24st-raw [⌂ EvenYrbk]

Attila Starčević

A Note on Common Germanic Enhancement and Vaux’s Law*

1. Introduction In their discussion of Verner’s Law, Iverson & Salmons (2003) make a particular claim about the systems of laryngeal oppositions where Vaux’s Law can be expected to work. Vaux’s Law (based on Vaux 1998) was coined by Avery & Idsardi (2001: 49), who claim that it only works in languages where fricatives are distinguished by GT ‘glottal tension’ (such as Spanish and Japanese). Vaux (1998), however, never formulated any law. He says that in Sanskrit and other languages voiceless fricatives are produced with [+spread glottis], voiced ones with [-spread glottis]. Sanskrit, for example, has no contrastive fricatives, but the unmarked ones (/s0/, /ʂ0/, /ʃ0/) were demonstrably [+spread] (aspirated), cf. Vaux 1998. Iverson & Salmons (2003: 49), however, take Avery & Idsardi’s assumptions for granted that fricatives phonetically enhanced with aspiration can only be found in systems where GW ‘glottal width’ is not used contrastively, i.e., in system where GT ‘glottal tension’ marks the two distinctive series of fricatives. In systems where fricatives are distinguished with GW (in ‘aspirating’ languages like English), enhancing the unmarked fricatives is not available given that enhancement cannot neutralise an underlying opposition (bath [θGW] and bathe [ð]/[θ0] still have different fricatives). Iverson & Salmons (2003) do not discuss languages with a single series of fricatives (and whether they may exemplify enhancement), but their analysis shows that enhancement presupposes a distinction based on GT. If there is a single series of fricatives, it follows thus that there is no enhancement, we may conclude on the basis of their analysis. Once an opposition develops based on GT in the fricatives, enhancement applies across the board in all positions in which unmarked fricatives are found. Verner’s Law is critically positioned in their analysis to a period when there was no enhancement (i.e., a period where there were only lenis fricatives in all positions). Had there been enhancement, Verner’s Law (essentially passive voicing unimpeded by enhancement) would have been rendered inoperative. We will continue to refer to the introduction of GW as enhancement as Vaux’s Law despite the historical or conceptual inaccuracy.

This publication was (partly) sponsored by the National Research, Development and Innovation Office (OTKA #142498). I would also like to thank the two reviewers for their insightful comments. The usual disclaimers follow.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 33

We will argue that (i) systems with a single series, such as Old English (OE) and early Germanic (EG), may also have enhancement and (ii) enhancement may only be found in (what counts as) strong positions. Of course, the question is how Verner’s Law (and the passive voicing it relies on) may be salvaged if such notions are entertained. We claim that in EG the word-internal pre-tonic position was a weak one (in contrast to its modern daughters), allowing Verner’s Law to work as expected. 2. What can be enhanced? GW can be found in the fricatives either underlyingly specified or derivatively introduced during the process whereby phonological representations become phonetic, i.e., ever more concrete. GW as a dimension thus exists both at the phonological and the phonetic levels. At the level of phonetics, it may be conceived of as a process which introduces GW into the laryngeally unmarked fricatives in systems where GW is not used contrastively, completed as [spread]. In systems where it is, phonologically specified GW is also completed as [spread] phonetically. This encapsulates the workings of Vaux’s Law, seen in (1), where the dashed line shows GW introduced at the level of phonetics. (1) Vaux Law Voice languages

Aspiration languages

(Japanese, Spanish) /t/

/d/

/s/

GT

(English, German) /z/

/t/

GT

GW

/d/

/s/

/z/

GW

GW [slack][spread] [slack]

phonol. dim. phonet. dim.

[spread]

completion

Spanish and English fricatives are thus [spread]. Proof for [spread] fricatives is found in debuccalisation, seen in Andalucian Spanish, where /s/ is [h], as in libros ‘books’ [liβrɔh] cf. Maza 2000a,b, Carvalho 2006. In Urban Scots, for example, /θ/ can be realized as [h], as in think, something (or deleted altogether), cf. Stuart-Smith (2008: 61). The presence of GW may explain why /s0/ is never found as [z] in Spanish in voice-friendly environments, such as the intervocalic. [spread] blocks the introduction of GT (whose default interpretation is [slack]/voiced) from being introduced phonetically into [sGW] from the leftadjacent spontaneously voiced sonorant/vowel. Sonorants/vowels are also

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 34

unmarked for GT phonologically, but receive this dimension phonetically by default (a dimension which can be introduced into unmarked obstruents in voice-friendly environments). English and Spanish fricatives are both [spread] phonetically, which accounts for why there is no ‘intervocalic voicing’ in passing, see (2a), or pasar ‘pass’, in (2b), in either language (GT cannot spread into something already specified for a feature, here GW), see (2), where the dashed pointed red arrow shows this schematically. In Iverson & Salmons (2003) phonetic spread of GT happens from left to right (and is gradient, depending on how voice-friendly the environment is). (2) GT-spread (a) English passing with lexical GW /p

a

GW

ɪ

ŋ/

GT

[slack]

s

GW GT

[spread] [slack]

[spread]

(b) Spanish pasar with derivative GW /p

a

s

a

r/

GT

GW

GT

[slack]

[spread]

[slack]

Vaux’s Law was originally formulated for fricatives (Avery & Idsardi 2001). However, as Iverson & Salmons (2003) argue, the unmarked stops can also be enhanced with GW phonetically. Such an enhancement is only possible in a

