Peripherality and Markedness in the Spread of the High German Consonant Shift*
Garry W. Davis, Gregory K. Iverson, and Joseph C. Salmons
[
[email protected],
[email protected]] Department of Foreign Languages and Linguistics University of Wisconsin–Milwaukee Milwaukee, Wisconsin 53201–0413 [
[email protected]] Department of German University of Wisconsin–Madison Madison, Wisconsin 53706
*An
earlier version of this paper was presented at the Second Germanic Linguistics Annual
Conference (GLAC 2, held at the University of Wisconsin–Madison, April 1996). We thank many members of that audience for their comments and questions, as well as Thomas Klein, Monica Macaulay and Marlys Macken; we especially thank Rob Howell for discussions on this topic and comments on an earlier version of the paper. The usual disclaimers apply.
Abstract
This paper applies Rice’s (1994) notion of “Peripherality” in Feature Geometry to sound change, employing it specifically to develop a markedness-based view of the spread of the High German Consonant Shift. Current work on the shift argues that the affrication/fricativization of voiceless stops began in position following short vowels, only later extending into other environments (Davis & Iverson 1995). The spread to initial, final, and post-consonantal positions was dialectally asymmetric with regard to place of articulation, however: Most commonly affected were the coronal stops, less the labials, least the velars. Rice has proposed a hierarchized geometry of place feature dependencies with respect to place assimilation, and we show here that the asymmetries in the High German Consonant Shift (as well as in the later Medienverschiebung, shifting lenis voiced stops to fortis or voiceless) reflect the same basic dynamic. Under the conventional assumption that marked spreads into unmarked structure (Kiparsky 1985), Rice’s Peripherality hypothesis holds that coronals are unspecified for Place, labials are marked only for Peripheral, while velars are marked for Peripheral plus its subordinate, Dorsal. This scheme of representational complexity accords exactly with the familiar pattern of the distinctly nonassimilatory High German Shift, too, because lesser feature-geometric structure correlates positively with broader application of the Shift. Specifically, changes in the unmarked stop, the coronal, reached the farthest north, while shift of the moderately marked labial extended into Middle German and affrication of the most marked velar was restricted to the far south.
0. Introduction. Current work on the genesis of the familiar High German Consonant Shift has adduced structural and dialectal evidence that the affrication of voiceless stops, with subsequent fricativization, began in position following short vowels. From there, it soon generalized to remaining postvocalic positions, then later extended into postconsonantal and word-initial environments to produce the pattern presented in (1a) (Davis & Iverson 1995). The shift was asymmetric with respect to place of articulation, however, as the dialect distributions shown in (1b) attest: Most commonly affected were the coronal (alveolar) stops, less the labials, least the velars; some evidence suggests that these asymmetries are reflected in the varied chronology of the shift, too, changing first coronals, then labials, last velars (Franz 1883; see also Sonderegger 1974:157–158).1 (1a) Unshifted (Old Saxon) and shifted (Upper German) cognates WestGmc. *t*-tt*-nt *-Lt *p*-pp*-mp *-lp
Unshifted Old Saxon tehan laÌtan lenten (MLG) herta plegan appul damp (MLG) helpan
*-rp
werpan
*k*-kk*-nk *-Lk
korn ackar thankon folk
Shifted Upper German zëhan laΩΩan lenzo herza pflegan apful dampf helfan helphan = [pf] werfan werphan = [pf] chorna acchar danchoÌn folcha
Gloss ‘ten’ ‘to let’ ‘spring(time)’ ‘heart’ ‘to care for’ ‘apple’ ‘steam’ ‘to help’ ‘to throw’ ‘grain’ ‘field, acre’ ‘to thank’ ‘people’
2 (1b) Overview of the High German Consonant Shift CORONALS
Old Saxon Middle Franconian Rhenish Franconian East Franconian Upper German
zzzz-
Old Saxon Middle Franconian Rhenish Franconian East Franconian Upper German
d d/t d/t t t
LABIALS
t
p -ΩΩ-ΩΩ-ΩΩ-ΩΩ-
pp/pfpfpf-
-ff-ff-ff-ff-
b b b b p
VELARS
k kkkch-
-hh-hh-hh-hh-
g g g g k
(C- = initial position; -C- = medial position) We show in the present paper that this pattern of spread and change correlates directly with the hierarchized geometry of place-feature dependencies laid out by Avery & Rice (1989) and Rice (1994) (cf. also Clements & Hume 1995). In brief, the Old High German (OHG) tenuesshift asymmetries are seen to reflect the amount of place structure present in feature-geometric representations, such that the strongest, or most place-structured, consonants are most resistant to the shift. These same asymmetries, for which previous generations of scholars have long aimed to identify an articulation-based explanation (cf. Braune 1874, Prokosch 1917, 1938, others), are also mirrored in the later Medienverschiebung.2
1. Place Geometry. The proposal for an asymmetric geometry among place features was founded primarily on a bias first noted in the phonology of Korean by Iverson & Kim (1987), where, under conditions of regressive place assimilation, coronals assimilate to labials and velars, and labials assimilate to velars; but velars do not assimilate to labials, and neither labials nor velars assimilate to coronals. Iverson & Kim lay out the relevant data, recently recapitulated as in (2) (within the framework of Optimality Theory) by Iverson & Lee (1995):
3 (2)
Place of Articulation Assimilation in Korean /pan+myën/ /sin-paL/ /han+kaÑ/ /pat+ko/ /os+pota/ /ëp+ko/ /kam+ki/ But: /nop+ta/ /nok+ta/ /kuk+muL/
→ → → → → → →
[pammyën] [sËimbal] [haÑgaÑ] [pakk’o] [opp’oda] [ëkk’o] [kaÑgi]
‘on the other hand’ (< ‘half+side’) ‘shoes(-foot)’ ‘Han River’ ‘receive-and’ ‘clothes-than’ ‘carry-and’ ‘cold’ (< ‘feeling’+’energy’)
→ → →
[nopt’a] [nokt’a] [kuÑmul]
‘high’ (DECLARATIVE) ‘melt’ (DECLARATIVE) ‘soup-broth’
This asymmetry derives conventionally from the spreading of marked into unmarked structure if, following Avery & Rice (1989) and Rice (1994), a node ‘Peripheral’ is posited in the feature geometry subordinate to Place; coronals are then unspecified for all Place nodes, labials are marked just for Peripheral, and velars contain Peripheral along with the subordinate articulator Dorsal, as shown in (3). (3)
VELARS
LABIALS
CORONALS
Place
Place
Place
Peripheral
Peripheral
Dorsal Thus, coronals take on the qualities of both labials and velars because coronals are less fully represented than either labials or velars. By the same consideration, labials take on the qualities of velars but not coronals, while velars resist assimilation to either of the other two place of articulation categories. This pattern is found not just in Korean, but also in English (in Kingston → i[Ñ] Kingston, in Plymouth → i[m] Plymouth, from Kingston → fro[Ñ] Kingston, but from Toronto → fro[m] Toronto, etc.), and Rice (1994) draws other support for
4 the Peripherality idea from historical developments in Romanian and a number of Algonquian (Arapaho, Atsina, Yurok) and Athabaskan languages (Ingalik). For assimilations, then, assuming an underspecification model as in (3) (along with the stipulation that only specified features can spread, and only into unspecified positions), the markedness of the representations alone takes care of what assimilates to what. But this says nothing about nonassimilatory changes, which the model would express through feature delinkings or simply feature additions, as in the creation of affricates from stops. In Sections 3 and 4 below, we show how the markedness relations deriving from this place of articulation hierarchy also bear, perhaps surprisingly, on the purely manner of articulation adjustments in the history of German known collectively as the High German Consonant Shift. But first we summarize a recent characterization of these changes which seeks to identify just how they may have begun.
