Language, Cognition and Neuroscience
ISSN: 2327-3798 (Print) 2327-3801 (Online) Journal homepage: http://www.tandfonline.com/loi/plcp21
Perceptual functionality of morphological redundancy in Choguita Rarámuri (Tarahumara) Gabriela Caballero & Vsevolod Kapatsinski To cite this article: Gabriela Caballero & Vsevolod Kapatsinski (2015) Perceptual functionality of morphological redundancy in Choguita Rarámuri (Tarahumara), Language, Cognition and Neuroscience, 30:9, 1134-1143, DOI: 10.1080/23273798.2014.940983 To link to this article: http://dx.doi.org/10.1080/23273798.2014.940983
Published online: 15 Jul 2014.
Submit your article to this journal
Article views: 95
View related articles
View Crossmark data
Citing articles: 2 View citing articles
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=plcp21 Download by: [University of Oregon]
Date: 25 March 2016, At: 01:10
Language, Cognition and Neuroscience, 2015 Vol. 30, No. 9, 1134–1143, http://dx.doi.org/10.1080/23273798.2014.940983
Perceptual functionality of morphological redundancy in Choguita Rarámuri (Tarahumara) Gabriela Caballeroa* and Vsevolod Kapatsinskib a
Department of Linguistics, University of California San Diego, 9500 Gilman Drive, #0108, La Jolla, CA 92093-0108, USA; b Department of Linguistics, University of Oregon, Eugene, OR 97403, USA
Downloaded by [University of Oregon] at 01:10 25 March 2016
(Received 2 July 2013; accepted 12 June 2014) A recent cross-linguistic survey suggests redundant marking of the same meaning by multiple morphological markers to be more widely attested than commonly believed. While this phenomenon (referred to as multiple (or extended) exponence in the morphological literature) has been examined within the context of morphological theory and diachronic research, little work has investigated the processing of morphological redundancy and synchronic motivations for its use. This paper reports a field speech-in-noise experiment to assess perceptual functionality of redundant markers in an agglutinating, morphologically complex language of Northern Mexico, Choguita Rarámuri (Tarahumara). This language possesses morphological patterns in which a meaning is redundantly cued by two consecutive suffixes, and where the second (outer) suffix is optional. We show that the effect of adding the optional suffix varies with the overall likelihood of recognising its meaning in context: cue redundancy helps when recognition of the cued meaning is difficult but hurts when recognition of the cued meaning is easy. The results are interpreted as support for the operation of Grice’s Maxim of Clarity in spoken word recognition and/or production: the listener expects the speaker to say only as much as is necessary to transmit the message. Keywords: redundancy; morphology; multiple exponence; Gricean inference; morphological processing
Typologically diverse languages are documented to have word structure patterns where a single meaning is redundantly cued by multiple morphological markers (Caballero & Harris, 2012; Caballero & Inkelas, 2013), a phenomenon known in morphological theory as multiple (or extended) exponence (defined in Matthews, 1974). Through examination of redundant morphological patterns in Choguita Rarámuri, we show that this phenomenon may bring a perceptual advantage and that, like other examples of redundancy in language, it is a fruitful testing ground for examining the interplay between production economy and clarity of expression. Redundant morphological marking is at odds with principles of economy in morphological theory (e.g., ‘[a]mong equally expressive expressions, the simplest is optimal’ (Kiparsky, 2005, p. 114) but consistent with a proposal regarding the role of clarity and transparency of form-meaning relationships in diachrony (‘hypercharacterisation’, where more ‘clearly’ or ‘overtly’ marked elements tend to be preferred in analogical change (Kurylowicz, 1947; see also Hock, 1991, p. 211). The role of clarity has also been emphasised in the cross-linguistic sentence processing tradition arising from work on the Competition Model (E. Bates, McNew, MacWhinney, Devescovi, & Smith, 1982; E. Bates & MacWhinney, 1989; MacWhinney, Pléh, & E. Bates, 1985).
*Corresponding author. Email:
[email protected] © 2014 Taylor & Francis
On the other hand, Grice (1975) proposed that the listener has certain expectations about the speaker’s behaviour in a conversation, which he called ‘maxims’. In particular, the Maxim of Clarity calls for the speaker to ‘avoid ambiguity’ while also avoiding ‘unnecessary prolixity’: the speaker is expected by the listener to say as little as possible without sacrificing communication (see also Zipf, 1949, for related ideas). The Gricean perspective provides the following prediction for the perceptual functionality of multiple exponence, which we test in this paper: multiple exponence, understood as morphological redundancy, should help the listener perceive the expressed meaning when help is needed, i.e., iff the meaning is not contextually predictable. When the meaning is not contextually predictable, multiple exponence provides an additional cue to the meaning and reduces the chance of miscommunication. On the other hand, when the meaning is predictable from context, multiple exponence constitutes unnecessary prolixity and may actually make recognition of the meaning harder. Does the speaker actually behave as the Gricean listener would expect him/her to behave? To some extent, this is still a question under discussion. Research on language production has investigated the extent to which speakers balance articulatory effort and the goal to be understood, thereby creating a trade-off between the amount and clarity
Downloaded by [University of Oregon] at 01:10 25 March 2016
Language, Cognition and Neuroscience of linguistic signal provided for an intended linguistic unit (e.g., a particular phoneme, word or structure) and predictability of that meaning in context (e.g., Aylett & Turk, 2004; Baker & Bradlow, 2009; Fowler & Housum, 1987; Hochberg, 1986; Jaeger, 2010; Lindblom, 1990; Lockridge & Brennan, 2002; Mahowald, Fedorenko, Piantadosi, & Gibson, 2013; Temperley, 2003; van Son & van Santen, 2005). For example, speakers are more likely to produce contracted morphology (e.g., he’s rather than he is, Frank & Jaeger, 2008) or to altogether omit morphemes (e.g., case markers in Japanese and Korean, Kurumada & Jaeger, 2013; Lee, 2007) when their meaning is contextually predictable (e.g., given the preceding words). Similarly, subject pronouns in Spanish are more likely to be omitted when the person/number of subject is unambiguously cued by the verbal inflexion (Abreu, 2009; Erker & Guy, 2012; Flores-Ferrán, 2004; Hochberg, 1986; Holmquist, 2012, Otheguy, Zentella, & Livert, 2007; Shin & Erker, in press). Plural marking in Spanish noun phrases is also more likely to be omitted when the number is redundantly marked on the verb (Poplack, 1980a, 1980b). Phonetic cues to phonemic contrasts are hyperarticulated when the contrast is non-redundant, being the only phonological difference between two words (Baese-Berk & Goldrick, 2009; Wedel, Sharp, & Jackson, 