structure. In those languages therefore, the phrase which expresses the time distance separating the event from the present is always a clause in itself and is therefore outside the scope of the present paper. Here I include an example from Babungo (Niger-Congo, Cameroon) which, apparently, cannot express the relevant meaning in a single clause: (37) ŋwә́ táa jc táa jwì

fáŋ vәshī vәbɔ̀ɔ shɔ̂ɔ iɔ̀ɔ shɔ̂ɔ shɔ̀ɔ shɔ̂ɔ jɔ̀ɔ shɔ̂ɔ

he FUT come [when days

(Haspelmath, 1997:55)

two pass.IPFV]

'He'll come in two days.' (lit. when two days have passed.) (38) ŋwә́ táa jc kû. ndwә́ táa jc lùu ŋú'sә bɔ̀ɔ shɔ̂ɔ iɔ̀ɔ shɔ̂ɔ he die

(Haspelmath, 1997:55)

now be years two

'He died two years ago' (lit. 'He died. It's now two years.') The reason why I have chosen to leave these constructions out of the typological study19 is that clausal elements have very different properties from nonclausal ones and they may not be comparable despite their identical semantics. For instance, sententials tend to have a freer syntax, which would have been problematic for the analysis of languages to which the linguist has only a limited access (in some cases a single sentence) and no knowledge of his own whatsoever. The Spanish examples below illustrate the different constituent order flexibility of the two constructions: Clausal (39) Hace diez días que

te

vi

it.makes ten days that you.ACC saw.1SG

'It has been ten days since I saw you'

Adverbial Hace diez días te ago

vi

ten days you.ACC saw.1SG

'I saw you ten days ago'

19 These biclausal constructions have not been left out of later sections 4 and 5. Quite on the contrary, they feature prominently there since, diachronically, they are very frequently the source of the adverbial, nonclausal strategies.

26

(40) Hace que

te

vi

diez días

*Hace te vi diez días

it.makes that you.ACC saw.1SG ten days

'It has been ten days since I saw you' (41) Diez días hace que

te

vi

*Diez días hace te vi

ten days it.makes that you.ACC saw.1SG

'It has been ten days since I saw you' With these examples I have tried to explain the decisions I have taken regarding what kind of expressions I analyze here and which ones I don't. In addition, I have exposed that the construction we are studying here is far from universal. This is meant to underline that the fact that Europe is over-represented in the present sample is not exclusively a result of the greater availability of information for these languages. It is probably also a result of the fact that, especially in many languages of hunter-gatherer societies20, the relevant structure is absent whatsoever.

3.3. Past Deictic Markers and their phrases cross-linguistically PDMs occur across languages in any possible word order with respect to their accompanying time NP. Thus, they may precede it, follow it or, as the third logical possibility, a discontinuous PDM may precede and follow it at the same time. I present below examples for each of the attested word orders with the PDP between brackets and the PDM in bold: (42) (Du-ela hamar) urte jaio zen

(Basque)

has-COMP ten year born was

'He was born ten years ago' (43) Kunit hunak (‘abil juma‘) I.was there before week

(Arabic)

(Kurzon, 2008:217)

(Punjabi)

(Haspelmath,1997:82)

'I was there a week ago (44) (Do saal páílãã) asĩĩ Multaan gae

two year before we Multaan went

'Two years ago we went to Multaan.' (45) Ya priyekhal syuda (tri nedeli nazad) I

came

to.here three weeks ago

(Russian)

'I came here three weeks ago' (46) Jag var i Stockholm (för tre år sedan) I was in Stockholm

(Swedish)

for three years since

'I was in Stockholm three years ago' (47) (Kә and sәat bәfit) ɨzzih nәbbәrә ago one hour ago

here was.he

(Amharic)

(Kurzon, 2008:217)

'He was here an hour ago'

20 The idea that language evolves to meet the communicative needs of its speakers (which is to say the needs of the society where the language is spoken) is not new (e.g. Deutscher, 2000; Bybee, 2010; Dixon, 2010:15-22) and is even expected if one accepts usage-based explanations of language change.

27

In the tetrachoric table below, I classify the languages in my sample according to whether their PDM is preposed or postposed to the time NP and according to whether the language in question is predominantly prepositional or postpositional 21. This is, logically, meant to analyze whether there is some correlation between the order of PDM and that of the adposition and to see what the cross-linguistic preference is for the placing of the PDM. Remember that the numbers have been adjusted to remove the effects of genetic relatedness:

Table 2 There are two things which are spotted here. On the one hand, it can be seen that, cross-linguistically, the PDM has a general tendency to be postposed to the time NP. This had been advanced informally by Plank (2011:457) for example, who mentioned that “of all adpositions, ago is universally among the most likely candidates (perhaps the most likely) for postposing even in languages where prepositions hugely predominate”. This seems to be largely confirmed here since almost 3 in 4 PDMs (74'4%) are postposed in our 'adjusted' sample. In total, 40 from the 56 languages for which I have the relevant data show a PDM postposed to the time NP, which is quite a robust tendency. On the other hand, more interestingly still, there seems to be a very clear correlation between the order of a PDM and the dominant order of adpositions in the language in question. Even if prepositional languages show preposed and postposed PDMs with roughly the same frequency, postpositional languages have a very strong preference for a postposed PDM. From the 25 postpositional languages in my sample for which I have the relevant data, only a single one (Basque) has a preposed PDM. And even this single exception may have an extra-linguistic explanation since the 'anomalous' order of the Basque PDM could well be a result of language contact since the language has for a very long time been a neighbor of both Spanish and French, languages where a preposed PDM is used. This hypothesis is supported by the fact that, up to the 17th century, the same PDM that nowadays appears preposed (42) was indeed postposed: (48) Beha egotu naiz ea zer erraiten zuten iendek nik (zenbait urthute duela) looking be AUX.1SG Q what

say

AUX.3PL people I.ERG some

eskiribatu nuen giristinoaren Dotrinaz write have.PST.1SG christian

doctrine

(Basque)

years

ago

(1617-23, EKC)

'I have been looking what do people say about the christian doctrine I wrote some years ago' 21 The two cases which, like Swedish, display a PDM which is both pre- and postposed have been counted as 0,5 for each of the two positions. In additon, as any typologist knows, the order of an adposition and its complement is strongly correlated with that of the verb and its object. Almost all of the prepositional languages in our sample therefore display the basic word order VO and almost all of the postpositional languages are OV. This will become important later on. The few cases where the two parameters do not co-occur as expected have been indicated in the annex 1.

28

Until approximately that time the word order of the Spanish PDM was predominantly postposed as well, so language contact had not exerted a pressure towards preposing. Only when later on, especially during the 18th century, preposing of the PDM became mainstream in Spanish through the replacement of haber by hacer22, the Basque language also came under pressure to switch the order of its PDM to fit the one found in the neighboring languages: (49) Orai (duela laur egun) othoitz-ean nindagoen... now

ago four day

pray-ing

AUX.PST.1SG

(Basque)

(1740, EKC)

'I was praying four days ago now...' As with many other things, it is impossible to prove beyond doubt that a certain change in a language is the result of contact, but the evidence I have presented quite strongly suggests that it may have been at least an important factor here. Back to the typological analysis, one factor which we have left out of the equation until now is the fact that PDMs may be monosemous (i.e. may have the location of an event at a certain point in the past as their only time-related semantic function) or may actually be polysemous. In fact, as argued by Haspelmath (1997:80-90) and as shown more graphically by Franco (2013:53), many PDMs are also put to use for the expression of other temporal relations. Very frequently they can also be used for 'retrospective' and 'anterior' uses: (50) Vor vierzig Jahren gab es hier eine wunderbare Landschaft before forty

years was there here a

wonderful

(German)

landscape

'Forty years ago there was a wonderful landscape here'

(51) Mein Vater war zehn Jahre vor Beginn des Zweiten Weltkriegs geboren (German) my

father was ten

years before beginning of Second

World.war

born

'My father was born ten years before the beginning of the Second World War' (52) Vor dem Essen war ich nicht hungrig (German) before the

meal was

I

not

hungry

'I wasn't hungry before the meal' (53) Dümdüz 250 jis idalaj wilik

(Lezgian)

(Haspelmath, 1997:82)

(Lezgian)

(Haspelmath, 1997:82)

exactly 250 years this before

'Exactly 250 years ago' (54) Däwedilaj wilik war

before

'Before the war'

Examples (50) and (53) have the semantics we have defined as our object of study in this paper, which Haspelmath (1997) called 'distance-past'. Example (51) has a different semantic contribution, since it does not make reference to the utterance time but rather the time distance is counted from another time reference. It is usually called 'distance-retrospective'. Examples (52) and (54) are more different still from 'distance-past' in that they neither refer to utterance time nor do they specify the time distance mediating between reference time and event time. Because of this, they are usually referred to simply as 'anterior': 22 Further details will be provided in section 5.

29

Figure 9 Cross-linguistically, therefore, PDMs can be specialized exclusively for its use in the 'distance-past' function or may have other temporal uses like the ones mentioned above. To see if the properties of the expressions are significantly different depending on whether they are devoted to 'distance-past' or not, I have analyzed again the correlation between their word order and that of adpositions. These are the results:

Table 3 The count shows that the word order of the two types of PDMs shows indeed remarkable differences. Those which are not confined in their use to the 'distancepast' function pattern very much like adpositions. Only 7'4% of languages in the adjusted sample deviate from the predominant order found there. All the 'extravagance' and 'maladaptiveness' observed in PDMs, therefore, is found in those used exclusively for the expression of that time relation. From this we can hypothesize two things. First of all, from the 'messy' properties and little aligned word order of dedicated PDMs vis à vis their non-dedicated counterparts one may propose that 'distance-past' is a time relation in which erstwhile free, discursive linguistic expressions first become a part of grammar. Such a fact could explain many of the characteristics of PDMs we have seen before such as their often synchronically recoverable lexical (frequently verbal) origin as well as the word order correlations analyzed here. Given the most frequent diachronic sources of adpositions and adpositional phrases23, the word order patterns that have been presented for PDMs are not unexpected. 23 Plank (2011:460) mentions that “the commonest sources of adpositions-in-adpositional-phrases are verbs-in-verb-phrases (primarily transitive, typically in some non-finite construction), head-nouns-inattributive-phrases (with body parts and other relational nouns as heads) and (local and temporal) adverbs gaining an obligatory complement”. When grammaticalizing from verbs, which is quite frequent among PDMs as has been shown, this predicts that in SVO languages both a preposition (out of VO) and a postposition (out of SV) are possible outcomes whereas in SOV languages the only possible outcome is a postposition. This agrees with the general correlations found here. Note that prepositional languages are frequently SVO whereas postpositional languages are in most cases SOV.

30

Secondly, Franco (2013) shows that the possible semantics of polysemous PDMs are subject to a constraint whereby a PDM cannot be used for the 'distance-past' and 'anterior' functions without being used as well to express 'distance-retrospective': Language Albanian Catalan Croatian English German Yukaghir Maltese Spanish Turkish

Distance-past parë fa prije ago vor tudä ilu hace önce

Distance-retrosp. Anterior para para abans abans prije prije before before vor vor tudä kieje qabel qabel antes antes önce önce (adapted from Franco, 2013:53)

Franco (2013) also includes the following information about the diachrony of Italian: Language Old Florentine Modern Italian

Distance-past addietro fa/addietro

Distance-retrosp. addietro prima

Anterior prima prima

As could be seen graphically in the above Figure 9, the semantics of 'timeretrospective' are indeed intermediate between the other two. It is thus not surprising that, when diachronic semantic extensions occur, a morph traveling from one of the extremes to the other necessarily will need to adopt the intermediate meaning as well. The constraint observed synchronically by Franco (2013) is thus the expected outcome of “normal” diachronic semantic extensions. It is at this point when our data come into play once again. The fact that the 'extravagant' properties of PDMs are limited to monosemous PDMs is an indication that semantic extensions among these three functions proceed (in agreement with Franco's diachronic example above) from the 'anterior' function towards the 'distance-past' via the 'distance-retrospective' 24 and rarely (or never) in the opposite direction:

Figure 10 Other interesting tendencies can also be extracted from the results of my typological survey of PDMs even if always with the necessary precautions because of the small numbers involved in the present sample. It looks remarkable, for example, that amongst prepositional languages, 50% have a devoted PDM whereas the percentage is only 10% among postpositional languages. I cannot see what could be the reason if 24 This involves, as can be seen in Figure 10, sucessive steps of interpretative enrichment, which constitutes a development “entirely expected from a Gricean point of view” (Haspelmath, 1997:84) and are the result of the natural tendency of the speaker to always assume more than has been said.

31

there would indeed be some correlation here. An alternative explanation to this is the observation that semantically devoted PDMs are almost totally restricted to Europe and are specially frequent in Western Europe, places where prepositional languages are the rule. Here I present the global distribution of dedicated PDMs:

Figure 11 As can be seen, only 2 out of 14 languages having a dedicated PDM are outside Europe. It may well be that this is another one of the cross-linguistically uncommon features of so-called 'Standard Average European'. It may also be hypothesized that, if this turns out to be a significant areal feature, the greater western-culture preoccupation with time we mentioned in section 2.2 may well be the reason behind it. If the different time constructions were used more frequently in European society and languages than in other parts of the world, this could have been a motivation for a greater precision in this respect (i.e. for the use of different morphs and structures for the different time relations). A further possibility mentioned by Haspelmath (1997:55) is that a greater frequency may also have triggered the grammaticalization of some biclausal structures (that would otherwise have been left out of this study) into monoclausal ones, thus increasing the frequency of dedicated PDMs.

3.4. Future Deictic Markers and their phrases cross-linguistically FDMs, the same as PDMs, also may occur preposed and postposed to the time NP. Unlike in the case of PDMs, however, I have not happened to find a FDM which both precedes and follows the time NP. Whether this is significant or just a product of chance is hard to tell because the numbers are not large enough. Conversely, I have found the 'distance-future' semantic function expressed by a particular grammatical case ending on the time NP, which was something I did not find to occur in 'distancepast'. Again the numbers are too small to be confident that these are general features distinguishing FDMs from PDMs, however, they seem to be in line with the overall higher grammaticalization of FDMs vis à vis PDMs. Here are a few examples of pre- and postposed, as well as grammatical case FDMs: (55) Una expedición an

viajará

a Marte (dentro de 15 años)

expedition travel.FUT.3SG to Mars

inside

of 15 years

(Spanish)

'An expedition will travel to Mars in 15 years' (56) Tha jiríz-o (se tris óres)

FUT return-1SG in three hours

(M. Greek)

(Haspelmath, 1997:90)

'I will return in three hours' 32

(57) (Bi urte barru) amaitu-ko ditu ikasketak two year inside

(Basque)

end-FUT AUX.3SG studies

'He/she will finish his/her studies in two years' (58) (Iki saat sonra) don-eceg-im

(Turkish)

(Franco, 2014:7)

(Finnish)

(Haspelmath, 1997:90)

(Georgian)

(Haspelmath, 1997:90)

two hour after return-FUT-1SG

'I will be back in two hours'

(59) Palaa-n (kahde-ssa tunni-ssa) return-1SG two-INESS hours-INESS

'I will return in two hours' (60) (Or saat-ši) davbrundebi three hour-LOC I.will.return

'I will return in three hours' In the tetrachoric table below, as was done in the previous section with PDMs, I classify the languages in my sample according to the position of their FDMs with respect to the time NP25 and according to whether the language in question is predominantly prepositional or postpositional. In this way, I will analyze whether some correlation exists between the order of FDMs and that of adpositions and I will see what the overall cross-linguistic preference is for the placing of the FDM. Remember once again that the numbers below have been adjusted to remove the effects of genetic relatedness:

Table 4 As can be seen in the data above, FDMs, unlike their mirror-images for the past are very unremarkable and pattern very much like adpositions. One of the two exceptions, a postposed marker in prepositional Persian, cannot be considered totally disharmonic either because, as a prepositional but SOV language, Persian exhibits mixed features regarding word order. As for the other exception, Indonesian FDM lagi, it has grammaticalized from the adverb 'still', (Haspelmath, 1997:165) so its word order, even if synchronically out of line in the language, is diachronically understandable. The same as we did in PDMs, we can also draw a distinction between those FDMs having 'distance-future' as their only time-related meaning and those polysemous morphs used with other meanings as well. The case of Turkish sonra is an example of a polysemous marker: (61) Iki saat sonra don-eceg-im

(Turkish)

(Franco, 2013:7)

(Turkish)

(Franco, 2013:7)

two hour after return-FUT-1SG

'I will be back in two hours'

(62) Sah-dan sonra bura-da ol-acağ-im Tuesday-ABL after here-LOC be-FUT-1SG

'I will be here after Tuesday'

25 I have decided to exclude here the FDMs which are grammatical case affixes on the time NP for reasons of homogeneity, however the five such cases in my sample are all case suffixes in postpositional languages and therefore would not have changed in anything the obtained results.

33

Since FDMs in general patterned as adpositions, we cannot expect this time much of an asymmetry between devoted and not-devoted markers and indeed there is none:

Table 5 This seems an indication that, unlike it happened in the case of the 'distance-past' meaning, 'distance-future' is usually not a locus for primary grammaticalization. The pattern for polysemous markers in the previous section also holds for FDMs and thus we can only find an identity of the 'distance-future' and the 'posterior' like the one of Turkish (61), (62) if the semantically intermediate function 'distance prospective' is also expressed with the same marker: Language Albanian Serbo-Croat. Arabic Hungarian Maltese Haitian Japanese

Distance-future pas do baʕda múlva fi nan go ni

Distance-prosp. Anterior pas pas poslije poslije baʕda baʕda múlva ultán wara wara apré apré go ni go ni (Adapted from Franco, 2013:52)

Therefore, regarding the diachronic semantic extensions I hypothesized for PDMs in the previous section, I have no reason to think they proceed differently here. Therefore, the 'posterior' is probably a frequent diachronic source for FDMs crosslinguistically but it is by no means the only one. We mentioned before that, given the unremarkable properties of even devoted FDMs, 'distance-future' was probably not a frequent locus for primary grammaticalization. It may be, however, for so-called secondary grammaticalization. As witnessed by the spatial inessive origin of many devoted FDMs (e.g. Sp. dentro de, Fr. dans, Basq. barru, Finnish inessive case...), as well as by the lack of identity of the expressions in closely related languages (see e.g. the FDMs of Romance and Slavic languages in annex 2), the use of such a spatial metaphor and the subsequent borrowing of the corresponding grammatical strategy from the domain of space may be quite frequent for expressing the distance-future function. That these inessive-based markers are semantically dedicated stroke Haspelmath (1997:100) as interesting. He commented: “'Within' markers never express both distance-future and distance-prospective, they are always purely deictic. It is not clear to me why this should be so”. I believe that Haspelmath's observation that 'within' markers are always dedicated to distance-future may constitute just an accidental gap in his sample. On the one hand, we may be suspicious of the pattern because, while

34

the verbs often giving rise to PDMs already incorporate a deictic meaning 26, a 'within'type expression in principle does not have deixis as an inherent part of its semantics. On the other hand and more importantly, we can see in the historical record that 'within' type FDMs which nowadays are restricted to distance-future (e.g. Sp. dentro de) had earlier not only the by Haspelmath expected 'within' meaning (63), but also a prospective (64) meaning independent of utterance time: (63) Está vacante una Prebenda la cual is

vacant

a

position

segun

los estatutos del mismo Colegio,

which according.to the

se debe proveer dentro de cincuenta dias desde

it.must.be

fill

within of

fifty

days

rules

of.the itself

el de la

college

vacante²⁷

from the.one of the vacant.position

'There is a vacant position which according to the rules of the college itself must be filled within fifty days' (64) Díxo-les que se

vistiessen; y dentro de poco tiempo bolvieron 27

said.3SG-them that REFL get.dressed.3PL and inside of little

time

came.back.3PL

'He told them to get dressed and a little later they came back'

The prospective use of dentro de is indeed attested from the earliest writings up to the 19th century, which suggests that it can be a stable feature and that, in principle, there is nothing barring 'within' markers from non-deictic uses. Note at this point that the later semantic change from distance-prospective to distance-future has been already hypothesized to be common and is not problematic for us here. Back to the results of my typological investigation and despite the big difference in the properties of the devoted markers for the past and those for the future, what we do find is a similar percentage of devoted FDMs and devoted PDMs (32,4% and 30,4% respectively28). These, in addition, have a geographical distribution very close to that of devoted PDMs. Here is the distribution of the devoted FDMs in our sample:

Figure 12 26 A present tense morphology in many cases anchors them to utterance time and disqualifies them for their use in any of the time relations presented here other than distance-past. 27 18th century examples from CORDE 28 Since it is much easier to find evidence that a certain marker is not devoted (finding a single other temporal use of it suffices) than to find evidence that the marker is devoted, the proportion of devoted markers may be underestimated. However I have no reason to believe that this would have any sizeable impact on either the roughly similar proportion of devoted PDMs and devoted FDMs nor in their geographical distribution, which appears too consistent to have arisen by chance.

