1.1. A Monogenetic or Polygenetic Shift?

While the facts regarding the history of pre-Old High German t, p, k under conditions of the OHG consonant shift (Tenuesverschiebung) are well known, traditional accounts struggle to explain both the mechanisms of the shift as well as the asymmetry of its distribution (see table 1). The outcome of the shift can be sketched as follows: Pre-OHG t, p, k in word-initial position became the voiceless affricates ts, pf, kx (as in ⁺tiohan, ⁺plegan, ⁺korn > OHG ziohan ‘to pull’—here the grapheme <z> presumably stands for /ts/—pflëgan ‘to tend’, chorn ‘grain’). Only the southern dialects of Old High German were affected by the shift of ⁺k > kx, however, and the more northerly dialects also were generally exempt from the shift of ⁺p > pf, particularly in word-initial position. The outcome of the shift after the sonorant consonants -l-, -r-, -m-, -n-, as well as in gemination, was generally the same as in word-initial position, at least in the Upper German dialects (pre-OHG ⁺swart > OHG swarz ‘black’, ⁺helpan > OHG hëlfan ‘to help’, ⁺dorp > OHG dorpf ‘village’, ⁺skalk > Old Upper German scalch ‘servant’, and ⁺skeppian > OHG ⁺sce(p)phen ‘to create’). However, after vowels the shift resulted in geminate fricatives rather than in affricates; for example, pre-OHG +opan > OHG offan ‘open’, ⁺etan > ëzzan ‘to eat’, ⁺makōn > mahhōn ‘to make’ (-hh- presumably indicated /xx/). Later, these geminate fricatives tended to reduce to single segments when the preceding vowel was long (slāffan > slāfan ‘to sleep’) (see Davis & Iverson Reference Davis and Iverson1995:111–112, also Sonderegger Reference Sonderegger2003:259–263).

Table 1.

Overview of the OHG Consonant Shift (after Sonderegger Reference Sonderegger2003:263).

The conventional monogenetic view of the shift holds that there was only a single shift event that spread geographically. Since the shift occurs in its most developed form in Old Alemannian, traditional scholarship postulates that the shift began there before gradually spreading both north (to the Franconian dialects) and south (to Langobardian) (Sonderegger Reference Sonderegger2003:267), losing some of its force and generality as it went.Footnote ¹ The complete absence of the shift of word-initial pre-OHG k- (to kx-) in the Franconian dialects, and a corresponding lack of the shift of initial p- to pf- in South Rhine Franconian and dialects to the north, is often cited as an example of this gradual ebbing of the shift's force (Rhenish fan) as it traveled north from its epicenter.

The monogenetic theory, however, lacks both a motivation for the origin of the shift—beyond pointing to the general effects of the aspiration of ⁺t, ⁺p, ⁺k—as well as a plausible explanation for its spread from Upper German to Franconian, a directionality that seems strange given the dominant position of the Franks at that time vis à vis the Bavarians and Alemannians.Footnote ² In addition, the assumption that the shift was fueled by heavy aspiration of ⁺t, ⁺p, ⁺k (Prokosch Reference Prokosch1917:16, Braune/Eggers 1987:84, Lerchner Reference Lerchner1971:32, and Zhirmunskii Reference Zhirmunskii1962:350ff) would predict the most extensive shift to occur in word-initial position, given that allophonic aspiration is generally strongest in that position. However, the shift advanced geographically further in word-internal position than elsewhere. This result—and the much more limited shift of initial ⁺k- (> kx) in comparison to initial ⁺t- (ts-) (despite the fact that both are more aspirated than p)—forces us to refine our notions of how aspiration influenced the shift (Davis, Iverson, & Salmons 1999:189). Questions regarding the relative ease of shift also pertain to long consonants (geminates), since the geminate +-tt- (> /ts/, often spelled <tz>) shifted over the entire OHG speech area, while the shift of the geminates ⁺-pp- and ⁺-kk- is geographically more limited, analogous to the shift of ⁺p- and ⁺k- in initial position.

1.2. Polygenetic Origin, External Influence, and Diastratic Variation

Clearly, the tidy scenario of a monogenetic origin of the shift is muddled by the existence of forms of an early age in the Rhineland that attest shifted ⁺t, ⁺p, ⁺k in postvocalic position. Schützeichel (1964:22, 1976:337) found evidence of shifted forms of early vintage in Middle Franconian in various sources including the Maihinger Evangeliar (Echtner, late eight century), the Bible Glossary of Paris (ninth century), the Cologne Glosses (ninth century), and the Lex Ribuaria (circa 800 AD). The Lex Ribuaria manuscript attests staffulus, staf(f)ulum ‘stone steps (upon which the royal court holds forth)’ (German, ‘Steinaufbau des Königsgerichts’), and similar forms with shifted postvocalic ⁺-p from Latin; compare Old English stapol, OHG staffal, staphal (Schützeichel Reference Schützeichel1976:370ff). The oldest glosses such as the Maihinger Evangeliar attest forms such as hase ‘hate’ and gan:os: ‘comrade’ (< ⁺-t). Place names containing the suffix -affa (< -apa-) that pre-date 800 AD are well known (Schützeichel Reference Schützeichel1976:309), and the general shift of word-initial ⁺t- in the Rhineland is recognized (Schützeichel Reference Schützeichel1964, 1968, Bruch Reference Bruch1953:129). Schützeichel's findings have been confirmed by others (for example, Bergmann Reference Bergmann1977:318), and most researchers now concur that a critical mass of evidence indicates that the shift was underway in the Rhineland at an early time (Schwerdt Reference Schwerdt2000:311). The shift of initial ⁺t- is especially problematic to those who advocate an authochthonous shift in the Rhineland, however, because it is difficult to explain why an autochthonous Rhenish shift would resemble the Upper German shift with respect to the shift of initial ⁺t (> ts-) but not with respect to the shift of initial ⁺p and ⁺k. For some, the solution is to posit the shift of ⁺t in the Rhineland as an independent development via assibilation or one influenced by similar developments in Romance (Schützeichel Reference Schützeichel1976: 239–240), or a combination of both (Heinrichs Reference Heinrichs1967). Finally, Venema (Reference Venema1997) and Pitz (1999:392) specifically emphasize the chronological plausibility of a connection between the shift of ⁺t in the Rhineland and the advent of the Romance assibilation in the sixth to the seventh centuries.