2014). Furthermore, phonemic contrasts are themselves more likely to be levelled if word identification rarely depends on perceiving the contrast (Wedel, Jackson, & Kaplan, 2013). While such evidence certainly is predicted if speakers aim to balance the competing goals of effort minimisation and communicative robustness (see Aylett and Turk, 2004; Jaeger, 2013; Lindblom, 1990; Zipf, 1949), alternative explanations exist. For example, one prominent view holds that the inverse relation between contextual predictability and the amount/clarity of the linguistic signal can be reduced to ease of production planning, for example, to facilitate fluency (e.g., for articulation see Arnold, 2008; Bard et al., 2000; Bell, Brenier, Gregory, Girand, & Jurafsky, 2009; for grammatical encoding see Ferreira and Dell, 2000; MacDonald, 2013; for discussion, see Jaeger, 2013; MacDonald, 2013). It is also a matter of ongoing debate to what extent speakers simulate the mental state of comprehenders (cf. the debates in Bard et al., 2000; Clark & Murphy, 1982; Galati & Brennan, 2010; Horton & Gerrig, 2005) and whether such simulation is even required for efficient communication (Jaeger, 2010, p. 50), given that interlocutors’ expectations tend to be closely aligned (e.g., Brennan & Clark, 1996; Pickering & Garrod, 2004). Regardless of the outcome of this discussion, there is broad agreement in research on language production that more predictable elements tend to be more likely to be omitted or reduced (see Jaeger & Tily, 2011, for a review). When this happens, a listener following Gricean assumptions would expect reduced forms to be used in contexts
1135
where they are in fact likely to be used by speakers. In fact, for listeners to have this expectation in such situations, no complex Gricean reasoning would even be needed: it would be sufficient to assume that listeners have expectations based on their past experience with what speakers produce in a given context. Indeed, there is much evidence that listeners do have such implicit knowledge and use it to anticipate upcoming speech material (e.g. Allopenna, Magnuson, & Tanenhaus, 1998; Dahan & Tanenhaus, 2004; Levy, 2008; MacDonald, 1994, 2013; Trueswell, Tanenhaus, & Garnsey, 1994). However, we should note that it is not known whether the choice Choguita Rarámuri speakers have between multiple and single exponence is a case in which the shorter expression is chosen precisely when the meaning is less predictable given the context. The choice of referring expression may be driven by factors other than predictability of the meaning, including phonological structure. For example, in the case of English auxiliary contraction examined by Frank and Jaeger (2008), the choice of variants is also driven by lexical category of the subject and the phonological context (see also Barth & Kapatsinski, in press; Labov, 1969; MacKenzie, 2013). In some cases, these structural factors can override contextual predictability of the meaning being expressed. For instance, English is is ineligible to contract after a noun ending in a sibilant no matter the predictability of its meaning (e.g., *Texas’s hot.; Barth & Kapatsinski, in press). In a report on her seminal study on the effects of ambiguity on Spanish number marking, Poplack (1980a, p. 66) notes that ‘functional factors affect [plural] marker deletion [in Puerto Rican Spanish] less than any other factor group studied’. Choice of referring expression in production may also be affected by associations between words and morphosyntactic constructions (Stefanowitsch & Gries, 2003). For example, Erker and Guy (2012) provide data suggesting that some common verbs are associated with pronominal subject expression while others are associated with subject omission. Similarly, Poplack (1992) finds that some high-frequency verbs when used in the main clause are associated with high rates of subjunctive expression in dependent clauses. For Choguita Rarámuri, it could be the case that some phonological characteristics of stems favour multiple exponence while others favour single exponence, or that multiple exponence production is simply associated with individual verb stems (e.g. because those stems are more commonly used in discourse contexts favouring multiple exponence, as argued by Brown & Raymond, 2012, and Bybee, 2002, for other variables) rather than being chosen precisely in cases when the to-be-expressed meaning is unpredictable. Lacking production data, it is not possible to determine whether the choice between redundant and single exponences in Choguita Rarámuri is driven by predictability of the meaning or any other factor. Thus, we
1136
G. Caballero and V. Kapatsinski
Downloaded by [University of Oregon] at 01:10 25 March 2016
can only say whether the listener behaves as if s/he expects the speaker to produce redundant exponence only in contexts where redundancy would be helpful because the expressed meaning is unexpected given the context. We cannot say whether the observed behaviour of the listener is based in general expectations about how speakers behave (pragmatic inference) or in specific experience with Choguita Rarámuri production (probability tracking). That is, we aim to determine whether the listener is descriptively Gricean. The mechanism by which the listener comes to behave in accordance with the maxim (pragmatic inference or probability tracking) remains an open question pending production data. Previous cross-linguistic work While only one paper (Harris & Samuel, 2011) explicitly focused on testing the perceptual functionality of redundant morphology, relevant work has also been conducted within the framework of the Competition Model (E. Bates et al., 1982; E. Bates & MacWhinney, 1989; MacWhinney et al., 1985). Within the Competition Model, morphological exponents are a subtype of ‘cues’,1 and morphological redundancy would be an example of cue coalition, a situation where multiple cues redundantly express the same meaning. Under the Competition Model, cue coalition is expected to be common; cross-linguistic work under this model has focused on exponence of agentivity (‘whodunit’), exploring the relative importance of cues like case marking, verbal agreement and word order in several, typological diverse languages (see MacWhinney, 2001, for the full list of citations). Most of the work in the Competition Model framework focused on cue competition, rather than the question of whether an additional redundant cue helps processing (‘cue coalition’), which is the focus of Harris and Samuel (2011) and the present paper. However, the work of Pléh and collaborators on Hungarian (MacWhinney et al., 1985; MacWhinney & Pléh, 1997; Pléh, 1989; Pléh, Jarovinskij, & Balajan, 1987); Kail (1989) on French; and Mimica, Sullivan, and Smith (1994) on Serbo-Croatian is directly relevant. The earliest version of the Competition Model (E. Bates et al., 1982) suggested that listeners rely on a given cue to a given meaning to the extent that the cue occurs whenever the meaning is expressed and vice versa. However, MacWhinney et al. (1985) pointed out that cues cannot be acquired unless they are detectable. The influence of cue detectability was demonstrated by MacWhinney et al. (1985), Pléh et al. (1987), Pléh (1989) and MacWhinney and Pléh (1997) in Hungarian. Hungarian has an accusative marker −t that lengthens a preceding [a]. Thus, when −t is attached to a noun ending in [a], accusativity is redundantly cued by [t] and the length of the preceding vowel. On the other hand, the presence of