2. The Core of the High German Consonant Shift. Davis & Iverson (1995) develop the view that the High German Consonant Shift began in word-internal environments, shifting /p t k/ to the affricates /pf ts kx/ (which then later become the geminate fricatives /ff ss xx/) when in position after a short, stressed vowel. As voiceless stops in pre-Old High German are held to have been aspirated, parallel to most Germanic languages today (Iverson & Salmons 1995), Davis & Iverson consider that a primary motivation for this otherwise puzzling change likely resided in a familiar prosodic generalization known as the ‘Syllable Weight Law’ (or ‘Prokosch’s Law’, Prokosch 1938:140). Vennemann (1988:30) expresses this as the syllable structure preference given in (4), noting that it often interacts crucially with the ‘Head Law’ defined in (5), again following Vennemann (1988:13–14).3
5
(4) Syllable Weight Law In stress accent languages an accented syllable is the more preferred the closer its syllable weight is to two moras, and an unaccented syllable is the more preferred the closer its weight is to one mora.
(5) Head Law A syllable head is the more preferred: (a) the closer the number of speech sounds in the head is to one, (b) the greater the Consonantal Strength value of its onset, and (c) the more sharply the Consonantal Strength drops from the onset toward the Consonantal Strength of the following syllable nucleus.
A preferred stressed syllable is thus one whose weight is two moras, i.e., which includes either a short vowel followed by a consonant (VC) or a long vowel followed by no consonant (VV). In the case of posited forms like pre-OHG [ó.phan] ‘open’, however, the initial stressed syllable would have contained a short vowel with no following consonant. Pursuing a suggestion first put forth by Braune (1874:1, 1987:87), such configurations could be brought into conformity with syllable structure preferences via “segmentalization” of the allophonically aspirated stop ([ph]), i.e., by converting it into a phonetically equivalent cluster of stop plus /h/ ([ph]). This interpretation of the phonetics would have been specifically favored, or triggered, by the prosodic considerations of syllable structure identified in (4) and (5). Coupled with the Head Law’s preference for onsets containing precisely one consonant, the Syllable Weight Law’s call for bimoraicity in stressed syllables then may be construed as the initial motivation for the range of obstruent changes that ultimately took place in Old High German, because the restructuring of words like [ó.phan] to [óp.han] accords bimoraic status to the stressed syllable while still retaining a consonantal onset in the head of the following syllable.4 Subsequently, the place of articulation features of the preceding stop spread forward
6 into the /h/, strengthening it to a fricative ([óp.han] → [óp.fan]), while the coda stop itself ultimately weakened to yield the attested OHG [óf.fan]. Later still, the pattern established by this trigger process factoring aspiration out as an individual segment would have generalized to the other positions in which the shift took place; but we see a distinct prosodic motivation in aspiration segmentation at the earliest stage of these changes in order to bring stressed short syllables into conformity with the Syllable Weight Law. The claim that the High German Consonant Shift originated after short vowels finds significant synchronic support in turn-of-the-century speech patterns of the village of Wermelskirchen, reported by Hasenclever (1905:42-44). In this North-Rhenish dialect, the shift indeed took place only after etymologically short vowels, i.e., precisely NOT “… im Anlaut, in der Gemination und nach Konsonanten … [oder] nach ursprünglich langem Vokal” (“ … in word-initial position, in gemination and in post-consonantal position … [or] after originally long vowels.”). Thus, short-vowel stems with Germanic /t/ undergo the shift, but long-vowel stems do not: vesën ‘to know’, èsën ‘to eat’, but ßmîÌten ‘to throw’, ßtròÌ:të ‘street’.5 And the same pattern holds with respect to Germanic /p/ and /k/: òfën ‘open’, lèfël ‘spoon’, but ßarp, ‘sharp’, pèfër ‘pepper’; brèçën ‘to break’, vèçë ‘week’, but zy:kën ‘to seek’,
ßrîÌkën ‘to shriek’. In the strong verb system, where short and long vowels alternate in ablaut, shifted and unshifted consonants even cooccur in the same paradigms, as seen in the principal parts (infinitive, 3.sg. preterit, past participle) of: ßîÌ:tën, ßòs, jëßòsën ‘shoot’, esën, ò:t, jèsën ‘eat’, brèçën, bròÌ:k, jëbrò≈ën ‘break’, ßtrîÌkën, ßtreç, jëßtreçën ‘strike’, etc. As Wermelskirchen is located just north of the Benrath Line between Düsseldorf and Cologne (the northern border of the High German consonant shift and therewith the traditional dividing line between Middle and Low German dialects), Davis & Iverson infer that the core of the shift process reached that dialect as the leading “wave” spread northward. But the more general forms of the shift did not extend this far, thus leaving behind dialectal reflexes pointing to the first, elemental form of the shift.6 In sum, the Wermelskirchen data are strong indication that the High German Consonant Shift originally was restricted to positions
7 following a short stressed vowel, and that motivation for this incipient stage lay in interaction between the Syllable Weight Law and the Head Law via exploitation of the well known phonetic (though not syllabic) equivalence between a single aspirated stop and a sequence of stop + /h/.7 Thus it appears that the High German consonant shift did not consist of a single, abrupt change in the phonology of pre-Old High German as might be inferred from traditional accounts in the handbooks. Rather, the shift must be understood as a series of phonological changes, each one resulting in a new stage of the grammar in each of the regional dialects affected by the shift. Table (6) illustrates the sequence of the various changes that occurred in post-vocalic position, which were shared by all the major regional dialects of Old High German.
(6) PRE-OHG
*o.phan
WEIGHT LAW…
*op.han
SEGMENTATION ASSIMILATION WEAKENING
*slaÌ.phan –––
*skep.phjan –––
*uÌph –––
–––
*slaÌ.phan
*skep.phen
*uÌph
*op.fan
*slaÌ.pfan
*skep.pfen
*uÌpf
of.fan
slaÌ.fan ~ slaf.fan
skep.fen
uÌf
The High German Consonant Shift: Postvocalic Developments (Davis & Iverson 1995:119).