Another line of research emphasizes possible diastratric (sociolinguistic) influences on the shift in the Rhineland. Heinrichs (Reference Heinrichs1967), for example, reconciles the claim of an autochthonous Rhenish consonant shift with the parallel existence of shifted and unshifted forms by suggesting that the shift may have in fact been in place in the Franconian upper social strata during the Middle Ages, while unshifted forms were still in use by the lower social strata (unverschobene Grundschicht). This supposes that the speech of both groups was to some extent represented in the written documents of the time. (See Meineke & Schwerdt Reference Meineke and Schwerdt2001:224 for an opposing view.) Lerchner (Reference Lerchner1971) documented the large number of unshifted relic forms in the Rhineland and developed a now popular view that the OHG shift was adopted into the Rhineland and West Middle German areas from the south based on shifted forms introduced by the Franconian upper classes. According to this view, the shift took place via the process of sound substitution (Lautersatz), but the lower socioeconomic classes were more resistant to the adoption of these new sounds, which Venema (Reference Venema1997) contends would explain, in part, the many unshifted relic forms in the Rhineland. After examining relic forms in the Rhineland with unshifted ⁺t, Venema (1997:462–465) concludes that the large number of instances of sound substitution that occurred from Upper German models must have brought about the extrapolation of a shift rule in the Franconian dialects that, again, varied according to social stratum. In Venema's model, shifted and unshifted forms could exist parallel at differing diastratic levels for long periods of time, and unshifted forms would sometimes eventually win out over their shifted counterparts, thereby (re)emerging as the dominant form in dialect use even at a comparatively late date.

Finally, Schwerdt (2000:346, 385–387) contends that the consonant shift in the Rhineland is not the same consonant shift that took place in the Old Upper German dialects, and that the two shifts did not result from a common trigger or via an analogous process, the similar outputs of both shifts being thus purely coincidental. As Schwerdt (p. 346) notes, this assumption is actually a retreat to a position once articulated as a theoretical possibility by Lerchner, but subsequently rejected by him. Lerchner (1971:193) wrote:

As for a possible motivation for the shift, Lerchner leaves the general impression that the shift (at least in the Old Upper German dialects) might have been triggered by prosodic (stress) conditions, but no evidence is provided in support of this view.

3.1. The Triggers: Stress, Bimoraicity, and Segment (Re-)Structuring.Footnote ⁴

The progression theory presented here maintains that the inherited system of stressing the root syllable of Germanic words steered monomoraic syllables in the pre-OHG dialects (indicated in our short-hand as (C)V) toward a preference for bimoraicity definied as either a long vowel ((C)VV) or as a short vowel followed by a tautosyllabic consonant ((C)VC). Syllable bimoraicity could thus be introduced in monomoraic syllables by invoking open syllable lengthening (OSL, (C)V) > (C)VV), as exemplified by pre-OHG ⁺o.pan ‘open’ > ⁺ō.pan where a dot (.) indicates a posited syllable boundary. While dialects in some regions would evolve in this way, the dialects that were to become Old High German selected an alternative form of syllable restructuring to achieve the needed bimoraicity in stressed syllables while simultaneously maintaining good syllable contacts. It was this syllable restructuring that set in motion the shift of ⁺t, ⁺p, ⁺k after short vowels, which was, in turn, the trigger for the rest of the shift to follow. (Following Davis, Iverson, & Salmons (1999), I refer to this development as the core shift.) It is thus no coincidence that non-shifting dialects such as Low Franconian developed OSL early (Paul Reference Paul, Wiehl and Grosse1998:58), since the operation of OSL by definition bleeds the environment in which the shift would otherwise operate (that is, in stressed, monomoraic syllables). It is thus to be expected that the isogloss between dialects that developed OSL early and dialects that developed it late frequently corresponds to the boundary between shifting and non-shifting dialects, a correlation that has been noted along the Benrath line (Hasenclever Reference Hasenclever1905, Bubner Reference Bubner1935) without recognition of its significance.Footnote ⁵

3.2. Core Shift

Following Davis & Iverson (Reference Davis and Iverson1995), the primary motivation for the restructuring that triggered the shift resides in a familiar prosodic generalization known as the Syllable Weight Law (or Prokosch's Law, Prokosch Reference Prokosch1938:140)—itself an outgrowth of the Germanic stress placement on the root syllable of a word. Vennemann (1988a:30) expresses this as the syllable structure preference given in 1, noting that it often interacts crucially with the Head Law defined in 2, again following Vennemann (1988a:13f).

1. (1) 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.

2. (2) Head Law

   A syllable 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 the 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, that is, which includes either a short vowel followed by a consonant (VC) or a long vowel followed by no consonant (VV). As we posit voiceless stops in pre-Old High German to have been aspirated, parallel to most Germanic languages today, we reconstruct the phonetics of pre-OHG ⁺opan ‘open’ as [‘o.p^han] (where a raised tick, ‘, indicates stress). That is, the initial stressed syllable would have contained a short vowel with no following consonant and the syllable head of the following syllable would have contained a voiceless, aspirated stop. Pursuing a suggestion first put forth by Braune (1874:1) (see also Braune/Eggers 1987:87), such configurations can be brought into conformity with syllable structure preferences via segmentalization of the allophonically aspirated stop ([p^h]), that is, 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 1 and 2 above. 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 [e.t^han], [o.p^han], [ma.k^hōn] > [et.han], [op.han], [mak.hōn] accords bimoraic status to the stressed syllable while retaining a consonant onset in the head of the following syllable. Subsequently, the place of articulation features of the preceding stop spread forward into the /h/, strengthening it to a fricative ([et.han], [op.han], [mak.hōn] > [et.san], [op.fan], [mak.xōn]. This is because the oral closure duration of postvocalic stops tends to be shorter than when they are in word-initial position, and the coda stop is thus susceptible to weakening (> [eʒ.ʒan], [of.fan], [max.xōn]).