[t] is hard to detect after another alveolar obstruent, namely [s]. Thus, when −t is attached to a noun ending in [s], the cues to accusativity are very weak. In a series of studies, MacWhinney, Pléh and collaborators demonstrated that a noun marked with −t was much more easily detected not to be the agent in a sentence context when the −t came after the lengthened [a] compared to when it came after [s]. As Pléh (1989, p. 160) writes, ‘some stems are inherently more difficult with regard to perceptual identification of certain endings, and some stem alternations – most notably linking vowels - may have been preserved in the language exactly to enhance perceivability’. While cue competition data has consistently supported perceptual functionality of cue redundancy, data obtained in the absence of strong competition from other cues has not (Kail, 1989; Mimica et al., 1994; Harris & Samuel, 2011). Kail (1989) found that an additional cue to agentivity (a clitic pronoun or subject-verb agreement) sped up reaction times to ‘whodunit’ questions in French and Spanish only if the other consistent cues to agentivity were weak on their own. If meaning recognition was at ceiling in the absence of the additional cue, adding the cue slowed down reaction times. Thus, redundant cues may not speed up processing if a single cue is expected in the context (E. Bates & MacWhinney, 1989, pp. 55–56). From a Gricean standpoint, this is particularly likely if a single exponent is strong enough to cue the meaning on its own: the listener can expect the speaker to make as little effort as possible while still successfully getting the meaning across. Evidence that this expectation on the part of the listener is reasonable comes from production experiments finding that speakers are indeed more likely to produce optional morphology when the meaning encoded by the morphology is unexpected in context (Kurumada & Jaeger, 2013; Norcliffe, 2009; see also Frank & Jaeger, 2008; Mahowald et al., 2013, for related evidence). For example, Kurumada and Jaeger (2013) find that speakers of Japanese are more likely to produce the optional object case marker for human referents, which are less likely to be the grammatical object, compared to inanimate referents (see also Lee, 2007, for similar evidence from Korean). We suggest that redundant cues help processing only when they are expected to occur (see also MacDonald, 2013), and that one influence on whether such cues are expected to occur is contextual predictability of the meaning expressed by the cues in question. When the meaning is expected, redundant cues are unexpected. Choguita Rarámuri Choguita Rarámuri (Tarahumara) ([tar]) is a Taracahitan language of the Uto-Aztecan (UA) family spoken in Northern Mexico by approximately 1000 speakers
Language, Cognition and Neuroscience
Downloaded by [University of Oregon] at 01:10 25 March 2016
(Casaus, 2008). Choguita Rarámuri is an almost exclusively suffixing language with complex verbal morphology featuring the following properties: (1) predominance of concatenative (readily segmentable) morphology resulting in a linear string of markers; (2) potentially long strings of suffixes; (3) large derivational paradigms (i.e., availability of several markers encoding meanings relevant to the lexical meaning of roots); and (4) optional marking. Crucially, while Choguita Rarámuri morphologically complex words are composed of mostly concatenative markers ((1) above), the degree of how tightly fused they are to the stem is gradient: markers closer to the stem display a fair amount of phonological cohesion, making them potentially difficult to parse out of the word (Caballero, 2008).
Why Choguita Rarámuri? Choguita Rarámuri has productive patterns of morphological redundancy where words containing an inner derivational marker that is either of limited productivity or phonologically reduced add a second, redundant outer exponent. Crucially, redundancy in Choguita Rarámuri is optional. Therefore, the speaker has a choice to use the redundant cue or not, allowing us to test whether the listener expects the speaker to be economical yet informative. Furthermore, the optional exponent follows the root and another exponent of the same meaning, allowing the listener to form expectations about whether it will occur. Note that we do not test whether speakers actually are economical yet informative in this paper, though see Kurumada and Jaeger (2013) and Norcliffe (2009) for evidence that they are in other languages. Choguita Rarámuri has several applicative constructions encoding the same function, exemplified in the contrast between a basic predicate and its applicative counterpart in (1): the applicative adds an object argument with the semantic role of benefactive (‘to do X for Y’) (1b)2: (1)
Basic predicate vs. applicative predicate a. ne ma ˈwi-ma suˈnu 1SG.NOM now harvest-FUT.SG corn ‘I’ll harvest corn now’ [LEL 06 4:151/el] b. ˈwi-ni-mo=n oˈla ne ˈje-ra suˈnu harvest-APPL-FUT.SG=1SG.NOM CER 1SG.SUBJ mom-poss corn ‘I will harvest the corn for my mom’ [BFL 06 5:146/el]
Some Choguita Rarámuri applicative constructions are of limited productivity, i.e., they appear with a limited set of roots. As exemplified in (2), an outer, fully productive applicative suffix (−ki) may be added to an applicative stem built using one of these constructions. In addition to their limited productivity, the inner applicative markers may be phonologically reduced, as in (2a–d), where the
1137
vowel of the suffixes (e.g., -ni (2a, b), -si (2c) and -wi (2d)) is often deleted: (2)
Choguita Rarámuri redundant applicative exponence Single Exponence ∼ Redundant Exponence a. ˈsu-n-ma ˈsu-n-ki-ma ‘sew-APPL(-APPL)-FUT.SG’ b. wasaˈra-ni-ma wasaˈra-n-ki-ra ‘plow-APPL(-APPL)-POT’ c. ˈpa-si-li ˈpa-s-ki-li ‘throw-APPL(-APPL)-PST’ d. riˈwi-wi-ma riˈwi-w-ki-ma ‘find-APPL(-APPL)-FUT.SG’ e. roˈn-e-ma roˈn-e-ki-ma ‘boil-APPL(-APPL)-FUT.SG’
Another redundant process in Choguita Rarámuri involves causative doubling. The causative construction, marked with a suffix with two allomorphs (−ri and −ti), introduces a causer argument to a verbal predicate (‘X causes Y to V’). The Choguita Rarámuri causative may occur with recursive semantics, i.e., where each causative suffix is associated with the introduction of a causer argument (e.g., biˈne-ri-ma ‘learn-CAUS-FUT.SG’ ‘X will make Y learn’ vs. biˈne-r-ti-ki ‘learn-CAUS-CAUS-PST.1’ ‘X made Y make Z learn’). In contrast, the same suffix may appear doubled in a redundant fashion: causative doubling (3b) is equivalent and optional to a construction with a single causative (3a; i.e., the construction in (3b) cannot be interpreted as meaning ‘the governor made me make somebody speak’ as would be expected if each casuative suffix were to introduce a causer argument): (3)
a.