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 35

system where GW is not used phonologically contrastively given that enhancement cannot destroy underlying oppositions (a language like English is predicted not to have enhancement of the unmarked/lenis stops). EG was a different matter: its system of obstruents developed out of the three series of Indo-European (IE) stops and a single fricative (*/s/), shown in (3). (3) Traditional set of IE obstruents (a) voiceless unaspirated: *p, t, k, kw, s (b) voiced unaspirated: *b, d, , w © voiced aspirated/murmured: *bh, dh, h, hw The typological oddity of this system is well known (cf. Salmons 1993 for a discussion), leading to the ‘glottalic’ reconstruction of IE. The technical details are irrelevant here, but the analysis developed by Iverson & Salmons (2003) works under the glottalic assumptions, too. EG thus inherits a single fricative form IE (*/s0/). Reconstructions of EG show that a more liberal system of fricatives developed out of IE stops: */f, θ, x, xw/, in addition to */s/, which remained unchanged. These are analysed by Iverson & Salmons (2003: 55) as having developed from IE */p t k kw/ through a process of Germanic enhancement in (4). (4) Germanic enhancement Laryngeally unspecified stops →GW The enhancement process in (4) can only apply to laryngeally unmarked stops, i.e. IE */p t k kw/, making these (phonetically) aspirated: *[pGW tGW kGW kwGW]. Although nothing is said explicitly about why enhancement is possible here, we must assume it is because there is (at least) two series of stops at this stage of EG marked for GT and GT+GW: unmarked vs voiced vs voiced aspirated. These phonetically aspirated stops, however, developed into fricatives over time through spirantisation, along the lines shown in (5). (5) Development of Common Germanic fricatives

ph th kh kwh > *(ɸ >) f θ x xw

Note that Iverson & Salmons (2003: 56, table 1) drop ‘h’ (GW) in (5) in the process of spirantisation with no discussion. We are left wondering what kind of fricatives we have now: were they phonetically voiceless at all? How did an unmarked fricative (?*/f0/) sound at this stage? This will be important for the argument to be developed.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 36

The development sketched out in (5) is not unique to Germanic. In the Dravidian languages, for example, the proto-Dravidian voiceless labial plosive

/p/ is found as /h/ (or ∅) in Kannada. The intermediate stages are widely

discussed, but seem to have involved a now lost fricative, which was later debuccalised, suggesting an initial *pGW for this lenition trajectory: *pGW > fGW > h > ∅ (Zvelebil 1970, Lass 1984, Foulkes 1997, Honeybone 2002). A similar process may have been at work in pre-archaic Japanese */p/ (or better still */pGW/), now found as /h/, /p/ and /b/ (Labrune 2012, 3.7). Similar changes are known from Proto-Celtic in which */p/ is lost both word-initially and medially, probably through *pGW > h > ∅ (e.g. Gaelic éan ‘bird’ vs English feather, Latin penna, Greek pterón; caora ‘sheep’ vs Latin caper ‘he-goat’, cf. MacAuley 1992: 3). Proto-Uralic */k/ before back vowels is found as /h/ (< ?*kGW) is Hungarian, */p/ as /f/ (cf. Sámmol Ánte 2022: 6) At this late stage of IE, */b d  w/ are still probably phonologically voiced (marked for GT), opposed to late IE */bGW dGW GW wGW/, marked for both GT and GW. As the change in (5) unfolds, the late IE/EG series of stops is unbalanced having voiced stops opposed to voiced aspirated ones. This may have provided the impetus for the late IE series of voiced stops to devoice (lose GT) to early Germanic */p t k kw/, which in this drag chain of events meant that the late IE voiced aspirated stops had to lose GW, leaving this series marked for GT only. This is, more or less, what standard textbooks have as Grimm’s Law (cf. Collinge 1985: 63–76). This gives us the EG series of obstruents in (6). (6) Early Common Germanic obstruents (as per Iverson & Salmons 2003) Fricatives: *f θ s x xw (= *f0 θ0 s0 x0 xw0) Voiceless (unmarked/lenis) stops: *p t k kw (= *p0 t0 k0 kw0) Voiced stops: *b d  w (= *pGT tGT kGT kwGT) At this moment, EG is still a voicing language, having a single series of phonologically neutral fricatives (shown here with ‘0’) and a twofold series of stops (lenis/unmarked vs GT). 3. Verner’s Law Verner’s Law claims that EG fricatives are found as voiceless immediately after EG stressed vowels (as well as word-initially), but voiced elsewhere (something reminiscent of a few modern English examples: lúxury [kʃ] vs luxúrious [ʒ], ánxious vs anxiety, a process, if at all, which has nothing to do with Verner’s Law). EG also inherited its vocalic and accentuation system from late IE. Iverson & Salmons (2003) argue that at this early stage, EG had vowels lexically marked for high tone running up to a system based on stress (prominence) which

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 37

developed in later EG (as witnessed by the modern daughter languages). Lexically marked GT (its [stiff] gesture) on vowels marked high tone in accented vowels (and accented syllabic consonants). By contrast, GT (its [slack] completion) marked phonetic spontaneous voicing in the unaccented vowels (and unaccented syllabic consonants), sonorant consonants and passive/perseverative voicing in obstruents. This is an important point for Verner’s Law (or what remains of it, as we will see): the [stiff] gesture of a high toned (accented) vowel carries over into the following laryngeally unmarked fricative as phonetic voicelessness. ‘Unmarked’ is of key importance here: the stiffness of the vocal cords in high toned vowels inhibits vibration of the following fricative under the low transglottal airflow conditions of obstruent articulations. The feature [stiff] of GT in the accented vowels spills over into the fricative barring it from becoming passively voiced. When a fricative at this stage is preceded by an unaccented vowel (i.e., a vowel not marked for [stiff]), the fricative is passively voiced, see (7) for the two commonly cited examples:

brṓθar ‘brother’ (Old English brōðor) and *faθár ‘father’ (fæder).

(7) Accented vowels and fricatives b

r

GT

ṓ

θ

r

GT

GT [slack] [slack] [ s t i f f ]

f

a

θ

á

[slack] [slack]

r

GT

GT [slack]

GT [stiff] [slack]

In *brṓθar, */b/ and */oː/ are phonologically marked for GT, as discussed earlier. The sonorant consonants receive GT derivatively, i.e. phonetically, and

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 38

are interpreted as [slack], shown above with dashed lines. The fricative */θ/, however, is voiceless because the phonetic completion of GT of the accented vowel ([stiff]) is carried along into the fricative and interpreted as voicelessness (= [stiff] in consonants). Note that at this early stage, EG is still a GT language (as seen for */b/). The importance of this is crucial in answering the question of why accented (high toned) vowels had no impact on right-adjacent obstruental stops. These were phonologically marked for GT, a feature which could not be ‘overshadowed’ by [stiff] from the preceding accented vowel. The voiced stops were allophonically spirantised in intervocalic position, something Iverson & Salmons (2003) call ‘Spanish-like’ passive spirantisation. Phonologically, however, they were stops marked for GT, see (8). (8) IE voiced aspirated stops > Germanic voiced stops á

d

GT

a

GT [stiff] [slack] [slack]