As noted at the outset, even the temporal implementation of the shift appears to reflect the distributional bias outlined in (1), which means that the first to be affected were the coronals, then the labials, last the velars. The most probable chronology appears to be that /t/ shifted in the south in the early 7th century, /p/ around 700, and /k/ in the 8th century (cf. Haubrichs 1987 and Buchmüller-Pfaff 1990, among others, for arguments in support of these dates using onomastic evidence). This chronology is still reflected to some extent in the dialect of Wermelskirchen,
8 too, where there are some apparently phonological exceptions to the shift of pre-OHG */k/ (as opposed to mere hypercorrections or borrowings which reintroduce /k/). Specifically, the shift of pre-OHG */k/ did not take place after originally short /a/, as in maÌkën ‘to make’, daÌk ‘roof’, ßvaÌk ‘weak’, zaÌkën ‘things’ (Hasenclever 1905:44). We infer that the sequence /ak/ (which, in open syllables, only much later lengthened to /aÌk/) proffered the last bit of resistance to the original form of the shift affecting first /t/, then /p/, and last /k/, of which some unshifted relics remain.8 To recapitulate our overall view, then, we consider that the High German consonant shift was primarily a prosodic event, triggered by interaction of the Syllable Weight Law and the Head Law in motivating the segmentalization of aspirated voiceless stops ([ph th kh]) into phonetically equivalent clusters of stop plus /h/ ([ph th kh]) when in position after a short, stressed vowel. Subsequently, but still in pre-Old High German, these reanalyzed clusters yielded affricates ([pf ts kx]) in the manner described above, probably changing first the coronal, then the labial, last the velar. We note again now that the Upper German dialects (by traditional assumption closest to the geographical point of origin of the shift) later affricated pre-OHG /p t k/ in word-initial position as well, even though the original motivation for the shift as provided in the Syllable Weight Law could not have played any role at the beginning of words. It is precisely in word-initial position—where affrication extended beyond the environment which originally motivated it—that major dialectal differences in the extent of the shift show up. In Upper German dialects, all three initial clusters of [ph th kh] were affricated, while the development in the Rhenish and East Franconian dialects (geographically somewhat farther out from the epicenter of the shift area) led to affricated coronals and labials, but not velars. The geographically still more remote Middle Franconian dialect failed to affricate either labials or velars while exhibiting a regular affrication of coronals, and, at the outer edge of the shift area—namely, in the dialect of Wermelskirchen—none of the stops affricated in word-initial position, not even coronals. Indeed, the shift might never have extended to word-initial environments in the first place had it not been for paradigmatic alternations which, we conclude, came about as a consequence
9 of the shift’s origin in medial environments. At a relatively early stage, when the shift had been generalized to affect voiceless stops following vowels irrespective of their length or stress (as in most of the dialects), numerous still unshifted words like [thaÌlen] ‘to pay’ would have alternated with shifted forms in such transparently inflected and derived words as (using Modern German examples) [getsaÌlt] past part., [betsaÌlen] ‘bezahlen’, etc., or in compounds (‘wiederzahlen’) and various procliticized words (such as with the negative proclitic OHG ni, MHG ne, or infinitival zu, often reduced to ze, assuming forms like *ze-zahlen). The pressure to regularize these paradigms forms a plausible impetus for extending the shift to word-initial environments as well, albeit with the now familiar reluctance exhibited by the dialect distributions exhibited in (7a, b) below.9
3. Peripherality and Extension of the Shift. Remaining manifestations of the High German Consonant Shift result from the extension of this original medial substitution to an increasing array of positions over time. From its inchoate instantiation after only short vowels, affecting /p t k/ equally, the shift generalized first to all intervocalic environments (pre-OHG [slaÌphan] > [slaÌffan] > OHG slaÌfan ‘to sleep’), then to postconsonantal positions ([helphan] > helpfan (= )’to help’, [dorph] > dorpf ‘village’), and, eventually, albeit far less consistently, to the beginnings of words ([phlegan], [thiohan], [khorn] > pflegan ‘to tend’, ziohan ‘to pull’, kxorn ‘grain’ (= )). These fully documented developments are summarized in the data set given in (7); it should be noted, however, that expansion of the shift to initial position appears actually to have “worsened” syllable structure patterns from the point of view of the Head Law,10 for the initial affricates may be presumed to have been disegmental in the same way as the affricates or fricatives which came about medially. That is, extension of the shift to initial environments appears to have made a simple, preferred onset more complex, contrary to the Head Law, whereas the original medial shift provided preferred mora structure (and simple onsets) to syllables which were otherwise less optimal. Perhaps this is simply the price that was paid in
10 order to extend the shift analogically to word-initial position, which already sanctioned two and even three-member consonant clusters (Straße, etc.); or, perhaps the segments that emerged word-initially were not construed as clusters at all (/pf/, etc.), but were rather perceived and structured as monosegmental affricates (/pf/, etc.), phonetically (if not prosodically) indistinguishable from their cluster counterparts. This possibility, of course, raises the familiar ‘one-phoneme-or-two’ question regarding the analysis of German affricates, which we will not try to answer here. But to begin with, at least, the postvocalic affricates were almost certainly disegmental, as motivated by the Syllable Weight Law, whereas in initial position they may have been interpreted as monosegmental, following the single-onset preference expressed by the Head Law.