b.
ˈne=mi raʔi ˈtʃa-ri-ma 1SG.NOM=2SG.ACC speak-caus-FUT.SG ‘I will make you speak’ [[speak] + Causative = make speak] ˈa biˈra taˈmi raʔiˈtʃa-r-ti-ri siˈriame AFF really 1SG.ACC speak-caus-caus-PST governor ‘The governor made me speak’ [[speak] + Caus = make speak]
In sum, redundancy is optional, the speaker has a choice in whether to be redundant, and the listener can therefore form expectations about whether the speaker will be redundant on any given occasion (for full discussion and data, see Caballero, 2008, 2011).
Methods Given the potential importance of perceptibility for use of redundant marking (MacWhinney et al., 1985), we decided to measure perceptibility directly by using a speech-in-noise gating task (Salasoo & Pisoni, 1985). From previously collected field recordings of naturalistic speech, we selected 13 minimal pairs of stems differing only in whether the causative or applicative meaning was expressed by one or two exponents. All items are morphologically complex, with tense/aspect marking and/or derivational morphology, and matched in terms of categories overtly marked in each individual pair. Crucially, all target forms were extracted from recordings of the same (male) speaker carefully uttering sentences in an
1138
G. Caballero and V. Kapatsinski
Table 1. Recognition accuracies as a function of stem context and exponence, ordered by overall probability of recognition. The higher accuracy cell in each row is italicised.
Downloaded by [University of Oregon] at 01:10 25 March 2016
Stimulus pair kiˈp-e-ba/kiˈp-e-ki-pa ˈpo-nu-ma/ˈpo-n-ki-ma siˈru-ni-ri/siˈru-n-ki-ri koˈʔi-ri-ma/koˈʔi-r-ti-ma ˈpa-si-ri/ˈpa-s-ki-ri ˈme-ti-ma/ˈme-r-ti-ma uˈk-e-ri/uˈk-e-ki-ma wikaˈra-ri-ma/wikaˈra-r-ti-ri suˈwe-ri/suˈwe-ki-ri wikaˈra-ni-ma/wikaˈra-ni-ki-ma riˈme-ni-ma/riˈme-ni-ki-ma riˈku-tu-ma/riˈku-r-ti-ma ˈub-ti-ri/uˈba-r-ti-pa
Recognition with single exponence
Recognition with redundant exponence
Probability of recognition per stem (%)
0/7 0/7 1/7 1/7 3/7 4/7 5/8 5/7 6/7 6/7 7/7 7/7 7/7
1/7 2/7 4/7 6/7 6/7 5/7 4/6 4/7 6/7 7/7 5/7 6/7 6/7
7 14 36 50 64 64 64 64 86 93 86 93 93
elicitation context, thus uttered in matching prosodic and syntactic contexts. Though at the expense of precise control of potential stimulus parameters and a larger set of items tested, these naturally produced minimal pairs allowed us to control for several factors while prioritising the ecological validity of the stimuli and keeping listening conditions as natural as possible in this particular field situation. We also worried that using made-up stem-suffix combinations would cause the participants to never guess that the word contained a causative or applicative suffix. As discussed below, the phoneme strings corresponding to causatives and applicatives have other functions in Choguita Rarámuri. Furthermore, causatives and applicatives are not particularly frequent in Choguita Rarámuri in comparison to other morphological constructions, and therefore relatively unlikely to be simply guessed by the listener, especially without sentential and discourse context. The pairs (five causative and eight applicative) are shown in Table 1. Participants were presented with one word from each minimal pair, six or seven featuring redundant exponence and seven or six single exponence. In addition, 12 multimorphemic filler items that did not feature causative or applicative suffixes were also presented to each subject. Trial number was controlled within the minimal pairs: participants who experienced the single-exponent member of the pair experienced it at the same point in the experiment at which other participants were exposed to the multiple-exponent member of the pair. Trial order was randomised once with the same random order presented to all subjects through headphones using a playlist in Praat. Each word was embedded in speech-shaped pink noise. The signal-to-noise ratio varied from very noisy (−10 dB) to non-noisy (+20 dB) in 2 dB increments. For each word, we presented each participant with the noisiest version and
then proceeded to reduce the noise level in 2 dB increments until the participant recognised the word or until we reached the lowest noise level. At each step on this continuum, the participant was asked to describe what the word meant, which would generally involve the participant producing a Spanish translation. When responses could be ambiguous, the first author would ask them to elaborate or describe a context in which that form would be uttered. Participants were asked for a Spanish translation rather than simple repetition of the word in order to assess how good redundant markers are at expressing the meaning. Because Choguita Rarámuri displays a high degree of phonological fusion, restricted phonotactics and a small phoneme inventory while also having a rich inventory of suffixes, Choguita Rarámuri suffixes display a high degree of ambiguity in the form-to-meaning mapping. The main difficulty in the task may not be perceiving whether, for instance, the auditory signal contained the phone sequence [ri] in a noisy environment, but rather perceiving that the [ri] was there to express causative meaning, as opposed to being a part of the stem, an allomorph of the past tense marker, or a part of the medio-passive suffix. The task yielded two measures for each word for each participant: (1) whether or not the participant recognised the word (in particular the causative or applicative meaning of the morphologically complex word), and (2) if the participant did recognise the word, then what was the noise level when the word was recognised. If the additional exponent helped access the meaning in a certain context in any way, we expected participants to be better at recognising the meanings when they were expressed by multiple exponents. If the additional exponent helped specifically by enhancing acoustic robustness of the cues to the meaning in the face of environmental noise, we
1139
Downloaded by [University of Oregon] at 01:10 25 March 2016
