You are on page 1of 16

The Glottalic Theory revisited: a negative appraisal

Part : The typological fallacy underlying the Glottalic Theory

Introduction

In part one of this study, I attempted to rebut the main arguments


proposed by Glottalicists against the Standard Model of PIE. - save
the following: the assertion that the Ejective Model is typologically
superior. Part Two addresses this proposition.

3.6 Jakobson9s Universal

Let us now return to the observation that is often cited äs the main
impetus to the development of the Glottalic Theory: Roman Jakob-
son's Universal (1958:23; cf. Salmons, 1993:11): An obstruent System
evincing the stops /T/ and /D/ will not add /D H / unless the System
already possesses /TH/. By implication, the Standard Theory of PIE.
is rendered highly suspect, given that the Standard Model contains
*T-*D-*DHbutnot*TH.
Note that Jakobson's Universal is well corroborated in both the
UPSID sample (Maddieson, 1984) äs well äs in Ruhlen's compilation
(1975). All 7 languages exhibiting DH in the UPSID sample are in
complete accordance with Jakobson's Universal and exhibit at least 4
series of stops: Kharia and Mundari (Austro-Asiatic, p. 321); Telugu
(Dravidian, p. 413); Bengali and Hindi-Urdu (Indo-lranian, p. 270);
!Xü (Khoisan, p. 422); Igbo (Niger-Kordofanian, p. 292).
Ruhlen (1975) lists 10 languages exhibiting DH not mentioned in
Maddieson (1984), 8 of which exhibit at least 4 series of stops and are
in complete accordance with Jakobson's Universal: Juang [p. 214]
and Santali [p. 263] (Austro-Asiatic); Javanese [p. 213] and Madurese
[p. 233] (Austronesian); Kurux [p. 228] (Dravidian [North]); Gujarati
[p. 202], Marathi [p. 237], Nepali [p; 247], Oriya [p. 252], and Sindhi
[p. 268] (Indo-lranian).
The only languages in Ruhlen that do not conform to Jakobson' s
Universal are the two belonging to the Austronesian family, Javanese

Indogermanische Forschungen, 108. Band 2003


Brought to you by | Cornell University Library
Authenticated
Download Date | 10/17/16 4:53 PM
2 Charles M. Barrack

(p. 213) and Madurese (p. 233), both of which are triseriaL Javanese
exhibits plain voiceless stops (T), voiced aspirates (DH), and voiced
prenasalized stops (ND). Since Madurese mirrors the Standard Mod-
el for PIE., it will be discussed more fully below.
In summary, Maddieson (1984) and Ruhlen (1975) present strong
corroboration of Jakobson's famous dictum. Nevertheless, Jakob-
son's Universal is not without exception: Two languages have been
documented with precisely the sarae triserial configuration of stops
posited by the St. Theory for PIE.: Kelabit, an Austronesian lan-
guage described by Blust (1974), and Madurese, an Austronesian
language described by Stevens (1968) and, in greater detail, by Ki-

On the basis of Stevens' brief, ambiguous description of the aspi-


rated stops of Madurese (1968:16,38), Hopper contends that they
are actually voiceless aspirates, i.e., TH, rather than DH (1977:43),
thereby denying that the obstruent System of Madurese mirrors the
Standard Model of PIE. However, Kiliaan's detailed description of
the Madurese obstruent System leaves little doubt that the aspirated
series does indeed coniprise voiced aspirates (translation mine) :

(10) Also peculiar to Madurese is the fact that the voiced [zackte
'soft'] occlusives are divided into either .unaspirated or aspi-
rated occlusives, viz., b and bh (labials), d and dh (dentals) ...
The voiced aspirates form the transition from the voiced unas-
pirated to the voiceless [scherpe 'sharp'] plosives, respectively

Further evidence supporting Kiliaan's Interpretation of the Mad-


urese stop System can be found on pp.65-71 where he notes that
the voiced aspirates occasionally alternate with voiced stops in Mad-
urese, e.g., biduri(h) ~ bidhurih 'good'.
In summary, although two triserial languages - Kelabit and Mad-
urese - have been documented exhibiting obstruent Systems which
mirror that posited in the St. Theory for PIE. (i.e., *T-*D-*DH), it is
clear that this triserial stop System must be viewed äs typologically
deviant. But does the Ejective Model fair any better in this regard?