35

As can be seen, FDMs are indeed restricted in our sample to European languages, a pattern already identified by Haspelmath (1997:100). This constitutes further evidence that our initial hypothesis to explain this could be on the right track. It may be that there existed a greater need for an unambiguous coding of time relations in Europe because of the bigger importance of time in European society. This would have favored more specific, monosemous ways of expressing the various time relations.

3.5. Conclusion In previous sections, the word order properties of TDMs were analysed and the a priori spotted asymmetry between PDMs and FDMs was confirmed to be a valid crosslinguistic generalization. PDMs appear to have a much more remarkable word order, since in prepositional languages preposing and postposing of the marker appear to be equally frequent. FDMs, on the other hand, pattern closely like adpositions. When the distinction between semantically dedicated and polysemous markers is introduced, another asymmetry appears to separate monosemous, much more remarkable PDMs from polysemous markers which, alike FDMs, are very similar to adpositions concerning word order. When this is considered together with the extra evidence from the patterns of polysemy and the lexical, mainly verbal sources of dedicated PDMs, we can hypothesize that the differences have a diachronic origin. I propose two main diachronic sources of TDMs. First, a secondary grammaticalization path which leads to TDMs from 'anterior' and 'posterior' time markers. This can easily be accommodated in theoretical models like for example FDG. Hengeveld & Mackenzie (2008:171) note that absolute location in time is a property of episodes, while relative location in time is a property of hierarchically lower states-of-affairs. An upwards development (from the latter to the former) is the expected direction for change. The other paths I propose here lead directly to TDMs from other sources. The 'distance-past' meaning where PDMs are used is a frequent locus of primary grammaticalization out of clausal strategies. The same is not the case of 'distancefuture', where mainly secondary grammaticalization takes place as grammatical strategies are borrowed from the spatial domain. Trying to be more precise about those grammaticalizations which feed directly into 'distance-future' and 'distance-past', I believe that both arise most frequently out of durative uses. An expression like within, which selects an interval at any point of which the event may take place, may become a dentro de type of expression through interpretative enrichment29. All it takes is that when hearing a sentence like (65) the hearer assumes that the speaker has been maximally informative and that the event will in fact take place towards the end of the three years' interval. (65) He will sure finish his career within three years 29 More investigation would be needed at this point, however, to clarify a few things. On the one hand, a synchronic cross-linguistic study would be needed to investigate to what extent Haspelmath's observation that inessive-based markers are synchronically deictic can remain a universal tendency. On the other hand, a diachronic in-depth quantitative analysis of the evolution of the semantics of an expression like dentro de could help illuminate the grammaticalization paths followed by these expressions when they give rise to FDMs. I leave these issues for future research.

36

Similarly, the verbs used in biclausal constructions to express 'distance-past', which are the most frequent sources of devoted PDMs, must ultimately at some point have selected their time NP interval as a period during which the event was taking place or not and not as the time interval separating the past event from the present. This is the only possibility we find in less grammaticalized time constructions (66) (67) and a meaning of which we still find some remnants in more grammaticalized PDMs (68). (66) Llevo dos años viniendo a esta playa / Ya van dos años que no vengo a esta playa take.1SG two years coming to this beach / already go.3PL two years that NEG come.1SG to this beach

'I've been two years coming to this beach' / 'I've two years without coming to this beach'

(67) *Llevo dos años que vine a esta playa / *Ya van dos años que vine a esta playa take.1SG two years that came.1SG to this beach / already go.3PL two years that came.1SG to this beach

'It has been two years since I came to this beach' (68) Trabajo en esta escuela hace 30 años work.1SG in this

school

ago 30 years

'I have been working in this school for 30 years' I believe, in addition, that the durative sources of distance-past and distance-future strategies are not limited to clausal or inessive sources respectively but these are rather just the most frequent implementations of the more general tendency to link duration of the event and distance from it and to diachronically derive the latter from the former. This can be observed for example in FDMs like Azerbaijani ərzində, which apart from this role also has a meaning of 'during'. The link can be spotted not only in grammatical constructions or elements but also in lexical items like Russian davno, which also shares durative, up-to-now uses (69) and distance-past ones (70): (69) Zapad perezhivaet etot krizis uzhe davno

the.west goes.through this crisis already long.time

(2003, Russian National Corpus)

'The west has been going through this crisis already for a long time' (70) Potomu chto ne tak davno umerla because

not so long.ago died.F

(2004, Russian National Corpus)

'Because she did not die so long ago' To conclude the discussion, I present here a graphic representation of the diachronic developments (primary grammaticalizations in gray, secondary grammaticalizations or semantic extensions in black) that have been proposed in this section:

Figure 13 37

I have presented in this section a broad picture of the synchronic word order properties of TDMs and of the diachronic developments leading to them. As could be seen, PDMs are especially interesting expressions, above all when they are the outcome of a recent process of grammaticalization. I will therefore concentrate on these and on the process represented by the gray arrow in the above figure 13 in the rest of the paper. In order to gain a deeper understanding of both their synchronic properties and their diachronic developments, I will largely abandon the typological perspective that characterized the present section and I will focus on the 'distancepast' function in Spanish.

38

4. THE SYNCHRONY OF HACER IN SPANISH TIME EXPRESSIONS 4.1. Introduction to hacer As has been suggested already by various earlier examples, the most widespread way of expressing the distance-past time relation in contemporary Spanish 30 is by means of hacer 'to make' preceding an NP containing a time measure: (71) Estuve en Vitoria hace dos años was.1SG in Vitoria makes two years

'I was in Vitoria two years ago'

The verb is also very frequently used with an up-to-now meaning, a meaning for which, unlike for the distance-past meaning in (63) various alternatives to hacer do exist: (72) Hace 2 años que soy socio / Llevo 2 años siendo socio / Tengo 2 años de ser socio 31 it.makes 2 years that am member / take.1SG 2 years being member / have.1SG 2 years of be member

'I have been a member for two years now' The verb hacer is very general in meaning and, even in its most core uses like (73), quite desemanticized. It occurs as well in impersonal constructions, were it is often semantically empty and forms something close to existentials like in (74) and (75): (73) Juan hace los deberes a las seis John makes the homework at the six

'John does his homework at six' (74) Hace falta más madera makes need more wood

'I/you... need more wood' (75) Hace un frío horrible makes a cold terrible

'It is terribly cold'

The hacer of the time constructions which we analyze here has also been called 'existential' by some authors (e.g. Rigau, 1999). The verb is at any rate also impersonal32 in these uses and, despite not being (72) in my opinion so semantically impoverished as in the other constructions of (74) and (75), it is clear that some parallelisms are bound to exist. Some authors have indeed analyzed all these constructions together (Pérez Toral, 1992). Here, however, I will focus on temporal hacer exclusively since I consider the links to other non-temporal uses too weak to demand a unitary approach. 30 Some archaizing uses of haber continue to appear sparsely as an stylistic device, for example: “personaje que fue, años ha, el responsable de que...” (Internet 25/7/2010) person

who was years ago the person.in.charge. of that

'who was years ago the person responsible for...' 31 Not used in Peninsular Spanish. 32 There is some diachronic evidence that the verb may have started as personal in this construction (Pérez Toral, 1992). In any case impersonal use is mainstream in contemporary Spanish.

39

4.2. Introduction to temporal hacer The more traditional grammatical descriptions of Spanish (e.g. Real Academia Española, 2009; Demonte & Bosque, 1999) in agreement with most linguists defend that, under hacer+time there are in reality two different constructions that need to be distinguished. Adopting (or rather translating) the terms used by the Real Academia Española (2009:1832-1840) to refer to each of them we will speak here of a clausal construction in the examples in (76) and (77)33 and an adverbial construction in (78): (76) Hace diez años que murió

Clausal construction

it.makes ten years that died.3SG

'It has been ten years since (s)he died' (77) Hace diez años de

su

muerte

Clausal construction

it.makes ten years from his/her/their death

'It has been ten years since (s)he died' (78) Murió hace diez años

Adverbial construction

died.3SG it.makes ten years

'(S)he died ten years ago'

Behind the big superficial similarity, the syntax of both constructions is very different. Their syntactic description is a much debated issue and little agreement exists in this respect in the linguistic community. I will enter that debate later on but for the moment I would just like to point out the immediately noticeable differences: Whereas in the clausal construction of (76) and (77) hace is the main verb of the sentence (since its phrase cannot be freely omitted) it is not the main verb in the adverbial construction (78), where the phrase it heads is fully optional. Adopting the terms used by RAE (2009) does not imply that I agree with the way in which they define the two constructions. Specifically, they mention that (RAE, 2009:1832) the clausal construction requires either a que-clause like (76) or a PP like (77) to be considered so. That constituent, I believe, may sometimes be left unexpressed without that implying, as it percolates from RAE's reasoning, that we are automatically dealing with the adverbial variant. For example, RAE (2009:1837) considers the following an example of an adverbial construction, which I believe is not: (79) Pronto

hará

cinco años, si no los ha hecho ya

soon make.FUT.3SG five years if NEG them has made already

'It will soon be five years if it is not five years already'

As can be seen, hacer is, despite the absence of a que-clause or an eventive PP, still the main verb in the sentence and according to our characterization of the expressions, a clausal and not an adverbial construction. Concerning the meaning (the truth conditions) of the clausal and adverbial constructions no difference can be found; however, there is a clear pragmatic distinction between both which I have tried to reflect in their English translations. Many authors have reflected on this (Brewer, 1987; Rasmussen, 1981 among others) and have attempted to explain it in different ways. A representative expression of this sentiment is that of Rasmussen (1981:131) who comments that “when the main verb is 33 Some authors (e.g. Rigau, 1999; Fábregas, to appear) do not believe (76) and (77) are comparable constructions.

40

hacer, the measurement of time prevails semantically; When it is subordinate, the action of the verb is more important” (translation mine). This constitutes a parallel with clefts and other focalizing constructions and may be suggestive of a derivational relationship between the clausal (cleft-like) and the adverbial constructions, which is something some authors (Rigau, 1999; Fábregas, to appear) have proposed. Many others (e.g. García Fernández, 1999), however, have noted as well that hacer is eventive in the clausal but not in the adverbial construction. This along with other differences in the syntactic properties of the constructions I will present later make it, in my opinion, difficult, to sustain a derived status for the clausal construction. The adverbial and the clausal time constructions with hacer can express not only the distance-past meaning in (76) to (78) but also, as was mentioned in the previous section 4.1., an up-to-now meaning where the event is not located on the time axis at some point in the past but rather extends from some point in the past until the utterance time. This meaning is fully legitimate in the clausal construction (80) but is nowadays less widespread in the adverbial construction (81), where speakers tend to use a preposition desde34 before hacer to unambiguously indicate the up-to-now meaning. Other prepositions, in use with other time adverbials, are also allowed with the phrase headed by hacer (82) (83): (80) Hace mucho que trabajo aquí it.makes a.lot

that work.1SG here

'I have been working here for a long time'

(81) Trabajo aquí (desde) hace mucho work.1SG here

since it.makes a.lot

'I have been working here for a long time' (82) Trabajaba aquí hasta hace poco work.IPF.3SG here until it.makes little

'He used to work here until recently' (83) Las ciudades de hace 100 años the

cities

of it.makes 100 years

'The cities of 100 years ago'

4.3. A competence-based approach to the synchronic properties of the adverbial construction with hacer We have earlier mentioned that the syntactic analysis of the hacer+time constructions is very controversial. One of the most common debates (and also, in my opinion, one of the most futile) concerns the nature (i.e. grammatical category) of hacer especially in the adverbial construction. The Real Academia Española (2009:1837) for example, argues that “even if it has been proposed that hacer has an adverbial or prepositional nature in these constructions, there are more arguments in favor of a verbal nature”. They go on to mention a series of verbal properties (most of which do not even agree with synchronic reality) that, in their opinion, prove beyond doubt that we are here 34 Some authors (e.g. Fábregas, to appear) directly assign an ungrammatical status to these up-to-now sentences lacking desde but, as Brucart (2014:5) shows, they continue to be a sizeable minority.

41

dealing with a bona fide verb. Among others, they mention the availability of time inflectional morphology (84), the possibility to have time adjuncts to hacer (85) or to negate it (86) or to use verbal periphrases (87): (84) Le

había

visto hacía

un año

him/her have.IMP.1SG seen make.IPF one year

'I had seen him/her one year before'

(85) Abandonó la ciudad pronto hará tres meses left.3SG

the city

early make.FUT three months

(RAE, 2009:1837)

'It will be soon three months since he left the city' (86) Se

casó

no hace ni un mes

REFL married.3SG NEG makes even one month

(RAE, 2009:1837)

'It is not even one month since he got married' (87) Se divorciaron debe de hacer dos años o así REFL divorced.3PL

must

make two years or so

(RAE, 2009:1837)

'It must have been two years or so since they got divorced' Many native speakers of Spanish may already be somewhat suspicious at reading these last sentences. Especially (85) to (87) sound very unnatural. For this reason, in order to know what the synchronic reality of the constructions is in the speech community, a questionnaire was designed that would check native speakers' intuitions on precisely these borderline cases and other unclear properties35. The results, based on 33 answers from speakers of the northern Peninsular Spanish speech community seem to support my initial reserves concerning the properties exemplified by the sentences (84) to (87). When evaluating (from 0 to 5) the wellformedness of a sentence containing a time adjunct to adverbial hace (88), speakers assigned it a mean of only 1.15, which shows that the grammaticality of those sentences is more than dubious synchronically. (88) Tuve el accidente mañana hace tres años had.1SG the accident tomorrow makes three years

'Tomorrow it will be three years since I had the accident' Still more interesting, there seem to be important differences between different age groups. Whereas speakers under 30 rated (88) with 0.83 on average, the sentence received 1.53 among speakers over 50. This might be suggestive of a diachronic change in progress. The adverbial construction with hace may be becoming compulsorily 35 I think that moving beyond the linguist's own grammaticality judgements is necessary, especially in less clear-cut cases such as the one which is analysed here. We see many disagreements in the linguistic literature concerning what is and what is not possible in this construction. Without trying to be exhaustive in this respect, Pérez Toral (1992) and the RAE (2009) believe pronominalization of the argument of hacer is possible whereas Fábregas (to appear) and Sáez del Álamo (1987) believe it is not. Fábregas (to appear) comments as well that in the clausal construction one finds only the subjunctive, the imperfective past, the future and the present, which is much less than the variety posited for example by the RAE (2009). Rasmussen (1981) mentions that time adjuncts cannot occur in the adverbial construction while García Fernández (1999) or RAE (2009) believe they are possible. All in all, many black-or-white judgements and contrasts are presented which arouse suspicion that they might be devised just to lend support to the individual linguist's structural analysis of the construction.

42

deictic, with the time distance counted always from the utterance time and thus allowing no place for further time modifiers36. The use of non-present forms of hacer in the adverbial construction was also found to be marked at best by speakers. Sentence (89) received an average rating of 2.12. (89) Bajó

la basura

hará

dos días

take.down.PST.3SG the rubbish make.3SG.FUT two days

'He took out the rubbish around two days ago' Similarly, a sentence like (90) was strongly preferred to a sentence like (91), especially by younger speakers which again points toward a progressive loss of the morphological possibilities of hacer in the adverbial construction, maybe as a result of its deictization. (90) Había estado contigo dos días antes had.1SG been with.you two days before

'I had been with you two days before' (91) Había estado contigo had.1SG been

hacía

dos días

with.you make.IPF.3SG two days

'I had been with you two days before' The results concerning the availability of the up-to-now interpretation in the adverbial construction also pointed toward a change in progress. The mean rating of a sentence like (92) was 2.49; that is, almost exactly on the equator between grammaticality and ungrammaticality. However, for younger speakers the mean was 1.84 compared to a 3.27 from the older speakers: (92) No estoy con tu hermana hace una semana not am

with your sister

makes one week

'I haven't been with your sister for a week' The results also showed in this respect an almost absolute preference from younger speakers to use desde in sentences like (93) whereas the preference was much weaker in older speakers. This points, as has been said, to a progressive loss of the ability of hacer by itself to denote a time interval up to the present in the adverbial construction . (93) No visito a tu familia (desde) hace dos meses not visit.1SG your family

since

makes two months

'I haven't visited your family for two months' Pronominalization is another one of the properties which are sometimes attributed to the adverbial construction. RAE (2009) the same as Pérez Toral (1992) believed it was possible there. This is, I believe, a confusion derived from the way the adverbial construction is defined vis a vis the clausal. Pérez Toral gives the following example: (94) _¿Estás esperando hace dos horas? _Sí, las hace. Are.2SG waiting

makes two hours

Yes them makes

'_Have you been waiting for two hours? _Yes, it has been two hours' 36 It might seem that taking synchronic age-dependent differences as evidencing diachronic change immediately presupposes that language change takes place (as is commonly assumed in the generative tradition) during language acquisition. This is, however, not necessarily the case. As Bybee (2010:21-22) puts it “for a child or language learner, each new token of experience can have a much larger impact on representation than it can for an adult, who has already built up a large store of exemplars. Thus changes in adults will be subtle and probably rather slow under most conditions”.

43

The answer provided here by Pérez Toral, I believe, does not match the question. It is a suitable answer to ¿Hace dos horas que estás esperando? It is thus an answer to the clausal construction and not to the adverbial, and should therefore classify as an instance of the clausal, not the adverbial construction. Attributing this property to the adverbial construction may thus be just the product of an unsuitable definition of the construction if only those constructions with overt que-clauses are classified as clausal. This cannot be the reason, however, for Pérez Toral's (1992:120) acceptance as grammatical of the following sentence: (95) Murió un año hace died.3SG a year makes

'He died a year ago' This, which she takes of course as evidence that the pre-posing of the time NP in the adverbial construction is acceptable, is an ungrammatical sentence to the vast majority (if not all) speakers of contemporary Spanish. Logically, it received a mean rating of a mere 0.15 in my questionnaire. It is thus difficult to know what the reason might have been for Pérez Toral to present this sentence as grammatical but this constitutes another argument against linguists relying exclusively on their own judgements. Negation, unlike many of the above mentioned properties, is a feature which has barely been addressed in relation with hacer+time in previous literature. It has been left untackled by most linguists (e.g. Pérez Toral (1992), García Fernández (1999), Rigau (1999), RAE (2009), Brucart (2014) among others) and when addressed, it has been suggested that it is hardly an interesting feature because there are scarcely any differences between the behavior of negation in the clausal and adverbial construction or between the possibilities for negation in hacer+time as opposed to other verbs. Rasmussen (1981:99) for example comments that in the adverbial construction negation introduces “very few changes in the syntax, combinatorial possibilities and the semantics”. This is clearly not true and even his own examples can be used to prove it. These include: (96) Vive usted allí no hace mucho tiempo live.3SG you there not makes a.lot

(Rasmussen, 1981:100)

time

'You haven't lived there for a long time' (97) Era usted de nuestra opinion no hace mucho tiempo were you of

our

opinion not makes a.lot

'You agreed with us not long ago'

(98) Lo habían maltratado no him had.3PL

abused

hacía

mucho

not make.IPF.3SG a.lot

time

(Rasmussen, 1981:100)

(Rasmussen, 1981:100)

'They hadn't abused him long before'

All of his examples involve the use of hacer+mucho, which is already revealing. In fact, if that expression mucho or mucho tiempo is replaced by any other time expression (e.g. 12 días, poco, años...) the result is an ungrammatical (or at least a very unnatural) sentence. Notice that without negation the sentences are again grammatical:

44

(99) *?Vive usted allí desde no hace 12 días37 live.3SG you there since not makes 12 days

'You haven't lived there for 12 days'

(100) *?Era usted de nuestra opinion no hace poco were you of

our

opinion not makes a.little

'You agreed with us not long ago' (101) *?Lo habían maltratado no him had.3PL

abused

hacía

años

not make.IPF.3SG years

'They hadn't abused him years before'

This shows that negation is in fact, unlike most (prescriptive) grammarians and linguists have assumed, extremely restricted in the adverbial construction. In my opinion, no hace mucho is actually just a cuasi-lexicalized, frequent word string in which negation has managed to survive longer. An internal linguistic pressure to get rid of negation in the adverbial construction with hace is easy to understand. Since hacer synchronically does not denote any event here and is just an ancillary grammatical element here, true polarity is simply excluded from it. In fact, no hace mucho is no longer a negation of hace mucho but it rather means hace poco which is maybe the reason that the expression has managed to survive. All signs indicate, however, that negation will eventually also lose this last stronghold, since the alternative word order hace no mucho has been introduced from the mid 20th century and is quickly gaining ground.