3.3. How the Core Shift “Progressed.”

The claim that the core OHG shift originated only in postvocalic position after short vowels finds support in an analysis of the dialect of Wermelskirchen known primarily from an early dissertation by Hasenclever (Reference Hasenclever1905) and a more recent dictionary of the dialect compiled by Joest (Reference Joest2003). An analysis of Joest's dictionary by Davis (Reference Davis2005) confirms Hasenclever's original description of a partial consonant shift in the dialect, in which pre-OHG t, p, k are shifted only after originally short vowels. Thus, modern Wermelskirchen dialect has shifted consonants after original short vowels as in äs n ‘to eat’, Nos ‘nut’, Jus ‘spout’, Päf r ‘pepper’, Jref ‘grip’, h fn ‘to hope’, Lo fl ‘spoon’, Bäch ‘stream’, Seche l ‘sickle’, Joch ‘yoke’, Bruchschteen ‘quarry-stone’ versus unshifted consonants in forms with old long vowels such as heet ‘hot’, Schtr t ‘street’, Foot ‘foot’, jruut ‘great, big’, riip ‘ripe’, Sch p ‘sheep’, Hoope n ‘pile’, roope n ‘to call’; Diik ‘pond’, week ‘soft’, Hals-dook ‘scarf’, and ruuke n ‘to smell, reek’. The alternation between shifted and unshifted consonants is starkest among the strong verbs where vowel length sometimes alternates according to apophony as in riite n, res, je resn ‘tear, tore, torn’ and äse n t, je jäsn ‘eat, ate, eaten’. The age of the Wermelskirchen shift is apparent from forms that contain originally stressed long vowels that underwent shortening in the post-Middle High German period. Consonants that follow recently shortened vowels do not undergo the shift but instead retain ⁺t, ⁺p, ⁺k (vit ‘white’, zupe n ‘to drink, booze’, ripe n ‘to ripen, mature’ kike n ‘to look’ < ⁺viit, ⁺riipe n, ⁺zuupe n, ⁺kiike n (Davis Reference Davis2005:266ff).Footnote ⁶

The present claim, therefore, is that the core shift originated in the above manner after short vowels in all dialects of Old High German and that Wermelskirchen merely retains much more than any other dialect the original state of affairs. Even in Wermelskirchen, the shift progressed somewhat beyond the core shift, however. I note that, in order to trigger the shift, the aspirated consonant [t^h, p^h, k^h] would have been in medial position so that the effects of the Head Law would be felt and thus trigger a reanalysis of the allophonic aspiration to a phonetically equivalent segment /h/ (as in op^han > op.han > op.fan > of.fan). In any case, word-final consonants would originally have remained unshifted—even in the paradigms of words with stressed short vowels.Footnote ⁷ Later, via paradigmatic analogy, the shifted consonants would have spread to word-final position after short vowels, as is preserved in Wermelskirchen forms such as res ‘torn’ (1st/3rd person preterite, riite n ‘to tear’) and Jus ‘spout, (German Guss)’ (< jiite n ‘to pour’). Thereafter, /ts ∼ t/, /pf ∼ p/, and /kx ∼ k/ would have continued to alternate in complementary distribution depending on the length of the preceding vowel. The tendency to level in favor of the affricates would have been particularly strong among the apophonically alternating strong verbs. Compare the affricate phase we reconstruct for pre-OHG ⁺rītan ‘to tear’: ⁺rit (1st sg. pret.) (> ⁺rits, after extension of shift to final position), ⁺ritsan (1st pl. pret.), ⁺garitsan ‘torn’ (pret. adjective).Footnote ⁸ As the leveling process gradually also replaced more and more voiceless stops after short vowels with the corresponding affricate, the process of replacement may have extended to operate after [+high] long vowels as well, since [ī] and [ū] are phonetically somewhat shorter than [-high] long vowels (Lehiste Reference Lehiste1977:18–27, Laver Reference Laver1994:435, Davis Reference Davis2008). The replacement rule could have later generalized to apply after all long vowels and then extended to postconsonantal position and to the geminate stops. The progression of the shift to word-initial position came as a chronological last step in the shift process of each of the affected stops. Thus, the shift of ⁺t, ⁺p, and ⁺k in initial position was contingent upon the earlier extension of the shift to the corresponding geminate, because the presence of a pre-existing affricate in word-medial position aided in the emergence of the same affricate in initial position (section 8).

Table 2.

Progression sequence of pre-OHG t, p, k (illustrated here only with the labial). Stress placement is indicated by a raised tick (‘).