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 3

3.6.1 Tlie co-occurrenceof*T'and *DHin theEjectiveModel

Glottalicists find the Ejective Model superior to the Standard


Model by virtue of the fact that they believe the former does not run
afoul of Jakobson's Universal. But äs we shall see, this belief is based
on specious reasoning. The more important question at this point is
how the Ejective Model compares with what is typically encountered
in ejective Systems. Here we find two anomalies both of which se-
riously undermine the Gl. Theory:
(11 a) 69.7% (221/317) of the languages in the UPSID corpus exhibit
affricates, whereas an examination of just those languages in
the sample with voiceless ejectives reveals that a füll 90.4%
(47/52) do so. Hence, the Ejective Model is anomalous be-
cause it lacks affricates, an anomaly of which at least some
Glottalicists are aware (cf. Vennemann 1989:108-114).
(l l b) Afar more serious problem for the Gl. Theory is that created
by the presence of voiced aspirates in PIE. To my knowledge,
no Glottalicist has addressed the question: What is the likeli-
hood that ejectives will co-occur with voiced aspirated stops?

3.6.1.1 Salmons' Quadraserial /D^/ Constraint

Before we examine the line of argumentation Glottalicists have fol-


lowed, a specious line of reasoning that skirts this important ques-
tion, let us review Jakobson's seminal Statement - what has come to
be known äs'Jakobson's Universal' (1958:23):
(12) To my knowledge, no language adds, to the pair /t/ - /d/, a
voiced aspirate /d h / without having its voiceless counterpart
/th/, while /t/, /d/, and /th/ frequently occur without com-
paratively rare /d h / ..., and such a stratification is easily ex-
plicable...
About three years earlier, Martinet had already explained why Sys-
tems with /D H / will also exhibit /TH/: Since /TH/ is only marked
once, viz., for aspiration, feature economy makes it likely that a sys-
tem will make use of/T H / before it increases its phonemic inventory
by adding /D H / - given that the latter is doubly marked, once for
voicing and once for aspiration (1955:114f.). Nevertheless, there is a

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
4 Charles M. Barrack

degree of ambiguity in Jakobson's Statement, which could loosely be


interpreted to assert the following (cf. Gam.-Iv., 1973:151, 153, 155):
(13) /D H /presupposes/T H A
AL stricter Interpretation is the following. Let us term this the
'Quadraserial /D H / Constraint' (cf. Salmons, 1993:55):
(14) /D H / presupposes three other series, /T/ - /D/ - /TH/.
This stricter Interpretation appears to be well founded theoretically
because it is based on the maximal use of the features [± voiced] and
[± aspiration], and is corroborated statistically: 88.3% (15/17) of the
languages exhibiting /D H / in the UPSID sample (7/7) taken to-
gether with those in Ruhlen's compilation (8/10) evince /T/ - /D/ -
/TH/ - /D H / - at least äs a proper subset - and, hence, are in com-
plete accord with this constraint.
Accordingly, both the triserial Ejective Model äs well äs the trise-
rial Standard Model should be regarded äs typologically deviant.
However Glottalicists will doubtless disagree with this assessment.
According to Gam.-Iv. (1986:94), "aspiration in [/TH/ and /DH/] is
phonologically irrelevant", for the corresponding phonemes of these
two series could also be represented by allophonic variants without
aspiration. In other words, the Ejective Model can simply be repre-
sented by underspecified Segments äs in (15):
(15) * T _* T '_* D

Moreover, Gam.-Iv. äs well äs Salmons argue that this Ejective


Model is in complete accord with the facts of synchronic typology.
Discussing their model Gam.-Iv. claim:
(16) Systems of this structural type (with oppositions betweeri glot-
talized and non-glottalized and voiced and voiceless) are
foünd in many historically attested languages (1986:95; em-
phasis mine).
In his attempt to revise the Ejective Model, Salmons makes a simi-
lar claim for his variable glottalic sei hypothesis by seekirig support
in the UPSID sample. I reproduce Salmons' claim here because of
the light it casts on the fallacy underlying this line of argumentation:

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 5

(17) The system described above [reflecting the variable glottalic


set hypothesis] represents an extremely unmarked system
across the languages of the world. As noted, Maddieson
(1984:28-29) fmds a contrast based on either aspiration or
voice plus one glottalic series the most common three-way se-
ries system... (1993:58; emphasis mine).