11 (7a) Distribution of shifted vs. unshifted labials (Sonderegger, 1974:158). *pGmc *plegan M Franc plëgan Rh-Franc plëgan S Rh-Franc plëgan E. Franc pflëgan Bav pflëgan pflëkan Alem (p)flëgan
*-pp*aplaappul appul apful apful apful
‘to tend’
*-lp *gelpagelpa gelpf/gelp gelpf gelpf gelpf gelf gelpf gelf
*-rp *scarpascarp scarpf/scarp scarpf scarpf scarpf
apful afful
*-mp *kampakamp kamp kampf; rare: kamp kampf champf chamf champf chamf
‘apple’
‘battle’
‘high-spirited’
‘sharp’
scarpf
(7b) Distribution of tenues generally (Sonderegger, 1974:159). Coronal Pre-OHG t- -tt- C+t -t-
Labial -t
p-
OSax t tt MFranc z z Rh-Franc z z
t z z
t t p ΩΩ t/Ω p ΩΩ Ω p
S Rh-Fr E Franc Bav Alem Langob
z z z z z
ΩΩ ΩΩ ΩΩ ΩΩ s(s)
z z z z z
z z z z z
Ω Ω Ω Ω s
-pp-
mp
pp pp pp
mp mp mp
lp
lp lp lp/ lpf p pf mpf lpf pf pf mpf lpf pf pf mf lf pf/f pf/ff mf lf p p(p) mpf lpf
Velar rp
-p(-)
k-
-kk- C+k -k(-)
rp rp rp/ rpf rpf rpf rf rf rpf
p f(f) f(f)
k k k
kk kk kk
k k k
k ch ch
f(f) f(f) f(f) f(f) p/f(f)
k k k≈ ch k
kk kk k≈ k≈ kk
k k k≈ ch k/k≈
ch ch ch ch ch
Remarkable, however—if seldom remarked on—is the fact that the High German Consonant Shift overall constitutes a nonassimilatory sound change that produces relatively marked segmental output (affricates) from unmarked input (simple stops). Yet it has long been recognized that nonassimilatory sound changes typically engender reductions in markedness, not increases (cf. Antilla 1972, Houlihan & Iverson 1979, Vennemann 1988, Forner et al. 1992, among others). For example, marked rounded front vowels ([ü, ö]) all became unrounded ([i, e]) in the history of English (my:si > mice), marked affricates ([ts, tsË]) all became less marked fricatives ([s, sË]) in the history of French (cf. chief < medieval French vs. chef < modern French), etc. The creation of an affricate series from plain voiceless stops in
12 Old High German thus represents a distinct oddity among sound changes, not just by virtue of having neither precedent nor parallel elsewhere (cf. particularly Vennemann 1985), but also because, qua nonassimilatory change, the output of the shift is invariably marked relative to its input. Now, however, a clear structural motivation for the beginning stages of the High German Consonant Shift has been identified, namely, the drive to satisfy general preferences of syllable structure as called for by interaction of the Syllable Weight Law and the Head Law. The segmental markedness which results can then be seen straightforwardly as a price that had to be paid in order to meet these evidently higher priority prosodic requirements. But the skewed geographical distribution of the shift in initial position continues to puzzle Germanic specialists. Articulatory motivations for this pattern, favored in traditional and handbook accounts, have proven particularly unsatisfying. For example, the widespread assertion (reviewed in Braune/Eggers 1987:84, Lerchner 1971:32–33, Schirmunski 1962:350–351) that the largely word-initial “expiratory accent” of Germanic was somehow conducive to affrication falters on the fact that the core of the shift appears not to have been introduced in the prosodically most prominent position, the beginning of a word with initial stress, but rather in the middle of such words, at the junction between a stressed short syllable and a following unstressed one. If expiratory accent—and with it the higher degree of aspiration that accompanies word-initial voiceless stops in Germanic (Iverson & Salmons 1995)—had been solely responsible for initiation of the shift, then we should expect the dialect distributions to reflect uniformity of initial, not medial affrication. Prokosch (1917:16, elsewhere), in fact, assumes this very position, speculating that initial aspiration increased in strength to the point that affrication became virtually unavoidable, first with respect to the coronal, then the labial, and finally the velar stop. He thus must suppose that while affrication was the result of the shift in initial (and postconsonantal) position, medially—where aspiration was likely weaker—the stops converted directly to fricatives. That is, the change must be supposed to
13 have gone farther in an environment without phonetic motivation than it did in an environment with phonetic motivation.11 But if varying duration of aspiration among initial stops is held to be the property which correlates with a propensity to affricate, then, in view of the differences in Voice Onset Time (VOT) lag which are known to obtain among the voiceless stops of English and other modern Germanic languages, it would be expected that the effect on Germanic /k/ should have been more extensive than on /p/. This is because, in the modern cases for which data are available, word-initial voiceless stops are more heavily aspirated at the coronal and velar places of articulation than at the bilabial; the table in (8), taken from Weismer (1979), lists average values in milliseconds for the duration of aspiration following release of word-initial /p/, /t/, and /k/ before each of six different vowels in American English.
(8) Group Mean VOT Values in Milliseconds (Weismer 1979)
/p/ /t/ /k/
/i/ 57.33 67.06 79.40
/e/ 56.73 67.86 67.40
/I/ 44.06 63.33 71.66
/ε/ 48.46 65.13 62.66
/u/ 57.60 66.00 70.60
/æ/ 52.80 63.93 66.53
The aspiration that accompanies /p/ is substantially less than that of either /t/ or /k/, irrespective of the quality of the following vowel.12 Weismer further observes, moreover, that the duration of closure is longer in /p/ than in either /t/ or /k/, so that in English an essentially constant interval of voicelessness emerges across place of articulation. The reason for this difference appears to be that build-up of intraoral air pressure during the closure phase of a relatively posterior /t/ or /k/ results in an early release, or burst, into the aspiration phase of the stop, whereas, in the maximally anterior stop /p/, pressure build-up during closure can be maintained longer because of the greater (and much more expandable) oral cavity which is formed by a labial occlusion. As English and other Germanic languages call for a more or less constant overall duration for initial voiceless stops, the period of aspiration is relatively long
14 after /t/ and /k/ (around 65 ms on average, because these have the shortest durations of closure at around 100 ms); but VOT lag is shortest after /p/ (around 50 ms, because /p/ has the longest closure duration at around 115-125 ms) (cf. Weismer 1980, Fourakis & Iverson 1985, 1987, Hutters 1985.)13 If extent of aspiration had driven the place-variable High German word-initial changes, in other words, then [ph], with the shortest duration of aspiration, would have shifted less consistently and over a smaller area than [kh], which has a longer period of aspiration. Alternatively, if resistance to affrication were a function of closure duration, which indeed is relatively short in the stop which was most susceptible (/t/), then it would also be expected that /k/ (with shorter duration of closure than /p/) should have been more affected by the shift than /p/. In brief, the key phonetic considerations of segment duration and VOT (which differ for Germanic stops by place of articulation) simply do not line up with the place-variable extent of the shift. Instead, we conclude, a strictly structural motivation lies behind the way in which the High German consonant shift extended from medial to initial position. The varying content of feature representation by place of articulation on the geometric model of (3) accords exactly with the attested pattern of the shift’s extension, such that less place-feature structure correlates positively with broader spread of the shift, while more structure inhibits spread. In the probable last step of the shift, which extended these developments into initial position, changes in the stop whose place of articulation is unmarked (the coronal) reached the farthest north, while shift of the next least marked stop (the labial) is distributed across Middle German, but with limitations, and shift of the representationally most complex stop (the velar) is restricted to the far south. The geographically more limited affrication of labials relative to coronals and the still more limited affrication of velars in the High German consonant shift correspond with Rice’s (1994) observation that increased place structure content correlates with resistance to assimilatory change, velars being most specified as to content and most resistant to change, labials somewhat less so, and coronals least marked and most susceptible to change. Under the general assumption that only marked features spread, therefore, and then only into less
15 marked positions, the bias in assimilation derives automatically from phonologically underspecified geometric representation. The High German consonant shift of (originally monosegmental) stops to (medially disegmental) affricates is not an assimilation, however; yet it exhibits the same asymmetries as found in place assimilation. Nonassimilatory changes like the High German Consonant Shift will be subject to the observed place-bias, too, if it is assumed that there is a principle such as the following governing operations whose outputs are marked relative to their inputs:
(9) Input Markedness Criterion Unmarked structures are affected preferentially in shifts to marked outputs.