Language, Cognition and Neuroscience additionally expected participants to recognise words with redundant exponence at higher noise levels. We did not find any significant effects on noise level sufficient for recognition. However, this null result may be due to insufficient power of the study, thus little can be concluded from it. Therefore, we focus on the effects of our predictors on recognition accuracy. Sixteen adult native speakers of Choguita Rarámuri (aged between 18 and 60) participated. Data for two participants were eliminated due to having observed accuracy of 0 in detecting causatives and applicatives. All participants were fully fluent in Spanish, the language in which instructions were given. Participants recruited were all who were available and who met this requirement at the time the experiment was conducted. Responses were coded by the first author for both recognition and level of noise of recognition. The majority of errors leading to lack of recognition involved misparsing the causative or applicative suffix as part of a (different) stem, with some marginal set of responses involving identifying the correct root but not detecting the causative or applicative meaning. The recognition accuracy data were analysed using logistic mixed-effects models (Jaeger, 2008) with byparticipant and by-stem random intercepts and slopes for all fixed effects using lme4 (D. Bates, Maeschler, & Bolker, 2013). The models included exponence (multiple vs. single, sum-coded) and average stem accuracy (scaled and centred) as fixed effects. Average stem accuracy was used to gauge how easy the meaning expressed by the single or multiple exponents is to guess in the context of the stem. We expected multiple exponence to help, compared to single exponence, for stem contexts in which recognition is difficult. In other words, we expected an interaction between the effect of exponence and the average stem accuracy. Difficulty of recognition in context is surely affected by many different factors, including perceptibility of that particular instance of the exponent, its confusability with other morphemes that could fit in the context, predictability given the context, etc. We do not know what all these factors are for our stimuli (which come from field recordings and are not a controlled set), in part because we are working on a language in which there are no large corpora that would allow for calculation of contextual predictability statistics. However, we hope that once we know recognition accuracy in context, we can predict the effect of redundancy in that context: the effect of redundancy should be facilitatory in contexts (in our case, wordinternal contexts) disfavoring recognition of the meaning that could potentially be expressed by multiple markers.3 Results The results are summarised in Table 1 and Figure 1. The mixed-effects model shown in Table 2 shows a significant
Table 2. Fixed effects from the mixed-effects model.
(Intercept) Exponence (multiple) Accuracy Exponence: accuracy
Estimate
SE
z value
Pr(>∣z∣)
0.7356 0.5106 9.6147 −4.8566
0.3573 0.2530 1.6768 1.9932
2.058 2.018 5.734 −2.437
0.0395 0.0436 9.81e-09 0.0148
Note: This model does not suffer from collinearity, as assessed by vif.mer (), downloaded from https://raw.github.com/aufrank/R-hacks/master/merutils.R. While it overfits the data somewhat, the upper bound of the 95% bootstrapping confidence interval for the interaction’s coefficient, as assessed by bootMer(), was –2.37, below 0. The reliability of the crucial correlation was also verified using Bayesian modelling. A reviewer has pointed out that treating accuracy per stem as a fixed effect, as we have done, ignores uncertainty in estimating accuracy and has very generously verified the reliability of the correlation using glmer2stan treating accuracy differences among stems as a random by-stem intercept and correlating these intercepts with the within-stem slopes of the effect of exponence (McElreath, 2013; glmer2stan uses Stan, a Bayesian inference engine; see http://hlplab.wordpress.com/2013/12/13/going-full-bayesianwith-mixed-effects-regression-models/). The highest posterior density interval for the correlation remained below 0, although it was considerably closer to 0 than the interval generated using bootMer().
main effect of exponence (multiple exponence leading to slightly higher accuracy) and, importantly, a significant interaction between the effect of exponence and stem accuracy, indicating that an additional cue helps recognise meaning when recognition rates are low and may hurt recognition when they are high, as predicted by the Gricean hypothesis. Follow-up analyses revealed that this effect was not reducible to order of presentation [there was no significant effect of trial order, and there was no significant correlation between the effect of trial order and the effect of exponence across items (r = −.14, p = .64); if trial order was added to the model, all other effects remain unchanged]. General discussion In sum, an additional exponent helps in recognition of the meaning when help is needed, but may hurt recognition when the context – in our case, word-internal context – is sufficiently predictive for recognition to occur without the extra exponent, although the additional help does not bring meaning recognition accuracy in non-predictive contexts up to the level observed in predictive contexts. These results are exactly what one would expect if perception obeyed Grice’s (1975) Maxim of Clarity: the listener expects the speaker to produce as little as possible while successfully transmitting the information. Given that redundant morphological marking in Choguita Rarámuri is optional, the listener expects the speaker to produce the second exponent only when the meaning would be unlikely to come across without it. We believe that the present results will not generalise to all instances of redundancy of exponence, but that they should generalise to situations where two conditions hold, as they do in
Downloaded by [University of Oregon] at 01:10 25 March 2016
1140
G. Caballero and V. Kapatsinski
Figure 1. The interaction between exponence and accuracy.
Choguita Rarámuri: (1) redundant marking is optional,4 so that the speaker has a choice in production and the listener can therefore take into account the speaker’s inherent tendency to be economical while transmitting their intended message; and (2) recognition of the optional exponent follows recognition of another (optional or obligatory) exponent,5 and can therefore be anticipated by the listener as s/he is incrementally processing the speech signal (Allopenna et al., 1998; Marslen-Wilson & Tyler, 1980). Neither of these properties holds in the one previous study that examined - and did not find - perceptual functionality of redundant exponence (in Batsbi, Harris & Samuel, 2011). In conclusion, we believe that redundancy in morphological marking can help recognition when the meaning is relatively difficult to recognise in context. It is in these circumstances that the diachronic process of hypercharacterisation (Kurylowicz, 1947) is likely to reinforce an existing exponent with an additional one. In contrast, when the expressed meaning is predictable from context, redundancy is unexpected and, if present in a language, diachronically unstable. Redundancy is neither always dispreferred (as predicted by principles of economy in morphological theory) nor always preferred (as predicted by most versions of the Competition Model, e.g., E. Bates & MacWhinney, 1989, p. 22; McDonald & MacWhinney, 1989). Rather, redundant expression may be preferred or dispreferred in perception because of the interplay between economy and clarity of expression in production (Grice, 1975; E. Bates & MacWhinney, 1989, p. 55; Jaeger, 2010; MacDonald, 2013; Piantadosi, Tily, & Gibson, 2012) and permitted by the grammar of the language.