In order fully to appreciate the fallacy underlying these attempts


to corroborate the Ejective Model, it is necessary to exam more
clbsely the phonological Status of the stop phonemes posited within
the Gl. Theory.

3.6.1.2 Underspecißed segments versus dominant allophones

There are two widely accepted but very different perspectives on


the nature of the phoneme: the phoneme viewed äs an underspeci-
fied segment versus the phoneme äs represented by its so-called
'dominant' allophone, the least environmentally conditioned allo-
phone of the respective phoneme.
As we shall see, a failure to heed this distinction has led in turn to
the misinterpretation of Jakobson's Universal (äs well äs to misrepre-
sentation of the UPSID sample) and to specious arguments purport-
ing to establish the typological superiority of the Ejective Model over
the Standard Model.
Let us begin by reexamining Jakobson's Universal, (12). Since he
implicitly bases his conclusion on the economical use of the features
of voice and aspiration, each phoneme must be specified for both of
these features in order to determine whether they are being used eco-
nomically, hence maximally. Nevertheless, let us assume for heuris-
tic reasons that Jakobson's intent is to configure underspecified Seg-
ments in his universal.
He begins by positing a simple System containing just the two se-
ries /T/ and /D/, hence a system in which voice is the sole contrast-
ing feature. He then addresses the question: If the system adds one
aspirated stop, will it be voiced (i.e., /D H /) or voiceless (i.e., /T H /)?
However, if Jakobson's intent is to set up underspecified segments,
the question is moot, for voicing becomes redundant in aspirated
stops if only one set of aspirates is added to the inventory. Hence, at

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
Charles M. Barrack

the level of underspecification, both /TH/ and /DH/ are equally


acceptable typologically (cf, Kurylowicz, 1973:68 f.).
08)
System A Voice Aspiration System B Voice Aspiration
/T/ - - /T/ - -
/D/ + - /D/ + -
/TH, + /D / H
+
<-) (+)
Comparison of a Natural to an Unnatural System with Aspirates

Hence, at the level of underspecification, the Standard Model of


PIE. (System B) is just äs adequate äs Jakobson's preferred model
(System A). Therefore, we can conclude that Jakobson's intent is to
compare the two Systems on the basis of their dominant allophones,
not on the basis of underspecified segments. Note that Gam.-Iv.
demonstrate a recognition of this fact by stating that *TH and *DH
are the dominant allophones respectively of the phonemes *T and
*D in their Ejective Model (1986:97) and that the presence of these
two dominant allophones satisfies Jakobson's Universal (1973:151,
153, 155).
Significantly, both Maddieson and Kuhlen employ the same type
of phonological analysis in their compilations. Concerning the
UPSID sample, Maddieson informs us: "In the database each seg-
ment which is considered phonemic is represented by its most char-
acteristic allophone ..." (1984:6; cf. Ruhlen, 1976:22-26). The fact
that all three - Jakobson, Maddieson, and Ruhlen - choose to com-
pare languages at the more phonetically specified level of the domi-
nant allophone is not unexpected: The patterns of phonological de-
tail languages preferentially select to flesh out the phonemic frames
of their segment inventory is at least äs important to typology äs the
skeletal frames themselves.
To summarize: Jakobson's Universal informs us that (1) a triserial
System of phonemes -äs represented by their dominant allophones -
is natural if it contains the following members, /T / - /D/ - /TH/;
(2) however, a triserial System comprising /T/ - /D/ - /D H / is un-
natural, for the aspirate /D H / will normally only occur in conjunc-
tion with the aspirate /TH/ in a quadraserial System comprising /T/

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 7

- /D/ - /TH/ - /DH/. These facts are captured in the Quadraserial


/DH/ Constraint and are corroborated by the UPSID sample äs well
äs by Ruhlen's compilation, both of which describe and compare
phonological segments in terms of dominant allophones.