4.1. Kraits-forms in Rhenish Dialects

As noted above, traditional reconstructions assume that postvocalic pre-OHG t, p, k shifted to affricates during the intermediate affricate phase. A number of researchers, particularly Bruch (Reference Bruch1953), have cited forms from modern Rhenish dialects that evince the word-final affricate −ts, the so-called kraits-forms, such as kre:ts/kre:s ‘circle’, mɛts(ər) ‘knife’, nɛtst ‘moisture’, ʃots ‘shot’, ʃwe:ts ‘sweat’ (see Goblirsch Reference Goblirsch2005:192), as relics of the intermediate affricate phase in the OHG shift.Footnote ⁹ Analogous forms with -ts in postvocalic position are not rare and occur along the Benrath line (Schützeichel Reference Schützeichel1976:249–256) and even along the shift line in Thuringia (jätze ‘alley’, NHG Gasse), as noted by Liesenberg (1890:60). I contend that most kraits-forms arose in isolated incidences when the postvocalic affricate ⁺-ts (< ⁺-t) resisted weakening to -ʒʒ-, or when -ts was perhaps confused with sequences of -t + -s that occurred across morpheme boundaries in pre-Old High German (compare the adverbial form OHG hīnahts ‘of this night, tonight’; hīnaht ‘this night’). Winteler (1876:46) records a number of kraits-like forms in the Swiss German dialect of Kerenz including grüətsə ‘to greet’, etsə ‘to feed’, wats ‘zeal’, guts ‘spout’, ruts ‘outbreak of anger’, and others.Footnote ¹⁰ Furthermore, Kögel (1887:111–112) argued for the reconstruction for early Old High German of a paradigmatic alternation of the word-final alveolar affricate ⁺-ts with its spirantized word-medial counterpart -ʒʒ-. Kögel claimed that this development left remnants in the form of doublets, such as OHG hiruz (with <z> spelling an affricate) and hiruʒ ‘stag’ (with <ʒ> spelling a fricative) that descended from the earlier paradigmatic alternation of hiruz with hiruʒʒes, comparable to the sort of mixed paradigms I recon-struct for an early stage of the shift in section 3. According to Kögel, forms with final labial or velar affricates were generally more prone to weakening than the coronal affricate.

4.2. Putative Evidence of an Intermediate Affricate Phase

Interestingly, an affricate stage in the progression of the shift was reconstructed for Langobardian by Bruckner (1895:146–147, 166) for both word-final postvocalic -p and -t. While acknowledging that post-vocalic -p in Langobardian generally shifted to -ff-, Bruckner maintained (pp. 146–147, note 2) that the variation in spelling of the extant examples of ⁺-p in Langobardian (anagrip, anagriph, and anagrif ‘attack’) may indicate that the word-final -p in these forms was an unshifted stop or an affricate and not a fricative (Bruckner Reference Bruckner1895:146–147):

Bruckner (1895:166, note 1) also made an analogous claim for unshifted -t in Langobardian:

Supporting evidence for the postvocalic extension of the shift to post-long vowel position and beyond comes from the West Franconian dialect via the OHG Pariser Gespräche (PaG) manuscript. Although it is traditionally claimed that the unusual postvocalic consonant graphemes found in the Pariser Gespräche result purely from the use of Romance graphemic traditions (Schützeichel Reference Schützeichel1976:122), Gusmani (Reference Gusmani1996) contends that the Pariser Gespräche may have originated in an area where the western dialect of the Franks had not yet totally died out and where ties still existed to cloisters in the OHG speech territory.

Gusmani thus argues that the graphemic representations in the Pariser Gespräche manuscript attest the shift of postvocalic ⁺t, ⁺p, ⁺k to affricates in West Franconian.Footnote ¹¹ Thus, PaG <ze> ‘to’, <zeni> ‘teeth’, <cit> ‘time’ (OHG zi, zeni; zīt,) attest the graphemes <z> and <c> in word-initial position for the affricate /ts/. <z> also occurs in postvocalic position in PaG <quaz> ‘what’ (64), with its affricate pronunciation being confirmed by the parallel occurrence of the spellings PaG <quats> (104) for the same form. Further examples of postvocalic shifted ⁺-t > -ts include forms with an original long vowel (the stage of the shift that I believe would have followed the stage attested in Langobardian), as in PaG <hutz> (40), <uthz> (46) ‘out’ (OHG ūʒ), PaG <atz> (97), <atst> (98), <ads> (103) ‘ate’ (OHG āʒ), and PaG <latz> (101) ‘let’, OHG lāʒ. The affricate -ts is even attested with the grapheme <z> in post-consonantal position in PaG <cherize> (58) ‘candle’ (OHG kerza/ cherza).

The occurrence of pre-OHG p is relatively rare in the Pariser Gespräche manuscript, and there appears to be not a single form attested that continues pre-OHG p- in initial position. Examples of postvocalic pre-OHG -p after long vowels occur, however, in PaG <sclaphen> (49) ‘to sleep’ and <ensclephen> (62, 63) (2nd pl. pret.) ‘slept’, (OHG slāfen). In the Pariser Gespräche, the pre-OHG clusters -rp and -lp appear to be in a state of flux. The labial stop appears to have already weakened to a fricative: PaG <durf> (102) ‘town’ (OHG dorpf, dorf ‘village’), PaG <elfe> (48) (3rd sg. pres. optative) ‘to help’, but the form <elpe> (13) (imperative) ‘to help’ without affricate is also attested. The shift of pre-OHG k > kx appears plausible in PaG <esprachen> (43) (pret. pl.) ‘to speak’ (OHG sprehhan), and in <roche> (47) ‘I care’ (OHG ruohhu ‘care, concern’). PaG <trenchen> (71) ‘to drink’ (OHG trinkan ‘to drink’) appears to attest the same velar affricate postconsonantally.Footnote ¹² Examples of pre-OHG geminates are rare in the PaG manuscript but, to the extent they occur, they are written in a fashion identical to the putative postvocalic affricates: PaG <luzzil> (pre-OHG luttil ‘little’) and PaG <trophen> (48, 74) ‘drop’ (OHG tropfo).

Although the spellings in the Pariser Gespräche must be approached with caution, it appears plausible that they attest the progression of the shift in West Franconian to the position after long vowels, to the position after sonorant consonants and perhaps even to the coronal and labial geminates, ⁺-tt- and ⁺-pp-. Table 3, slightly modified from Davis & Iverson Reference Davis and Iverson1995:119), provides an overview of the stages of the shift we reconstructed based on relic evidence.

Table 3.

Overview of the posited stages of the shift.