Acknowledgements We are grateful to our Choguita Rarámuri teachers for their collaboration and patience. This paper benefited from feedback on several versions of this paper from two anonymous reviewers and Florian Jaeger. For valuable discussions, we would like to thank Roger Levy, Scott DeLancey and audience members at the American International Morphology Meeting (UMass), the International Conference on the Mental Lexicon (Université de Montréal/McGill University), the UCSD and U of Oregon Linguistic Fieldwork Working Groups. Any errors and omissions are our sole responsibility.
Funding This work was supported by the Documenting Endangered Languages (DEL) Program/National Science Foundation (NSF) [under grant no. 1160672] and a Hellman Fellowship (2011– 2012) awarded to Gabriela Caballero.
Notes 1. Where ‘cues’ are understood as the dimensions (phonological, morphological and syntactic) within the formal (expressive) level that allow hearers to infer the functional content of utterances, plus any functional or extralinguistic content that may affect the inference process. For instance, animacy is considered a lexical semantic cue to agentivity (see Bates & MacWhinney, 1989). 2. Data are transcribed using the International Phonetic Alphabet. Tone is left unmarked. Abbreviations used include: APPL – applicative; CAUS – causative; CER – certainty; FUT – future; NOM – nominative; PASS – passive; PST – past; POSS – possessive; POT – potential; SG – singular. 3. It is worth pointing out that recognition accuracy is agreed to be a function of predictability of what is being recognised given the available context and acoustic/sensory robustness of the set of cues to what is being recognised (e.g., Broadbent, 1967; Norris & McQueen, 2008). 4. This may obtain more broadly for cue coalition. We expect the results to extend beyond morphological cues. For
Language, Cognition and Neuroscience instance, phonetic cues associated with ‘clear speech’ might hurt recognition of a meaning that is predictable in context (instead being interpreted as a cue that some other message is intended, as in sarcasm (‘He is [s:oʊ:] smart’) or exasperation at the listener: ‘I … do … not … know’ vs. [ə˜ ə˜ ə˜], Hawkins & Smith, 2001). 5. We assume that the order of recognition usually tracks order of occurrence in the speech signal though some exponents might be harder to recognise than others or might be superimposed (fully or partially) on the other exponents, e.g., tone.
Downloaded by [University of Oregon] at 01:10 25 March 2016
References Abreu, L. (2009). Spanish subject personal pronoun use by monolinguals, bilinguals, and second language learners (Unpublished doctoral dissertation). University of Florida, Gainesville. Allopenna, P. D., Magnuson, J. S., & Tanenhaus, M. K. (1998). Tracking the time course of spoken word recognition with eye movements: Evidence for continuous mapping models. Journal of Memory & Language, 38, 419–439. doi:10.1006/ jmla.1997.2558 Arnold, J. E. (2008). Reference production: Production-internal and addressee-oriented processes. Language and Cognitive Processes, 23(4), 495–527. doi:10.1080/01690960801920099 Aylett, M., & Turk, A. (2004). The smooth signal redundancy hypothesis: A functional explanation for relationships between redundancy, prosodic prominence and duration in spontaneous speech. Language & Speech, 47, 31–56. doi:10.1177/00238309040470010201 Baese-Berk, M., & Goldrick, M. (2009). Mechanisms of interaction in speech production. Language and Cognitive Processes, 24, 527–554. doi:10.1080/01690960802299378 Baker, R. E., & Bradlow, A. R. (2009). Variability in word duration as a function of probability, speech style and prosody. Language and Speech, 52, 391–413. doi:10.1177/ 0023830909336575 Bard, E. G., Anderson, A. H., Sotillo, C., Aylett, M., DohertySneddon, G., & Newlands, A. (2000). Controlling the intelligibility of referring expressions in dialogue. Journal of Memory and Language, 42(1), 1–22. doi:10.1006/jmla. 1999.2667 Barth, D., & Kapatsinski, V. (In press). A multimodel inference approach to categorical variant choice: construction, priming and frequency effects on the choice between full and contracted forms of am, are and is. Corpus Linguistics and Linguistic Theory. Bates, D., Maeschler, M., & Bolker, B. (2013). lme4: Linear mixed-effects models using S4 classes. R package (version 0.999999-2). Retrieved from http://CRAN.R-project.org/ package=lme4 Bates, E., & MacWhinney, B. (1989). Functionalism and the competition model. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 3–73). Cambridge: Cambridge University Press. Bates, E., McNew, S., MacWhinney, B., Devescovi, A., & Smith, S. (1982). Functional constraints on sentence processing: A cross-linguistic study. Cognition, 11, 245–299. doi:10.1016/0010-0277(82)90017-8 Bell, A., Brenier, J. M., Gregory, M., Girand, C., & Jurafsky, D. (2009). Predictability effects on durations of content and function words in conversational English. Journal of Memory & Language, 60, 92–111. doi:10.1016/j.jml.2008.06.003 Brennan, S. E., & Clark, H. H. (1996). Conceptual pacts and lexical choice in conversation. Journal of Experimental
1141