The IE voiced aspirates are found as voiced stops in EG. The feature GT in this series of stops makes it impossible for [stiff] to be carried into */d/, which is phonologically marked for voice. It is only the (phonologically and phonetically) unmarked fricatives, as seen in (7), which were voiceless after an accented vowel in EG. The devoicing effect of [stiff] on the following obstruent was not a neutralising one. If it had been, the opposition between *áta and *áda would have merged into *áta. In *áθa the spread of [stiff] did not jeopardise any opposition in the language. Note here that the reason Iverson & Salmons (2003) do not claim *ada was in fact áða (< IE *adGWa) is because they believe, as we have seen, Vaux’s Law can only work in a system with an existing opposition in the fricatives not based on GW. Once a system has an opposition of the áða-áθa-kind, the argument goes, enhancement will (or is expected to) appear. In such a system *faθár would have had a voiceless fricative, violating Verner’s Law. The assumption that a pre-existing contrast is required for enhancement to affect the fricatives is needed because otherwise Verner’s Law would not have worked. Consider now the well-known examples in (9) with the (false) assumption that all fricatives should be enhanced in all the positions in a language with a single series of fricatives such as EG.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 39

(9) An alternative (false) version of EG b r ṓ θ a r GT

GT [stiff] GT

?GW

[slack][slack]

f

a

GT

[spread] [slack] [slack]

θ

á

r

GT [stiff] GW

GT

GW

[spread] [slack] [spread]

GT [slack]

In *brṓθar and *faθár now every fricative is (phonetically) enhanced with GW. In *brṓθar the spread of [stiff] to the fricative (making it voiceless, shown with the red dashed line) from the accented vowel would have been impossible to distinguish from independent enhancement of fricative. Crucially, in *faθár we can see two instances of enhancement, one on the word-initial fricative */f/, the other on */θ/. */θ/ is now phonetically aspirated, bearing [spread] of the derivatively accrued GW, and cannot be passively voiced (shown by the crossed red dashed line highlighting the fact that GT is now blocked from being interpreted as [slack] in the fricative). Had this been the case, Verner’s Law could never have existed in EG: after accent was reinterpreted as stress and moved to the initial syllables of morphemes, late EG *brṓθar and *fáθar would have shared the same history in the daughter languages and expected to surface as brōðer and fæðer in Old English. It must have been this alternative scenario that made Iverson & Salmons (2003) claim that there was no enhancement of the fricatives in any of the positions in a system with a single series of fricatives, such as EG. To this is added the misconception that enhancement can only be found in languages that have a laryngeal contrast in the fricatives not based on GW.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 40

The phonological reality of EG was different though, as we have seen. As Iverson & Salmons (2003) argue, Verner’s Law can be understood as mundane passive voicing of the unmarked fricatives opposed to their voiceless peers in identical environments after accented vowels were reinterpreted as stressed, followed by fixation of stress on initial syllables. This brought with itself the phonologisation of the once allophonic distribution of the fricatives. In words like *brṓθar the fricative must have acquired GW from the reinterpretation of [stiff] stemming from the once high-toned vowel to its left even before accent was reinterpreted as stress. Otherwise, with the demise of [stiff] (and the tonal system) the interpretation of the post-vocalic fricatives as voiceless would have been wiped away leaving the unmarked fricative susceptible to passive voicing. This did not happen, suggesting that */θ/ in *brṓθar had already become phonologically marked for GW (*/θGW/) even before lexically marked high tone disappeared from EG, see (10). (10) [stiff] phonologised as GW in the fricatives b

r

ṓ

θ

a

r

GT

[sl] [spr]

[sl]

GT

GW

GT [sl] [sl]

In (10) we can see EG after the shift of stress where (i) stressed vowels have relative prominence (but no GT), (ii) fricatives show GW as a way of ‘legacy marking’ originally superimposed on them by the preceding accented vowel, now completed as [spread] and opposed to passively voiced fricatives in identical environments: *brṓθar vs *fáðar, (iii) the stops may still show GT as a laryngeal specification. At this stage, we have a fricative subsystem specified for GW vs unmarked and a system for the stops still showing GT vs unmarked. Although such a system is not inconceivable across the two subsystems of obstruents, the further trajectory of oppositions took a different path. Alongside fricatives phonologically marked for GW, there must have been a stage when stops were marked for GT (as a residue of late IE). As Iverson & Salmons (2003) argue, the unmarked stops (from IE voiced stops) were also undergoing a change: they were enhanced phonetically with GW (the result of Germanic enhancement), producing *[pGW], *[tGW], etc. At this stage there are thus oppositions like *[b] (fully voiced/phonologically

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 41

marked for GT) vs *[pGW] (aspirated through enhancement/phonologically unmarked) for the stops, *[ð] (passively voiced/phonologically unmarked) vs

[θGW] (phonetically voiceless/phonologically marked for GW) for the

fricatives. The system was ‘overmarked’ and eventually streamlined with aspiration in the stops reinterpreted as GW, bringing these in line with the fricatives. The two classes of obstruents are now uniformly distinguished as GW vs unmarked with [spread] as phonetic completion for both subclasses (as the lenis fricatives were the allophones of the lenis stops, the two could not have conflicting specifications anyway), see (11). (11) EG obstruents1 stops: */pGW/ /tGW/ /kGW/ /kwGW/ vs */b0/ /d0/ /0/ /w0/ (e.g. West Germanic *pōl- ‘pool’ vs *bion- ‘bee’) fricatives: */fGW/ /θGW/ /sGW/ /xGW/ vs */f0/ /θ0/ /s0/ /x0/ (e.g. *ofn- ‘oven’ with */fGW/ vs *laivjan- ‘leave’ with */f0/) As far as the stops are concerned, they could be found in all of the positions (word-initial/final, intervocalic).2 The fricatives, however, were more restricted (owing to their history): the lenis ones were not found initially. We have now arrived at a point where the following question becomes pertinent: what exactly was the pronunciation of word-initial fricatives in EG, as in *faθár and how does it bear on enhancement (and Verner’s Law), something we turn to in the next sections. 4. A short detour into Old/Middle English and back to Common Germanic The fricatives in the VFricV environment had two pronunciations in EG, as we have seen: (i) voiceless after accented high-toned vowels and (ii) (passively) voiced before such vowels. We must turn to word-initial fricatives now, to what 1