5.1. Rhenish Unshifted ⁺-t, ⁺-p, ⁺-k After Long Vowels

The progression of shifted consonants after long vowels is least con-sistent in the Rhineland. Lerchner (Reference Lerchner1971) lists scores of putatively unshifted Rhenish forms in all positions. The etymological background of many of the words listed is uncertain, however. Tables 4, 5, and 6, below, make use of only the etymologically most reliable forms from Lerchner's data, which were culled by selecting only those words that occur (according to Lerchner) throughout the entire Rhenish and West Middle German area and have a reliable Old High German or old Germanic etymology as confirmed by standard etymological sources.Footnote ¹³

Table 4.

Unshifted ⁺-t- in Lerchner's data.

Table 5.

Unshifted ⁺-p- in Lerchner's data.

Table 6.

Unshifted postvocalic ⁺-k- in Lerchner's data.

Lerchner records 73 relic words with unshifted postvocalic ⁺-t, of which only 9 are etymologically certain and occur throughout the Rhenish-West Middle German area.

Lerchner records 45 cases of relic words with postvocalic unshifted ⁺-p. Of these, only four cases are etymologically certain and simultaneously occur throughout the entire Rhenish-West Middle German area.

Lerchner records 124 putative relic words with postvocalic unshifted ⁺−k. Of these, however, only 13 are etymologically certain and occur throughout the entire Rhenish-West Middle German area.

In this etymologically reliable subset of Lerchner's data, the majority of cases of unshifted ⁺t, ⁺p, ⁺k (19 out of 26 cases, approximately 70%) occur after originally long vowels (Davis Reference Davis2008).Footnote ¹⁴ This result supports the prediction made by the progression theory that there would be more relic forms with unshifted ⁺t, ⁺p, ⁺k after long vowels, because the shift progressed to that environment only secondarily. The more tenacious retention of unshifted ⁺t, ⁺p, ⁺k after long vowels in the Rhineland vis à vis the apparently seamless extension of the shift to that position in other OHG varieties requires explanation. In Davis Reference Davis2008, I suggested that the pretonemic (that is, not yet phonemicized) Rhenish tone accents—as reconstructed by Schmidt (2002:220)—played a role in delaying the phonological extension of the shift after [-high +long] vowels.

In early work, Frings (1916:22) described the geographic extent of the tone accent traditionally known as Schärfung as following a line from Saarhölzbach southwest of Trier to Vianden in Luxemburg, then to Birkenfeld in the south of the Rhine Province, on to the Netherlands-Limburg area, and then east to Wermelskirchen and Mühlheim-Ruhr, thus encompassing Rhine, Middle, and Low Franconian territory. This area thus includes unshifted dialects above the Benrath line and in the Netherlands. The rest of the tone accent area overlaps in large measure with the Rhenish fan, including Wermelskirchen, where the consonant shift is known to have developed in a more limited fashion than in other HG dialects.

Map 1.

Map of the Rhenish fan (Schützeichel Reference Schützeichel1976:184). The superimposed dotted line circles the approximate original area of the Rhenish tone accents according to Schmidt (2002:208).

5.2. Mixed Blocking Effects

In pre-Old High German, the tone accents were pretonemic or non-distinctive accents. Sometime after the Middle High German period, however, two accents, both of which occurred mostly in bimoraic syllables, were phonemicized in the more northerly Rhenish dialects. In modern parlance, these are referred to as TA (tone accent) 1, traditionally called Schärfung, and TA 2, traditionally known as as Zweigipfeligkeit. The pretonemic tone accents that Schmidt reconstructs for the OHG period do not correspond in tone contour to the phonemicized TA 1 and TA 2 accents that now operate in most modern Rhenish dialects, how-ever. Crucially, Schmidt finds good reasons to reconstruct pretonemic TA 1 as having been associated with [−high +long] vowels during the pretonemic stage, and he believes this accent to have had a long, falling tone contour.

Following Schmidt's reconstructions, Davis (Reference Davis2008) contends that pretonemic TA 1 with its long, falling tone contour would have emphasized the length of the (already long) vowels in question. A long, falling tone accent on a long vowel that preceded ⁺-t, ⁺-p, ⁺- k might have delayed the extension of the shifted consonants in these forms and may thus account for the high number of unshifted relic forms in the Rhenish fan, particularly in the more northerly areas where the effects of the tone accents were apparently more robust.

Tentative confirmation comes via Venema's (1997:275–301) study of a set of frequently occurring Rhenish words with unshifted ⁺t, including the Rhenish forms corresponding to OHG grōʒ ‘large’, klōʒ ‘clod’, lāʒʒan ‘to allow, let’, and muoʒan ‘to have to, must’. These forms would have had an originally [-high +long] vowel with Schmidt's pretonemic TA 1 when they occurred in words that were under tone accent. However, as Frings noted early on, vowels that are capable of exhibiting the effects of tone accent do not do so at all times. Instead, only a word in a sentence that is emphasized (“hervorgehoben”) exhibits tone accent (Frings Reference Frings1916:13), as in Frings’ examples: dat øs diχ mār en štī:f ‘that's a stiff/formal one’ (person), with tone accent on the em-phasized word štī:f, versus a sentence such as doə štēt en šot l štīv rīs ‘there's a bowl of stiff rice pudding’ where the emphasis is on rīs, not štīv .Footnote ¹⁵

The effects of sentence-level or phrase-level intonation on the occur-rence of the tone accents in some modern Dutch dialects have been studied recently by Gussenhoven & van der Vliet (1999) and Fournier et al. (Reference Fournier, Verhoeven, Swerts and Gussenhoven2006). In the Roermond dialect, for example, Fournier et al. (p. 32) showed that the contrast between TA 1 and TA 2 is neutralized in minimal pairs of words if the words occur in a nonfocused or syntactically nonfinal context.Footnote ¹⁶ To the extent that this system reflects earlier stages of the tone accent dialects, it seems reasonable to expect the effects of the tone accents on the shift of ⁺t, ⁺p, ⁺k after [-high +long] vowels to be inconsistent, resulting in doublets and mixed paradigms that contain both shifted and unshifted consonants that would need to be sorted out by subsequent generations of language learners, leading to the mixture of forms now extant in the Rhenish Franconian dialects.Footnote ¹⁷ Thus, I contend, the less consistent consonant shift in the Rhenish fan was likely caused by the interplay of an autochthonous shift with the effects of tone accent, not the gradual weakening of a south to north geographic spread of the shift.