3.6.1.3 The Ejective Model äs a typological isolate

Now let us return to the question posed earlier: Does Gam.-Iv.'s


version of the Ejective Model run afoul of the Quadraserial /DH/
Constraint? To answer this question we must first determine the
dominant allophones of the phonemes in Gam.-Iv.'s Ejective Model.
Regarding the two non-glottalic series of stops in their model, they
inform us: "[T]he dominant allophone of these sets [viz., /*T(HV
and /*D(HV] is the aspirated one - for it appears in most phoneti-
cally independent positions ..." (Gam.-Iv., 1986:97). In terms of
dominant allophones, therefore, Gam.-Iv.'s Ejective Model com-
prises the triserial System given in (19).
(19)
Standard Ejective
Series
Model Model
*jH
L *T
II. *D *T>
III. *DH *DH
Dominant Allophones

If we examine the Standard Model and Gam.-Iv.'s version of the


Ejective Model in light of the Quadraserial /D H / Constraint we find
that both models violate this Constraint (1) by exhibiting the pho-
neme /*DH/ in a triserial, not qüadraserial, system and (2) by not
exhibiting /*T/ and /*TH/ äs two separate phonemic sets: Further-
more, the Ejective Model violates the Constraint by not exhibiting
the /*D/ series äs a separate set from /*DH/..
It is important to note, however, that the Standard Model,
although typologically deviant, does not constitute a typological iso-

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
8 Charles M. Barrack

late: As we saw in 3.6.1. above, two languages, Kelabit and Madu-


rese, have been documented .which instantiate precisely the same
configuration of stops äs that posited in the Standard Model.
However, Gam.-Iv.'s Ejective Model is more problematic, and, in
fact, appears to be a typological isolate. As noted earlier, the prob-
lem arises from the fact that a series of voiceless ejectives (T) appears
to be incompatible with a series of voiced aspirated stops (DH). This
incompatibility is corroborated by the UPSID sample (Maddieson,
1984) and by Ruhlen's (1976) larger but less carefully controlled
compilation of languages (raany of which are also listed in Maddie-
son).
(20) An examination of the 317 languages of the UPSID sample -
indeed an examination of Ruhlen's ca. 700 languages - reveals
the remarkable fact that there are no triserial languages that in-
stantiate Gam.-Iv.'s Ejective Model in terms of dominant allo-
phones.
In other words, no triserial language contains both voiceless ejec-
tives (/T/) and voiced aspirated stops (/DH/). Indeed, one will
search in vain for the co-occurrence of these series in languages with
from two to five stop series. It is only when one examines !Xü - one
of only two languages in UPSID with six series of oral stops - that
one encounters the co-occurrence of /T/ and one member of /DH/,
viz., /gh/. (See the Appendix for a list of languages in UPSID con-
taining T'.)
Moreover, !Xü - a member of the Khoisan family - can lend little
support to the Gl. Theory, for !Xü is atypical of the languages of the
world along several dimensions and represents the typological isolate
par excellence in the UPSID inventory: Not only does !Xu exhibit 6
series of non-click oral stops, it also exhibits at least a dozen series of
affricated and non-affricated clicks. In fact,, its phonological inven-
tory contains the largest number of segments of any language in the
entire UPSID corpus. Whereas the mean number of segments per
language in this corpus is between 31 and 32; !Xü exhibits 141 (Mad-
dieson, 1984:7, 421 f.).
To my knowledge, in none of the proposals extolling the triserial
ejective model for its alleged typological superiority have any of its
proponents cited a single language that instantiates that model on
the phonologically detailed level of dominant allophones. (Paradoxi-

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 9

cally, many Glottalicists claim that one of the strengths of the Gl.
Theory is its attention to phonetic detail in reconstruction [e.g.,
Gam.-Iv., 1973:155; Salmons, 1993:15,70].)
For example, Hopper Claims that the Ejective Model is "both com-
mon, and importantly, has a wide geographic and cultural spread".
As examples, he cites Georgian from the Caucasus, Hausa from Afri-
ca, and Quechua from the Americas. However, not one of these three
languages exhibits the voiced aspirate series /DH/ äs dominant allo-
phones (cf. Maddieson, 1984:318,402, 417 respectively). Hence,
Hopper is citing these three languages in support of the Ejective
Model on the broad phonological scale of underspecification. See
(21).