The assimilatory biases noted previously conform to this criterion because a shift from coronal to peripheral (or from labial to dorsal) place of articulation constitutes an increase in the representational markedness of the segments affected by the change, hence coronals are more liable to assimilate than peripherals (and labials more than dorsals). Extension of the nonassimilatory High German shift into word-marginal positions meets this criterion in the same way, with the same kind of bias, because the shift’s affricate output is marked relative to its simple stop input. Increased place structure content thus correlates with resistance to assimilatory as well as nonassimilatory change under the terms of (9).14 This principle further authenticates the fundamental role that markedness plays in sound change, for it complements Vennemann’s (1988) ‘Diachronic Maxim’ interpreting sound change generally as causing decreases in markedness.15 Indeed, the Diachronic Maxim can be rephrased in language exactly parallel to (9): ‘Marked structures are affected preferentially in shifts to unmarked outputs’.16 While this “fix-the-worst-first” principle holds for changes which happen to result in markedness reductions, many of them context-free, it misses the extensive class of context-sensitive changes which produce relatively marked outputs with a bias to affect unmarked structures preferentially over marked ones. Together, the Diachronic
16 Maxim (governing changes resulting in less marked outputs) and the Input Markedness Criterion (governing changes resulting in more marked outputs) cover a fuller range of markedness-based sound changes and contribute to explanation of the observed asymmetries.17
4. The Shift of the Voiced Stops. The later and more geographically restricted Medienverschiebung, or shift of the (traditionally lenis) voiced stops to voiceless, displays similarities to the Tenuesverschiebung described above with respect to the skewing of its geographical spread by place of articulation. Specifically, while the coronal lenis stop /d/ changed throughout the Upper and Central German dialects, the peripheral stops in this series occur as fortis largely only in Upper German documents. Handbooks often treat /b/ and /g/ as having basically the same distribution, but a number of texts reflect more shift of b > p than of g > k. For example, Penzl (1970:160) concludes: “Im Oberdeutschen zeigt sich in frühen Quellen p für die Media in allen Stellungen …” [“In early written sources, Upper German has p for the mediae [i.e. b] in all environments …”], offering examples from the early Bavarian Exhortatio, the St. Gall Paternoster and the St. Gall Vorakte. Penzl continues (1970:161): “Die Lage bei den Gaumenlauten gleicht in mancher Hinsicht der der Lippenlaute … Im Oberdeutschen jedoch weisen die Schreibungen mit k, c für vorahd. *g besonders frühahd., zwar vielfach auf stimmlosen Fortisverschlußlaut, doch bleibt die Opposition zu vorahd. *k im Anlaut und nach Konsonant (ausser nach s) allgemein bestehen.” [“The situation of the velars is in some ways similar to that of the labials … In Upper German, however, the graphemes k and c that continue pre-OHG *g indeed frequently represent voiceless, fortis stops especially in early Old High German. The phonemic contrast [of k and c] to pre-OHG *k generally remains intact, however, in word-initial position and post-consonantally (except after s).”] Similarly, Sonderegger (1974:161) observes: “nur auf das Oberdeutsche (mit breiter Vertretung im Bairischen) beschränkt ist die Verschiebung von b > p … Selbst oberdeutsch nicht durchgehend ist die Verschiebung von g > k.” [“The shift of b > p (with wide attestation in
17 Bavarian) is limited to Upper German … The shift of g > k is not complete even in Upper German.”] But if these changes had been driven by a general push to devoice what are traditionally analyzed as the Germanic voiced stops, we expect that [g] would have shifted to [k] first and most broadly, and [b] last and least. This is because greater physiological effort is required in order to maintain voicing in a velar than at points farther forward due to the small size of the chamber between the velum and larynx available for passive expansion while sustaining airflow over the vocal folds (Ohala 1983). This property is reflected typologically in the tendency of gaps in a simple voiced stop series to occur at the velar place of articulation, as in Dutch, Modern Standard Arabic, many dialects of Mixtec, and elsewhere (cf. Gamkrelidze 1975, Maddieson 1984). With bilabials, on the other hand, air-flow over the vocal folds is rather easily maintained during closure because of the greater space available for expansion of the vocal tract in accommodation of continued airflow up from the larynx. In fact, the propensity to voice a bilabial stop is sufficiently strong to preclude the phonemic existence of voiceless bilabial stops in a number of languages, Arabic among them. The correlation which holds with respect to spread of the Medienverschiebung is the same as obtains in the High German Consonant Shift proper: The presence of less place structure correlates positively with broader geographical spread. We follow Iverson & Salmons (1995) in representing the Germanic mediae, or lenis stops, as laryngeally unmarked (literally lacking specification) relative to the tenues, or fortis stops (identified by the feature [spread glottis]). The structural interpretation which results then is that the Medienverschiebung was effected in the form of addition of the feature [spread glottis], which means that this conversion of unmarked (lenis) inputs into marked (fortis) outputs was also subject to the Input Markedness Criterion laid out in (9). Accordingly, the strengthening of Germanic lenis /d/ to fortis /t/ was most widespread because its segmental representation is less than that of /b/ and /g/, and the shift of /b/ is better instantiated than that of /g/ because /b/ is representationally less complex than /g/.