6.1. Place Bias and Post-Consonantal Shifts

As shown, evidence indicates that the shift progressed first to the position after long vowels, then to postconsonantal position (see table 2). This is supported by the status of the shift in the dialect of Wermelskirchen, which had neither a shift after long vowels nor one after consonants, and by the situation in the Middle Franconian dialects where shifted consonants occur after vowels (except in the relics noted above with unshifted ⁺t, ⁺p, ⁺k after long vowel), but where the shift after sonorant consonants is limited largely to ⁺-t. The relative chrono-logy of the shift after consonants is established by the fact that in South Rhine Franconian the shift of ⁺-p occurs both in gemination and post-consonantally (OHG skep(p)fen ‘to create’, dorpf ‘village’), but in Rhine Franconian where there is an incipient shift of the clusters ⁺-lp, ⁺−rp > −lpf, -rpf, there is no shift of geminate ⁺-pp- (see table 1 above). From the standpoint of the naturalness of the extension, it is obvious that the consonants -m, -n, -l, and -r are highly sonorous and thus rank close to vowels in sonority value, as in 3 (adapted from Vennemann Reference Vennemann1986:36).

1. (3) Sonority Hierarchy

   

In addition, a sequence of short vowel (V) + sonorant consonant (C) would count as bimoraic (VC) in the same way that a long vowel (VV) would. Thus, the progression of the shift from V + ⁺-ts, -pf, -kx (pre-OHG rits ‘torn’ 1st/3rd sg. pret.) to VV + ⁺-ts, ⁺-pf, ⁺-kx (pre-OHG slāpfan ‘to sleep’) created a template for shifted stops to occur in the sequence V + l, r, n, m + ⁺-pf, ⁺-ts, ⁺-kx (OHG hërza ‘heart’, dorpf ‘village’, wërcch/wërch ‘work’). Important for the analysis presented here is that in the dialects of the Rhineland ⁺-t shifts postconsonantally most extensively, while the postconsonantal shift of ⁺-p is graduated with shifted -lpf and -rpf occurring only incipiently in Rhine Franconian, with a full postconsonantal shift of -Cp > -Cpf (including -mpf) occurring only on the cusp of the primary tone accent area in the southern part of the Rhenish fan.

6.2. Place Bias and the Distribution of Affricates

Near the Benrath isogloss we note both the greater incidence of unshifted ⁺t, ⁺p, ⁺k after long vowels and especially the greater prevalence of unshifted ⁺-k in that environment. As Schützeichel (1976:268) noted, individual lexemes containing ⁺-k after long vowels sometimes have isoglosses that run far south of the Benrath line (such as reiken ‘to reach’ and s ken ‘to seek’).Footnote ¹⁸ Hasenclever (1905:44) records that ⁺-k is the only consonant in the dialect of Wermelskirchen that escapes shift on a sporadic basis after originally short vowels (māk n ‘to make, do’, dāk ‘roof’, ʃvāk ‘weak’), which begs the question whether -k is more resistant to shift in general, whether triggered by phonological rule or by analogical tendencies.

Evidence from many quarters suggests that place hierarchies operate in language. This is certainly the case for the position of articulation (PoA) hierarchies posited for many languages. Such hierarchies hold that dorsals (velars and uvulars) are more marked than are labials, which are, in turn, more marked than coronals (with glottals being still less marked than coronals, at least in some hierarchies that have been proposed) (De Lacy Reference De Lacy2006:34ff.). In a PoA hierarchy, a dominance relation would presumably operate that could be observed in synchronic alternations in the language. Thus, in Kelantan Malay, /k/, /p/, and /t/ all become [] in codas, establishing the greater markedness of non-glottals over glottals. As De Lacy (pp. 424, 42–43) notes, furthermore, /k/, /p/ appear as [t] in a Taiwanese secret language, showing the relation of dorsals and labials to coronals.

Assimilation in various languages also supports the dominance of velars over labials. Iverson & Kim (Reference Iverson and Kim1987) note this bias in the phonology of Korean 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.Footnote ¹⁹ With reference again specifically to assimilatory sound change, Rice (Reference Rice1994) likewise observed that increased place structure content corresponds to resistance to such change. She concludes that velars are the most specified as to content and are simultaneously the most resistant to assimilatory change. Labials are somewhat less resistant to such change, and coronals are the least marked and simultaneously the least resistant to change. Rice accordingly posits a model of feature geometry that assigns both labials and velars the feature peripheral subordinate to place, while coronals are unspecified for all place nodes. Labials are marked just for peripheral, and velars contain peripheral along with the subordinate articulator dorsal.

Building on Rice's generalizations about the connection between place content and the resistance to assimilatory change, Davis, Iverson, & Salmons (1999:191) argued that the segments p and k have a “heavier” feature geometry or place content that resists a non-assimilatory change like the shift more robustly than does a coronal stop such as t. Under this assumption, then, t would tend to shift more readily than either p or k, thus providing a theoretical framework for the traditional assumption that the shift of ⁺t in Old High German predates the shift of ⁺p and ⁺k (Sonderegger Reference Sonderegger2003:261–262).Footnote ²⁰