2

We have indicated the unmarked stops here with what would usually be used for voiced stops, followed by the superscript ‘0’ indicating the unmarkedness of the segment (e.g. /b0/), a practice not used for the fricatives. Here unmarked fricatives are shown using the voiceless symbols (e.g. /f0/). The reason for this lies in the fact that English spelling practices are (more or less) consistent in their orthographic representation of underlying differences (e.g. tin vs din), whereas those for the fricatives are not (e.g. thin [θh] /θGW/ vs thine [ð] /θ0/, Asia with both /ʃGW/ and /ʒ0/, etc.). This, however, has no bearing on the analysis. Post-nasal fricatives before stress (e.g., *faŋxV́ - > *faŋɣV́ > *faŋV́ ) appear as stops after Verner’s Law, and here they merge with the reflexes of IE voiced aspirates. The intermediate step in the example between *x and * cannot be proved, but the reflex of *x as * shows its predecessor must have undergone passive voicing.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 42

the pronunciation of the initial fricative of *faθár was: [f] or [v/β]? The answer based on (traditional) comparative linguistics is simple: word-initially there cannot have been voicing. If the fricative had been [v/β], it would have been reinterpreted as EG */b0/, merging with */b0/ from IE */bGW/, giving OE bæder, a word of no extraordinary phonology (cf. bīdan ‘wait’). The problem is complex and revolves around the phonetic interpretation of phonologically unmarked segments in a single series of obstruents, here the fricatives. The pronunciation of such unmarked fricatives may be either voiceless or voiced. No phonological opposition would be jeopardised. That a word-initial unmarked fricative may be rendered phonetically voiced is supplied by Southern English Fricative Voicing (SEFV, Honeybone 2005), which, strictly speaking, is a misnomer, discussed below, see (12). (12) SEFV (a) Modern English

father

freedom

sin

sword

(b) Middle English

uader

vridom

zenne

zuord

The process in (12) has also been called Middle English Fricative Voicing (Lass 1991–93) and Voicing of Initial Fricatives in Middle English (ME) (Fisiak 1984). Although the exact boundaries of the dialects where this process was innovated into are difficult to pinpoint, Fisiak shows that the whole of the South of England and some parts of the West Midlands were affected by it. ‘Voicing’ suggests an active process whereby a segment acquires a feature it did not have previously, such as GT which accrues to a segment phonetically in voicefriendly environments, such as the intervocalic (as in OE hæsel ‘hazel’), as analysed by Lass (1994: 72) in a traditional framework. This is not SEVF, however, unless one is adamant that the word-initial fricatives were voiced in an ‘across-the-word’ voice-friendly environment, i.e. post-lexically (in phrases like what a uader < what a fader), and then lexicalized as such. Word-initial voiced fricatives in words like thumb, shilling, etc. are still heard in southern non-reference varieties of English, as recorded for the Survey of English Dialects (e.g. Orton 1962). Many of the details need to be assessed about SEFV, such as: was this voicing found in Middle English words like smale ‘small’ and standen ‘stand’, and can voiced fricatives (still) be heard in modern English in this environment, etc.? Although the environments in small and stand may be put aside as presenting more complex examples (note that SEFV did work in vridom ‘freedom’ though), the examples usually adduced for this voicing beg the question of how they are to be accommodated typologically. If the fricatives in (12b) are voiced, this variety of English would have voiced fricatives only

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 43

(but not voiceless ones), which goes against typological considerations: if there are voiced fricatives (let us say, word-initially), by implication there should also be voiceless ones (word-initially) in the same environment, something not borne out by the data in (12). Lass (1991–93: 29) claims that a system with a single non-derived voiced series of fricatives should not be legal. In laryngeal phonology, however, the question can be approached from a different perspective. The question is whether the ‘change’ in (12) involved a change in the laryngeal specification of the affected consonants at all. In a system with no contrastive fricatives (no [f] vs [v] in any position), such as the one in (12b), the phonetic interpretation of an unmarked fricative can be either voiceless or voiced (or anything variably voiced on a phonetic scale ranging from completely voiced to totally voiceless) given that there is no contrast to be compromised (in a SEFV-system sin and zin will never contrast). A phonetically voiced fricative seen in SEFV shows the pronunciation of a phonologically unmarked fricative ([v], [z], etc.) with no instructions to the glottis to make it voiceless, that is, [spread] (as for [fGW], [sGW], etc.). A fricative marked for GW will be voiceless as it has lexical instructions (in the form of an underlying GW) to keep it such.3 In other words, zin with [z] shows the phonetic interpretation of a fricative with no specified GW to ensure its voicelessness. It is not phonologically voiced, however, and as such it does not jeopardise some cherished notions of typological implications. The reference accents of modern English do have a word-initial opposition between the two series of fricatives (fine vs vine, seal vs zeal), but this is a different system: phonetically voiceless fricatives are marked for GW (fine, seal), the (variably) voiced ones are not (they are unmarked). In a system like this, the fricatives of vine and zeal cannot be phonetically enhanced with GW as Vaux’s Law cannot destroy underlying oppositions. We have now arrived at a point where the phonetic interpretation of EG unenhanced fricatives may be looked at from another perspective. The wordinitial fricative of *faθár/fáðar is not likely to have been (produced or perceived as) phonetically voiced, as expected from a laryngeally unmarked fricative (vaθár/váðar). Had this been the case, it would have led to the reanalysis of such fricatives as word-initial stop allophones of IE voiced aspirated stops, here

/bGW/, producing another source for word-initial (voiced) stops. This scenario

would also have extinguished all the word-initial fricatives save */z/ (< IE */s/), see (13).