4.4. A usage-based approach to the synchronic properties of hacer+time The competence of native speakers has already been seen to show important deviations from previous theoretical analyses. How does this competence translate into usage? Howe (2011) is a corpus-based approach to present-day spoken Spanish which attempts to show that hacer+time construction in Spanish is neither purely clausal nor adverbial but rather displays mixed features. 'Older' more clausal properties are retained better in non-present forms of hacer while the most frequent hace has advanced further along the grammaticalization path. Howe (2011) finds that, on his corpus, present hace constitutes an impressive 96% of the total instances of temporal hacer while imperfective hacía, future hará or past hizo contribute an insignificant 3%, 0.6% and 0.3% of the total respectively. This shows, in his opinion, that hacer+time is not a normal clausal, verbal construction, which I believe few linguists would be willing to propose in any case. His corpus analysis, despite many useful observations, in my opinion, has a major shortcoming in the absence of a consistent separation of the data from the clausal and the adverbial constructions. This might be a conscious 'commitment not to be committed' to previous theoretical analyses but I believe is detrimental for the meaningful interpretation of the various data he presents. 37 Notice that in this sentence desde has been inserted in Rasmussen's example to avoid the possible interference from an ungrammaticality derived from the up-to-now meaning of an adverbial construction without a preposition as was analyzed for sentence (93).

45

Regarding the previous percentages for instance, the fact that he often lumps together adverbial pre-verbal and clausal constructions (by virtue of their both being pre-verbal) obscures the frequency of each of the constructions. Only with a rule of three can one recover for example the actual percentages he assembled for the clausal construction. From his results we infer that only 15.6% of the cases with hace were clausal constructions compared to 46% in the case of hacía. This indicates that the proportion of non-present forms of hacer in the clausal construction is much higher than that mere 4% which was true for the hacer+time as a whole, thus providing a weaker evidence to regard the clausal construction as non-verbal and highly grammaticalized. He mentions that (Howe, 2011:276) “the predominance of the present tense collocate in the data suggests that the verbal features of hacer+time have been largely neutralized”. For the reasons given I believe this may be true of the adverbial construction but not necessarily of the clausal. Other novel observations by Howe (2011:276) include for instance an “increased compatibility of the present tense collocate of hacer+time with a past-shifted right boundary modification”. That is, he finds sentences like: (102) Le había visto hace dos días him had.1SG seen makes two days

'I had seen him two days before' Those occurrences are interpreted as suggesting a leveling of the properties across the whole hacer+time. I do not reject that possibility out of hand since the adverbial and the clausal constructions are indeed connected in the mind of the speaker and some convergence would not be unexpected, especially given the vast outnumbering of the clausal construction by the adverbial one, which is about ten times as frequent as will be shown later. There is, however, an alternative possibility that in fact sentences like (102) above illustrate an increasing difference between the clausal and the adverbial constructions as the latter increasingly rejects non-present forms of hacer while they continue to occur in the autonomous clausal construction. Possible levelings aside, I consider the grammatical properties and semantic contribution of the clausal and the adverbial constructions with hacer show considerable differences which demand a separate look at their usage. Given that Howe's (2011) approach to hacer+time does not 'sort out' data consistently on that premise, I have carried out a corpus search on the oral, Peninsular Spanish section of CREA to see what are the emerging patterns. The next paragraphs present the most relevant data. Of the total number of tokens of hace (N=410) which were inspected from the years 1996-99, 36.8% (N=151) of them were temporal. Of these one third (N=51) were preceded by a preposition38, most frequently desde. Of the remaining 100 tokens (which had been decided beforehand to be the goal number) only 8 occurred in clausal constructions while the remaining 92 occurred in the adverbial (31 pre-verbal, 61 postverbal). This shows in my opinion convincingly that adverbial hace constitutes the overwhelmingly most frequent use. Estimating 39 the raw frequency of hace+time as a whole I have reached a total number of 418 times per million words. 38 All instances of hace preceded by a preposition naturally belong to the adverbial construction.

46

When the same procedure was followed for hacía, the total goal number of 100 tokens of bare (i.e. without preposition) hacía could not be attained as a result of its much lower frequency. Of the total number of tokens of hacía (N=545) which were inspected from the years 1980-2000, only 7.3% (N=40) of them had a temporal meaning. Note the dramatic difference with respect to hace. Of these around one sixth (N=7) occurred preceded by a preposition. From the remaining 33 tokens, 66.7% (N=22) occurred in clausal constructions. This is where the difference from hace is more marked. Whereas the clausal construction constituted only a marginal proportion of temporal hace, it constitutes the majority of temporal hacía. The estimated raw frequency of hacía+time is 12.5 per million words. This difference in raw frequency of temporal hace and hacía is the one which Howe (2011) regarded as revealing of a loss of the verbal properties of the expression; however, as I am trying to show here, this overall number hid two very different realities: Clausal construction hace constitutes just 5.3% (N=8) of hace+time (N=151) and its estimated frequency is thus only of 22 per million words. Clausal construction hacía, in turn, constitutes 55% (N=22) of hacía+time (N=40) and therefore its estimated frequency is 7 per million words. This is very much in line with the figures which Pérez Toral (1992:94) provides for the clausal construction for the twentieth century40. We see that when the clausal construction is analyzed in isolation from the adverbial, a very different pattern emerges whereby the present tense hace continues to be the most frequent (as it naturally is in most verbs anyway) but overall, hacer shows a much more balanced distribution of its TAM morphology, which points towards a full verbal status for hacer in the clausal construction with que. The other side of the coin is of course that, once the clausal construction has been removed from hacer+time, the adverbial construction is found to be still more unbalanced toward the use of the present tense, which would then occur in almost 99% of the adverbial constructions. There are other marked differences in the properties of the constructions however, apart from their morphology. For instance, I found that 93.5% (N=86) of the adverbial 39 I say estimating because CREA does not provide direct information regarding this and frequency had to be calculated independently. When only a subsection of CREA is explored (in this case the intersection of the oral and the Spanish-of-Spain subsections) the size of the portion of the corpus which is being searched is unknown. For that reason it had to be estimated by searching for some of the most common words in Spanish from different word classes (de, que, es, la) in the whole CREA (for which the size is known). These words combined amounted to a total of 12.82% of the corpus. Assuming that they would constitute a similar percentage of any given subsection of that corpus one can estimate the total size of any subsection by adding the number of tokens of de, que, es and la and making a rule of three. By doing this I estimated that the oral+Spain subsection of CREA had a total of 3 263 151 words. It has to kept in mind that this is an estimation and the raw frequencies therefore estimated as well but this in any case will not affect the relative frequencies of a given token vs another. 40 She finds in her count a total of 72 instances of present hace vs. 27 for imperfect hacía in the clausal construction during the 20th century. For the adverbial construction (page 172), during the 19 th and 20th centuries, she finds 133 tokens of hace vs. 13 of hacía, which is already a sizeable difference to the clausal even if nothing comparable to the vast predominance of the present found in presentday Spanish, especially in the oral register.

47

constructions had a distance-past punctual meaning while 68.2% (N=45) 41 of the clausal constructions had a durative, up-to-now meaning. Thus, despite both meanings being attested and possible in both constructions, each of them seems specialized in a different one. A last a priori unexpected significant difference which emerged from the data concerned the negation of the main event. By this I mean the negation of the verb other than hacer, either when this is the main verb (which will be in the adverbial construction) or when it is in the que-clause subordinate to hacer (which will be in the clausal construction). Thus, the main event is negated in only 5.9% of the cases in adverbial constructions but in 58.8% of the cases in the clausal construction. This suggests the existence of a very frequent pattern hace tiempo que no... / hacía tiempo que no... Clausal hacer thus shows a predilection for a negated verb in its que-clause, which, if the clausal construction were derived from the adverbial, would be difficult to explain (i.e. why would an adverbial select any property of the main verb?). I would like to finish this section with a cautionary tale regarding the use of corpora to investigate the grammatical properties of constructions. As we have been showing, corpora may be very useful ways of spotting and quantifying the trends of frequent constructions, but are not so well suited to gain a knowledge of more infrequent constructions or variants. As is frequently expressed, absence of evidence is not evidence of absence. This must always be kept in mind when working with corpora, especially in more infrequent constructions. Howe (2011:257) noted for instance that despite the possibility of typical verbs to pre-pose their object in non-interrogative cases, this construction variant did not occur in his corpus analysis of spoken Spanish. This is, in my opinion, nothing but an accidental gap in Howe's data like the grammaticality of sentence (103) below shows. This gap has been made possible (and even likely) by the combined effect of two factors: 1) The relatively small proportion of clausal constructions (which is where pre-posing is possible) within the general hacer+time he studies and 2) the low frequency of preposing in non-interrogative contexts in general even in prototypical verbs. It is therefore not unexpected that in his limited sample (which is what a corpus is to language) he finds no instances of preposing like the perfectly grammatical (103): (103) Dos años hace que no le veo two years makes that NEG him see

'It has been two years since I last saw him' A corpus approach to the constructions analyzed here has thus allowed us to discover and quantify their most frequent patterns of use but can hardly inform us about less frequent or marginally grammatical properties which is why data about competence was also gathered in section 4.3.

41 Given that the total number of clausal constructions found initially was, as a result of their lower frequency, not big enough to allow a relative confidence in their extrapolation to the construction as a whole, I searched for additional instances of the clausal construction. A total number of 62 was eventually reached.

48

4.5. The relationship between the adverbial and clausal constructions. Along with the verbal or prepositional nature of hacer in adverbial contructions, a much debated issue is that of the relationship between the clausal and the adverbial constructions. Some like Brucart (2014) consider they are independent constructions while others like Rigau (1999) or Fábregas (to appear) argue they are derivationally related and that the phrase headed by hacer is, in clausal constructions, a constituent fronted from its position in the adverbial construction and thus not the main verb. The fact that the que-clause appears to contain the most salient predicative content in clausal sentences and that it does not have an immediately clear role in the main sentence seems to support the derived status. Many in my opinion unsatisfactory proposals have been made about the syntactic role of the que-clause (Rebollo Torío, 1979; García Fernández, 1999). If we compare the clausal construction to other semantically synonymous sentences, however, these problems seem solve themselves: (104) Han pasado dos años desde que se casaron = Hace dos años que se casaron

have.3PL elapsed two years since that REFL married.3PL makes two years that REFL married.3PL

'It has been two years since they married'

(105) Lleva diez días sin

salir = Hace diez días que no

takes ten days without go.out

sale

makes ten days that NEG go.out.3SG

'He hasn't been out for ten days'

Sentence (104) is aimed at showing that many verbs, especially those which do not have animate or agentive arguments, may be not too prominent regarding its predicative content without this (hopefully) casting doubt upon its role as the main syntactic verb in a sentence. Sentence (105) in turn, shows that, once the formal difference of finiteness is removed from the equation, the semantic contribution and syntactic role of the que-clause in the main sentence can be accommodated more easily into a more classical schema. The fact that que no sale does not constitute a suitable answer to any particular question asked to the verb while sin salir does constitute a suitable answer to cómo (how)42 is probably just a result of the greater elaboration and clausality of the first vis a vis the second. In any case, the phenomenon of hypotaxis, sometimes called cosubordination, has been repeatedly observed in the literature (Hopper & Traugott ,2003:177-181); Lehmann ,1988; Van Valin, 1984 among others) and therefore finding a construction which is not semantically embedded but which is still syntactically dependent should not be considered so striking nowadays. It is widely known (Haegeman, 2012); Aelbrecht et al., 2012) that subordinate, especially adverbial clauses may display so-called main clause phenomena (illocutionary force, speaker-related adverbs...) which could, albeit weakly, have supported a main clause status for the que-clause. However even these we do not find: (106) *Hace tres días que sí le

vi

makes three days that yes him saw.1SG

'*It has been three days since I did see him'

42 This was not necessarily so at earlier periods as witnessed by these examples from the 12 th century: Lorando de los oios que non viestes atal; Cortol el yelmo que lego a la carne (1140, CORDE)

49

(107) *Hace tres días que seguramente le makes three days that

probably

vi/viera

him saw.1SG/saw.SBJV.1SG

'*It has been three days since probably I saw him'

It looks thus difficult to postulate a main clause status for the que-clause rather than for the clause with hacer on these accounts. Many linguists, however, seem not to pay any attention to these issues but still consider the clausal construction derived from the adverbial. Fábregas (to appear) seems to base his analysis on the fact that the clausal construction, like the adverbial, rejects NPs other than measure phrases: (108) *Hace todo el día que no tengo señal makes whole the day that NEG have.1SG signal

(Fábregas, to appear:1)

'I haven't had a signal for the whole day'

This time NPs which are not measure phrases can apparently yield a grammatical sentence in a Spanish variety from Argentina, in which case Fábregas analyzes the clausal construction not as derived like in most other varieties which reject (108) but rather as an independent construction. Other semantically similar constructions like the one with llevar (109) which does not have a comparable adverbial construction do allow that kind of definite time NPs, which seems to support the derived analysis: (109) Llevo todo el día sin señal

take.1SG whole the day without signal

'I haven't had a signal for the whole day' I believe, however, that the observation is not solid enough. On the one hand, if the raison d'être of the clausal construction is, as many authors have argued, to measure time, the observed restriction may be semantic in nature and analogous to the one found in other verbs of physical measurement: (110) *Este tomate pesa todo el kilo this tomato weighs whole the kilo

'*This tomato weighs the whole kilo' On the other hand, in other languages, time clausal constructions which lack a corresponding adverbial variant from which they could possibly derive show a similar dislike for NPs other than measure phrases: (111) ?*It has been the whole day since I last saw you These last two sentences just show that it is possible for the clausal and the adverbial constructions to have the same constraint involving measure phrases without making it unlikely that they are not derivationally related. To argue that the two constructions are synchronically unrelated, Brucart (2014) presents sentences like (112): (112) Hace tres meses de

su

muerte

makes three months from his/her/their death

'It has been three months since (s)he/they died' To get around the problem posed by sentence (112) where there is no clause in which hace tres meses could have been base-generated, Fábregas posits that the clausal with de is indeed base-generated but that it is a construction different from the clausal construction with que altogether. I consider this unlikely since their properties are very similar. With the intention of presenting evidence that the two constructions (the one with de and the one with que) are different, Rigau (1999:324) presents this sentence: 50

(113) *Ayer

hizo tres meses de su licenciatura y que tú te caíste al río

yesterday made.3SG three months of her graduation and that you REFL fell to.the river

'Yesterday it was three months since she graduated and you fell into the river'

This does suggest, like Rigau herself mentions, that the two constituents are different, which is something evident from the beginning from their external composition alone (PP vs CoCl). This, however, does not prove the unrelatedness of both constructions as these examples suggest: (114) *Ayer

vi

a mi madre y la película

yesterday saw.1SG to my mother and the film

'Yesterday I saw my mother and the film'

(115) Ayer

hizo

tres meses de su licenciatura y de que tú te caíste al río

yesterday made.3SG three months of her graduation and of that you REFL fell to.the river

'Yesterday it was three months since she graduated and you fell into the river'

Sentence (114) shows that two constituents may be of a different kind (impossible to coordinate) while still appearing each of them in the same position and structure in a bigger construction. I am assuming here that vi a mi madre (I saw my mother) and vi la película (I saw the film) are two instances of the same construction. Sentence (115) in turn suggests that the clash presented by Rigau (1999) in (113) might be rather superficial, since it can be overcome just by the insertion of a preposition de. Even after all this argumentation, I still consider that the problem posed by (112) for a derivational analysis of the clausal construction with hacer may not be insurmountable. Time adverbials are indeed possible in non-clausal (at least nonverbal) constituents which like su muerte in (112) are eventive and possibly therefore more clausal-like than could be initially thought43: (116) El atraco

al banco el pasado martes supuso el décimo en lo que va de mes

the robbery to.the bank the last

tuesday

was the tenth

in

'Last tuesday's robbery of a bank was the tenth this month'

what goes of month

It could therefore still be maintained that there is a derivational relationship, but more evidence yet is available which suggests otherwise: (117) *Hace tres meses que no le

vi

makes three months that NEG him saw.1SG

Intended: 'I didn't see him three months ago'

(118) Hace tres días

hizo un año que murió mi abuelo

makes three days made.3SG one years that died.3SG my grandfather

'Three days ago it was one year since my grandfather died'

Sentence (117) is problematic from a derivational position because it would be expected to be grammatical, since the equivalent adverbial sentence (No le vi hace tres meses) is indeed possible44. 43 For a discussion on the topic see Deutscher (2009). 44 Note that this sentence has two readings: No le vi hace tres meses = hace tres meses no le vi (the interesting meaning here) with a negation of the event denoted by ver and No le vi hace tres meses = No hace tres meses que le vi. In any case, sentence (117) above would be expected to be grammatical if the clausal construction were derived from the adverbial.