18 This parallel between the change of the tenues and the change of the mediae has induced many scholars to speculate about the possibility of a chain shift, though some are unwilling to assume a direct connection between the High German consonant shift per se and the Medienverschiebung. Wells (1984:422), for instance, speaks of a “possibly related shift of other pre-OHG consonant phonemes.” Far more usual, however, is the assumption of a chain shift, with some arguing for a push chain (Martinet and many others, reviewed in Lerchner 1971:150–51) and others for a pull chain (Kufner, Moulton; for review of this literature, see again Lerchner 1971). King (1969:194–200) pursues this latter position on largely theoretical grounds, arguing against the possibility of push chains in general based on a rejection of the idea that phonological change is necessarily gradual.18 Under our view, the High German consonant shift and the Medienverschiebung share a basic structural motivation, namely, the tendency of stops with more place structure to be more resistant to shift. Were these two changes directly connected in a drag chain—where the gap of the shifted /t/ had pulled a /d/ to /t/ in its wake—we would expect a dialect like Wermelskirchen to show consistent shifting of /d/ to /t/ after etymological short vowels, where a void existed by virtue of /t/ having shifted to /ts/ (> /ss/). Hasenclever, however, describes exactly the opposite situation, i.e., the retention of Germanic /d/ throughout the dialect. While Germanic /†/ did shift to /d/ throughout the broader set of continental dialects, this resulted in merger instead of precipitating a further shift of Germanic /d/ to /t/ (dax ‘day’, not *tax). In a chain shift consisting of three steps, then, Wermelskirchen provides evidence for the first and the third, but specifically lacks the crucial middle link. This obtains throughout Middle Franconian, in fact, even in areas which exhibit more of the shift generally. For example, Cologne and other Middle Franconian dialects show affrication of initial /t/ to /ts/, cf. Zick ‘time’ (Modern Standard German Zeit, with typical velarization of the final stop in the Middle Franconian form), but have merged old /†/ with /d/, as in deck ‘thick’ (MSG dick), and did not participate at all in the Medienverschiebung of the coronal, cf. Dier ‘animal’ (MSG Tier). Thus, though these dialects created a substantial gap among the traditional voiceless stops with
19 the shift of /t/ to /ts/, they did not fill that gap by further shifting /d/ to /t/, whether that /d/ was original or the product of merger with /†/.19 The relevant developments are schematized in (10). (10) Evidence against a pushchain from Wermelskirchen Stage Change: Wermelskirchen:
†
1 > yes
d
2 > no
t
3 > ts partial
Such evidence is fatal to a push chain view, for if the drive to maintain distinctions in the obstruent system (combined with the goal of ridding the system of highly marked /†/) had caused /t/ to shift to /ts/, it would have necessarily been by way of the previous shift of /d/ to /t/—otherwise there would have been no “push”. In a drag chain approach, though, the picture is more complex. The “slot” traditionally regarded as the most unmarked in the system is partially vacated by the shift of /t/ to /ts/ in Wermelskirchen and many dialects to the south without movement of more marked /d/ into that position. Germanic /d/ did shift to /t/ throughout most (but not all) of the High German area, and this served to fill a gap created by the affrication of /t/ to /ts/; but remaining elements of the Medienverschiebung were far more restricted.20
5. Conclusion and Summary. We conclude that the long-recognized but largely unexplained asymmetries in the extent of the High German consonant shift reflect the hierarchical, apparently universal scheme of representational markedness for place of articulation properties given in (3). The High German shift thus affected /t/ more broadly and with greater regularity in post-consonantal and wordinitial positions than it did /p/, and /p/ more than /k/, because this ranking reflects the featuregeometric markedness of these stops relative to each other. In short, a segment’s resistance to the High German consonant shift correlates with the complexity of its representation, just as does
20 its resistance in other languages to place of articulation assimilation.21 The prosodically motivated aspect of the shift ([ophan] > offan) in Wermelskirchen was subsequently extended elsewhere into post-long-vowel environments ([slaÌphan] > slaÌfan). This first extension of the shift could have served no clear prosodic function, for both the Syllable Weight Law and the Head Law would have been fully satisfied by the status quo ante of Germanic forms like [slaÌ.phan]; however, the original absence of shift after long vowels gave rise to numerous paradigmatic alternations of the sort still extant in turn-of-the-century Wermelskirchen, e.g.,
ßîÌ:tën, ßòs, jëßòsën ‘shoot’, esën, ò:t, jèsën ‘eat’. The first generalization of the prosodically motivated core shift from “position following short vowel” to “position following any vowel” removed these alternations throughout (most of) High German, making the paradigms of strong verbs uniform with respect to their shifted consonantism. The drive to reduce intraparadigmatic variation, or allomorphy, is among the strongest forces in language change (cf., e.g., Kiparsky 1971), and is seen here as the progenitor of the extension of the High German shift from a strictly prosodically determined core change to one triggered, in environments where it could be generalized, by paradigmatic or morphophonemic factors as well. The first and most regular of these extensions, then, was from a preceding short vowel environment to all postvocalic positions. The dialect distributions suggest that the next level of extension was to the next level of sonority, to position following liquids; hence [helphan] > helpfan, [dorph] > dorpf, etc., as detailed above in (7a, 7b). At about the same time, (postvocalic) geminates were affected as well, e.g., [skep.pian] > sche(p)phen ‘to create’. Likely last (if at all) to be affected among the medials, in view of the distributions cited in (7), were stops following even less sonorous segments, viz., the nasals (modern German Dampf ‘steam’, etc.). Following obstruents, no shift at all took place, hence /t/ remains unshifted in oft ‘often’, acht ‘eight’, etc., as well as in word-initial clusters (Straße ‘street’). In fact, the word-initial environment was the least hospitable to shift even of single stops, and the velar, the most resistant to change among the three places of articulation, remains unshifted word-initially in the standard language even though it did shift in postvocalic and post-liquid environments (Kirche
21 ‘church’). This word-initial context represents a greater relaxation of the original post-short vowel environment than does its successive generalization to all preceding vowels and subsequently to any preceding sonorant. The analogical leap from progressively more general medial and then final contexts to word-initial position constitutes the end stage of the tenues shift, and here its sensitivity to place-markedness is greatest of all. Moerover, the shift of the mediae reveal the same bias according to place-markedness, with coronal /d/ most likely to undergo change to /t/, /b/ less likely to shift to /p/, and /g/ least likely to shift to /k/. Rice (1994) concludes her exposition of phonological place-asymmetries with this assessment: “The major point that I hope to have made, then, is that empirical evidence provides support for a hierarchical hypothesis over a flat one, requiring that labials and dorsals be ranked with respect to each other and not be equally ranked with respect to coronals.” The present paper has sought to provide additional support for just that hypothesis, along with an expanded understanding of the central role of markedness in sound change.
22 Notes
1
While Rice (1996) distinguishes between ‘velar’ and ‘dorsal’ place of articulation, we use the
terms synonymously in this paper.
2
In Optimality Theory, Padgett (1995) has proposed a non-geometrical approach, “Feature Class
Theory”, to handle similar patterns of place assimilation. Cahill & Parkinson (forthcoming) and Parkinson & Cahill (1997), however, have raised serious questions about Feature Class Theory, arguing that its purported advantages over geometrical accounts are illusory and showing that it predicts many unattested assimilations that geometric accounts do not.
3
‘Head’ in Vennemann’s terminology refers to the syllable onset. For further discussion,
including of syllable contact, see Davis & Iverson (1995).
4
Other Germanic languages, faced with this same prosodic challenge, responded with open
syllable lengthening, as in Dutch and other Germanic languages.
5
The products of this shift in Wermelskirchen must have been, at least originally, geminates.