3

Cases of lenition are special, of course, and may show the deletion of such lexically specified instructions.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 44

(13) Putative merger of IE voiced aspirated stops with unmarked fricatives word-initially IE #bGW+GW, #p0 > EG #b0, #pGW > #b0, fGW > (loss/absence of [spread])

b0, v > EG #b0

IE #dGW+GW, #t0 > EG #d0, #tGW > #d0, θGW > (loss/absence of [spread])

d0, ð > EG #d0

IE #s0 (assuming it was unenhanced) > EG #z IE #GW+GW(w), #k0(w) > EG #0(w), #kGW > #0(w), xGW > (loss/absence of [spread]) #0(w), ɣ > EG #0(w) The examples in (13) show what would have happened in this alternative EG: word-initial unenhanced fricatives would have been phonetically interpreted as voiced and merged with the reflexes of IE voiced aspirated stops (with the exception of IE */s/, which has no ‘pair’ to merge with, found as EG */z/).4 A scenario in (13) would have led to a totally different late EG, one in which voiceless fricatives were to be found word-medially in what used to be the position after the high-toned accented vowels (e.g. *brṓθar ‘brother’ vs bádar ‘father’). The fate of these fricatives in this alternative scenario is open to speculation: they may have lost their GW (a feature they accrued from their positional legacy, explained above) and merged with */d/ (with stop and continuant allophones). We would like to suggest, however, that it was enhancement with phonetic GW in EG that prevented word-initial fricatives from being reinterpreted as reflexes of IE voiced aspirated stops. The possibility for our analysis is found within Iverson & Salmons’ (2003) approach itself, discarded because of the misconception that GW as enhancement can only be found in systems where GW is not used contrastively, but not in languages with a single series of fricatives (EG). Languages with a single series of unmarked fricatives also do not use GW contrastively. The reason for not considering enhancement as a possibility in EG may be rooted in the belief that enhancement, if it is found in a language, must be applied to all (unmarked) segments in all phonological positions (weak 4

That fricatives voiced by Verner’s Law merged with the reflexes of IE voiced stops is shown by word-internal developments, e.g. *θurfán- > *θurván- > *θúrvan- ‘need’ (OE þurfan) with [v/β] merging with that of IE *nebh- > *neb/neβ- (e.g. OE nifol ‘dark’, German Nebel ‘fog’). A similar putative merger may have affected the word-initial unenhanced fricatives in our account.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 45

and strong). Iverson & Salmons (2003) cannot apply enhancement to unmarked fricatives because that would make their analysis of Verner’s Law impossible: in *faθár the unmarked fricative /θ0/ would accrue GW derivatively, making it impossible for GT (from the preceding non-accented vowel) to propagate into such segments making them passively voiced. Passive voicing is only possible if a segment has no (underlying or derivative) GW. This also means that Iverson & Salmons’ Germanic enhancement of the IE unmarked stops popped up at a time, led to spirantisation, was lost in the fricatives in EG, only to reappear on the fricatives after the spirantisation of post-vocalic voiced stops (*/bGT/ > [β], etc.) and the appearance of a voiced-voiceless opposition ([f] vs [β]), later still to be reinterpreted in late EG as a phonological dimension after the loss of phonological GT, continued into the descendants of EG (such as English), see (14). (14) Germanic enhancement and the phonologisation of GW (1) IE *p t k(w) > (2) EG (via Germanic enhancement) *pGW tGW kGW(w) > (3) EG (spirantisation) *fGW θGW xGW(w) > (4) EG *f0 θ0 x0(w) (+ s0 < IE *s) (loss of GW/enhancement after spirantisation) (5) EG voiced stops spirantised post-vocalically (e.g. */bGT/ > [β]) (6) enhancement reappears as there is a GT vs ∅ opposition in the fricatives now: [βGT] vs [pGW] (via enhancement), etc. (7) after loss of GT, GW becomes a phonological dimension in late EG:

fGW θGW sGW xGW (w)

Iverson & Salmons do not explicitly discuss whether GW is found in step (3), but it must have been present in step (2) for fricativisation to produce what it did (fricatives). Step (4), however, lacks GW to make sure passive voicing can accrue to them in voice-friendly environments (and make it possible for Verner’s Law to work). If the fricatives had no GW in (4), their interpretation (word-initially) is also expected to have been voiced (as segments lacking phonological instruction in the form of GW to ensure their voiceless). Assuming that they were voiced word-initially, however, gives us a totally new kind of EG, a system which is at odds with the data. In step (6) enhancement reappears

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 46

as now there is a (transitional) opposition in the fricatives in the form of GT vs ∅. In step (7) GW reappears as a phonological feature (following the loss of high tone, the retraction of stress to initial syllables, loss of GT as a phonological dimension and GW percolating into phonology). The question, put simply, is whether enhancement in step (4) in (14) was ever lost.5 The answer is that it was not, and the effects of Verner’s Law can be explained even with enhancement. 5. A short detour into Old/Middle English and back to Common Germanic (again) SEFV, as discussed above, shows the orthographic imprint of a laryngeally unmarked fricative pronounced as voiced. Why this change makes its appearance in spelling in ME has to do with Old French orthography that brought the letters z and v for [z] and [v], respectively, into Middle English spelling. In OE z was used for [ts], but never for [z], and v was almost nonexistent. The letter u was not used for the consonant [v] in OE either, only for the vowel [u(ː)]. This allowed for some of the voiced fricatives to be orthographically visible for the first time in history. The absence of a spelling imprint of [ð] may be explained by the absence of dh in Old French orthography. Although French orthography may be credited for the orthographic visibility of Middle English (perceived) voiced fricatives, there is no reason to discard the possibility that [v] and [z] (and also [ð]) had existed word-initially even before the arrival of French orthography. The OE f or s says very little about the pronunciation these letters were used for. After all, OE had a single series of fricatives, so distinguishing the allophones had never been attempted or indeed necessary. Honeybone (2005: 344), who reinterprets SEFV in terms of laryngeal realism for the first time, assumes that this process shows a change in the laryngeal specification, of ‘h’ (a feature similar to GW) being lost from the specification of OE fricatives, see (15). (15) Loss of Old English ‘h’ (producing SEFV) fh > f0 [v], θh > θ0 [ð], sh > s0 [z], ʃh > ʃ0 [ʒ]

5

Let us quote Iverson & Salmons (2003: 55) in full: “Vaux’s Law […] itself could not be brought into play at this late stage of Indo-European (or early stage of Germanic), due to the absence of laryngeal contrasts among fricatives.” Also observe the following: “postvocalic spirantisation of /b d / creates the contrast that is necessary in order to enable Vaux’s Law to come into play, which in turn enhances the class of voiceless fricatives with the dimension of Glottal Width at this still early stage of Germanic.” (ibid: 59)