51

Sentence (118) shows simultaneously in a single sentence the clausal and the adverbial constructions. From a derivational perspective, with only one true clause (murió mi abuelo) one would in principle be able to generate only one hace-phrase. The simultaneous presence of the base-generated and the derived structures with different time measures is, in my opinion, impossible to explain derivationally. For (118) we need to acknowledge that hacer+time+que-clause is a verbal template in its own right that denotes an event which can, like any other, be located on the time axis with adverbial expressions of any kind. This along with the following list of grammatical properties which differ from the adverbial to the clausal construction of hacer lead me to argue here in favor of a synchronic independence of the two constructions: Clausal Construction

Adverbial Construction

a) Up-to-now meaning grammatical (119) Hace un año que trabajo aquí

Up-to-now meaning ungrammatical⁴⁵ ?Trabajo aquí hace un año

makes a year that work.1.SG here

work.1SG here makes a year

'I have been working here for a year'

b) Time adjuncts to hacer grammatical (120) Mañana hace un año que te vi

tomorrow makes a year that you saw.1SG

Time adjuncts to hacer ungrammatical45 ?Te vi mañana hace un año you saw.1SG tomorrow makes a year

'Tomorrow it'll be a year since I saw you'

c) Negation of hacer always available (121) No hace 10 días que te vi

Negation of hacer restricted ?*Te vi no hace 10 días

NEG makes 10 days that you saw.1SG

you saw.1SG NEG makes 10 days

'It hasn't been 10 days since I saw you'

d) Subjunctive mood grammatical (122) Puede que haga un año que te

vi

Subjunctive mood ungrammatical *Puede que haga un año te vi

may.be that make.SBJV a year that you saw.1SG may.be that make.SBJV a year you saw.1SG

'It may be one year since I saw you'

e) Compound tenses grammatical (123) Va a hacer un año que te vi

It.is.going.to.make a year that you saw.1SG

Compound tenses ungrammatical *Te vi va a hacer un año you saw.1SG it.is.going.to.make a year

'It will soon be a year since I saw you'

f) Split of hacer and time NP allowed (124) Hace que murió Pedro dos años makes that died.3SG Peter two years

Split of hacer and time NP ungrammatical *Hace murió Pedro dos años makes died.3SG Peter two years

'It has been two years since Peter died'

g) Pre-posing of time NP grammatical (125) Dos años hace que murió Pedro two years makes that died.3SG Peter

Pre-posing of time NP ungrammatical *Pedro murió dos años hace Peter died.3SG two years makes

'It has been two years since Peter died'

45 For most speakers. See section 4.3. and annex 3.

52

h) Negation of the main event in distance-past ungrammatical (126) *Hace dos años que no te

Negation of the main event in distance-past grammatical Hace dos años no te vi

vi

makes two years that NEG you saw.1SG

makes two years NEG you saw.1SG

'Two years ago I didn't see you'

This of course does not preclude the existence of many other properties which are shared. Among these we can mention: i) The rejection of NPs other than measure phrases j) The availability of the distance-past meaning k) The preference for a pre-position of hacer with respect to the time NP l) The continued use of non-present forms of hacer (e.g. hacía) The properties or the clausal and adverbial temporal constructions with hacer have been shown to be quite different, both in usage as well as in native speakers' intuitions. Their sets of properties constitute disjoint albeit overlapping groups:

Figure 14 This is the reason why I consider the most parsimonious explanation of these differences to regard both constructions as synchronically independent. Note that the properties of the expression in a truly derived construction (e.g. in the cleft Fue hace dos años que/cuando te vi) do not change with respect to the adverbial source construction. Based on all the properties we have seen throughout this section and the previous ones, I propose for the two constructions, structures along these broad lines:

Figure 15 Clausal hacer on the one hand, is nearer to being a prototypical verb than to being any other thing, since, as far as we have seen, the only typical verbal property it lacks is the ability to take a subject. On the other hand, on account of its synchronic properties, the adverbial construction hace should be better placed in the continuum between verbs and prepositions but, in my opinion, nearer to prototypical adpositions than to prototypical verbs. 53

With this, however, I do not want to imply that in the mind of the speaker there is absolutely no connection between both constructions. I do not even dismiss the possibility that some speakers or variants do have a clausal construction derived by movement from the adverbial. Examples like (127) could suggest it might be so, as the subjunctive mood of the clausal construction appears on the verb other than hacer, thus suggesting for it a hierarchical position higher than that of hacer: (127) Tiene dinero a montones que le viene has

money

loads

del marqués aunque hace rato que se

that her comes from.the marquess despite makes a.while that REFL

hayan divorciado a causa de la marihuana y otras razones parecidas46 have.SBJV divorced

because.of the marijuana and other reasons

similar

'She has lots of money which she got from the marquess even if it is long since they got divorced as a result of marijuana and similar reasons' I believe, however, that examples such as this are isolated and not representative of the most frequent usage and native speaker linguistic competence which has been analyzed here so far. In addition, this example could also be an on-line production error from the author and need not necessarily represent an underlying syntactic structure different from that which has been proposed so far. Such examples do show, however, that the two constructions do influence one another, unsurprisingly on account of their semantic and formal similarity. I believe, for instance, that it is precisely the analogy with the clausal construction that has prevented the complete loss of the verbal properties of hacer in the adverbial construction. On the one hand we see that Spanish and French still have clausal constructions with hace and il y a respectively where their verbal properties are prominent. It is these languages that have also managed to preserve some of the verbal characteristics of the expressions in adverbial position, where features like TAM morphology, negation of the verb, time adjuncts... continue to appear, albeit infrequently, in the modern language. On the other hand we have Italian, which lost the clausal construction with fare. Thus, without the moderating influence of her mother, the grammaticalization process in the adverbial construction with fare was unchecked and in modern Italian all traces of verbality (negation, TAM morphology, word order flexibility...) have been lost. The degree of grammaticalization of the Italian expression fa (or of English ago for that matter) is thus synchronically greater than the one of Spanish adverbial hacer. We have also presented throughout this section, however, that adverbial hacer does show important differences from typical verbs. Is it then synchronically grammatical (i.e. a part of grammar) or lexical?47 46 Documented in Ceremonias:230 by Julio Cortázar and cited by Rasmussen (1981:49). This sentence is ungrammatical in Peninsular Spanish and I believe in most other varieties of the language. The more common construction would involve a subjunctive in hacer rather than in divorciarse: i.e. Tiene dinero a montones que le viene del marqués aunque haga rato que se han divorciado a causa de la marihuana y otras razones parecidas. This would support a higher hierarchical position for hacer in the clausal construction. In the case of the adverbial construction, given that here hacer is indeed subordinate to the other verb, the subjunctive would be assigned like in Cortazar´s example: Tiene dinero a montones que le viene del marqués aunque se hayan divorciado hace rato a causa de la marihuana y otras razones parecidas.

54

(128) Hacer hace una semana que se fue pero llevaba make makes a

week

meses sin hablar-me

that left.3SG but took.IPF.3SG months without talk-to.me

'It has been only a week since he left but he hadn't talked to me in months' (129) Hacer hacía una semana nada más que se había ido pero ya ha vuelto make made.IPF a

week nothing else that had.left.IPF.3SG but already has come.back

'It was only a week only since he had left but he has come back already' (130) *Se fue hacer hace una semana pero llevaba left.3SG make makes a

week

meses

sin hablar-me

but took.IPF.3SG months without talk-to.me

'It has been only a week since he left but he hadn't talked to me in months' (131) *Se había ido hacer hacía una semana nada más pero ya ha vuelto had.left.IPF.3SG make made.IPF a

week nothing else but already has come.back

'It was only a week only since he had left but he has come back already'

Boye and Harder's (2012) test for synchronic grammaticalization thus classifies the hacer of clausal constructions as a lexical element whereas hacer in adverbial constructions appears to be a grammatical element. This is in line with the greater rigidity we have observed in this section in the adverbial construction in comparison with the clausal (see Figure 14) and constitutes further evidence of the distance which synchronically separates the two constructions. It is likely that this is the result of a diachronic process of grammaticalization targeting the adverbial construction while sparing the clausal. In the next section I will try to analyze and explain the diachronic developments attested in each of the two constructions to see what evidence can be found of a process of grammaticalization in the adverbial construction and I will hypothesize, in addition, which mechanism could have been responsible for the split of what must originally have been a single construction into two.

47 This question makes more sense on the assumption that there is synchronically a sharp distinction between the grammar and the lexicon (see section 2.4.), but even if one believes there is not, the synchronic grammaticalization of different expressions is worth investigating in search of differences.

55

5. THE DIACHRONY OF HACER/HABER+TIME 5.1. Introduction The Spanish hacer+time construction is a relatively recent one. With its first documented instances in the 16th century, it was not until the 19 th century that hacer+time replaced the earlier strategy haber+time. The latter had been present in the language since the first writings and appears to be quite ancient, since it can be found in other Romance languages as well (e.g. Portuguese, French or Sicilian). It may have been present, therefore, in Late Latin (132) but was not present in Classical Latin (133) where the verb esse (be) was used instead of habere (have): (132) Pater eius Appollonius, ex quo hinc profectus est, habet annos XIV (Late Latin, father her

Appollonius from whom where

left

is

has

years fourteen

3 c. AD) (Historia Apollonii Regis Tyri 31,7) (Howe & Ranson, 2010:204) 'It is fourteen years since her father Appollonius set out from here' rd

(133) Hanc domum iam multos annos est cum possideo this

(Classical Latin, 2nd c. BC)

house already many years is that inhabit.1SG

(Plautus, Aulularia, 3-4) (Howe & Ranson, 2010:204) 'It is many years that I have been occupying this house' As explained by Pérez Toral (1992:151) there was, still in the earliest writings in Old Spanish, a great overlap in the uses and constructions allowed by the two verbs 48, which was probably the reason for the replacement of esse by habere and for the loss of the subject in the latter. Some isolated examples do appear in the written record, however, which still preserve traces of the verb's earlier ability to take a subject: (134) Esso deve de aver más de dos mill y quinientos años that

must

have more than 2000 and

500

'It must be more than 2500 years since that'

(Pérez Toral, 1992:61)

years

(135) dize que avie Mahomat XL annos de su hedat e IX que regnava (1320-22, CORDE) says that had.3SG Mahomet 40 years of his age and 9 that reigned.3SG

'It is said that Mahomet was 40 years of age and had been reigning for 9 years'

Along haber+time there were, of course, other constructions to express the same time relations (distance-past and up-to-now). We find, for example: (136) (…) fue tornado a senyorio de cristianos 700 annos aca was turned.into an estate

(1305-28, CORDE)

of christians 700 years here

'Was turned into a christian estate 700 years ago'

(137) (…) no se les ha pagado de tres annos aesta parte

(1500, CORDE)

NEG REFL them have paid from three years to.this part

'They haven't been paid for three years'

48 Pérez Toral (1992:51) mentions Cuervo's proposal that the impersonal existential habere (Hubo guerras en España) could have emerged out of the mix of two different constructions: the possessive (España hubo guerras) and the existential (Fueron guerras en España). As evidence of the diffuse borders between these, Pérez Toral mentions early examples such as: Grand alegría es entre todos essos christianos vs El pueblo e la villa hovo grant alegria; Quantos que allí ha vs Quantos que y son. It is well known that there is indeed a semantically motivated cross-linguistic affinity between existence, location and possession (see e.g. Langacker,2004 or Heine 2006).

56

All in all, the impersonal haber+time construction appears well established in Spanish by the time of its earliest attestations in both its clausal and in its adverbial variants and for both the up-to-now meaning and for distance-past.

5.2. About the durative and the puntual meanings As has been just mentioned, the punctual and the durative meanings of the haber+time constructions appear in Spanish from the earliest writings: (138) ¿Quanto

i

a que la vistes?

(12th century)

(Rasmussen, 1981:13)

How.much there has that her saw.2SG

'How long has it been since you saw her?' (139) XLVII annyos ha, al 47

mi cuidar, que de ti no

years has to.the my mind



fablar

(1215, CORDE)

that from you NEG heard.1SG talk

'I haven't heard from you for 47 years, I think'

However, like I argued in section 3.5 and in agreement with Howe & Ranson (2010), I believe the durative up-to-now meaning had to precede the punctual distance-past. On the one hand, many less grammaticalized temporal clausal constructions like the ones with llevar+time can only be used with a durative meaning. On the other hand, we saw in section 3.5 that the main diachronic direction for change was from the durative to the punctual meaning, which is in agreement to the documented trend in Spanish: whereas in the oldest stages of the language the durative meaning was predominant in haber+time, that proportion has declined progressively until the present, where the punctual is predominant. In addition, in section 4.3 we showed that in Spanish nowadays we were experiencing the loss of the ability of the PDM hace to be used with an up-to-now meaning. This same development from durative to punctual has taken place in the past in many other languages as well. These examples show some PDMs which allowed the up-tonow durative meaning earlier but have nowadays become confined to distance-past: (140) trois jours a, ne dormi

(Old French)

(Díez Itza & Pérez Toral, 1991:49)

three days ago NEG sleep.1SG

'I haven't slept for three days' (141) I woot it by myself full yore agon

(ME)

(Chaucer, The Knight's Tale:1813)

I know it by myself many years ago

'I have known that myself for a long time' (142) han er hos Vorherre for snese Aar siden

(Danish)

(Rasmussen, 1981:90)

he is with our.Lord ago tens year ago

'He has been dead for tens of years'

How can an expression used with an up-to-now interpretation come to acquire the distance-past meaning? I cannot provide an answer at this point to how the change was operated since the construction which is analyzed here already had both interpretations from the first documents. There are two observations which are interesting, however: (143) Hace un año que no le veo

makes a year that NEG him see.1SG

'I haven't seen him for a year'

=

Hace un año que le

vi

makes a year that him saw.1SG

'It has been a year since I last saw him' 57

If maximal informativeness is assumed for the sentence containing a negation of an up-to-now event (which is indeed the most usual use of such an expression) it immediately is equivalent to the assertion of the event in distance-past. If the time of the event, rather than the subsequent time interval without the event wants to be emphasized, this may provide a motivation for the change from up-to-now to distancepast. Another this time language-specific link between the two time relations may be: (144) Poco timpo a que es nacida little

(12th century)

(Rasmussen, 1981:13)

time has that is born.F

'She was born a short time ago' The verbal tense illustrated by (144) above, much like English present perfect (e.g. has studied), has undergone an interesting shift in meaning. From having a present resultative meaning it became a perfect and cross-linguistically it often eventually comes to develop a perfective meaning. The change in the temporal relation described by this tense means that, while (144) above may have had an up-to-now meaning initially (i.e. She is in a state of being born; that is, alive and outside the mother's womb), it would not have had the same meaning in later periods. If the time construction with haber+time continued to appear with this tense it would have come to have a distance-past meaning here (i.e. She was born) which could have acted as a bridge to its further use for distance-past in combination with other past tenses.

5.3. About the clausal and the adverbial constructions The same as the durative and punctual meanings of the expressions, the clausal and the adverbial constructions of haber+time and of hacer+time appear in writing approximately at the same time: (145) Pocos dias ha, rey, que una lid ha arrancado few

days has king that a fight has begun

(1140, Poema de Mio Cid, CORDE)

'A fight broke up a few days ago'

(146) Fuera el rey a San Fagunt aún poco ha go.PLUP the king to San Fagunt still little has

(1140, Poema de Mio Cid, CORDE)

'The king had gone to San Fagunt only a short time ago' It is not too problematic, however, to assume that the biclausal construction must have been chronologically prior to the adverbial. There are many reasons to believe so. First of all we have the verbal morphology (and therefore likely verbal origin) of the expression. Despite some proposals for a different, non-verbal origin of hace (Elerick, 1989), the verbal morphology found in the equivalent, cognate expressions in many older Romance languages clearly seems to support a verbal, clausal origin. Second, the lower degree of grammaticalization found in the clausal construction compared to the adverbial suggests that it may be the original, since the same relationship is found in other well-known cases of grammaticalization where there has been a divergence (e.g. Fr. homme > on, Eng. one > a(n)). Third, the earliest documented stages of Spanish still show a quantitative predominance of the clausal over the adverbial construction which only later is reverted, when the adverbial construction increased in frequency probably hand in 58

hand with its grammaticalization and loss of clausal status. These diachronic developments will be analyzed in-depth in section 5.7. Last but not least, all languages have some biclausal construction to indicate either a distance-past or an up-to-now meaning whereas not all languages have the possibility of encoding those time relations in a single clause as was explained in section 3.1. This along with the verbal etymology of many PDMs argues for a chronological precedence of biclausal over mono-clausal strategies.

5.4. Time biclausal constructions: Sources I provided in section 3.3 a cross-linguistic overview of the PDMs in monoclausal constructions and I have argued that biclausal constructions are often the origin of monoclausal strategies involving PDMs but which are the sources of these biclausal constructions? Time biclausal constructions are very diverse across languages. As argued by Haspelmath (1997:136-138) most are based upon light, desemanticized verbs like 'make', 'exist', 'become', 'pass'... but others do not even include a verb and may be based on presentatives: (147) Vot uzhe pjat' let kak ja zhivu v Parizhe PREST already five year that I live in Paris

(Russian)

(Haspelmath, 1997:138)

'I have been living in Paris for five years'

Other biclausal constructions are based on copulative verbs: (148) Noh sâl ast ke dar Bamberg zendegi mi-kon-am (Persian) (Haspelmath,1997:136) nine year is that in

Bamberg

living

IPF-do-1SG

'I have been living in Bamberg for nine years'

I have been here defending the biclausal origin of many monoclausal structures with PDMs. This of course does not preclude the possibility that some biclausal constructions actually are in turn derived from monoclausal constructions. Analogously to Fábregas' (to appear) or Rigau's (1999) synchronic analysis of the Spanish biclausal construction, some biclausal constructions like (148) could well be derived from monoclausal ones by a means of a cleft-like focalizing construction. Another obvious possible source for biclausal constructions is clause fusion; that is, the gradual combination of two erstwhile independent clauses into one: (149) ŋwә́ táa jc kû. ndwә́ táa jc lùu ŋú'sә bɔ̀ɔ shɔ̂ɔ iɔ̀ɔ shɔ̂ɔ he die

(Babungo)

(Haspelmath, 1997:55)

now be years two

'He died two years ago' (lit. 'He died. It's now two years.') Two independent but semantically connected clauses like these could well combine into a complex biclausal sentence by the subordination or loss of syntactic independence of one of the clauses. These processes have received a fair degree of attention (e.g. Hopper & Traugott, 2003:177; Fischer, 2007:222) and I will not comment further upon this possibility. I would like to focus instead on a third, less obvious possible source of temporal biclausal constructions. Clause elaboration (Lehmann, 1988; Deutscher, 2000), also labeled clause expansion (Heine, 2009), is the term used for the gradual acquisition of clausal features like adjuncts, direct objects, subjects, finite verb morphology etc. by erstwhile non-clausal constituents. 59

A development along these lines has taken place in temporal clausal constructions like the ones with llevar (to take) in Spanish, which are equivalent to the clausal constructions with hacer with an up-to-now meaning. The initial sparse instances of temporal llevar (150) (151) (152) differed importantly from modern usage (153) (154): (150) en quince años que lleva la fundación ha adelantado poco in fifteen years that takes the foundation has advanced

(1803, CORDE)

little

'In its 15 year existence, the foundation has achieved little' (151) en los 22 años que lleva de continua residencia en dicha Iglesia in the 22 years that takes of continuous

stay

'in the 22 years (s)he has stayed in that church' (152) (…) cuarenta años que llevo de andar por forty

years that take.1SG of

(1828, CORDE)

in the.said church

el mundo

(1834, CORDE)

walk through the world

'I have been forty years roaming through the world' (153) Lleva diez años viviendo en Londres takes ten years

living

in

London

'He has been living in London for ten years' (154) Lleva días que me trae por la calle de la amargura (1984, CREA) takes days that me brings through the street of the sorrow

'(S)he has been troubling me for some days'

In modern examples (153) and (154) there is a semantically “main event” (vivir en Londres and traer por la calle de la amargura) which is being located in the time axis with an up-to-now meaning. In contrast, the earliest appearances of temporal llevar (150) and (151) did not include any sort of event located in the time axis. The only clause and event were those denoted by llevar. A small-scale corpus research in CORDE and CREA confirmed that these are valid generalizations. For the period 1800-1850 (Spain), out of a total of 674 tokens of llevar, only 20 (2.9%) had a temporal meaning. Of these, not a single one was found where the event was coded by a gerund like (153) or a participle and only a modest 15% was represented by an infinitive like (152). In the remaining 85% of the cases, the event, if there was any, was a deverbal noun, usually preceded by a preposition like in (151). For the period 1990-2000 (Spain), on the contrary, a much greater proportion of temporal tokens of llevar was found. Out of 196 inspected cases, 20 (10.2%) had a temporal meaning. In 60% of these, the event located in the time axis had the form of a gerund and in 10% of a participle. As can be seen, there has been a syntactic elaboration of a predominantly nominal constituent into a more verbal, clausal one. The process of clause elaboration was probably operated in the following fashion. Initially, like any other NP, the time phrase which was the object of llevar (e.g. 22 años) could take a modifier providing further information about the nature of that time extent (e.g. 22 años de sufrimiento; 22 years of suffering) but this (151) was still a monoclausal structure. The frequent inclusion in these positions of deverbal, eventive nouns made it possible to assign a new structure to the construction whereby the modifier of the time phrase became a modifier of llevar instead: (155) Lleva ((diez años) de viaje) takes

ten years of travel

--------------->

(Lleva (diez años)) de viaje takes

ten years of travel

'He has experienced ten years of travelling' 'He has been travelling for ten years' 60

Once it was understood that way, there was nothing preventing more verbal structures like infinitives (152) and later gerunds (153) or participles from appearing in this construction. Given that the main predicative content is that expressed by the subordinate verb (gerund, infinitive or participle) rather than by llevar, it is not surprising that speakers want to provide more information about it. The non-finite verb morphology is thus an obstacle, so some speakers have further elaborated the clausality of that phrase to allow finite clauses like (154). In any case, with its up-tonow meaning, the time construction with llevar is already competing with the clausal construction with hacer, which would not have been possible before elaboration. Only time will tell whether the construction with llevar and a finite clause will spread through the community to become standard, whether it will give rise to a monoclausal strategy or whether it will develop distance-past uses.