Viewed another way, the present paper claims that the greater resistance to shift exhibited by the heavier segments ⁺p and ⁺k cor-responds to the synchronic dominance relations observed in PoA hierarchies in many languages. Following Davis, Iverson, & Salmons (1999), then, the present paper contends that the geographic extent of the shift of ⁺t, ⁺p, ⁺k falls out in part from the place content of the segments involved. I also contend that analogous predictions of place bias hold for the same consonant segments when geminate so that place bias also favored the earlier shift of the coronal geminate, before the labial and the velar. I thus speculate that the shift of postconsonantal ⁺-t progressed to geminate ⁺-tt- even before the shift of postconsonantal ⁺-Cp and ⁺-Ck occurred and that geminate ⁺-tt- shifted earlier and more completely than the corresponding labial or the velar geminates. In fact, the coronal was so susceptible to shift that geminate ⁺-tt- in the Rhenish dialects appears to have begun to shift before the completion of the shift of ⁺-p and ⁺-k after long vowels (where tone accents delayed this progression, see section 5). This would explain how the shift of geminate ⁺-tt- (settian ‘to set’ > ⁺-tts-) could occur, while the shift of geminate ⁺-pp- (skeppian ‘to create’) > ⁺-ppf- and ⁺-kk- (wecckan ‘to wake’) > ⁺-kkx-) never developed in the Rhineland, thereby producing a particularly asym-metrical shift of the geminate affricates in the Rhenish dialects.

7.1. Cross-Linguistic Markedness

As noted in the previous section, heavier place content correlates with greater resistance to the shift, predicting—as a general proposition—that p and k will shift later and less robustly than t. However, place bias predicts nothing about the diachronic viability of the affricates ts, pf, and kx after they emerge. The greater susceptibility of ⁺-t- to shift compared to its labial and velar counterparts appears to correlate to the statistical frequency of the affricate ts across languages. In a study of a repre-sentative sample of the world's languages, Maddieson (1984:38) showed that /tʃ/ and /ts/ (palato-alveolar and dental/alveolar sibilants) are the affricates that occur most widely. In fact, /tʃ/ or /ts/ (or their aspirated counterparts /tʃ/ and /ts/) occur in an overwhelming 311 of the 317 languages in Maddieson's database of languages. Maddieson also notes (p. 39):

The rarity of the affricates /pf/ and /kx/ in his database is pronounced. Besides Standard German, Maddieson (Reference Maddieson1984) records only Beembe and Teke as having the affricate /pf/ or /pf/.Footnote ²² Tavgy, Chipewyan, Nama, Tlingit, and Zulu are recorded with the velar affricate /k/ or /k/, and Maddieson lists only three languages with postvelar affricates: Wolof /qχ/, Nez Percé /qχ/, and Cabardinian (see Luschützky Reference Luschützky1992:154.) Furthermore, as Luschützky (p. 161) shows in a further analysis of Maddieson's data, languages in his database that have either a labiodental or velar (including postvelar) affricate also have an affricate that falls somewhere within the dental to postalveolar range. This distri-bution lends support to Melikischwili's claim (1970:71), based on an analysis of a number of Caucasian languages, of an implicational hier-archy among affricates whereby languages with a “peripheral” affricate must also have a “medial” one but not necessarily the reverse (with “peripheral” corresponding to both labial and velar and “medial” corresponding to the alveolar and alveopalatal range). Thus, place bias favored the shift of ⁺t > ts (more than the shift of the labial and the velar stops) and the cross-linguistic frequency of the alveolar affricate, coupled with the implicational hierarchy suggested by Melikischwili, predicted its survival in all varieties of Old High German.

Relevant to the present discussion is the position in which affricates occur. Unfortunately, Maddieson (Reference Maddieson1984) and Luschützky (1992:161) fail to reference this information for the affricates pf and kx in the languages they analyze. Lavoie's (2001:23, 45–46) study of consonant strength, however, based on data presented in the scholarly literature on some 147 languages, records this information for each language in her database that exhibits velar or labial affricates. In most cases (in Burushaski, Island Carib, Turkana, and Maori), the velar and labial affricates in question occur only in initial position (word-initial or syllable-initial). Of these languages, only Maori has an entire series of affricates at the coronal, labial, and velar places of articulation, although these affricates appear to be allophones of aspirated /t, p, k/ (Bauer Reference Bauer1993:530). Most of the other systems have only non-velar affricates.Footnote ²³ Interestingly, Lavoie ponders whether affrication is best considered a form of consonant strengthening or weakening (pp. 45–46). She notes that if affrication represents the addition of phonological specification or structure, such addition of structure can indicate greater consonantal strength. However, she adds that in historical change affricates often “represent a step on the way from stop to fricative so in that sense it [affrication] could be [considered a form of consonant] weakening.” Since the examples of affrication she cites from her database occur primarily in initial position, she decides to take them to be instances of consonantal strengthening. The present paper interprets these findings to mean that aspirated stops can sometimes give rise to affricates, especially in initial position, and that affricates in the palato-alveolar and dental/alevolar range are the cross-linguistically most common type of affricates to arise in this fashion, with labiodental and velar affricates lagging far behind. An affricate in postvocalic position might be expected to weaken more readily than the same affricate in initial- or postconsonantal position given the shorter closure duration of postvocalic stops in general (Ohala Reference Ohala, Kingston and Beckman1990:438), but the pattern of weakening would be the same in either case.

As claimed above, cases in which aspirated (initial) stops are reanalyzed as affricates may result when affricated allophones of highly aspirated initial voiceless stops are perceived by speakers as the best exemplars of these segments. This is even more likely to occur, I would claim, if the affricate in question happens to already exist in word-internal position in the language because speakers could easily identify the affricated allophone of the initial consonant with the preexisting affricate (see section 8).

7.2. Markedness Reduction and the Trisegmental Affricates

Lavoie's observation that affricates often weaken to fricatives raises the question of why affrication of the pre-OHG postvocalic singletons ⁺-t, ⁺-p, ⁺-k (that is, the intermediate affricate phase) usually resulted in weakening (except in isolated relic forms), while the affricates that arose from the geminate consonants remained mostly intact into the modern dialects. Furthermore, we must address the question of why the OHG shift occurred more widely in medial rather than word-initial position (readily visible in South Rhine Franconian which has the shift of geminate ⁺-pp- (> ⁺-ppf- > -pf-) but no shift of word-initial ⁺p-). This point is of interest since most of the languages in Lavoie's database exhibited the affrication of voiceless stops in initial rather than in medial position.