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 47

As Honeybone (2005: 344) explains for (15) to work we have to assume that (i) unspecified segments exist, (ii) the input to the process is characterized as

spread

which this change was innovated. The |spread| feature in Honeybone’s account is ultimately of Germanic origin (and immediately of OE origin, we must add). Of course, unspecified fricatives (in at least some of the varieties of OE) may always have had a voiced pronunciation word-initially (given the fact that there was only a single series of fricatives in OE). Before we return to EG, let us ask a further question in connection with (15). If we suppose for the time being that those ME varieties into which delaryngealisation shown in (15) was innovated must originate in OE varieties in which there were aspirated fricatives, the question of the origin of this aspiration must be answered given that OE only had a single series of (unspecified) fricatives. The answer is that aspiration must come from enhancement in a system in which fricatives are unspecified for laryngeal dimensions. In a system like this, GW must accrue to such fricatives derivatively. If OE was a language with enhancement, EG may also have been one given the existence of only as single series of fricatives. What phonological evidence do we have for this? Before we look at some of the phonological processes of OE, the distinction between strong and weak positions needs to be made. Not all consonants are equally susceptible to melodic decomposition (lenition). Their position (as well as melody) defines how they behave with respect to decomposition, something that can be observed if one looks at how languages change with time. The phonological strength of a position can be measured in terms of how much decomposition is allowed with respect to the other positions. There is an implication: if a certain process is observed in a strong position (e.g., wordinitially /b/ > [w]) it must also occur in a weak position (e.g., word-finally), but not the other way around (cf. Scheer 2004, for example). In the present-day Germanic languages in general, consonants sitting word-initially and wordinternally followed by a stressed vowel are strong, i.e., less likely to undergo lenition (e.g., tap, potáto, where the underlined consonants are aspirated in MoE). The post-consonantal position (either followed by a vowel or not, as in party and part) is subject to language-specific constraints on both the consonant preceding the consonant (/tGW/ in this example) and the vowel after it (whether there is one phonetically or not). Intervocalic consonants followed by an unstressed vowel (as in city) and word-final consonants (pat) are also in a weak position, as demonstrated by their propensity to undergo lenition to a glottal stop in some varieties of the modern reference varieties of English: [siʔij], [paʔ] (as opposed to tap, in which the ‘t’ is aspirated, and cannot be replaced by a glottal stop: [ʔap]).

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 48

The OE obstruental stops generally show no neutralisation in the weak position (intervocalic, word-final, etc.), see (15). This is different to what we encounter in Danish: in the weak positions, the stops are all voiceless lenes for all places of articulation, cf. Haugen (1982: 81), Page (1997): tyk ‘thick’, lække ‘leak’, and tyg ‘chew’, lægge ‘lay’ all have [ß]/[k]). We can explain this neutralisation as the delinking (deletion) of the GW dimension in the weak positions, see (16) for some examples for the various positions in OE. (16) Non-sonorant stops in the weak positions in OE opposi- (post-consonantal) tion/po- word-final sition

intervocalic

post-consonantal pre-vocalic

sib(b) – scip ‘relation, ship’, camp – comb ‘field, comb’

(no intervocalic singleton [b])

climban – hrimpan ‘climb, twist’

brȳd – fēt ‘bride, feet’, bord – wyrt ‘board, wort’

blēdan – blǣtan ‘bleed, bleat’

swindan – sprintan ‘vanish, emit’

(d)dʒ vs brycg – dīc ‘bridge, tʃ ditch’, seng – drync ‘singe Imp, drink’

secges – īcan ‘man GSg, add’

sengan – þyncan ‘burn, think’

dāg – lāc ‘dough, play’, lang – wlanc ‘long, splendid’

magan – lācan ‘have the power to, jump’

stingan – stincan ‘pierce, smell’

b vs p

d vs t

/ɣ/x vs k

crypel ‘cripple’

As we can see in (16), oppositions in the weak positions show no neutralisation (lācan, for example, is not found as lāgan, brȳd is not generally brȳt, brycg is not bryc). Some oppositions, for historical reasons, can only be found word-initially (e.g., /kGW/ vs /0/, cund ‘origin’ vs gund ‘matter’). The distribution of OE singleton fricatives seems open-and-shut, regulated by the (phonetically) voiced environment (Hogg 2001, §7.54). Put simply: in

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 49

voiced environments (e.g. between sonorants/vowels, or next to passively voiced lenis stops provided these are couched in a voiced environment, too, as in oven, scythe, hazel, law (< lagu), wisdom, but not if the first vowel belongs to an unstressed prefix, as in beseech), voiced fricatives are predicted in OE, while in voiceless environments (word-initially and finally, as well as in geminates), voiceless ones (see, grass, moth < moððe). There are many problems with this seemingly straightforward formulation of what appears to have been a case of passive voicing in voice-friendly environments (discussed extensively by Fulk 2002, and more recently by Starčević 2024). Hogg (2011, §7.54), for example, does not discuss why fricatives are found voiceless in intervocalic position after an unstressed vowel, followed by a stressed one, as in in befáran ‘go round’, geþánc ‘mind’, aséndan ‘send forth’. This is added as one of the exceptions to the voicing rule affecting the fricatives with no explanation along different (morphological, lexicophonological?) lines. Let us try to explain this along phonological lines. The phonological answer to this may be sought in enhancement, the addition of GW to the unmarked fricatives sitting in the strong position (as in fore/afore), producing [fGW] for /f0/. The difference between OE ofer [v] (= [fGT]) and (be)faran [fGW] falls out from the difference between strong and weak positions: GW was added derivatively to fricatives found in the strong positions only. In intervocalic (weak) position, there was no enhancement with GW, allowing for passive voicing (enhancement with GT) to run its course in voice-friendly environments. The difference between strong and weak positions was played out in the rest of the positions, too, in which the fricatives were found, something we do not attempt here, but cf. Fulk (2002), who shows that wordfinal fricatives (sitting in another weak position) are found voiced (i.e. not enhanced with GW) in the first elements of obscured and quasi-compounds. Enhancement is never found for the unmarked stops given that enhancement can never destroy an opposition (here GW vs unmarked). This leads us to propose the following enhancement rule for OE, shown in (17). (17) Enhancement in Old English Fric in strong position (word-initially or before a stressed vowel) → GW (in weak positions there was no enhancement) The rule in (17) accounts for the difference between (be)faran [fGW] vs ofer [v] ‘over’ in the northern varieties of OE (in the southern varieties there was no such enhancement, giving us words like vixen, van, vial, etc., a handful of those