5.5. Time biclausal constructions: Outcomes A monoclausal construction containing a PDM may be a frequent diachronic outcome of time biclausal constructions like we have seen in previous sections, but it is surely not the only possibility. Biclausal constructions based on copulative and other very frequent verbs are probably less likely to grammaticalize into PDMs and lose their clausal status because of their sheer frequency of occurrence in other uses: (156) Noh sâl ast ke dar Bamberg zendegi mi-kon-am (Persian) (Haspelmath,1997:136) nine year is that in

Bamberg

living

IPF-do-1SG

'I have been living in Bamberg for nine years'

As was mentioned in section 4.6, the loss of the source construction (as was the case in Italian fa) or of the lexical item (like the English verb ago) may be a big step towards the grammaticalization of a PDM and the loss of the clausal status of a PDP. This can hardly happen if the verb or construction in question is extraordinarily productive in other non-temporal contexts. Despite the mentioned limitations, biclausal constructions are clearly a source for monoclausal strategies, as was already noticed by Haspelmath (1997:55). For some languages with de-verbal PDMs, proposals along these lines have already been made. Franco (2012) for example, notes that, unlike in Spanish, in Italian, the clausal construction is attested earlier than the adverbial and, thus, he derives the much more grammaticalized PDM fa from the clausal construction with fare ('to make', cognate to Spanish hacer): (157) Oggi fa

l'anno che nel

ciel

salisti

(Old Italian)

(Franco, 2012:67)

today makes the.year that to.the heaven went.2SG

'It is one year since you went to heaven'

Similarly, Bourdin (2011:48) notes briefly that the English PDM ago might have arisen from a clausal construction where the Old English verb āgān (to leave) was the main verb in the sentence. The following is an illustrative Middle English example: (158) Ye woot youreself wel how that ye cam heere into this hous, it is nat longe ago You know yourself

well how that you came here into this house it is not long ago

(Middle English) (Chaucer, The Clerk's Tale:477) 'You know yourself well that it is not long since you came here into this house'

61

Note that a sequence like it is longe ago(n) that... is structurally ambiguous. On the one hand it could represent a clausal structure with a “normal” verb ago (to leave) in a structure parallel to that used with be in modern English it has been seven years since... On the other hand, it might represent a cleft-like structure in which longe ago(n) would be a fronted, focalized constituent like next month is in it is next month that I´ll see him. The two different interpretations are illustrated by the parallel translations provided below: (159) it is a long time agone that I came ---------> it is a long time agone that I came it has a long time been since I came it was a long time ago that I came Note at this point that a participle like agone would have appeared finally in older stages of English and that a verb of motion like ago would have taken be rather than have as an auxiliary. In fact I believe that the change in these two aspects during the Middle English period might have contributed to the reanalysis of the original clausal construction into a cleft from which, by syntactic back-formation (García-Castillero, 2014), the modern adjunctival ago-phrases could then start to appear. We have seen that, in some languages with deverbal PDMs like Italian or English, proposals have been made to explain the change from a biclausal to a monoclausal construction. Given the chronological precedence that I have also posited for the Spanish time (bi)clausal construction over the adverbial, I will propose in the next section some ways in which the latter might have arisen from the former.

5.6. The emergence of the adverbial construction: Two proposals 5.6.1. Preliminary observations A corpus research was conducted in CORDE for the earliest documented stages of Spanish. This was aimed at obtaining some knowledge of the earliest properties of haber+time which might help shed some light into how the adverbial construction emerged from the clausal. Two strong tendencies were discovered in the earliest instances of the adverbial construction which I believe demand attention. The first is that, in the earliest adverbial constructions, the phrase headed by haber appeared postposed to the main verb in the vast majority of cases (85-90%) 49. This might seem at first hardly surprising, since adverbial elements tend to be placed predominantly towards the end of the sentence in any case. This postposed position becomes more interesting once we see that, in later periods, these clauses start to appear preposed with a progressively greater frequency until becoming roughly equally frequent to postposition. I believe that the postposed position might therefore constitute the original place in which the adverbial construction first appeared. This renders for example the simple omission of que from the clausal construction an infelicitous explanation for the implementation of the change, since it would predict a predominance of pre-posed adverbial constructions in the earliest stages: 49 In this section I provide only rough numbers obtained from the corpus research in CORDE because the diachronic documented developments of the constructions will be analyzed in depth in section 5.7. In any case, the precise numbers behind the percentages and tendencies mentioned throughout this section can be seen in annex 5.

62

(160) Ha dos años (que) te

vi

has two years that you saw.1SG

'It has been two years that I last saw you' Another interesting tendency is that, in the earliest adverbial constructions, the matrix verb appeared, exceedingly frequently, embedded into a higher clause, very frequently by means of the complementizer or relative pronoun que. I refer to sentences like: (161) están perdidas, que non se labraron grant tienpo ha are.3PL lost.F.PL because NEG REFL plowed.3PL big

time has

(1303, CORDE)

'They (the fields) are ruined because they haven't been plowed for a long time'

(162) respuesta avréis la que no vos pudiera

dar oy ha diez días (1300-5, CORDE)

answer have.FUT.2PL which NEG you can.PLUP.1SG give today has ten days

'You will receive the answer I could not give you ten days ago' (163) querie

yr ha ayudar al Rey don ferrando que yazie sobre coynbra

wanted.3SG go to

siete años auja

help to.the king Mr. Ferrando who laid.3SG above Coimbra

(1325, CORDE)

seven years have.IPF

'He wanted to go to help King Ferrando, who had been in Coimbra for seven years' In 69.4% of the earliest (up to 1400) postposed adverbial constructions (N=34), the main verb appears in an embedded clause, subordinated by que. This proportion is progressively reduced in later periods, reaching for example only 32.2% (N=19) in the 18th century. A fraction of that higher frequency in the earliest periods is explained by the fact that, as mentioned for example by González (1984), Cano (2004:466) or Cervera Rodríguez (2007), the complementizer que enjoyed in Old Spanish a wider array of uses than in later stages of the language. This, however, is insufficient in my opinion to explain the strong preference of the adverbial construction for subordinate, que-containing contexts. This preference for que is obviously especially suspicious because the complementizer is the only linguistic element overtly distinguishing the adverbial from the clausal constructions and, whereas it is expected to occur in the latter it is not in principle expected in the former. I believe this preference of the adverbial construction for que-subordination might be a residue from the grammatical context in which the adverbial construction first arose. Therefore, a finely-tuned proposal for the mechanism responsible for the emergence of the adverbial construction must, in my opinion, be able to account for the initial predominance of post-posed PDPs and for their early preference for que-subordinated matrix clauses. These considerations have guided the two proposals I present next. 5.6.2. Morphological haplology + reanalysis Morphological haplology refers to the phenomenon whereby a morpheme is deleted in the presence of a phonologically identical, most often adjacent phonological sequence or morpheme. The remaining morph or sequence becomes then the simultaneous realization of both elements (i.e. of the preserved and of the deleted, haplologized morpheme). The phenomenon has been dealt with among others by Stemberger (1981), De Lacy (1999) or Nevins (2012). It arguably results from a desire of the speaker to avoid redundancy and from a general cognitive deafness to repetition (Nevins, 2012:10763

108). Despite the general cognitive abilities involved, haplology is not regularly applied to every sequence of morphemes that could potentially trigger it: (164) Gel-miş-miş-Ø

(Turkish)

(Hengeveld & Mackenzie (2008:287)

come-RES-EVID-3SG

'He is said to have come'

It is, nonetheless, very common and one could even say general among the world's languages. Illustrative examples of haplology are: (165) mice+GEN = mice's / *mice' (166) dago+COMP = dagoen be.3SG

'(s)he is'

that.be.3SG

'that (s)he is'

cats+GEN = cats' / *cats's

(Stemberger, 1981)

zegoen+COMP = zegoen / *zegoenen

be.PST.3SG

'(s)he was'

that.be.PST.3SG

(Basque)

'that (s)he was'

The first involves the interaction between the English plural and genitive /s/ suffixes. The second involves the interaction between the Basque past and complementizer /n/ suffixes. As can be seen, they are mutually exclusive in that only one of them can appear phonologically but this single affix will behave as representing both identical affixes simultaneously. It was noted in the definition of the phenomenon that it most often applies in contexts of strict adjacency of the morphs in question. There are also cases, however, where two identical morphemes are banned from occurring proximate to one another: (167) *Raam-ko baccõ-ko samhaalna paḍaa50

(Nevins, 2010:105)

Raam-DAT children-ACC take.care fell.3SG

'To Raam fell the taking care of the children' (168) Raam-ko

kal

baccõ-ko samhaalna paḍaa

Raam-DAT yesterday children-ACC take.care fell.3SG

(Nevins, 2010:105)

'Yesterday to Raam fell the taking care of the children' Going back to our Spanish constructions, it was noted that in a big proportion of adverbial constructions with haber, the matrix verb was subordinated by que, as was in clausal constructions. The obvious question is whether (some of) these ques might be performing a double duty within their sentences. Cases of morphological haplology (169) and other strategies for the avoidance of repetition (170) involving the complementizer que are synchronically attested in Romance languages: (169) Je préfère que tu restes, plutôt que (*que) tu t'en ailles (French) (Nevins, 2010:92) I

prefer that you stay

rather than that you go.away

'I would like you to stay rather than leave' (170) Mejor que te

ayuden que no que te peguen

better that you help.SBJV.3PL than Ø that you hit.SBJV.3PL

(Spanish)

'I'd rather get helped than hit'

In the Spanish adverbial constructions embedded with que (161-163) it can be seen that, if those ques, the same as that in sentence (169), were performing a double duty, 50 This is in fact quite similar to the interaction found in Spanish between the a which marks animate direct objects and the a which marks indirect objects. Note that a sentence like Ayer le robaron a una madre a su bebé is grammatically degraded whereas if the two a-headed constituents are not contiguous the sentence improves: A una madre le robaron ayer a su bebé

64

those sentences would in fact cease to be instances of the adverbial construction and become clausal constructions in disguise: (171) están perdidas, que (*que) non se labraron grant tienpo ha are.3PL lost.F.PL because that NEG REFL plowed.3PL big

(1303, CORDE)

time has

'(The fields) are ruined because it has been a long time since they were last plowed'

The problem posed by this analysis is that at present, a word order where the queclause precedes hacer is in Spanish rare at best and most likely ungrammatical for most speakers. The said word order was, however, more common in earlier Romance: (172) Che no-llo

avìe veduto già

fa grande tempo (1375, Tuscan) (Franco, 2012)

that NEG-him had.1SG seen already makes big

time

'It is already a long time since I last saw him' (173) (…) todos los dela mi tierra me otorgaron los diezmos delos puertos por tres everyone the of.the my land me.DAT gave.3PL the

tithe

of.the

ports

for three

annos, non lo auiendo de ffuero, et que sson passados los tres annos tienpo á years NEG it

having compulsory and that are.3PL elapsed the three years

time ago

(Old Spanish) (1329, CORDE) 'Everyone in my land gave me the tithe for three years without having to and it has been a long time since the three years elapsed'

The infrequency of the word order above in the clausal construction in the earliest stages of Spanish is still more remarkable if we take into account that, for some speakers at least, the que-clause might have been the subject of hacer: (174) pero quanto tiempo a esto

(1334, CORDE)

but how.much time has this

'But how long has it been from this?' (175) esto ha bien 22 años

(1338, CORDE)

this has good 22 years

'It has been a good 22 years from this' (176) Esso deve de aver más de dos mill y quinientos años that

must

have more than 2000 and

500

'It must be more than 2500 years since that'

(Pérez Toral, 1992:61)

years

If some sentences like 152-154 are indeed clausal constructions in disguise, preposed que-clauses were not as infrequent as they appear to be initially. The first proposal I present here for the initial emergence of the adverbial construction, therefore, is the following: It might be that, at some point, some speakers decided to avoid a sequence of two consecutive instances of que by the phonological deletion of one of them (i.e. by haplology) 51. The remaining que, as in other cases of morphological haplology, would have come to perform the functions of both morphs simultaneously, thus still subordinating the following clause to haber. Later on, some speakers reanalyzed 52 the underlying structure of those sequences as not involving haplology at all and therefore as a construction structurally different 51 Alternatively it could be proposed that the sequence que que never came into being because of this will to avoid repetition and that haplology was a synchronically operative rule in the language. 52 I do not imply that this reanalysis had to be a drammatic, one-step event. Reanalyses may be gradual, as put forth by some scholars (e.g. Bybee, 2010).

65

from the clausal. Unsurprisingly, when faced with a sentence with two clauses but no markers of subordination whatsoever, speakers decided to make the clause containing time information subordinate to that containing the event information:

Figure 16 The stages proposed here would therefore be: 1st: Morphological haplology to avoid the cacophonic sequence que que. 2nd: Reanalysis of the underlying structure. 3rd: Actualization: structure spread to other syntactic (e.g. non-subordinated) contexts The reason therefore, that, in the earliest periods, we find such a high proportion of que-embedded adverbial constructions could thus be the product of some speakers still preserving a variant of the grammar where the adverbial construction was nothing but a haplologized version of the clausal construction and thus different in surface realization but identical in underlying structure. That there exist also numerous uses where the adverbial construction is not subordinated by que proves, however, that this was not the grammatical system of all speakers. The grammar of those which could produce sentences like (177) below must inevitably have already undergone reanalysis, thus having definitely split the adverbial from the clausal construction: (177) pasaron aesta tierra de españa tienpo auja passed.3PL to.this land of

Spain

ya

(1344, CORDE)

time have.IPF already

'They had already come to Spain some time before'

As was deemed necessary in the introduction, the present proposal can account for the two tendencies found for the earliest adverbial constructions. On the one hand, it can easily explain the great proportion of que-subordinated adverbial constructions as an intermediate stage in the emergence of that construction from the clausal. On the other hand, it explains successfully the initial predominance of the postposing of the haber-clause in adverbial constructions. This would just be a result of the word order in which the construction first emerged: Preposed haber-clause: No haplology can occur: Remains clausal (178) Creo que ha tiempo que te vi ---> Creo que ha tiempo que te vi Postposed haber-clause: Haplology can occur: Becomes adverbial (179) Creo que que te vi ha tiempo ---> Creo que que te vi ha tiempo53 53 Diachronic explanations relying like this one on morphological haplology are uncommon but by no means nonexistent. An example of a proposal in that direction, involving the emergence of complementizer that out of the demonstrative pronoun, can be found in Fischer (2007:223).

66

5.6.3. Reanalysis of surface structure + syntactic back-formation The second proposal I make concerning how the adverbial construction with haber might have emerged from the clausal involves the observation that there existed, in Old Spanish, subordinate topic constructions like the following: (180) Rogando al Criador quanto ella meior sabe, que a Mio Çid el Campeador begging to.the Maker

que Dios le

curias

as

she

de

best knows that to Mio Cid the reconnoitrer

mal

(1140, CORDE)

that God him protect.SBJV.3SG from misfortune

'Begging the Lord as well as she could to protect the Cid from misfortune' (181) Vio-lo Myo Çid que con los averes que

auien tomados, que si-s pudiessen yr,

saw.3SG-it Mio Cid that with the goods that have.IPF.3PL taken.PL that if-REFL can.SBJV.3PL go

fer lo yen

de grado

(1140, CORDE)

make it would.3PL gladly

'The Cid saw that if they could leave with the goods they had taken, they would do it gladly' (182) Ordenaron assi que los germanos que fincasen en sus tierras (Wanner, 2002:32) order.PST.3PL thus that the Germans that settle.SBJV.3PL in their lands

'Thus they ordered that the Germans settle on their lands'

This subordinate topic construction,54 which for obvious reasons often receives the name of 'complementizer doubling', was also frequent in other Romance varieties: (183) Le aveva ditto che se sua maistà

voleva

lo stato suo che se llo venesse

him had.3SG told that if his majesty want.IPF.3SG the state his that him it come.SBJV.3SG

a ppigliare co la spata in mano

to fetch

(Old Neapolitan)

(Munaro, 2015)

with the sword in hand

'He had told him that if the king wanted his state he should come and take it by force'

As authors like Wanner (2002:42) mention, the construction was quite popular in Old Spanish but later became more infrequent. It is marginal, but still possible, in the modern language: (184) Le dijeron que si

quería

que

podía

ir con ellos

him tell.PST.3PL that if want.IPF.3SG that can.IPF.3SG go with them

'They told him than he could go with them if he wanted to' (185) Me

dijo

que el sábado que no contáramos con él

me tell.PST.3SG that the saturday that NEG count.SBJV.1PL with him

'He told me to count him out on Saturday'

As can be seen, the subordinate topic construction consists on the fronting to the left of some constituent from the subordinate clause and its separation from that clause by a complementizer. The result is that the fronted, topicalized element appears 'sandwiched' between two ques even if semantically it is part of the subordinate. The proposal I make here for the mechanism responsible for the change from the clausal to the adverbial construction with haber capitalizes from the similarity of a queembedded time clausal construction with the subordinate topic construction:

54 Note that this construction should not be confused with other similar and possibly not unrelated constructions like the so-called 'que de racchochage' (Pusch, 2001:220) or with syntactic prolepsis.

67

(186) Viste a abraham que ha mas de mjll a[ñ]os que es muerto see.PST.2SG Abraham that has more than 1000 years that is

(1293, CORDE)

dead

'You saw Abraham, who has been dead for more than 1000 years'

(187) Susanna sepas

que ha grand tienpo que somos enamorados de ti (1293, CORDE)

Susanna know.SBJV.3SG that has big time

that are.3PL

in.love.PL

of you

'Know, Susanna, that we have been in love with you for a long time' (188) Commo quier que tienpo auja que sanna les as

tenja

wants that time have.IPF that hatred them have.IPF.3SG

(1350, CORDE)

'Since he had hated them for a long time'

(189) (...) decia que habia quince años que se say.IPF.3SG that have.IPF

15

habia perdido alli

years that REFL have.IPF.3SG lost there

(1527, CORDE)

'He was saying that he had got lost there 15 years ago'

(190) Las barbas de plata, que no ha veinte años que fueron de oro (1613, Cervantes) the beards of silver that NEG has twenty years that were.3PL of gold

'A silver-like beard that was gold-like less than 20 years ago'

Given the superficial similarity and semantic compatibility of these sentences with the subordinate topic constructions, these sequences may have been reanalyzed by some speakers as instances of the subordinate topic constructions. As discussed e.g. by Langacker (1977:110), the ideal, transparent situation in language, would be that a given surface string can be assigned a consistent function. In this case, the surface manifestation que... que... when the second que is neither completive nor relative, would ideally have a single associated meaning. Thus, the assimilation of those strings to subordinate topic constructions would be an attempt towards transparency. Note that in subordinate topic constructions, the element which appears between the complementizers is a member of the subordinate clause which has been fronted. If some speakers came to analyze these sentences (186-190) as subordinate topic constructions they would immediately have had the possibility of producing novel utterances like (191) in which the phrase headed by haber is not topicalized: (191) Decía que se

había perdido allí había quince años (adapted from 189)

say.IPF.3SG that REFL have.IPF.3SG lost there have.IPF 15

years

'He was saying that he had got lost there 15 years ago'

The present proposal thus hypothesizes a reanalysis of the syntactic structure of quesubordinated clausal constructions with haber, which were equated to subordinate topic constructions. This is in line with De Smet's (2009) observation that surface ambiguity is not a sufficient explanation for why a certain syntactic reanalysis takes place, since ambiguity logically only exists as a consequence and not as a cause of the reanalysis itself. Here, it is the analogy to the common subordinate topic construction that makes reanalysis possible. This reanalysis would have enabled speakers to produce the corresponding non-topicalized construction, thus giving rise to the adverbial construction as we know it:

68

Figure 17 The present proposal thus involves these stages: 1st: Reanalysis of the que-subordinated clausal construction as a subordinate topic construction. 2nd: Syntactic back-formation of the derived subordinate topic construction into the corresponding underived construction. 3rd: Actualization: Extension of the new construction to other syntactic contexts. The process of actualization would have later spread the new structure to other contexts like subordinate sentences without que or to main clauses. Note, thus, that for some time or some speakers, the clausal and the adverbial constructions would have actually been derivationally related under this hypothesis. However, probably when in the late Middle Ages the subordinate topic construction started to lose ground, the two constructions eventually parted ways since the synchronic connection between them was no longer evident. To conclude, it has to be mentioned that like the previous proposal, this one also accommodates smoothly the initial observations in 5.6.1. It can explain why is it that adverbial constructions initially occur most often in clauses subordinated by que, since, in the clausal constructions which were not embedded, reanalysis did not take place since no analogy could be established there to the subordinate topic constructions. 5.6.4. Discussion Even if both of the proposals I have presented are consistent with the initial predominant properties of the adverbial construction they do differ in important respects. The most important disadvantage of the haplology scenario is, in my opinion, that it requires an infrequently attested word order to be the locus for change. This is a serious drawback that leads me to favor the reanalysis in 5.6.3 as a likelier scenario for the emergence of the adverbial construction. This second proposal, however, is not without its problems. For example, after the relevant sequence had been reanalyzed as a subordinate topic construction, the base-generated place of the haber-clause would not necessarily have automatically been the end of the sentence. Sentence initial constituents such as subjects could also appear in the subordinate topic construction (see 182). A preference from speakers to infer a sentence-final original position for the constituent might have been an analogy to other optional constituents bearing time information. All in all, and even if I am initially inclined to prefer the second scenario (5.6.3) I will let each of the proposals speak for themselves. 69

There is still an important point, independent of the actual implementation of the change, which merits attention. It is remarkable by itself that this change has actually taken place in a similar direction in a number of different languages. In both English, Italian, Spanish etc. what were originally main sentences containing time information became optional constituents within a clause while the originally subordinate clauses were promoted to a main clause status by the deletion of the complementizer. The persistence of that development seems to indicate that there is a strong pressure favoring that development and thus that the change cannot be explained merely by an 'accident' like an isolated punctual reanalysis. In any case that would not explain why the adverbial construction tends to prevail over the clausal once the two constructions are synchronically present in the language 55. I believe, therefore, that there has to be some pressure inherently favoring the emergence of adverbial constructions and their posterior prevalence over the clausal. In this respect, it has to be noted that in the clausal construction the main sentence provides exclusively time information while the subordinate is the one providing the information about the main event. That is arguably cognitively quite anomalous and opposed to the principle of functional stability (see e.g. Heine & Reh, 1984:28 or Hengeveld & Mackenzie, 2008:286) whereby constituents with the same specification tend to be placed in the same structural position. Constituents bearing time information overwhelmingly tend to be optional elements in their clause. It is therefore not unexpected that the main and subordinate roles of these biclausal constructions tend to reverse in order to make the clause containing the main event the matrix clause and to demote time information to an adjunct position. The main obstacle for that development is of course the complementizer (that, che, que...) which in the clausal construction overtly signals the subordination of the que-clause56. Any development which manages to remove it will thus be strongly favored.