(21)
Ejective
Series Hausa1 Quechua Georgian
Model
L *TH T T T
II. *T» T/D 1
T r
*DH •pH
III. D yH

Three Ejective Systems Compared to Ejective Model

To summarize: Whereas the Standard Model of PIE. is marginally


acceptable, given that Kelabit and Madurese have similar stop Sys-
tems, no such corroboration can be found for any triserial model ex-
hibiting both ejectives and voiced aspirated stops.
Counter to the Claims of Glottalicists, therefore, any version of the
Ejective Model that posits the co-occurrence of *T and *DH must be
looked upon äs alypologica! isolate rather than äs a model of typolo-
gical adequacy.

1
In Hausa, ejectives and implosives constitute a variable glottalic set: Ejectives
occur dorsally; implosives have frontal articulation.

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
10 Charles M. Barrack

4. The Standard Model versus the Neogrammarian Model

Nevertheless, the fact that the Standard Model is preferable to the


Ejective Model does not completely exonerate the formen Even
though the phonological Systems of Kelabit and Madurese are proof
that the Standard Model does not constitute a typological isolate, the
fact that one can find no corroboration in the UPSID sample and
only one in Ruhlen's compilation (viz., Madurese) should give one
pause, for such a lacuna is prima facie evidence that Jakobson's Uni-
versal represents a valid generalization: triserial Systems exhibiting
/T/ - /D/ - /TH/ are well represented; Systems exhibiting /T/ - /D/
- /D H / are rare.
Even more troublesome is the highly suggestive fact that virtually
all direct evidence of the presence of *DH in PIE. comes from dia-
lects that are in perfect accord with Jakobson's Universal. In the In-
dic subfamily, not only Sanskrit but apparently every modern dialect
that preserves /D H / exhibits this aspirated series in Jakobson's pre-
ferred quadraserial configuration, at least äs a proper subset: /T/ -
/D/ - /TH/ - /DH/.' Kuhlen (1975) provides the following exam-
ples: Bengali (p. 168), Gujarati (p. 202), Hindi-Urdu (pp. 204f.),
Marathi (p. 237), Nepali (p. 249), Oriya (p. 252), Sindhi (p. 268).
If we accept Garrett's assessment (1991:794-798) that the series
/D H / exhibited by a number of Armenian dialects represents a direct
descendent of PIE. *DH, we find yet more corroboration of Jakob-
son's Universal, for although these dialects are triserial, they never-
theless all exhibit /TH/.
If one assumes that the original PIE. obstruent System is in fact
better represented by the Neogrammarian quadraserial configuration
*T - *D - *TH - *DH - hence in complete accordance with Jakob-
son's Universal - then another important correlation emerges which
also corroborates Jakobson's thesis: When the four stop series of
Indo-European were reduced to three through the deaspiration of
PIE. *TH in Greek and Proto-Italic, *DH devoiced rendering a new
series, *TH, thereby creating Jakobson's preferred triserial configura-
tion containing /T/ and /D/ when accompanied by an aspirated se-
ries: *T - *D - *TH. The same process has taken place in the modern
triserial Indic dialects of Kashmiri and Punjabi with the devoicing of
*DH (Maddieson, 1984:271).
Comparatists finally rejected the two voiceless stop series, viz., *T

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 11

and *TH, in the Neogrammarian quadraserial reconstruction in favor


of the present triserial Standard Model without *TH and derive these
aspirates from a sequence of *T followed by a laryngeal (see Sal-
mons, 1993:6-9 and Mayrhofer, 1986:91 f.). The claim is occasion-
ally made that the Neogrammarian quadraserial reconstruction was
originally postulated out of reverence for an imagined primacy of Ve-
dic among the older Indo-European sister dialects with the result
that Vedic TH was given undo weight in the reconstruction of the
PIE. obstruent System (cf. Gamkrelidze, 1990:6).
Doubtless such bias did play a role in the acceptance of the quad-
raserial reconstruction of PIE. What is not recognized, however, is a
more subtle bias in favor of the triserial over the Neogrammarian
quadraserial configuration: the unexamined bias on the part of lin-
guists toward formally 'simpler' Systems. The possibility of recon-
structing a triserial obstruent System for PIE. in place of the Neo-
grammarian quadraserial configuration must have been ipso facto at-
tractive to many Indo-Europeanists.
Nevertheless, with the advent of compilations like the UPSID
sample, it is obvious now that a quadraserial System containing a
dual series of aspirates, TH and DH, is far more common than any
configuration - no matter how formally 'simpler' on the surface -
that contains the sole aspirate series DH. Clearly the Neogrammarian
quadraserial configuration deserves another look and the arguments
of its proponents such äs Szemerenyi (1999:143 f.) and Rasmussen
(1989:153-176; 249-254) should be granted careful scrutiny.