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 50

that survived in the modern reference varieties, in addition to the dialectal forms that still exist). The question of lenis fricatives in unstressed function words such as the, with, is, of, as well as the suffixes that contained -s found in an unstressed syllable (plural, genitive, etc.) are independent of a northernsouthern division of OE and ME: these fricatives are expected not to show enhancement and were (variably) voiced. This, however, needs further refinement. Although there is no readily available evidence for what the loss of GW in (14) between steps (3) and (4) would have been perceived as in EG. As we argued in Section 4, if they had been perceived as voiced, they would have merged with the word-initial stop allophones of EG voiced stops. This may be interpreted as evidence for their voicelessness, which leads us to the following section. 6. Back to Common Germanic If (the northern varieties of) OE had enhancement (in a system where there was a single series of fricatives), let us entertain the idea that EG had it, too. Iverson & Salmons (2003) do not entertain this possibility because they assume enhancement cannot be found in languages with a single series of fricatives. If enhancement had been embraced by Iverson & Salmons’ (2003) account, it would have made Verner’s Law difficult to formulate (the fricative in faθár would then be expected to turn out voiceless, giving OE fæðer, rather than attested fæder). As there is no serialism in this account, a solution relying on late-activated GW (as opposed to GT coming from the left-adjacent vowel prior to GW) is ruled out. If we, however, entertain the idea that enhancement was found in EG, how can Verner’s Law be salvaged? If we want to maintain that word-initial fricatives were voiceless owing to derivative GW, we can simply explain the presence of passive voicing in faθár as the result of the weak position in which the fricative was found. For this to work, we must assume that this EG was a language in which there was no difference between V́_V and V_V́. It is well-known that with respect to lenition processes the place of stress is relevant in the Germanic languages, but generally not so in the Romance languages (where stress plays no relevance in the definition of intervocalic, cf. Harris 1994, 1997, Scheer 2004, §203, Balogné Bérces 2011, among many others). In this (intervocalic) weak position, there was no enhancement. Enhancement in EG can be formulated in (18). (18) Enhancement in early Common Germanic Fric in strong position (word-initial) → GW

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 51

Passive voicing in the fricatives in words like *faθár can be ascribed to the failure of implementing (18). The fricative (similarly to OE) was passively voiced. The same effect is found when the fricative is both preceded and followed by non-accented vowels, as in *bhérontV- ‘carry, present participle’ (cf. Latin ferent-) > pre-Verner EG *béranθV- > post-Verner EG *bérand- (with

/d0/ from passively voiced */θ0/) > OE berende.6

We can see a marked difference between the strong positions of modern Germanic languages and early Common Germanic. Although more research is needed into this aspect of early Germanic, it may not be far-fetched to conclude that through the prism of weak vs strong positions early Germanic receives another corroborating fact for its transitional state between late IE tone-based accentuation and post-EG prominence-based stress. At the time when Verner’s Law was in operation, stress played no role in the ‘splitting’ of intervocalic into weak intervocalic and strong intervocalic given that stress as a category was yet to develop. If anything, Verner’s Law is here to give us another piece of evidence for the slow development of Germanic phonology from late IE into something that will come to categorise its daughter languages (e.g. OE), see summary in (19). (19) Strong and weak positions in Common Germanic Early Common Germanic Strong # (*faθár [fGW] /f0/) Weak

V(Son)V (*bérenθV- [ð] /θ0/, *faθár [ð] /θ0/)

Old English

, VV́

(fæder [fGW] /f0/, befaran [fGW] /f0/) V́V (hæsel [z] /s0/, ofer [v] /f0/)

The table in (19) shows only a subpart of the positions in which a fricative can appear. We do not analyse positions such as V́SonFricV for OE here or any of the other positions in (early) EG. This is left for further research.7

6

7

The OE suffix -aþ ‘PresIndPl’ < Germanic *-anþ < IE *-ont(i) is expected to show voicing of the fricative, but here one must assume that at least sometimes the suffix was stressed (*-ánþ). A similar solution is suggested by Hogg & Fulk (2011, §2.6) regarding the genitive suffix *-asa (OE -es), which must be assumed (at least sometimes) to have been *-ása or otherwise it would have been voiced and lost. Kümmel (2016) discusses cases of lenition of EG post-nasal stops in a Verner-environment dependent on successive rounds of accentual shift and consonant shift, showing that stops, too, were prone to lenition in EG (even after Grimm’s Law).

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 52

7. Conclusions We have looked anew at the workings of early Common Germanic from the perspective of Verner’s Law, an early Common Germanic regularity, as described by Iverson & Salmons (2003), who claim that the law essentially describes mundane passive voicing which came to be phonologised in a striking way following a number of phonological processes of early Common Germanic. Their analysis crucially hinges on two different enhancement laws, one enhancing the late Indo-European (unmarked) voiceless stops, the other known as Vaux’s Law which they misinterpret as a regularity which can only work in languages that have a system of oppositions in the fricatives based on GT. They do not discuss whether fricative enhancement can also work in languages with a single (unmarked) system of fricatives. The reason for their discarding of such a possibility may be sought in how they conceptualise of enhancement, a process only at work in languages in which there is a GT vs opposition ∅. We show data from Old English that show enhancement of fricatives in a system in which there is only a single series of fricatives. We show that word-initial fricatives were enhanced in early Common Germanic, a feature which precluded them from being analysed as allophones of the early Germanic developments of the Indo-European voiced aspirated stops. An interesting aspect of Common Germanic (as opposed to Old English and the modern Germanic languages in general) is revealed by the absence of the pre-stressed vowel position as a strong position for consonants (the fricatives in both VFricV́ and VFricV positions are weak, as witnessed by their inability to be enhanced with GT). This is expected given the fact that early Germanic did not yet have a system of accentuation based on relative prominence (stress), but tone (shown by the lexical marking of GT on accented vowels). Early Germanic thus represents a transitory state between late IndoEuropean and later Germanic with a stress system known from the modern daughter languages, GW (originally a feature of phonetic enhancement) percolating into the phonological representation of GW-marked fricatives and stops (with concomitant loss of GT as a laryngeal dimension). The early Common Germanic treatment of unmarked fricatives in the intervocalic positions shows ever more decidedly that Verner’s Law can only have worked in early Germanic before GW became a phonological feature of the newly established series of fricatives.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 53