55 Note that in English and Italian the original time clausal source constructions (containing the verbs ago and fare respectively) dissapeared while the adverbial constructions (containing adpositional ago and fa) have survived. Similarly, in Spanish, where the two constructions have long been in competition, the frequency of the adverbial construction has been constantly increasing in comparison with that of the clausal (see 5.7). 56 This development is in fact quite parallel to that studied by Thompson & Mulac (1991) or Brinton (2008) for English expressions like I think or I guess. In sentences like I think (that) he came, they initially constituted the matrix clause, which hosted a complement clause. The main propositional content of those sentences was to be found in the complement clauses, however, while the main clause would have only provided epistemic information. Unsurprisingly, there was a change of syntactic roles here as well which promoted the erstwhile complement clause to a main clause status while the earlier main clause was demoted to an epistemic modifier of the clause. The complementizer has logically been a frequent victim of the process, since, as Thompson & Mulac (1991) and Brinton (2008) show, the expressions most often occur now without it and frequently as parentheticals in typical adverbial positions: He came, I think.

70

5.7. A corpus-based quantitative analysis of the diachronic changes in time constructions with haber and hacer Quantitative corpus studies are becoming more and more popular in linguistic research. This is the result of the greater availability of computer-readable corpora on the one hand and of an increasingly scientific approach to Linguistics on the other. Quantitative corpus-based analyses allow for an increased objectivity, extrapolability and replicability of the results. In addition, as mentioned for example by Hilpert & Gries (to appear), the increased observational detail that quantitative analyses provide is necessary to understand the internal dynamics of linguistic change, which might in turn provide information about why a given change happened. Moreover, in corpusdriven studies the observed patterns may provide information the linguist had not necessarily anticipated. 5.7.1. Methodology The present corpus study has been carried out in the Corpus Diacrónico del Español (CORDE) of the Real Academia Española. For reasons of homogeneity I have limited my study to the Spanish spoken in Spain but I do not believe that the diachronic developments that will be presented here differ substantially from those found in most other Spanish speaking regions of America. The time constructions involving haber and hacer as a whole have been the research target of this study. The search, however, has been limited to the present (ha, hace) and the imperfect (había, hacía) forms of the verbs. This is something largely imposed by the fact that, given the small frequency of other verb forms like the past (hubo, hizo) or the future (habrá, hará)57, a significant number of tokens could not have been reached for them. In addition, I believe that the most relevant difference (present vs non-present) can still be successfully observed by comparing present and imperfect. The mentioned verb forms were tracked in CORDE in five different periods: 14 th century, 16th century, 18th century, 2nd half of the 19th century and 2nd half of the 20th century. The exact time span which has been mined for data depended on the possibility to reach the beforehand established number of tokens. Given the lower frequency of the imperfect compared to the present verb forms and given the lower proportion of texts from the earliest periods, the time intervals are longer in the earliest periods and for the imperfect forms of the verbs. The exact time intervals have been specified in annex 5. Inside each of those intervals, a search was conducted for each of the two verb forms and, in the periods where temporal haber and hacer coexisted (18th century and 2nd half of the 19th century), for both verbs. No concordances or collocations were used in the search. On the one hand, because CORDE is not a syntactically tagged corpus. On the other, having searched for specific collocations like ha años or años ha, hace días or días hace would have probably biased the data in unexpected directions. 58 Because of this, every single verb form 59 was inspected “manually” and classified as temporal or non-temporal and as preceded 57 Remember that Howe (2011) for example found that past and future forms of hacer together amounted to less than 1% of the total hacer+time. The present represented in his data 96% and the imperfect 3% of the total instances of hacer+time.

71

by a preposition or not preceded by a preposition. 100 tokens 60 of the temporal construction not preceded by a preposition was set as the goal number of tokens. Each of those them was further classified according to different parameters which were: -Type of construction instantiated (clausal, adverbial preverbal or adverbial postverbal) -Time relation expressed (durative/up-to-now or punctual/distance-past) -Negation of the temporal verb (Presence or absence of negation in haber or hacer) -Negation of the main event (Presence or absence of negation of the verb other than the temporal haber or hacer) -Time adjunction (Presence or absence of time modifiers to temporal haber or hacer) -Position of the temporal verb (haber/hacer preposed or postposed to its time NP) In addition, for the earliest periods, information was also gathered for adverbial constructions on whether they were subordinated by que or were not.61 It needs to be acknowledged that not every single instance of temporal haber or hacer could be classified for these parameters: (192) _¿Conque estáis casado? _Seis años ha. so

are.2PL tired

(1855-95, CORDE)

six years has

'_So are you tired? _For the last six years/It has been six years' For some like (192) it was impossible to decide whether they constituted examples of the clausal or adverbial construction for lack of a sufficient context. For others there was some particular parameter which was problematic: (193) días ha muchos que busco su amor (1505-17, CORDE) days has many that search.1SG her love

'I have been seeking her love for a long time' 58 Note for instance that with the searches specified above, instances like hace 3 días, hace muchos días etc. would have been left out while others like 3 días hace, muchos días hace etc. would have been included. This would have resulted in an overrepresentation of postposed instances of hacer. 59 It has to be noted at this point that, because of the spelling inconsistencies of the earliest periods, ha and había could appear in different forms. Thus, ha appeared variously as ha, a, há or á. More remarkably, había could appear in a vast number of different spellings. The most frequent were había, avia, avía, havía, avíe, auja, habia, hauja, avie, hauie, abía, hauía, hauje, abia, auje... These various spelling differences have been ignored and subsumed under the verbal forms they represent. 60 When the goal number was reached for a certain verb form for a certain period, the search was finished. In three cases (18th century había, 18th century hacía and 2nd half of the 19th century había) the goal number could not be reached and the search was halted when the period mined for data reached the whole century. These data have been ignored later on when the achieved number was below 20, since it was deemed that such a small number would not allow for extrapolation. 61 The purpose of this last bit of information was only to explore whether some residue of an earlier grammatical conditioning might have been preserved up to the earliest documented stages. As was already advanced in the previous section, 69.4% (N=34) of 14 th century postverbal adverbial constructions were subordinated by que. This figure is 59% (N=13) in the 16 th century, 45.7% (N=16) in the 18th century and 32.2% (N=19) in the second half of the 19th century. Given that this parameter was used only in section 5.6, it is not included in annex 5.

72

In this example, the time phrase occurs both before and after ha, so this can hardly be classified as preposed or postposed to its time NP. Those few tokens of the time constructions under study which were not amenable to classification for all the relevant parameters were ignored and have not been included in the figures which will be presented below. It is therefore possible that the frequency of time constructions has been slightly underestimated as a result. 5.7.2. Results The total number of verb forms inspected in the present study has been 31965 of which 16240 correspond to imperfect and 15725 to present verb forms. 3.99% of them (N=1275) appeared in the temporal constructions here studied. Of these, 504 include imperfect verb forms and 771 present verb forms. 89.9% (N=1146) of these were not followed by a preposition. Of these 446 are imperfect and 700 present verb forms. These 1146 tokens are the ones which have been classified for the six parameters presented in the previous section. These are the most important patterns discovered: 5.7.2.1. The replacement of haber by hacer One of the things which makes the period under study special to us is that in these years we witness the transition from an older time construction formed with haber to a new strategy with hacer. This took place most prominently during the 19 th century. Whereas haber had been in these time constructions almost three times more frequent than hacer in the 18th century, the situation was reversed by the end of the 19th century by which time hacer had become the dominant choice: 62

Graphic 1 62 The raw frequency of the forms has been estimated as was explained in Footnote 35. The estimations for the temporal construction as a whole (clausal and adverbial with or without preposition) have been: había, 14th century: 26 pmw; había, 16th century: 26pmw; había, 18th century: 13 pmw; hacía, 19th century: 16 pmw; había, 20th century: 10 pmw; ha, 14th century: 47 pmw; ha, 16th century: 57 pmw; ha, 18th century: 88 pmw: ha, 19th century: 63pmw; hace, 18th century: 33 pmw; hace 19th century: 112 pmw; hace 20th century: 161 pmw. Note that the last figure is still less than half the frequency estimation I made for the temporal hace in contemporary oral Spanish. It is quite probable that most of that difference can be explained because of the different (written vs spoken) source of the data.

73

Concerning the relationship between the two constructions, we find strong reasons to believe that they strongly influenced one another, which is in agreement with most authors like Díez Itza (1992) and Pérez Toral (1992). In the two periods for which we have data from both strategies, haber and hacer feature a similar proportion of clausal constructions, adverbial preverbal and adverbial postverbal constructions:

Graphic 2 There are also important differences, however. With respect to their order relative to the time NP, ha and hace show similar properties in the clausal construction but not in the adverbial, where hace showed a preference for preposing (65,5% N=38 in the 18 th century) and ha for postposing (68,6% N=35 in the 18 th century). In addition, focusing on the 18th century, adverbial haber shows a greater proportion of punctual uses (76,6% N=49) than hacer (51,5% N=35) and a total lack of time adjuncts compared to a 12,1% presence in the case of hacer. In addition, while the two expressions co-existed in the language, some lexemes showed a preference for haber and some for hacer. Años (years) for instance seems to collocate more strongly with haber. Thus, whereas between 1800 and 1860 hace años occurs 19 times, años ha occurs 50. On the other hand poco (a little) seems to prefer the construction with hacer, since, whereas between 1800 and 1880 temporal hace poco occurs 175 times, the combined ha poco and poco ha amount to only 66 times. More remarkable still is the preference of different lexemes for different word orders. Between 1800 and 1880 ha mucho occurs 93 times63 and mucho ha only 4. On the contrary, años ha occurs 73 times while ha años does not occur at all in CORDE between those years. This suggests that there may be several sub-patterns and collocational uses within the time constructions as a whole, which might be relevant for the diachronic development of the structures here studied. A collostructional analysis of hacer+time (Stefanowitsch & Gries, 2003) could be a valuable contribution at this point but exceeds the scope of the present research. 5.7.2.2. Durative and punctual meanings We have mentioned in earlier sections that the durative meaning of the time constructions had to precede the punctual, distance-past meaning. The pattern which emerges from the corpus data seems to support that direction for change, since the 63 Remember at this point the quasi-lexicalized use of no ha mucho and no hace mucho we have been mentioning in earlier sections.

74

proportion of distance-past uses of the constructions tends to increase over time. This trend is common to both the clausal and the adverbial construction and to both present and imperfect verb forms. The trend is most clear, however, in the adverbial construction and with present verb forms. Here, as was already shown in the synchronic section 4.3, the punctual meaning has become virtually compulsory nowadays despite both meanings having been roughly as frequent earlier. Distancepast amounted for example to 60,6% (N=20) of the uses in the 14 th century, to 68% (N=79) in the 18th century and to 94,9% (N=74) in the second half of the 20 th century:

Graphic 3 5.7.2.3. Clausal and adverbial constructions Another previous contention of ours which was that the clausal construction had chronological precedence over the adverbial has also found quantitative diachronic support. During the studied period, for both imperfect and for present verb forms, the tendency has been towards a greater use of the adverbial construction with respect to the clausal. The biggest change in this respect happened from the 16 th century to the second half of the 19th century. In the case of the present forms, the gathered data show that the adverbial construction increased from 20,6% (N=21) in the 16 th century to 81% (N=179) in the second half of the 19 th century. Even if the tendency runs parallel in the present and in the imperfect verb forms, it is worth noting that the imperfect always 'lags behind' the present forms and shows a lower proportion of the adverbial construction in all the periods:

Graphic 4 75

This difference in the behavior of the present and the imperfect forms may be caused by the fact that, as we mentioned earlier in 4.5, the verbs in clausal constructions show a more balanced TAM morphology than the verbs/PDMs in adverbial constructions. The big change which is observable here in the intermediate periods could potentially be explained both by a decline in the use of the clausal construction or by an increase in the use of the adverbial. Our data seem to indicate that it is actually the result of a sharp increase in the frequency of the adverbial construction:

Graphic 5 The frequency did not increase in all the verb forms, however: while that of the present tense forms soared, that of the imperfect forms remained relatively stable and moderate over time. The change, therefore, affected the raw frequency of the adverbial constructions with ha or hace, which exploded from slightly above 10 per million words in the 16th century to nearly 140 per million words in the 19th century. Many scholars like for example Hopper & Traugott (1993:126) mention that an increase in “textual frequency has long been recognized informally as a concomitant of grammaticalization” which should put us under alert. Bybee (2003a), Bybee (2006) or Bybee & Thompson (2000) among others mention that high frequency may have conserving effects in that it may prevent the application of analogy and preserve irregular forms and in that, in constructions, it may preserve features otherwise lost. Extremely high frequency, however, results in the erosion of the inner structure, semantics and properties of linguistic expressions. Whereas some scholars (Bybee, 2010) are optimistic concerning a direct relationship between frequency and grammaticalization, others (Hay, 2001; Gries et al, 2005; Peng, 2012) believe that the correlation between them may be less direct. Hay (2001) for example believes that relative frequency of a particular construction relative to the frequency of its etymological/lexical base might also be a relevant parameter to explain the potential for autonomy of a construction from its source. This was also hinted here previously in section 5.5. In the case of the constructions under study, this would ease the autonomy and the grammaticalization of the hace+time adverbial construction (which in our current, oral usage data amounts to 22,4% (N=92) of the total instances of hace) over that of the adverbial ha+time (which in the 19th century represented only 2,1% N=51 of the total number of tokens of ha) due to the higher frequency of haber over hacer. 76

Grammaticalization and loss of inner structure, in the case of the phrases headed by ha or hace, implies desentencialization, a process largely opposed to the one of clause elaboration I presented in 5.4. Lehmann (1988) mentions that, in this event, a subordinate clause is compressed into a nominal or adverbial constituent of the matrix clause. In the process, the clause will lose the semantic components and categories that make up a full-fledged sentence. Lehmann (1988) mentions among others a loss of illocutionary force (capacity of the sentence to be negated, questioned or asserted), TAM morphology, actants and circumstants and word order freedom. We will look at the rest of the patterns which have arisen in the present corpus research in order to see if there is evidence of other changes of the ones which usually accompany grammaticalization or desentencialization. 5.7.2.4. Word order of the past deictic phrase Regarding the word order of the temporal verb and its accompanying time NP, the overall tendency is toward preposing of the verb or PDM. In clausal constructions very little difference has been found between the word order of haber and that of hacer and their slow progression towards a preposed position is probably parallel to the more general increasing preference for verbs to precede their objects in the language as a whole. Regarding adverbial constructions, however, more substantial differences have been found, since ha, as we have mentioned earlier, tended to follow the time NP whereas hace tended to precede it:

Graphic 6 It is interesting to note that, at its earliest studied periods, in both haber and hacer the word order found in the clausal and in the adverbial constructions was very similar. It was only later that the two diverged. In the case of adverbial construction haber, it tended to preserve a position postposed to its complement despite the change in the clausal construction. In the case of adverbial construction hacer, it advanced faster than the clausal construction toward a rigid position before its complement. Quite in agreement with previous developments, the intermediate periods (from the 16th to the second half of the 19 th century) were the ones which witnessed the biggest leap towards a rigid pre-posing of the PDM, first with the introduction of a predominantly preposed alternative to the earlier postpositional haber and second with the swift progress towards pre-posing from the 18 th century to the second half of the 19th century and by the disappearance of haber as a productive strategy. By the 77

second half of the 20th century hace in adverbial constructions had become compulsorily preposed to its complement time NP. This reduction of word order freedom is a typical consequence of grammaticalization. Lehmann (1992:403) for example mentions that whereas in German weakly grammaticalized adpositions like wegen, entlang or nach allow both pre- and postposing, more grammaticalized ones (e.g. von, zu, in...) can only precede their complements. In Lehmann's (1992:413) words, “Reduction of syntagmatic variability includes fixation of word order. This is why grammaticalization goes hand-in-hand with the loss of word-order freedom”. 5.7.2.5. Time adjuncts The presence of time adjuncts to haber or hacer in the temporal constructions also shows an interesting, quite parallel evolution. I refer to cases such as (194), were there is a time adjunct in the past deictic phrase which overtly specifies its reference time: (194) Agora ha un año no hezistes assí

(1500, CORDE)

now has one year NEG do.PST.2SG so

'You didn't do it like that one year ago'

In the earliest periods (up to the 16 th century), time adjuncts to haber occurred with roughly the same frequency in the adverbial and in the clausal constructions. In the 14th century, for example, we find a 13,2% (N=19) of time adjuncts in the clausal constructions and a 12,5% (N=7) in the adverbial. This changed later as time adjuncts started to be more infrequent in the adverbial construction and more frequent in the clausal. For example in the second half of the 19 th century I found only a 4.5% (N=10) of time adjuncts in adverbial constructions opposed to a 19,5% (N=17) of time adjuncts in clausal constructions:

Graphic 7 This evolution is probably not unrelated to the huge increase in frequency experienced by the adverbial construction (especially with present verb forms of haber and hacer) between the 16th and the 19th century. In the case of the ha in adverbial constructions, the drop from the 16th to the 18th century is remarkable: from a 15.8% to zero. In the case of hace there is also a drop from 12% in the 18 th century to 8,6% in the second half of the 19th century. I take this development to be once more linked to the desententialization of the adverbial construction. Locational time adjuncts are modifiers of states of affairs. The desemantization and loss of eventivity of haber and hacer (especially of their present tense forms) in 78

adverbial constructions meant that there was less motivation to try to locate them in time. The increased use of time adjuncts in the clausal construction could be seen as a side-effect of the reduced clausality of the ha(ce)-phrases of the adverbial constructions. When a time reference, maybe different from the utterance time, wanted to be specified it would have to be done predominantly in a clausal construction, since the adverbial hardly allowed time adjuncts to appear any more. 5.7.2.6. Adpositional uses Since finite clauses without a complementizer do not usually combine with adpositions in Spanish, another interesting feature to look at in order to get further evidence for desententialization is the possibility of the phrases headed by ha or hace to be preceded by a preposition. In Lehmann's (1988:13) words, “the more a subordinate clause is nominalized, the more easily it combines with adpositions and case affixes”. In agreement with what we could expect from our previous findings, whereas in the earliest stages (up to the 18th century) the phrases introduced by ha or hace could hardly be introduced by a preposition, in later periods this has become increasingly frequent:

Graphic 8 This once again seems to confirm that a major change has been operated, in the case of the adverbial construction, in the phrases headed by hace, which could not easily function as complements of adpositions in the earliest periods but have acquired that property later on. I believe the desentencialization of the phrases with hace in adverbial position is also the reason for the present development. When those phrases ceased to be perceived as subordinate finite clauses there was no reason not to extent to them the prepositions that could already occur with other non-clausal adverbial constituents e.g. hasta ahora, desde la semana pasada, de ayer... 5.7.2.7. Negation The picture is more complicated in the case of negation. One reason for this is the existence of the frequent expression no ha mucho/no hace mucho which, probably as a result of its high frequency, as we have argued earlier, has managed to preserve negation while most other combinations have lost it. For example, out of 29 times that the expression ha mucho appears in CORDE between 1860 and 1880, 21 appear negated. This frequent fixed expression distorts our data by increasing the proportion of negated uses of haber/hacer. The result is that our diachronic data do not show this 79

time a clear trend. As was shown in the synchronic section, however, in the modern language (probably also in earlier stages), negation was barred from occurring in most contexts in the adverbial construction. This development is not unexpected. As Lehmann (1988:13) points out “at some stage of strong desentencialization, the polarity of the subordinate clause is also affected. This usually means that it can no longer be independently negated”. 5.7.2.8. Preverbal vs postverbal adverbial construction A pattern which has emerged from the present corpus study and which had not expected initially concerns the evolution of the word order of the adverbial constructions with ha or hace. The Past Deictic Phrases in these constructions can appear initially (195) or towards the end of their matrix clause (196): (195) Y si Dios me concede lo que días ha le pido

and if God me give.3SG what that days has him ask.1SG

(1725, CORDE)

'And if God gives me what I have been asking him for days' (196) Va cumpliendo lo que prometió hace pocos años goes

fulfilling

what that promised.3SG makes few years

(1734, CORDE)

'He has been keeping what he promised a few years ago'

As we have already advanced earlier, it was found that the Past Deictic Phrases (PDPs) appeared in most cases (81.8% N=27) postverbally in the earliest period. The proportion of preverbal PDPs increases gradually until the 18 th century, when it became even more frequent that postverbal position (52,3% N=57). Later it decreased again up to the present, when postverbal position is again predominant (82% N=64). This trend has never been spotted before as far as I know. The figures provided by Pérez Toral (1992:115) for instance show rather that the postverbal position was always predominant and that the proportion (around 70% postverbal vs 30% preverbal) was indeed quite stable from the Middle Ages to the present. There are two things, however, which make me trust by numbers over Pérez Toral's. On the one hand the present study is based upon a bigger number of analyzed cases. Thus, for the crucial 18th century, she relies on 34 observations whereas the present study has been based on 109. On the other hand, the same development is observed in my data for both ha and hace, which increases my confidence that the observation is indeed significant and not a product of chance:

Graphic 9 80

A different thing is how to account for the trend itself and what could explain the initial increase of preverbal PDPs and the posterior decrease. We see once again that the turning point is to be found in the intermediate periods, so that it may be again linked to the previously presented diachronic developments. If the predominantly postverbal position of the earliest periods was still a residue of the grammatical context in which the adverbial construction arose in the first place, we might explain the initial increase in frequency of preverbal PDPs as a normal adjustment and as part of the actualization after the initial reanalysis, but what would be the pressure for the clauses headed by haber or hacer to occur initially in a great number of cases? Diessel (2008) for instance notes that, overall, time adverbial clauses are preferably placed after the main clause and in a chronologically iconic word order. The phrases headed by ha or hace would thus be doubly expected to occur postposed; on the one hand, because of the general preference to place time adverbials sentence-finally and on the other because the event they present takes place chronologically later than the event of the matrix clause: (197) Te

vi

hace dos años

you saw.1SG makes two years

'I saw you two years ago'

In the situation described by (197), the event of me seeing you is chronologically prior to the event of having elapsed 2 years since that. It is actually the canonical word order of the clausal construction which is anti-iconic: (198) Hace dos años que te

vi

makes two years that you saw.1SG

'It has been two years since I saw you'

The event of elapsing the two years is posterior to that of me seeing you but is presented first. It might be, therefore, that it is this word order of the clausal construction which is having an impact on that of the adverbial, favoring a sentence initial position by analogy between the two constructions. In the 18 th and especially in the 19th century, however, as we have seen, the adverbial construction became extraordinarily frequent and largely lost its clausal status, which made earlier sentences like (199) no longer possible: (199) Murió Ana Le-Fevre pienso que ha tres, o quatro años died.3SG Ana Le-Fevre believe.1SG that has three or four

(1726, CORDE)

years

'Ana Le-Favre died three or four years ago, I believe' As the properties of the two constructions became increasingly different, the analogical link between them would have been loosened, leading to an incipient emancipation of the adverbial construction from the clausal and to a divergence between a verbal hace in the clausal construction and a more prepositional-like hace64 in the adverbial construction. In the latest periods, therefore, the influence from the clausal construction upon the adverbial would have been much more reduced, which could have contributed to reestablishing the preference for sentence final PDPs we observed in the data. 64 As many authors mention, however, (e.g. Bybee, 2003:162) “when grammaticalization is occurring, it may not be possible to uniquely assign elements to particular grammatical categories or structures”.

81

5.7.3. Conclusion Through the different parameters which have been presented in this section it has been shown that the diachronic development of the time adverbial constructions with ha and hace shows many of the characteristics typical of grammaticalization. During the 17th, 18th and 19th centuries adverbial constructions with ha or hace increased their frequency exponentially and underwent changes in many of the parameters identified by Lehmann (1995) as diagnostic of grammaticalization. We observe a strong erosion of the inflectional morphology of haber/hacer which become in the adverbial construction almost fixed in its earlier present tense forms (integrity). In this construction, hace has acquired many adpositional characteristics, advancing on its path from verb to adposition and thus from a major to a minor word class (paradigmaticity). The expression has become compulsory for the expression of a distance-past time relation (paradigmatic variability) and has acquired a rigid word order since hace must now in these positions occur before its time NP (syntagmatic variability). The past deictic phrase in addition has lost or severely restricted many of its earlier clausal features like negation, assertion, time adjunction or the possibility to include expressions that make reference to the speaker. None of these developments or the previous ones have taken place in the clausal construction however, and thus we are faced with a divergence between the adverbial and the clausal construction hace. As Hopper & Traugott (2003:118) mention, the phenomenon known as divergence “is a natural outcome of the process of grammaticalization, which begins as a fixing of a lexical form in a specific potentially grammatical environment where the form takes on a new meaning”. The progressive differentiation of the adverbial and the clausal constructions which began around the 17th century is, however, probably still underway, which may be the reason for the uncertain grammaticality of many of the sentences provided by prescriptive grammars as illustrative of specific properties or for the inter-generational differences in the grammaticality judgements presented in section 4.3. This would constitute an excellent opportunity to investigate grammaticalization and desententialization in real time. Additional research would be greatly appreciated here.

82

6. CONCLUSION The present work has focused on so-called Past Deictic Markers (PDMs), expressions like Eng. ago, Sp. hace, German vor, Rus. nazad, Basq. duela etc. which are used for the introduction of time adverbials for the location of an event at some specific point in the past. The general aim has been to analyze those expressions in synchrony and diachrony from a cross-linguistic perspective, with a focus on Spanish hacer+time as one of the most attractive of the strategies under study. In a typological introduction to the expressions, it has been shown that there exists a big asymmetry between the word order found in these constructions and that of the equivalent constructions for the future. Whereas in the latter the word order is most often harmonic with the rest of the parameters in the language, in the former a postposed position is the most frequent option, even among prepositional languages. More precisely, PDMs which are exclusively dedicated to the expression of a single time relation have been shown to be the ones which show dis-harmonic features whereas polysemous PDMs are actually perfectly well-behaved. This evidence, along with the synchronically observable etymologies of many PDMs and their patterns of polysemy have led me to hypothesize two different diachronic sources of PDMs. A primary grammaticalization path leads to PDMs directly from verbs in biclausal time constructions. The expressions arising in this way are the ones which display more idiosyncratic word order properties and grammatical behavior. A secondary grammaticalization path leads to PDMs from adverbials of anteriority. The expressions with this origin are usually less remarkable within their languages because they have been 'inside grammar' for a longer time. Two similar grammaticalization paths have been proposed for Future Deictic Markers (FDMs): the secondary grammaticalization from time expressions for posteriority (entirely symmetrical to the one found for the past) and another path which consists in the borrowing of spatial inessive strategies for the expression of the relevant time relation. The difference is that in this case it is also a secondary grammaticalization, which leaves FDMs in general more homogeneous and little noteworthy. Looking for a deeper knowledge of the primary grammaticalization path presented in the typological section and of the grammatical properties exhibited by those expressions, the Spanish past deictic constructions with the verb hacer (to make) have been analyzed in synchrony and diachrony. Native speaker grammaticality judgements and synchronic oral usage seem to agree that the grammatical properties of the clausal and the adverbial construction are much more different than most previous studies have suggested. The adverbial construction has been found to be much more restrictive than the clausal, since corpus data reveal that the present tense amounts to 99% of the instances of the adverbial construction whereas the TAM morphology is more evenly distributed (like in more prototypical verbs) in the case of the clausal construction. Regarding specific grammatical properties, the Spanish adverbial construction seems to be synchronically little tolerant of time adjuncts, negation or word orders other than hacer+time NP. These, however, are shown to be perfectly possible in the clausal construction. These and other contrasts have led me to propose a synchronic independence of both constructions as the most parsimonious structural description of the constructions. 83

Concerning diachronic considerations, it has been argued that the clausal construction must be chronologically prior to the adverbial. Two different proposals have been presented for how the change from the former to the latter might have proceeded, the first one involving morphological haplology and the second one a reanalysis on the basis of an earlier subordinate topic construction. In addition, some proposals have been presented for the origin of the temporal biclausal constructions out of which past deictic adverbials frequently originate. The most developed of them involves clause elaboration of an erstwhile non-clausal modifier and has been illustrated by the diachronic development of the Spanish construction llevar+time. Last but not least, the diachronic development of the Spanish time constructions with hacer and earlier haber has been investigated in a quantitative corpus research. The most relevant finding has been a huge increase in the frequency of the adverbial constructions with the present verb forms ha and hace between the 16th and the 19th century. This along with other diachronic patterns such as the fixation of word order or an increasing rejection of time adjuncts by the adverbial construction suggest that a process of grammaticalization might have affected those constructions while sparing the clausal, increasing as a consequence the grammatical distance between the two constructions. Synchronic findings such as the variability found between different age groups or the unclear grammatical status of some sentences suggest that the grammaticalization of hace might still be underway, which would constitute an excellent opportunity for the study of grammaticalization in real time. Further research should focus on similar cases where the grammaticalization of the PDM in question has been completed (e.g. Eng. ago, It. fa...) in order to see if their initial stages can compare to the developments discovered here for hacer+time. Analyzing the evolution of the textual frequencies of the relevant expressions during grammaticalization would constitute a necessary step toward answering whether a certain frequency threshold (whether absolute or relative) can be identified on a crosslinguistic basis as a trigger (or outcome) for these processes. Subsequent grammatical changes could be compared to see if some pattern can be detected which is common to all instances of grammaticalization from a verb to an adposition. It might also be profitable to analyze in depth the secondary grammaticalizations initially presented within different time relations or in the inessive space-to-time metaphor to gain a deeper, quantitative knowledge of other frequent grammatical developments of time adverbials.

84

ANNEX 1. TDMs' word order, their polysemy vs. monosemy and the preferred word order of adpositions for the sample languages Language

Past Deictic

PD dedicated?

Fut. Deictic FD dedicated?

Adposit.

English

NP PD

Yes

FD NP

No

Prep

German

PD NP

No

FD NP

No

Prep

Swedish

PD NP PD

Yes

FD NP

Yes

Prep

French

PD NP

Yes

FD NP

Yes

Prep

Spanish

PD NP

Yes

FD NP

Yes

Prep

Italian

NP PD

Yes

FD NP

Yes

Prep

Romanian

PD NP

Yes

FD NP

Yes

Prep

Latin

PD NP

No

FD NP

No

Prep

Russian

NP PD

Yes

FD NP

Yes

Prep

Polish

NP PD

No

FD NP

Yes

Prep

Serbo-Croat

PD NP

No

FD NP

Yes

Prep

Bulgarian

PD NP

No

FD NP

No

Prep

Hindi

NP PD

Punjabi

NP PD

Persian

NP FD

Post

No

NP FD

Post

NP PD

No

NP FD

Prep/SOV

Armenian

NP PD

No

NP FD

Post

Albanian

PD NP

No

FD NP

Welsh

NP PD

Yes

FD NP

Irish

NP PD

Yes

FD NP

Yes

Prep

Lithuanian

PD NP

No

FD NP

No

Prep

Latvian

PD NP

No

FD NP

No

Prep

Mod. Greek

PD NP

No

FD NP

No

Prep

FD NP

No

Prep

Anc. Greek

No

Prep Prep

Finnish

NP PD

No

NP-FD

Yes

Post

Estonian

NP PD

Yes

NP FD

No

Post

Hungarian

NP PD

No

NP FD

No

Post

Udmurt

NP PD

No

NP FD

No

Post

Arabic

PD NP

No

FD NP

No

Prep

Maltese

NP PD

Yes

FD NP

Yes

Prep

Hebrew

PD NP

No

FD NP

PD NP PD

Yes

FD NP

Hausa

Prep No

Prep 85

Igbo

NP PD

FD NP

Swahili

Prep

FD NP

No

Prep

Georgian

NP PD

No

NP-FD

No

Post

Lezgian

NP PD

No

NP-FD

No

Post

Chechen

NP PD

No

NP FD

No

Post

Hunzib

NP PD

No

NP FD

No

Post

Abkhaz

NP PD

No

NP FD

Turkish

NP PD

No

NP FD

No

Post

Azeri

NP PD

No

NP FD

No

Post

Japanese

NP PD

No

NP FD

No

Post

Korean

NP PD

No

NP FD

Post

Lao

NP PD

FD NP

Prep

Thai

NP PD

FD NP

Prep

Kannada

NP PD

NP-FD

Post

Malayalam

NP PD

No

NP FD

Post

Tamil

NP PD

No

NP FD

Post

Maori

NP PD

FD NP

Prep

Tagalog

NP PD

FD NP

Prep

Indonesian

NP PD

NP FD

Prep

Basque

PD NP

Yes

NP FD

Yes

Post

Chinese

NP PD

No

NP FD

No

Post/SVO

Nanai

NP PD

Slave

NP PD

No

NP FD

No

Post

Haitian

PD NP

No

FD NP

No

Prep

NP FD

Quechua Eskimo

NP-FD NP PD

No

Hopi NP PD

Abui

PD NP

Post

Post Post

NP FD

Evenki

Post

Post Post

Yes

nd/SOV

86

ANNEX 2. Examples of PDPs and FDPs Language

Past Deictic Phrase

Future Deictic Phrase

English

Three hours ago

In one year

German

Vor zwei Jahren (two years)

In zwei Monaten (two months)

Swedish

För fjorton år sedan (14 years)

Om två dagar (two days)

French

Il y a deux ans (two years)

Dans deux jours (two days)

Spanish

Hace un año (a years)

Dentro de un año (a year)

Italian

Due anni fa (two years)

Tra due anni (two years)

Romanian

Acum paisprezece ani (14 years)

Peste o lună (a month)

Latin

Ante annos 14 (14 years)

Post biduum (two days)

Russian

Dva goda nazad (two years)

Cherez god (a year)

Polish

Sześć miesięcy temu (six months)

Za tydzień (a week)

Serbo-Croat

Prije četrnaest godina (14 years)

Do dva dana (two days)

Bulgarian

Predi edna sedmica (a week)

Sled edna sedmica (a week)

Hindi

Bīsa minața pahalē (20 minutes)

Bīsa minața bāda (20 minutes)

Punjabi

Do saal páílãã (two years)

do kàṇṭe vicc (two hours)

Persian

Do sâ’at-e piš (two hours)

Se sâl-e digar (three years)

Armenian

Erku žam aṙaǰ (two hours)

Erku taru-c‘ heto (two years)

Albanian

Ka dy vjet (two years)

Pas një jave (a week)

Welsh

Ddwy flynedd yn ôl (two years)

Ymhen mis (a month)

Irish

Bliain ó shin (a year)

I gcionn trí lá (three days)

Lithuanian

Prieš tris dienas (three days)

Po kelių minučių (a few minutes)

Latvian

Pirms gada (a year)

Pēc divām stundām (two hours)

Mod. Greek

Prin apó dhió óres (two hours)

Metá apó tris óres (three hours) Metá pollàs hēméras (many days)

Anc. Greek Finnish

Kolme vuotta sitten (four years)

Kahde-ssa tunni-ssa (two hours)

Estonian

kaks tundi tagasi (two hours)

Kahe tunni pärast (two hours)

Hungarian

Három hét elött (three weeks)

Három hét múlva (three weeks)

Udmurt

Odig ar taleś aźlo (one year)

Odig ćas bere (one hour)

Arabic

Munðu ʔarbaʕati ʔayyaamin (four days)

Baʕda yawmayni (two days)

Maltese

Erbat ijiem ilu (four days)

Fi ftit minuti (few minutes)

Hebrew

Lifney šloša yamim (three days)

ʕod yomayim (two days)

Hausa

Cikin awàa biyun dà sukà wucèe (two hours)

Baayan shèekaràa ukkù (three years)

Igbo

Afo abuo gara aga (year two)

N'ime afo abuo (year two)

Swahili

Baada ya siku mbili (two days)

87

Georgian

Or saatis c’in (two hours)

Or saat-ši (two hours)

Lezgian

250 jis idalaj wilik (250 years)

Q’we wacra-laj (two months)

Chechen

Pxi šo ħalxa (five years)

Ill minot jälča (ten minutes)

Hunzib

Q'anu anƛ'i art'o (two weeks)

λaʕel muğáƛ (year)

Abkhaz

Y°ә-sàat-k’ r-àpx’a (two hours)

Y°ә-sàat-k’ rә́ táa j§-la (two hours)

Turkish

Iki yıl önce (two years)

Iki gün sonra (two days)

Azeri

Üç il əvvəl (three years)

Üç il ərzində (three years)

Japanese

Ni-zi-kan mae-ni (two hours)

Ni-zi-kan go-ni (two hours)

Korean

Twu sikan cen-ey (two hours)

Sam nyen twi-ey (three years)

Lao

Songpi konnani (two years)

Nai songpi (two years)

Thai

S¬xng pī thì læd w (two years)

Nı sxng pī (two years)

Kannada

Ardha gaṇṭeya hinde (half an hour)

Aidu nimiṣad-alli (five minutes)

Malayalam

Raṇṭ varṣaṁ mump (two years)

Raṇṭ varṣa-ttinuḷḷil (two years)

Tamil

Muuṇu maṇi-kki munnaale (three hours)

Raṇṭu maṇi neerattle (two hours)

Maori

Rua haora noa atu raa (two hours)

A te rua haora (two hours)

Tagalog

Pitong taong nakaraan (seven years)

Sa loob ng dalawang taon (two years)

Indonesian

Empat hari yang lalu (four days)

Dua hari lagi (two days)

Basque

Duela ehun urte (100 years)

Bi urte barru (two years)

Chinese

Liǎng-ge zhŏngtóu yǐqián (two hours)

Liǎng-ge zhŏngtóu yǐhòu (two hours)

Nanai

ǯuer ajŋaniwa xamasi (two years)

ǯuer ajŋani-du bipie (two years)

Slave

Tǫ dzene t'ǫh (many day)

Tai dzene ndah (three day)

Haitian

fè kat jou (four days)

Nan kèk jou ankò (a few days)

Quechua Eskimo

Ishkay uras-pi (two hours) Nalunaaquttap akunniri pingasut matuma siurnagut (three hours) Hikis taala-t ang (few days)

Hopi Evenki Abui

Ilanma tyrganilva amaski (three days) Afe hetung nuku (year one)

Most of the data above are taken from Haspelmath (1997) and Franco (2013). For individual languages I am indebted to van den Berg (1995:64-65) for Hunzib, Rice (1989:295-97) for Slave and to personal communication from various native speakers.