Psychology: Learning, Memory, and Cognition, 22(6), 1482–1493. doi:10.1037/0278-7393.22.6.1482 Broadbent, D. E. (1967). Word-frequency effect and response bias. Psychological Review, 74, 1–15. doi:10.1037/h0024206 Brown, E. L., & Raymond, W. D. (2012). How discourse context shapes the lexicon: Explaining the distribution of Spanish f-/h- words. Diachronica, 29, 139–161. doi:10.1075/dia.29. 2.02bro Bybee, J. (2002). Word frequency and context of use in the lexical diffusion of phonetically conditioned sound change. Language Variation and Change, 14, 261–290. doi:10.1017/ S0954394502143018 Caballero, G. (2008). Choguita Rarámuri (Tarahumara) phonology and morphology. (Unpublished doctoral dissertation). University of California, Berkeley. Caballero, G. (2011). Multiple exponence and the phonologymorphology interface. Proceedings of the North East Linguistics Society, 39, 177–190. Caballero, G., & Harris, A. C. (2012). A working typology of multiple exponence. In F. Kiefer, M. Ladányi, & P. Siptár (Eds.), Current issues in morphological theory: (Ir)regularity, analogy and frequency. Selected papers from the 14th International Morphology Meeting, Budapest (pp. 163–188). Amsterdam: John Benjamins. Caballero, G., & Inkelas, S. (2013). Word construction: Tracing an optimal path through the lexicon. Morphology, 23, 103– 143. doi:10.1007/s11525-013-9220-x Casaus, M. (2008). Quantitative ethnobotany and acculturation among the Rarámuri of Choguita, Chihuahua, México. Unpublished manuscript. Clark, H. H., & Murphy, G. L. (1982). Audience design in meaning and reference. Advances in Psychology, 9, 287–299. Dahan, D., & Tanenhaus, M. K. (2004). Continuous mapping from sound to meaning in spoken-language comprehension: Immediate effects of verb-based thematic constraints. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 498–513. doi:10.1037/0278-7393.30. 2.498 Erker, D., & Guy, G. (2012). The role of lexical frequency in syntactic variability: Variable subject personal pronoun expression in Spanish. Language, 88, 526–557. doi:10.13 53/lan.2012.0050 Ferreira, V. S., & Dell, G. S. (2000). Effect of ambiguity and lexical availability on syntactic and lexical production. Cognitive Psychology, 40, 296–340. doi:10.1006/cogp. 1999.0730 Flores-Ferrán, N. (2004). Spanish personal pronoun use in New York City Puerto Ricans: Can we rest the case of English contact? Language Variation & Change, 16, 49–73. doi:10. 10170S0954394504161048 Fowler, C. A., & Housum, J. (1987). Talkers’ signaling of “new” and “old” words in speech and listeners’ perception and use of the distinction. Journal of Memory & Language, 26, 489– 504. doi:10.1016/0749-596X(87)90136-7 Frank, A., & Jaeger, T. F. (2008). Speaking rationally: Uniform information density as an optimal strategy for language production. Proceedings of the Annual Meeting of the Cognitive Science Society, 30, 939–944. Galati, A., & Brennan, S. E. (2010). Attenuating repeated information: For the speaker, or for the addressee? Journal of Memory & Language, 62, 35–51. doi:10.1016/j.jml.2009. 09.002 Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Speech acts (pp. 41–58). New York, NY: Academic Press.
Downloaded by [University of Oregon] at 01:10 25 March 2016
1142
G. Caballero and V. Kapatsinski
Harris, A., & Samuel, A. (2011). Perception of exuberant exponence in Batsbi: Functional or incidental? Language, 87, 447–469. doi:10.1353/lan.2011.0068 Hawkins, S., & Smith, R. (2001). Polysp: A polysystemic, phonetically-rich approach to speech understanding. Rivista di Linguistica, 13, 99–188. Hochberg, J. (1986). Functional compensation for /-s/ deletion in Puerto Rican Spanish. Language, 62, 609–621. doi:10.1353/ lan.1986.0041 Hock, H. H. (1991). Principles of historical linguistics. Berlin: Mouton de Gruyter. Holmquist, J. (2012). Frequency rates and constraints on subject personal pronoun expression: Findings from the Puerto Rican highlands. Language Variation & Change, 24, 203–220. doi:10.1017/S0954394512000117 Horton, W. S., & Gerrig, R. J. (2005). The impact of memory demands on audience design during language production. Cognition, 96, 127–142. doi:10.1016/j.cognition.2004. 07.001 Jaeger, T. F. (2008). Categorical data analysis: Away from ANOVAs (transformation or not) and towards logit mixed models. Journal of Memory & Language, 59, 434–446. doi:10.1016/j.jml.2007.11.007 Jaeger, T. F. (2010). Redundancy and reduction: Speakers manage syntactic information density. Cognitive Psychology, 61, 23– 62. doi:10.1016/j.cogpsych.2010.02.002 Jaeger, T. F. (2013). Production preferences cannot be understood without reference to communication. Frontiers in Psychology, 4, 230. doi:10.3389/fpsyg.2013.00230 Jaeger, T. F., & Tily, H. (2011). On language ‘utility’: Processing complexity and communicative efficiency. Wiley Interdisciplinary Reviews: Cognitive Science, 2(3), 323–335. Kail, M. (1989). Cue validity, cue cost, and processing types in sentence comprehension in French and Spanish. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 77–117). Cambridge: Cambridge University Press. Kiparsky, P. (2005). Blocking and periphrasis in inflectional paradigms. In G. E. Booij & J. van Marle (Eds.), Yearbook of morphology 2004 (pp. 113–135). Dordrecht: Springer. Kurumada, C., & Jaeger, T. F. (2013). Communicatively efficient language production and case-marker omission in Japanese. Proceedings of the Annual Meeting of the Cognitive Science Society, 35, 858–863. Kurylowicz, J. (1947). The nature of the so-called analogical processes. (Margaret Winters, Trans. 1995), Diachronica, 12, 113–145. Labov, W. (1969). Contraction, deletion, and inherent variability of the English copula. Language, 45, 715–762. doi:10.2307/ 412333 Lee, H. (2007). Case ellipsis at the grammar/pragmatics interface: A formal analysis from a typological perspective. Journal of Pragmatics, 39, 1465–1481. doi:10.1016/j.prag ma.2007.04.012 Levy, R. (2008). Expectation-based syntactic comprehension. Cognition, 106, 1126–1177. doi:10.1016/j.cognition.2007. 05.006 Lindblom, B. (1990). Explaining phonetic variation: A sketch of the H&H theory. In W. J. Hardcastle & A. Marchal (Eds.), Speech production and speech modelling (pp. 403–439). Dordrecht: Springer. Lockridge, C. B., & Brennan, S. E. (2002). Addressees’ needs influence speakers’ early syntactic choices. Psychonomic Bulletin & Review, 9, 550–557. doi:10.3758/BF03 196312