The weakening of the postvocalic affricates of the intermediate affricate phase fits well into the framework of cross-linguistic marked-ness sketched in section 7.1, because the general paucity of labial and velar affricates among the languages catalogued by Maddieson suggests that such affricates—once they emerge—may be more susceptible to weakening than the cross-linguistically less marked ones. Within the context of the OHG shift of ⁺t, ⁺p, ⁺k, this predicts that once the affricates pf and kx emerged they would be more susceptible to markedness reduction than would ts, because the weakening (spirantization) of the affricates ⁺-pf- and ⁺-kx- > -ff- and -xx- amounts to a reduction of markedness in the phonological system in line with the prediction that the more marked affricates will be lost first and the less marked -ts- will be lost least (or not at all). As noted above, this outcome is also reflected in the wider survival in the dialects of the affricate -ts- from the intermediate affricate phase.Footnote ²⁴ But the development of word-internal long affricates from ⁺-tt-, ⁺-pp-, ⁺-kk- in the OHG shift differed from the earlier affrication of the singletons ⁺t, ⁺p, ⁺k cited above, because the geminates were long (disegmental) consonants (see table 3 above) that had developed via WGmc. consonant gemination (Denton Reference Denton1998). The long consonants in this environment gave rise to the corre-sponding long (trisegmental) affricates ⁺-tts-, ⁺-ppf-, ⁺-kkx-. In Lavoie's terms, the affrication of the geminate consonant added phonological structure (making it trisegmental), and it is this heavier structure that presumably allowed them to survive as affricates in postvocalic position by weakening only to disegmental affricates (-ts-, −pf-, -kx-) rather than to fricatives.

The assumption that forms in the Rhenish dialects such as kraits ‘circle’ are indeed frozen relics of the intermediate affricate phase of ⁺-t begs the question of why we fail to find analogous relic forms with ⁺-pf and ⁺-kx in modern Rhenish dialects that go back to the postvocalic singletons ⁺-p and ⁺-k. The answer is, I would contend, linked to the general place bias (section 6) evident in the shift's progression and to the greater susceptibility of the affricates pf and kx to reduction. Thus, relics of the affricate phase of ⁺-p (> ⁺-pf) and ⁺-k (> ⁺-kx) are lacking in Rhenish varieties because the geminate stops ⁺-pp- and ⁺-kk- failed to shift to long (trisegmental) affricates there. In Rhenish dialects, then, there were no long affricates (⁺-ppf-, ⁺-kkx) to shorten to disegmental ones (⁺-pp-, ⁺-kk- > ⁺-ppf-, ⁺-kkx- > -pf-, -kx-) with which any postvocalic labial and velar relics of the intermediate affricate phase could merge and thus survive.Footnote ²⁵ Deprived of the opportunity to gravitate into the orbit of the new affricates, any isolated relics of ⁺-pf- and ⁺-kx- from the intermediate affricate phase would have likely succumbed to the general tendency to weaken to -ff- and -ʒʒ- without leaving any frozen relic forms behind that contain these affricates in postvocalic position. Thus, isolated relic forms with the postvocalic affricate -ts- (that failed to weaken to the fricative -ʒʒ-) can exist in all OHG dialects because geminate ⁺-tt- underwent shift (⁺-tt- > ⁺-tts- > -ts-) in all dialects but the affricates ⁺-pf- or ⁺-kx- could only be preserved in those dialects where the velar and labial geminate affricates also shifted (as in a number of forms in South Alemannian varieties cited by Goblirsch (1997:273, 2005:185). Examples include ʃots ‘shot’, gryətsə ‘to greet’, flø:tsə ‘to float’, flo:ts ‘raft’, bœitsə ‘to bite’, byətsə ‘to atone’, ʃmŒtsə ‘to throw’, ʃwɛitsə ‘to weld’ gots ‘spout’, b ts ‘(small) bit’, r ts ‘rip’; sœipfe ‘soap’, ʃlœ pfə ‘to grind’, hɛpfe ‘yeast’, ʃla:pfə ‘to sleep’, ʃtɛipfə ‘line’, ʃtrapfœl ‘step (German Staffel)’, riepfu ‘to call’, ripf ‘call’, xaupfe ‘to buy’, reipfu ‘to mature, ripen’; blɛikxə ‘to bleach’, and wæ kxə ‘to yield (German weichen)’.Footnote ²⁶

To summarize, in the Rhineland, place bias and tone accents delayed the shift of ⁺-p and ⁺-k after long vowels and blocked the shift of geminate ⁺-pp- and ⁺-kk- entirely. Only the relics with -ts- survived into modern Rhenish dialects, while the remaining OHG dialects underwent a more general shift. The same pattern of weakening that reduced the postvocalic affricates of the intermediate affricate phase to fricatives (−ʒʒ-, -ff-, -xx-) is likewise observable at later stages in the dialects outside the Rhineland. Weakening occurs in the case of the initial affri-cate kx- > x- in southern Alemannian as in xalt ‘cold’, xįnd ‘child’, xraŋkx ‘sick’, etc. (Zhirmunskii Reference Zhirmunskii1962:297), as does weakening of initial pf- > f- in East Middle German varieties (fund ‘pound’, fennig ‘penny’, flaumə ‘plum’, fęf r ‘pepper’) and in sporadic weakening in Swabian (flõm ∼ pflõm ‘plum’) (see table 1 above and Kauffmann Reference Kauffmann1890:220–225 and Zhirmunskii Reference Zhirmunskii1962:273). By contrast, the weakening of the affricate ts- > s- in the German dialects is limited (Zhirmunskii, p. 293).