6
Our view that the Wermelskirchen dialect (as recorded by Hasenclever 1905) preserves an
incipient phase of the OHG consonant shift is consistent with the known settlement history of the region. Onomastic evidence in the vicinity of Wermelskirchen indicates settlement between the 7th-8th centuries with a majority of settlers Saxons, not Franks, given many place names in -inghausen (Dittmaier 1956). As the OHG consonant shift radiated out from its area of origin in the south, it would have taken some time to reach its northern point of termination; so the first phase of the shift may not have reached this area until after settlement. Indeed, the linguistically
23
mixed area on the border between Frankish and Old Saxon is a likely place for this southern innovation to have ground to a final halt. We conclude that the Wermelskirchen *p, *t, *k reflexes unveil the margins of an innovative area, for here not only is the shift of /p t k/ limited to word-internal position after short vowels, but the shift of /k/ is incomplete even in this syllabically triggered environment (cf. below). 7
Wermelskirchen has been characterized as phonologically innovative by Wiesinger (1995:33-
34), who suggests that the dialect’s vowel system was in flux in the late nineteenth and early twentieth centuries, influenced by nearby dialects. Like most other Ripuarian dialects, however, Wermelskirchen was also affected by the development variously known as the Rhenish Accentuation (Rheinische Akzentuierung), Schärfung, or correption. The results of correption are not uniform in all dialects, but it brings about changes in vowels followed by /l r m n / (cf. Newton 1990:153). Like any instability in the dialect dating from around the turn of the century, of course, correption is of post-Middle High German origin and so did not affect the fundamental patterns of short and long vowels found in the earlier system. In short, the environment in which the shift of pre-OHG */p t k/ began (*-VÃp/t/k) has remained stable with respect to vowel length in the dialect since Old High German despite other developments that may have occurred subsequently. 8
The failure of /ak/ sequences to shift consistently in the expected fashion in Wermelskirchen
may be explained in terms of the Input Markedness Criterion (IMC, discussed below and stated in (9)). According to the IMC, unmarked structures are affected preferentially in shifts toward marked outputs. Thus, relatively marked velars are less likely to succumb to change than either coronals or labials in nonassimilatory changes like those under discussion. The sequence /ak/ was more resistant to the shift than were sequences of other short vowels + /k/ (e.g. /ik/, /ek/,
24
/ok/, /uk/) in the original Wermelskirchen system because the sum of the marked properties in /ak/ is less than in sequences of other short vowels+ /k/, given that /a/ is the least marked vowel. We expect, then, that in cases where more marked segments are beginning to change (after the lesser marked segments have already done so), the tendency is for the least marked among them (e.g. /ak/) to change last and least. In some East Middle German areas, including parts of Thuringia, the isogloss for shifted initial
9
p- runs to the north of the line for medial -pp-. That is, there are areas showing fund/pfund but appel, cf. Schirmunski (1962:296). Since these communities were formed originally by a mixture of dialect speakers from the west, we concur with the traditional assumption that these are dialect borrowings of shifted forms limited to (prosodically prominent) word-initial position.
10
Postconsonantal affrication would have a similarly deleterious effect on syllable codas (cf.
the ‘Coda Law’, Vennemann 1988) unless affricates in that position too were analytically monosegmental.
11
Without the differentiation by place, this view is essentially the standard handbook
interpretation of the shift, namely, that in ‘strong’ positions—initial, postconsonantal or geminate—aspiration triggered affrication, while in ‘weak’ position—intervocalic—the stops went directly to fricatives; cf. Paul, Wiehl, Grosse (1987:116–117). 12
Before /i/, aspiration in /k/ is especially long, perhaps “…due to the unique pre-constriction
volume conditions associated with dorsal stops” (Weismer 1979:202; cf. also Klatt 1991). 13
Indeed, as Iverson and Salmons (1995) have argued, the trigger for spirantization of
voiceless stops under the earlier Germanic Consonant Shift (/p, t, k/ > /f † x/) was also aspiration (or rather a spread glottis articulation at the point of closure release), hence Grimm’s
25
Law took effect in PIE *tak- ([thakh-]) > Gothic óah-an ‘be silent’, but not in PIE *ist ([ist]) > Gothic ist ‘is’.
14
Accordingly, we find consistent shift across the High German area with respect to the affricate
which is found most commonly around the world (the coronal), while the least common affricate is also the most geographically restricted (the velar).
15
In Vennemann’s words, “… linguistic change on a given parameter does not affect a language
structure as long as there exist structures in the language system that are less preferred in terms of the relevant preference law” (1988:2).
16
Vennemann proposes the Diachronic Maxim for ‘syllable structure preference laws’, but the
principle appears to hold for markedness reductions in general. 17
The preferences noted lead to the broader, synthetic generalization that input markedness
varies inversely with output markedness. 18
In more recent times (see especially Labov 1994), of course, the empirical study of sound
change in progress has firmly established both the general gradualness of sound change and the existence of phonological push chains.
19
The markedness oddity of increasing the stock of /d/’s in the system while /t/’s are being
moved out is merely apparent, according to Iverson & Salmons (1995), who view /d/ as unmarked in the /d/ - /t/ opposition.
20
Thomas Klein points out to us (by electronic communication) that the evidence in South
Rhine Franconian speaks for the beginning of the Medienverschiebung with /d/ to /t/ in medial
26
position; for example, Otfrid shows initial (unshifted) d-, but medial and final –t(-). In view of support for the claim that the tenues shift started with the coronal (/t/ to /ts/) before moving on to labials and velars, a drag chain analysis may well be defensible. 21
For the original form of the shift triggered by the Syllable Weight Law and Head Law
considerations, we have far less information on place-sensitive variability. Still, even there, the Wermelskirchen reflexes show some exceptions involving the velar, particularly after /a/.
References
Antilla, Raimo 1972: An Introduction to Historical and Comparative Linguistics. New York: Macmillan.
Avery, Peter & Keren Rice 1989: Segment Structure and Coronal Underspecification, Phonology 6.179–200.
Braune, Wilhelm 1874: Zur Kenntnis des Fränkischen und zur hochdeutschen Lautverschiebung. PBB (Beiträge zur Geschichte der deutschen Sprache und Literatur) 1.1–56.
— 1987: Althochdeutsche Grammatik. 14th ed., ed. by Hans Eggers. (Sammlung Kurzer Grammatiken Germanischer Dialekte; A. Hauptreihe, 5.) Tübingen: Niemeyer.
Buchmüller-Pfaff, Monika 1990: Siedlungsnamen zwischen Spätantike und frühem Mittelalter. Die -(i)acum-Namen der römischen Provinz Belgica Prima. Tübingen: Niemeyer.
Cahill, Mike & Frederick Parkinson Forthcoming: Partial Class Behavior and Feature Geometry: Remarks on Feature Class Theory, Proceedings of NELS 27.
Clements, George N. & Elizabeth Hume 1995: The Internal Organization of Speech Sounds, The Handbook of Phonological Theory, ed. John Goldsmith. Cambridge, Mass.: Blackwell, pp. 245–306.
28 Davis, Garry W. & Gregory K. Iverson 1995: The High German Consonant Shift as Feature Spreading, American Journal of Germanic Linguistics & Literatures 7.111–127.
Dittmaier, H. 1956: Siedlungsnamen und Siedlungsgeschichte des Bergischen Landes, Neustadt a.d. Aisch.