References Avery, Peter and William J. Idsardi. 2001. Laryngeal dimensions, completion and enhancement. In: Tracy Alan Hall (ed.), Distinctive feature theory. Berlin: Mouton de Gruyter. 41–70. https://doi.org/10.1515/9783110886672.41 Balogné Bérces, Katalin. 2011. Weak and semiweak phonological positions in English. Journal of English Studies 9: 75–96. https://doi.org/10.18172/jes.160 Carvalho, Ana Maria. 2006. Spanish (s) aspiration as a prestige marker on the UruguayanBrazilian border. Spanish in Context 3(1): 85–114. https://doi.org/10.1075/sic.3.1.07car Collinge, Neville Edgar. 1985. The laws of Indo-European. Amsterdam/Philadelphia: John Benjamins. https://doi.org/10.1075/cilt.35 Fisiak, Jacek. 1984. The voicing of Initial fricatives in Middle English. Studia Anglica Posnaniensia 17: 3–23. http://hdl.handle.net/10593/12509 Foulkes, Paul. 1997. Historical laboratory phonology - investigating /p/ > /f/ > /h/ changes. Language and Speech 40: 249–276. https://doi.org/10.1177/002383099704000 Fulk, Robert Dennis. 2002. Conditions for the voicing of Old English fricatives, II: Morphology and syllable structure. English Language & Linguistics 6(1): 81–104. https://doi.org/10.1017/S1360674302001053 Harris, John. 1994. English sound structure. Oxford: Blackwell. Harris, John. 1997. Licensing Inheritance: an integrated theory of neutralisation. Phonology 14: 315-370. https://doi.org/10.1017/S0952675798003479 Haugen, Einar. 1982. Scandinavian language structure: A comparative historical survey. Minneapolis: University of Minnesota Press. Hogg, Richard Milne and Robert D. Fulk. 2011. A grammar of Old English. Volume 2: Morphology. Oxford: Wiley-Blackwell. https://doi.org/10.1002/9781444327472 Hogg, Richard Milne. 2011. A grammar of Old English. Volume 1: Phonology. Oxford: Wiley-Blackwell. https://doi.org/10.1002/9781444341355 Honeybone, Patrick. 2002. Germanic Obstruent Lenition: some mutual implications of theoretical and historical phonology. PhD thesis, University of Newcastle upon Tyne. Honeybone, Patrick. 2005. Diachronic evidence in segmental phonology: The case of obstruent laryngeal specifications. In Marc van Oostendorp & Jeroen van de Weijer (eds.), The internal organization of phonological segments. Berlin & New York: Mouton de Gruyter. 319–354. https://doi.org/10.1515/9783110890402.317 Iverson, Gregory K. and Joseph C. Salmons. 2003. Laryngeal enhancement in early Germanic. Phonology 20 (1): 43–74. https://doi.org/10.1017/S0952675703004469 Kümmel, Martin Joachim. 2016. Das dünkt mich dunkel: Germanische etymologische Probleme [It seems dark to me: Etymological problems in Germanic]. In Bjarne Simmelkjær Sandgaard Hansen, Benedicte Nielsen Whitehead, Thomas Olander & Birgit Anette Olsen (eds.), Etymology and the European lexicon. Proceedings of the 14th Fachtagung der Indogermanischen Gesellschaft, 17-22 September 2012, Copenhagen. Wiesbaden: Reichert. 219–233.

Starčević, A Note on Common Germanic Enhancement and Vaux’s Law 54 Labrune, Laurence. 2012. The phonology of Japanese. Oxford: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199545834.001.0001 Lass, Roger. 1984. Phonology. Cambridge: Cambridge University Press. Lass, Roger. 1991–1993. Old English fricative voicing unvisited. Studia Anglica Posnaniensia 25–27: 3–46. MacAuley, Donald. 1992. The Celtic Languages. Cambridge: Cambridge University Press. Maza, Maria. 2000a. Phonetic data and functional explanation in phonology: the case of laxness vowel harmony in Granada Spanish. Newcastle: University of Newcastle upon Tyne, Manuscript. Maza, Maria. 2000b. Ambiguity and Laxness Vowel Harmony in Granada Spanish. Durham: University of Durham, Manuscript. Orton, Harold. 1962. Survey of English dialects: Introduction. Leeds: E. J. Arnold. Page, B. Richard. 1997. On the origin of preaspiration in Scandinavian. Journal of Germanic Linguistics 9(2): 167–190. https://doi.org/10.1017/S1040820700002857 Salmons, Joseph C. 1993. The glottalic theory: survey and synthesis. McLean, Va.: Institute for the Study of Man. Sámmol Ánte, Luobbal Sámmol. 2022. Proto-Uralic. In: Marianne Bakró-Nagy, Johanna Laakso and Elena Skribnik (eds.). The Oxford Guide to the Uralic Languages. Oxford: Oxford University Press. https://doi.org/10.1093/oso/9780198767664.003.0001 Scheer, Tobias. 2004. A lateral theory of phonology. What is CVCV, and why should it be?. Berlin & New York: Mouton de Gruyter. https://doi.org/10.1515/9783110908336 Starčević, Attila. 2024. Another fortis-lenis language: A reanalysis of Old English obstruents. Acta Linguistica Academica 71(3): 259–308. https://doi.org/10.1556/2062.2023.00705 Stuart-Smith, Jane. 2008. Scottish English: phonology. In: Bernd Kortmann, Clive Upton and Edgar W. Schneider (eds.), The British Isles. Berlin, New York: De Gruyter Mouton. 48– 70. doi.org/10.1515/9783110208399.1.48 Vaux, Bert. 1998. The laryngeal specifications of fricatives. Linguistic Inquiry 29(3): 497– 511. https://doi.org/10.1162/002438998553833 Zvelebil, Kamil. 1970. Comparative Dravidian phonology. The Hague: Mouton. Attila Starčević Eötvös Loránd University ORCID 0000-0003-3055-1404 starcevic.attila@btk.elte.hu