5. Summary and conclusion

When the Neogrammarian quadraserial reconstruction of the PIE.


stop System (viz., *T - *D - *TH - *DH) was rejected in favor of the
present triserial model (viz. *T - *D - *DH), Roman Jakobson re-
sponded with his famous dictum noting that such a triserial config-
uration was typologically deviant because it did not pair *DH with
*TH. Jakobson's observation helped give impetus to the Glottalic
Theory, which in the best known Version, that of Gamkrelidze and
Ivanov (1973), replaces PIE. *D with the ejective series, T: Glot-
talicists find support for the Ejective Model in a number of anom-
alies, among the most important of which are the following:

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
12 Charles M. Barrack

(22a) The dearth of the labial element *b in the Standard Model


By interpreting the labial to be the ejective *p', Gamkrelidze-
Ivanov claim to account for its dearth in PIE., for p' is excep-
tionally rare in ejective Systems according to Gamkrelidze-Iva-
nov. However, äs the UPSID sample demonstrates, the labial
ejective is much more likely to be present in an ejective System
than not. Hence replacing *b with *p' will not account for the
dearth of the labial, *b.
(22b) The absence or dearth of roots of the structure *DVD
By inte eting these roots to be sequences of ejectives, i.e.,
*T'VT, Gam.-Iv. claim to be able to account for their absence
because, according to Gam.-Iv., such sequences are often dis-
allowed in ejective languages. However, a statistical analysis
corroborates Iverson and Salmons' (1992:295) claim that this
dearth is an artifact of the overall rarity of *D and the tendency
for stops to be avoided in codas.
(22c) The dearth of PIE. *D äs compared to *DH in inflectional af-
fixes
By reconstructing Series II äs *T instead of *D, Glottalicists
claim to have explained this dearth because marked segments
are typically uncommon in inflectional affixes, and *T is more
marked than *D. But this argument is faulty because *DH is
more marked than either *D or glottalic *T. Hence the re-
placement of *D by *T would not account for the dispropor-
tionate frequency of *DH in inflections.
The putative change *T > *D is postulated in the Gl. Theory in
order to account for the correspondences between *T in Anatolian,
Armenian, Germanic and a few others, on the one hand, and the *D
of Indo-Iranian, Greek, Italic, Celtic, Balto-Slavic, and Albanian,
on the other.2 However, this is a very poorly attested change; much
more common is simple deglottalization without voicing, *T > *T.
Other noteworthy objections to the glottalic voicing hypothesis

Kortlandt asserts that the deglottalized voiceless stop *T (< *T) was only pre-
served in Anatolian and Tocharian (1988:391).