Forner, Monica, Jeanette K. Gundel, Kathleen Houlihan & Gerald Sanders 1992: On the Historical Development of Marked Forms, Explanation in Historical Linguistics (Current Issues in Linguistic Theory 84), ed. by G. Davis & G. Iverson. Amsterdam & Philadelphia: John Benjamins, pp. 77–93.
Fourakis, Marios & Gregory K. Iverson 1985: On the Acquisition of Second Language Timing Patterns, Language Learning 35.431–442.
— &—1987: More on Second Language Timing Patterns, Language Learning 37.297–302.
Franz, W. 1883: Die lateinisch-romanischen Elemente im Althochdeutschen. PhD dissertation, Straßburg.
Gamkrelidze, Thomas V. 1975: On the Correlation of Stops and Fricatives in a Phonological System, Lingua 35.231–61.
Hasenclever, Max 1905: Der Dialekt der Gemeinde Wermelskirchen. Marburg: Elwert.
Haubrichs, Wolfgang 1987: Lautverschiebung in Lothringen. Zur althochdeutschen Integration vorgermanischer Toponyme der historischen Sprachlandschaft zwischen Mosel und Saar.
29 In: Rolf Bergmann u.a. (Hrsg.), Althochdeutsch. Bd. II: Wörter und Namen. Heidelberg:Winter, 1350-1400.
Houlihan, Kathleen & Gregory K. Iverson 1979: Functionally Constrained Phonology, Current Approaches to Phonological Theory, ed. by D. Dinnsen. Bloomington: Indiana University Press, pp. 50–73.
Hutters, Birgit 1985: Vocal Fold Adjustments in Aspirated and Unaspirated Stops in Danish, Phonetica 42.1–24.
Iverson, Gregory K. & Kee-Ho Kim 1987: Underspecification and Hierarchical Feature Representation in Korean Consonantal Phonology, Proceedings of the Chicago Linguistics Society 23:2.182–198.
— & Shinsook Lee 1995: Variation as Optimality in Korean Cluster Reduction, Proceedings of the Eleventh Eastern States Conference on Linguistics, ed. by J. Fuller, H. Han, & D. Parkinson. Ithaca, NY: DMLL Publications, pp. 174–185.
— & Joseph C. Salmons 1995: Aspiration and Laryngeal Representation in Germanic, Phonology 12.369–396.
Kenstowicz, Michael 1994: Phonology in Generative Grammar. Cambridge, Mass: Blackwell.
King, Robert D. 1969: Historical Linguistics and Generative Grammar. Englewood Cliffs, NJ: Prentice-Hall.
30 Kiparsky, Paul 1971: Historical Linguistics, A Survey of Linguistic Science, ed. by W. O. Dingwall. College Park: University of Maryland Linguistics Program, pp. 576–649.
— 1985: Some consequences of Lexical Phonology, Phonology 2.85–138.
Klatt, Dennis H. 1991: Voice Onset Time, Frication, and Aspiration in Word-Initial Consonant Clusters, Readings in Clinical Spectrography of Speech, ed. by R. J. Baken & R. G. Daniloff. San Diego: Singular Publishing Group, pp. 226-246.
Labov, William 1994: Principles of Sound Change, volume 1: Internal factors. Cambridge, Mass.: Blackwell.
Lerchner, Gotthard 1971: Zur II. Lautverschiebung im Rheinisch-Westmitteldeutschen. Halle: Niemeyer. (Mitteldeutsche Studien, 30.)
Maddieson, Ian 1984: Patterns of Sound. Cambridge: Cambridge University Press.
Newton, Gerald. 1989. Central Franconian, The Dialects of Modern German: A linguistic survey, ed. by Charles V. J. Russ. Stanford: Stanford University Press, pp. 136–209.
Ohala, John J. 1983: The Origin of Sound Patterns in Vocal Tract Constraints, The Production of Speech, ed. by P. MacNeilage. New York: Springer, pp. 189–216.
— 1995: The perceptual basis of some sound patterns. Phonology and Phonetic Evidence: Papers in laboratory phonology, IV, ed. by Bruce Connell and Amalia Arvaniti. Cambridge: Cambridge University Press, 87–92.
31 Padgett, Jaye 1995: Feature Classes, Papers in Optimality Theory, ed. by Jill N. Beckman, Laura Walsh Dickey and Suzanne Urbanczyk. (University of Massachusetts Occasional Papers, 18.) Amherst: Graduate Linguistic Student Association, 385–420.
Parkinson, Frederick & Mike Cahill 1997: Overgeneration in feature class theory, presented to the Annual Meeting of the Linguistic Society of America, Chicago.
Paul, Hermann 1987: Mittelhochdeutsche Grammatik. 23rd ed., ed. by Peter Wiehl & Siegfried Grosse. (Sammlung Kurzer Grammatiken Germanischer Dialekte; A. Hauptreihe, 2.) Tübingen: Niemeyer.
Penzl, Herbert 1970: Lautsystem und Lautwandel in den althochdeutschen Dialekten. München: Hueber.
Prokosch, Eduard 1917: Die deutsche Lautverschiebung und die Völkerwanderung, Journal of English and Germanic Philology 16.1–26.
— 1938: A Comparative Germanic Grammar. William Dwight Whitney Linguistic Series. Baltimore: Linguistic Society of America.
Rice, Keren 1994: Peripheral in Consonants, Canadian Journal of Linguistics 39.191–216.
— 1996: Default Variability: The Coronal-Velar Relationship, Natural Language and Linguistic Theory 14.493–543.
Schirmunski, Viktor M. 1962: Deutsche Mundartkunde. Berlin: Akademie Verlag.
32 Sonderegger, Stefan 1974: Althochdeutsche Sprache und Literatur. Berlin: Walter de Gruyter.
Vennemann, Theo 1985: The Bifurcation Theory of the Germanic and German Consonant Shifts: Synopsis and Some Further Thoughts, Papers from the Sixth International Conference on Historical Linguistics, ed. by J. Fisiak. Amsterdam: John Benjamins Publishing Company. pp. 527–547.
— 1988: Preference Laws for Syllable Structure and the Explanation of Sound Change—With Special Reference to German, Germanic, Italian, and Latin. Berlin: Mouton de Gruyter.
Weismer, Gary 1979: Sensititivity of Voice Onset Time (VOT) Measures to Certain Segmental Features in Speech Production, Journal of Phonetics 7.197–204.
— 1980: Control of the Voicing Distinction for Intervocalic Stops and Fricatives: Some Data and Theoretical Considerations, Journal of Phonetics 8.427–438.
Wells, Christopher J. 1984: German: A History to 1945. Oxford: Clarendon.
Wiesinger, Peter 1975: Strukturgeographische und strukturhistorische Untersuchungen zur Stellung der bergischen Mundarten zwischen Ripuarisch, Niederfränkisch und Westfälisch, in: Neuere Forschungen in Linguistik und Philologie. Festschrift für Ludwig Erich Schmitt, Wiesbaden.