MacDonald, M. C. (1994). Probabilistic constraints and syntactic ambiguity resolution. Language & Cognitive Processes, 9, 157–201. doi:10.1080/01690969408402115 MacDonald, M. C. (2013). How language production shapes language form and comprehension. Frontiers in Psychology, 4, 226. MacKenzie, L. (2013). Variation in English auxiliary realization: A new take on contraction. Language Variation & Change, 25, 17–41. doi:10.1017/S0954394512000257 MacWhinney, B. (2001). The competition model: The input, the context, and the brain. In P. Robinson (Ed.), Cognition and second language instruction (pp. 69–90). Cambridge: Cambridge University Press. MacWhinney, B., & Pléh, Cs. (1997). Double agreement: Role identification in Hungarian. Language & Cognitive Processes, 12, 67–102. doi:10.1080/016909697386916 MacWhinney, B., Pléh, Cs., & Bates, E. (1985). The development of sentence interpretation in Hungarian. Cognitive Psychology, 17, 178–209. doi:10.1016/0010-0285(85) 90007-6 Mahowald, K., Fedorenko, E., Piantadosi, S. T., & Gibson, E. (2013). Info/information theory: Speakers choose shorter words in predictive contexts. Cognition, 126, 313–318. doi:10.1016/j.cognition.2012.09.010 Marslen-Wilson, W. D., & Tyler, L. K. (1980). The temporal structure of spoken language understanding. Cognition, 8, 1–71. doi:10.1016/0010-0277(80)90015-3 Matthews, P. (1974). Morphology. Cambridge: Cambridge University Press. McDonald, J., & MacWhinney, B. (1989). Maximum likelihood models for sentence processing. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp. 397–421). Cambridge: Cambridge University Press. McElreath, R. (2013). glmer2stan: RStan models defined by glmer formulas. R package (version 0.994). Retrieved from https://github.com/rmcelreath/glmer2stan Mimica, I., Sullivan, M., & Smith, S. (1994). An on-line study of sentence interpretation in native Croatian speakers. Applied Psycholinguistics, 15, 237–261. doi:10.1017/S01427164 00005348 Norcliffe, E. J. (2009). Head-marking in usage and grammar: A study of variation and change in Yucatec Maya (Unpublished doctoral dissertation). Stanford University, Palo Alto. Norris, D., & McQueen, J. M. (2008). Shortlist B: A Bayesian model of continuous speech recognition. Psychological Review, 115, 357–395. doi:10.1037/0033-295X.115.2.357 Otheguy, R., Zentella, A. C., & Livert, D. (2007). Language and dialect contact in Spanish in New York: Towards the formation of a speech community. Language, 83, 770–802. doi:10.1353/lan.2008.0019 Piantadosi, S. T., Tily, H., & Gibson, E. (2012). The communicative function of ambiguity in language. Cognition, 122, 280–291. doi:10.1016/j.cognition.2011.10.004 Pickering, M. J., & Garrod, S. (2004). Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences, 27, 169–190. doi:10.1017/S0140525X04000056 Pléh, Cs. 1989. The development of sentence interpretation in Hungarian. In B. MacWhinney & E. Bates (Eds.), The crosslinguistic study of sentence processing (pp.158–184). Cambridge: Cambridge University Press. Pléh,Cs., Jarovinskij, A., & Balajan, A. (1987). Sentence comprehension in Hungarian-Russian bilingual and monolingual preschool children. Journal of Child Language, 14, 587–603. doi:10.1017/S0305000900010308 Poplack, S. (1980a). The notion of the plural in Puerto Rican Spanish: Competing constraints on (s) deletion. In W. Labov
Downloaded by [University of Oregon] at 01:10 25 March 2016
Language, Cognition and Neuroscience (Ed.), Locating language in time and space (pp. 55–67). New York, NY: Academic Press. Poplack, S. (1980b). Deletion and disambiguation in Puerto Rican Spanish. Language, 56, 371–385. doi:10.1353/lan. 1980.0033 Poplack, S. (1992). The inherent variability of the French subjunctive. In C. Laeufer & T. A. Morgan (Eds.), Theoretical analyses in Romance linguistics: Selected papers from the 19th Linguistic Symposium on Romance Languages (pp. 235–263). Amsterdam: John Benjamins. Salasoo, A., & Pisoni, D. B. (1985). Interaction of knowledge sources in spoken word identification. Journal of Memory and Language, 24, 210–231. doi:10.1016/0749-596X(85) 90025-7 Shin, N. L., & Erker, D. (In press). The emergence of structured variability in morphosyntax: Childhood acquisition of Spanish subject pronouns. In A. M. Carvalho, R. Orozco, & N. L. Shin (Eds.), Subject pronoun expression is Spanish: A crosslinguistic perspective. Washington, DC: Georgetown University Press. Stefanowitsch, A., & Gries, S. Th. (2003). Collostructions: Investigating the interaction of words and constructions. International Journal of Corpus Linguistics, 8, 209–243. doi:10.1075/ijcl.8.2.03ste
1143
Temperley, D. (2003). Ambiguity avoidance in English relative clauses. Language, 79, 464–484. doi:10.1353/lan.2003.0189 Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing: Use of thematic role information in syntactic ambiguity resolution. Journal of Memory and Language, 33, 285–318. doi:10.1006/jmla. 1994.1014 Van Son, R. J., & Van Santen, J. P. (2005). Duration and spectral balance of intervocalic consonants: A case for efficient communication. Speech Communication, 47(1), 100–123. doi:10.1016/j.specom.2005.06.005 Wedel, A., Jackson, S., & Kaplan, A. (2013). Functional load and the lexicon: Evidence that syntactic category and frequency relationships in minimal lemma pairs predict the loss of phoneme contrasts in language change. Language and Speech, 56, 395–417. doi:10.1177/002383091 3489096 Wedel, A., Sharp, R., & Jackson, S. (2014, January 2–5). Phonetic cues distinguishing minimal pairs are hyperarticulated in natural speech. Paper presented at the 88th Annual Meeting of the Lingusitic Society of America, Minneapolis, MN. Zipf, G. K. (1949). Human behavior and the principle of least effort. Oxford: Addison-Wesley.