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 13

have been raised. For example, Meid (1987:10 f.) has noted that nati-
vization of loanwords containing donor *D in Germanic regularly
produces *T from donor D, indicating that Germanic did in fact pass
through a period of devoicing and that such devoicing corroborates
Grimm's postulated change of PIE. *D > PGmc. *T rather than
Gamkrelidze-Ivanov's putative change of PIE. *T > PGmc. *T.
Gamkrelidze (1990:7f.) counters that this change of donor *D to
PGmc. *T is merely a case of phonological Substitution. However, if
phonological Substitution were indeed the underlying cause, one
would expect PGmc. *D to be just s likely to be substituted for do-
nor D s PGmc. *T, for although the latter uniquely shares [-con-
tinuant] with donor D, the former uniquely shares [+voice] with do-
nor D. The fact that loanwords into Germanic uniformly exhibit
PGmc. *T for donor D is prima facie evidence of a prehistoric de-
voicing stage, not sound Substitution.
Faced with the substantial amount of evidence that Series II must
have been voiced, some Glottalicists have proposed that this series
constituted not *T' but rather voiced implosives, *ΟΊ. Haider's at-
tempt (1985:10-25) at such a solution fails in that it ignores the fact
that implosives tend to have frontal rather than dorsal articulation,
and, hence, cannot account for the normal distribution of Series II,
*D, at the (labio)velar point of articulation.
Salmons' (1993:24,58) s well s Kortlandt's attempts (1977:319,
1978:107-110, 1988:391) to obviate this problem by postulating a
variable glottalic set with *T and *Dn allophones runs into a similar
problem: If their accounts were correct, one should expect *ΟΊ >
*D to be uncommon in Segments exhibiting dorsal articulation. Ac-
cordingly, Anatolian, Armenian, and Germanic3 should evince more
instantiations of *T > *T in roots containing dorsal Segments than
Indic-lranian, Greek, etc., evince of *Dn > *D in those same roots -
which, of course, is not the case.
However, the fundamental weakness in the Glottalic Theory is re-
vealed in a more fundamental discrepancy: its claim to typological
superiority over the Standard Theory. For this superiority is an illu-

J
Kortlandt would exclude Germanic from this group.' In other words, he con-
tends that Series II sufiered the following fate in Germanic: PIE. *T > *Ό^
> *D > «T. See note 2.

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
14 Charles M. Barrack

sion fostered by the lack of clarity on the phonological Status of the


segments under discussion.
The criticism of the traditional triserial model of PIE. is based on
Jakobson's Universal, which implicitly declares the Standard Model
to be typologically deviant in terms of its dominant allophones, not
in terms of its underspecified segments. Arguments extolling the pu-
tative typological superiority of the Ejective Model, on the other
hand, are formulated not in terms of its dominant allophones, viz.,
*TH - *T - *DH, but rather in terms of these phonemes äs under-
specified segments, viz., *T - *T - *D.
Clearly, it is not legitimate to compare the typological adequacy of
the Standard Model with that of the Ejective Model using different
measures of phonological specificity. Moreover, an examination of
the dominant allophones of the Ejective Model against the UPSID
sample and Ruhlen's data base reveals the remarkable fact that the
Ejective Model is very deviant from a typological perspective: There
are no triserial languages in either database exhibiting the co-occur-
rence of voiceless ejectives (/T/) with voiced aspirated stops (/DH/).
Indeed, only one language, !Xü - a veritable typological isolate -
with its six series of oral stops and more than a dozen series of clicks,
exhibits the co-occurrence of/TV and only one member of the /D H /
series, the velar. Hence Gamkrelidze-Ivanov's Ejective Model repre-
sents a typological isolate.
The lasting contribution of the Glottalic Theory is not a superior
model of the PIE. obstruent System. Rather, the Glottalic Theory
compels us to reexamine not only the adequacy of the Standard
Model but to take a closer look at the typologically superior quadra-
serial configuration that preceded it: Neogrammarian *T - *D - *TH
-*DH.
Appendix'
Languages in Maddieson (1984) with voiceless ejective stops (n =
52)
Only one, !XO, also exhibits 'voiced aspirated' stops, viz., /gh/.
The column of single digits indicates the number of stop series.
Acoma(p. 387) 4 Mazahua (p. 376) 5
Amharic(p. 311) 3 . Nambiquara, South, (p. 401) 4
Armenian, East. (p. 273) 3 Navaho (p. 369) 3
Berta(p.309) 4 Nez Perce (p. 371) 2

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
The Glottalic Theory revisited 15