88

ANNEX 3. Results of the native speaker competence questionnaire + than 50 years old

- than 30 years old

total mean

a) Lleva una semana que no para de llover

4.33

3.17

3.70

b) Tuve el accidente mañana hace tres años

1.53

0.80

1.13

c) Bajó la basura hará dos días

1.80

1.28

1.52

d) Estuve con tu padre dos días hace

0.33

0,00

0.15

e) No estoy con tu hermana hace una semana

3.27

1.83

2.48

f) Lleva una temporada que no doy abasto

2.20

1.44

1.79

g) Aprobó la oposición hace que me lo dijo un año

0.27

0.28

0.28

h) Hace dos generaciones que aquí no había agua corriente

2.80

1.94

2.30

i) Hace una semana que tu padre bajó la basura

3.87

4.50

4.21

j) Juan vino hace tres días, no una semana

3.40

3.72

3.57

k) No te veía hacía mucho tiempo

3.07

3.06

3.06

l) Juan no vino hace tres días sino hace dos

4.27

4.28

4.28

i) 0) No visito a tu familia hace dos meses 5) No visito a tu familia desde hace dos meses

3.73

4.61

4.21

ii) 0) Juan vino hace cinco días, no una semana 5) Juan vino hace cinco días, no hace una semana

3.27

3.28

3.28

iii) 0) Había estado contigo hacía dos días 5) Había estado contigo dos días antes

3.20

3.83

3.54

In an open answer question the most repeated answer to describe a distance-past situation was an adverbial construction hace 5 días. This was received six times by the younger speakers and three times by the older. To express an up-to-now meaning there was a greater variety of responses. The most repeated answer by the younger speakers (only three times) was the adverbial construction desde hace 5 días. Among the older speakers the clausal construction hace cinco días que was the most frequent answer (only twice). Excluded from this annex are distractors and other questions unrelated to the present research. Also excluded are control questions which were included to detect and discard random or careless answers. Four questionnaires were left out of this study as a result. Information regarding gender and social class of the speakers was also collected but has not been included here since it has not been used in the present work.

89

ANNEX 4. Synchronic usage data for hacer+time. From CREA, oral, Spain. Hace 1996-1999, oral, Spain Tokens inspected: 410. Of these temporal: 151. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

8

6

2

0

8

3

5

1

7

8

0

31

1

30

1

30

2

29

1

30

31

0

ADV. (post-verb) 61

5

56

3

58

0

61

4

57

61

0

ADV. (pre-verb)

Hacía 1980-2000, oral, Spain Tokens inspected: 545. Of these temporal: 40. Of these not preceded by preposition: 33 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

22

15

7

0

22

12

10

2

20

21

1

ADV. (pre-verb) 3

0

3

0

3

0

3

0

3

3

0

ADV. (post-verb) 8

3

5

0

8

0

8

0

8

8

0

Hace 1980-2000, oral, Spain (additional search for clausal constructions) Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

32

24

8

0

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

32

19

13

1

31

29

3

ANNEX 5. Diachronic usage data for hacer/haber+time. From CORDE, Spain. Ha 1300-1360 Tokens inspected: 3569. Of these temporal: 100. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

67

40

27

2

65

19

48

6

61

25

42

ADV. (pre-verb) 6

1

5

1

5

1

5

1

5

3

3

ADV. (post-verb) 27

12

15

0

27

6

21

4

23

9

18 90

Ha 1500-1520 Tokens inspected: 3069. Of these temporal: 102. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

81

53

28

7

74

18

63

7

74

44

37

ADV. (pre-verb) 7

3

4

0

7

1

6

1

6

3

4

ADV. (post-verb) 12

7

5

0

12

2

10

2

10

3

9

Ha 1720-1750 Tokens inspected: 2414. Of these temporal: 100. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

49

22

27

7

42

2

47

3

46

32

17

ADV. (pre-verb) 27

4

23

2

25

0

27

0

27

8

19

ADV. (post-verb) 24

3

21

5

19

1

23

0

24

8

16

Ha 1850-1880 Tokens inspected: 3420. Of these temporal: 102. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

23

18

5

1

22

2

21

2

21

16

7

ADV. (pre-verb) 29

8

21

3

26

0

29

0

29

20

9

ADV. (post-verb) 48

6

42

10

38

2

46

0

48

21

27

Hace 1700-1800 Tokens inspected: 1884. Of these temporal: 103. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

42

24

18

2

40

10

32

10

32

30

12

ADV. (pre-verb) 30

13

17

0

30

2

28

5

25

20

10

ADV. (post-verb) 28

10

18

3

25

0

28

2

26

18

10

91

Hace 1860-1870 Tokens inspected: 717. Of these temporal: 119. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

19

6

13

5

14

2

17

5

14

13

6

ADV. (pre-verb) 24

6

18

3

21

3

21

3

21

23

1

ADV. (post-verb) 57

17

40

7

50

0

57

4

53

51

6

Hace 1960-1970 Tokens inspected: 653. Of these temporal: 145. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

22

9

13

2

20

4

18

6

16

20

2

ADV. (pre-verb) 14

1

13

1

13

1

13

1

13

14

0

ADV. (post-verb) 64

3

61

4

60

1

63

5

59

64

0

Había 1300-1400 Tokens inspected: 4050. Of these temporal: 100. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

CLAUSAL

77

51

26

4

73

9

68

13

64

57

20

ADV. (pre-verb)

1

0

1

0

1

0

1

0

1

0

1

ADV. (post-verb) 22

11

11

1

21

1

21

2

20

3

19

Había 1500-1550 Tokens inspected: 3989. Of these temporal: 101. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

86

58

28

5

81

18

68

12

74

63

23

ADV. (pre-verb) 3

1

2

0

3

1

2

0

3

3

0

ADV. (post-verb) 11

8

3

0

11

1

10

2

9

1

10

92

Había 1700-1800 Tokens inspected: 3901. Of these temporal: 39. Of these not preceded by preposition: 39 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

26

17

9

2

24

3

23

0

26

23

3

ADV. (pre-verb) 2

0

2

0

2

0

2

0

2

2

0

ADV. (post-verb) 11

8

3

0

11

0

11

1

10

0

11

Hacía 1855-1875 Tokens inspected: 1120. Of these temporal: 102. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes CLAUSAL

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

45

30

15

5

40

10

35

10

35

35

10

ADV. (pre-verb) 15

8

7

2

13

2

13

0

15

14

1

ADV. (post-verb) 40

29

11

0

40

3

37

3

37

33

7

Hacía 1955-1975 Tokens inspected: 2680. Of these temporal: 153. Of these not preceded by preposition: 100 Meaning

Neg. hace

Durat. Punct. Yes

Neg. event

Time Adjunc. Hace preposed to hace to time NP

No

Yes

No

Yes

No

Yes

No

CLAUSAL

48

28

20

6

42

14

34

16

32

41

7

ADV. (pre-verb)

15

4

11

3

12

2

13

1

14

14

1

ADV. (post-verb) 37

8

29

1

36

1

36

1

36

36

1

93

REFERENCES Aelbrecht, L., Haegeman, L. & Nye, R. (2012). Main Clause Phenomena: New Horizons. Amsterdam/Philadelphia: John Benjamins. Baker, M. C. (2003) Lexical categories: verbs, nouns and adjectives. Cambridge: Cambridge University Press. Bourdin, Philippe. (2011). Ten years ago and ten years since: competition and standardization in Early Modern English. Groninger Arbeiten zur Germanistischen Linguistik 53. 45-70. Brewer, W. B. (1987). New and old information in Spanish sentences containing hace+ (TIME). Hispania 70, 4. 895-899. Brinton, L. (2008). The Comment Clause in English: Syntactic Origins and Pragmatic Development. Cambridge: Cambridge University Press Bybee, J. (2003a). Mechanisms of change in grammaticization: the role of frequency. In Joseph, B. & Janda, R. (eds), The handbook of historical linguistics. 602-623. Oxford: Blackwell. Bybee, J. (2003b). Cognitive processes in grammaticalization. In Tomasello, M. (ed.), The new psychology of language: Cognitive and functional approaches to language structure. 145-167. Brighton: Psychology Press. Bybee, J. (2006). From usage to grammar: The mind's response to repetition. Language 82, 4. 711-733. Bybee, J. (2010). Language, usage and cognition. Cambridge: Cambridge University Press. Bybee, J., Perkins, R., & Pagliuca, W. (1994). The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press. Bybee, J. & Thompson, S. (2000). Three frequency effects in syntax. Berkeley Linguistics Society 23. 65-85. Cano, R. (2004). Historia de la Lengua Española. Barcelona: Ariel. Cervera Rodríguez. A. (2007). Variedad gramatical y polivalencia pragmática de que. In Puigvert Ocal, A. & Delgado Cobos, I. (eds.), Ex admiratione et amicitia. Homenaje a Ramón Santiago. 205-242. Madrid: Ediciones del Orto. Comrie, (1985) Tense. Cambridge: Cambridge University Press. Croft, W. (2001). Radical Construction Grammar. Oxford: Oxford University Press. De Lacy, P. (1999). Morphological haplology and correspondence. University of Massachusetts, Amherst. De Smet, H. (2009). Analysing reanalysis. Lingua 119. 1728-55. Demirdache, H., & Uribe-Etxebarria, M. (2004). The syntax of time adverbs. In Guéron J. & Lecarme J. (eds.), The syntax of time. Boston: MIT Press. 143-179. Demonte, V. & Bosque, I. (1999). Gramática Descriptiva de la Lengua Española. Madrid: Espasa-Calpe. 94

Deutscher, G. (2000). Syntactic Change in Akkadian: The Evolution of Sentential Complementation. Oxford: Oxford University Press. Deutscher, G. (2009). Nominalization and the origin of subordination. In Givón, T. & Shibatani, M. (eds.), Syntactic Complexity: Diachrony, acquisition, neurocognition, evolution. Typological Studies in Language 85. 199-209. Amsterdam/Philadelphia: John Benjamins. Di Sciullo, A. M. (ed.). (2003). Asymmetry in grammar Vol. 1. Amsterdam/Philadelphia: John Benjamins. Diessel, H. (2008). Iconicity of sequence: A corpus-based analysis of the positioning of temporal adverbial clauses in English. Cognitive linguistics 19. 457-482. Diez Itza, E. (1992). Ha, hay, hace temporales en el siglo de oro. In Ariza Viguera, M. (ed.), Actas del II Congreso Internacional de Historia de la Lengua Española. Vol 1. 373380. Díez Itza, E. & Pérez Toral, M. (1991). Las formas verbales de ciertas expresiones temporales en las lenguas románicas: un estudio de morfosintaxis histórica comparada. Oviedo: Universidad de Oviedo. Dixon, R. M. W. (1972). The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Dixon, R. M. W. (2010). Basic Linguistic Theory. Methodology. Volume 1. Oxford: Oxford University Press. Dixon, R. M. W. & Vogel, A. R. (2004). The Jarawara Language of Southern Amazonia. Oxford: Oxford University Press. Elerick, C. (1989). Latin *abhinc annos (cum): Spanish hace anos (que). Hispanic Linguistics 3. 89-98. Elvira González, J. (1984). Sobre procedimientos de formación de relativas en español antiguo. Cuadernos de Investigación Filológica 10. 99-104. Evans, N. D. (1995). A Grammar of Kayardild: with historical-comparative notes on Tangkic. New York/Berlin: Walter de Gruyter. Everett, D. L. & Kern, B. (1997). Wari: the Pacaas Novos language of western Brazil. Brighton: Psychology Press. Fábregas, A. (to appear). On the structure and variation of ‘hace’ as a temporal expression Fischer, O. (2007). Morphosyntactic Change: Functional and Formal Perspectives. Oxford: Oxford University Press. Franco, L. (2012). Movement triggers and the etiology of grammaticalization: The case of Italian postposition fa. Sintagma: revista de lingüística 24. 65-83. Franco, L. (2013). Before strikes back: An *ABA constraint on temporal expressions. Acta Linguistica Hungarica 60, 3. 265-301.

95

García Fernández, L. (1999). Los complementos adverbiales temporales: La subordinación temporal. In Bosque, I. & Demonte, V. (eds.), Gramática descriptiva de la Lengua Española, Vol 3. 3129-3208. Madrid: Espasa-Calpe. García Castillero, C. (2014). De-adjectival preverbs in the Old Irish verbal complex: a synchronic and diachronic study. Zeitschrift für celtische Philologie 61. 89-93. Gries, S. T., Hampe, B. & Schönefeld, D. (2005). Converging evidence: bringing together experimental and corpus data on the association of verbs and constructions. Cognitive Linguistics 16, 4. 635-676. Haegeman, L. (2012). Adverbial Clauses, Main Clause Phenomena, and Composition of the Left Periphery: The Cartography of Syntactic Structures. Oxford: Oxford University Press. Hagège, C. (2010). Adpositions. Oxford: Oxford University Press. Haspelmath, M. (1997). From space to time: Temporal adverbials in the world's languages. Munich: LINCOM Europa. Haspelmath, M. (1998). Does grammaticalization need reanalysis? Studies in Language 22.2: 315-351. Haspelmath, M. (2000). Why can't we talk to each other? Lingua 110, 4. 235-255, Haspelmath, M. (2008). Creating economical morphosyntactic patterns in language change. In Jeff Good (ed.), Linguistic Universals and Language Change. 185-214. Oxford: Oxford University Press. Haspelmath, M. & Dryer, M. S. (Eds.) (2008) The World Atlas of Language Structures Online. München: Max Planck Digital Library. Accessed April 2015. Hay, J. (2001). Lexical frequency in morphology: is everything relative? Linguistics 39. 1041-1070 Heine, B. (1997). Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, B. (2006). Possession: Cognitive sources, forces, and grammaticalization. Cambridge: Cambridge University Press. Heine, B. (2009). From nominal to clausal morphosyntax: complexity via expansion. In Givón, T. & Shibatani, M. (eds.), Syntactic Complexity: Diachrony, acquisition, neurocognition, evolution. 23-51. Amsterdam/Philadelphia: John Benjamins Heine, B., & Kuteva, T. (2002). World lexicon of grammaticalization. Cambridge: Cambridge University Press. Heine, B. & Reh, M. (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Buske. Hengeveld, K. & Mackenzie, J. L. (2008). Functional Discourse Grammar: A typologically-based theory of language structure. Oxford: Oxford University Press. Hilpert, M. & Gries, S. T. (to appear). Quantitative approaches to diachronic corpus linguistics. Hopper, P. J. & Traugott, E. C. (2003). Grammaticalization. Cambridge: Cambridge University Press. 96

Howe, C. (2011). Structural autonomy in grammaticalization: Leveling and retention with Spanish hacer+time. Probus 23. 247-282. Howe, C. & Ranson, D. L. (2010). The evolution of clausal temporal modifiers in Spanish and French. Romance Philology 64. 197-207. Karlsson, F. (1991). Gramática básica del finés. Madrid: Universidad Autónoma. Klein, W. (2009). How time is encoded. In Klein, W., & Li, P. (eds.), The expression of time. 39-81. New York/Berlin: Walter de Gruyter. Kurzon, D. (2008). “Ago” and its grammatical status in English and other languages. In Kurzon, D. & Adler, S. (eds.), Adpositions, pragmatic, semantic and syntactic perspectives. Typological Studies in Language, 74. 209-227. Amsterdam/Philadelphia: John Benjamins. Langacker, R. W. (1987). Foundations of cognitive grammar: Theoretical prerequisites. Palo Alto: Stanford university press. Langacker, R. W. (2004). Possession, location and existence. In Soares da Silva, A., Torres, A. & Gonçalves, M. (eds.), Linguagem, Cultura e Cognição. Estudos de Linguïstica Cognitiva, Vol 1. 85-120. Coimbra: Almedina. Lehmann, C. (1988). Towards a typology of clause linkage. In Haiman, J. & Thompson, S. A. (eds.) Clause Combining in Grammar and Discourse. 181-225. Amsterdam/Philadelphia: John Benjamins. Lehmann, C. (1992). Word order change by grammaticalization. In Gerritsen, M. & Stein, D. (eds.), Internal and external factors in syntactic change. 395-416. New York/Berlin: Walter de Gruyter. Lehmann, C. (1995). Thoughts on Grammaticalization. Munich: LINCOM Europa. Lehmann, C. (2004). Synchronic variation and diachronic change. In Katamba, F. (ed.), Morphology: its place in the wider context 6, 3. 444-458. London/New York: Routledge. Lenz, F. (ed.). (2003). Deictic conceptualisation of space, time and person. Amsterdam/Philadelphia: John Benjamins. Miestamo, M. (2005). Standard Negation: The negation of declarative verbal main clauses in a typological perspective. New York/Berlin: Walter de Gruyter. Moore, K. E. (2006). Space-to-time mappings and temporal concepts. Cognitive Linguistics 17, 2. 199-244. Munaro, N. (2015). Complementizer doubling and the structure of the topic field in Italo-Romance. Talk devivered at the conference Formal Approaches to MorphoSyntactic Variation. Vitoria-Gasteiz, 24-25 June 2015. Nevins, A. (2012). Haplological dissimilation at distinct stages of exponence. In J. Trommer (ed.), The morphology and phonology of exponence. 84-116. Oxford: Oxford University Press. Nichols, J. (1992) Linguistic diversity in space and time. Chicago: University of Chicago Press.

97

Peng, R. (2012). Critical frequency as an independent variable in grammaticalization. Studies in Language 36. 345-381. Pérez Toral, M. (1992). Sintaxis Histórica Funcional del Español: el verbo 'hacer' como impersonal. Oviedo: Universidad de Oviedo. Pusch, C. D. (2001). Morphosyntax, Informationsstruktur und Pragmatik: Präverbale Marker im gaskognischen Okzitanish und in anderen Sprachen. Tübingen: Gunter Narr. Rasmussen, P. (1981). El verbo hacer en expresiones temporales: Estudio sintáctico y semántico. Copenhagen: Akademisk Forlag. Real Academia Española. (2009). Nueva Gramática de la Lengua Española. Madrid: Espasa-Calpe. Real Academia Española. CORDE: Corpus diacrónico del español. Accessed JanuaryAugust 2015. Real Academia Española. CREA: Corpus de referencia del español actual. Accessed June-August 2015. Rebollo Torío, M. A. (1979). Otra interpretación de 'hace' y 'ha'. Anuario de Estudios Filológicos 1. 149-162. Rice, K. (1989). A Grammar of Slave. New York/Berlin: Walter de Gruyter. Rigau, G. (1999). Temporal existential constructions in Romance. In D'Hulst, Y., Rooryck, J. & Schroten J. (eds.), Romance Languages and Linguistic Theory. 307-333. Amsterdam/Philadelphia: John Benjamins. Russian Academy of Sciences. Russian National Corpus. Accessed May 2015. Sáez del Álamo, L. A. (1987). Caracterización de 'hace-expresión temporal' en el marco de la Rección y Ligamiento. In Martín Vide, C. (ed.), Lenguajes Naturales y Lenguajes Formales. Actas del III congreso de lenguajes naturales y lenguajes formales. 713-722. Barcelona: Promociones y Publicaciones Universitarias. Stefanowitsch, A. & Gries, S. T. (2003). Collostructions: Investigating the interaction of words and constructions. International journal of corpus linguistics 8, 2. 209-243. Stemberger, J. P. (1981). Morphological haplology. Language 57. 791-817. Tabor, W., & Traugott, E. C. (1998). Structural scope expansion and grammaticalization. In Giacalone-Ramat, A & Hopper, P. J., The Limits of Grammaticalization. Typological Studies in Language 37. 229-272. Amsterdam/Philadelphia: John Benjamins. Taylor, J. R. (2003). Linguistic categorization. Oxford: Oxford University Press. Tenbrink, T. (2007). Space, time, and the use of language: An investigation of relationships. New York/Berlin: Walter de Gruyter. Thompson, S. A. & Mulac, A. (1991). A quantitative perspective on the grammaticization of epistemic parentheticals in English. In Traugott, E. C. & B. Heine (eds.), Approaches to grammaticalization II. Focus on types of grammatical markers. 313-329. Amsterdam/Philadelphia: John Benjamins. University of the Basque Country. Euskal Klasikoen Korpusa. Accessed May 2015. 98

Van den Berg, H. (1995). A Grammar of Hunzib: with texts and lexicon. Munich: LINCOM Europa. Van der Auwera, J. & Baoill, D. Ó. (eds.). (1998). Adverbial constructions in the languages of Europe. New York/Berlin: Walter de Gruyter. Van Valin Jr, R. D. (1984). A typology of syntactic relations in clause linkage. Annual Meeting of the Berkeley Linguistics Society 10. 542-558. Wanner, D. (2002). The corpus as a key to diachronic explanation. In Pusch C. D. & Raible, W. (eds.), Romanistiche Korpuslinguistik II. 31-44. Tübingen: Gunter Narr.

99