Chipewyan (p. 369) 2 Nootka (p. 378) 2


Chontal (p. 373) 3 Otomi (p. 376) 5
Dakota (p. 391) 4 Porno, Southeastern (p.384) 3
Dizi(p.317) . 3 Füget Sound Salish (p. 380) 3
Georgian(p.417) ,3 Quechua (p. 402) 3
Gununa-Kena (p. 403) 3 Quileute (p. 379) 3
Haida (p. 368) 3 Shasta (p. 386) 2
Hamer(p.318) 4 Socotri(p. 312) 3
Hausa(p.318) 4 Squamish (p. 380) 2
Hupa(p.370) 3 Tigre(p.311) 3
Ik (p. 304) 4 Tiwa (p. 383) 4
Itonama (p. 393) 3 Tlingit (p. 368) 3
Jaqaru (p. 403) 3 Tolowa (p. 370) 3
Kabardian(p.417) 4 Tzeltal (p. 374) 4
Kefa (p. 317) 3 Wappo (p. 393) 3
Klamath (p. 371) 4 Wichita (p. 390) 3
Koma(p.310) 4 Wintu (p. 372) 4
Kullo(p.316) 4 !Xü(p.421) 6
Kwaw'ala (p. 379) 4 Yana (p. 385) ' 4
K'ekchi (p. 375) 3 Yuchi (p. 391) 4
Lak(p.418) 5 Zulu (p. 297) 4
Maidu (p. 372) 3 Zuni (p. 387) 3

References

Blust, R. 1974. A double counter-universal in Kelabit. Papers in Linguistics 7,


309-324.
Gamkrelidze, T. 1990. The Indo-European glottalic theory in the light of recent
critique. Folia Linguistica IX/ l, 3-12.
Gamkrelidze T. and .V. Ivanov. 1973. Sprachtypologie und die Rekonstruktion
der gemeinindogermanischen Verschlüsse. Phonetika 27, 150-156.
-, 1986. On the reconstruction of the Proto-Indo-European stops. In: V. Shevor-
oshkin and T. Markey (eds. & trans.), Typological relationship and time: a col-
lection of papers on language change and relationship by Soviet linguists, 87-
108.
Garrett, A. 1991. Indo-European reconstruction and historical methodologies.
Language 67, 790-804.
Haider, H. 1985. The failacy of typology: remarks on tbe PIE. stop System. Lin-
gua65, 1-27.
Hopper, P. 1977. The typology of the Proto-Indo-European segmental inventory.
Journal of Indo-European Studies, 41-53.

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM
16 Charles M. Barrack

Jakobson, R. 1958. Typological studies and contributions in historical compara-


tive linguistics. In: Proceedings of the Eighth International Congress of Lin-
guists, 17-25. Oslo: Oslo University Press.
Kiliaan, H. 1911. Madoereesche spraakkunst. Eerste stuk: inleidingen klankleer.
Semarang, Indonesia: H. A. Benjamins; Semarang-Drukkerij en boekhandel.
Kortlandt, F. 1977. Historical laws of Baltic accentuation. Baltistica 13:319-330.
-, 1978. Proto-Indo-European obstruents. Indogermanische Forschungen
83:107-118.
-, 1988. Remarks on Winter's Law. In: Dutch contributions to the lOth Interna-
tional Congress of Slavists. Sofia. Linguistics (= Studies in Slavic and general
linguistics 11), 387-396. Amsterdam: RODOPI.
Kurylowicz, J. 1973. Internal reconstruction. In: T. Sebeok (ed.), Current trends
in linguistics. 11:63-92. The Hague: Mouton.
Maddieson, J. 1984. Patterns of sounds. Cambridge: Cambridge University
Press.
Martinet, A. Economie des changements phonetiques: traite de phonologie dia-
chronique. Berne: Francke.
Mayrhofer, M. 1986. Lautlehre (segmentale Phonologie des Indogermanischen).
Zweiter Halbband. In: M. Mayrhofer (ed.), Indogermanische Grammatik.
Heidelberg: Carl Winter.
Meid, W. 1987. Althochdeutsch I: Grammatik, Glossen und Texte. Heidelberg:
Carl Winter.
Rasmussen, J. 1989. Die Tenues Aspiratae: Dreiteilung oder Vierteilung des in-
dogermanischen Plosivsystems und die Konsequenzen dieser Frage für die
Chronologie einer Glottalreihe. In: T. Vennemann (ed.), The new sound of
Indo-European: essays in phonological reconstruction. Berlin: Mouton de
Gruyter.
Ruhlen, M. 1975. A guide to the languages of the world. Stanford, CA: Stanford
University.
Salmons, J. 1993. The glottalic theory: survey and synthesis. Journal of Indo-Eu-
ropean Studies. Monograph series number 10.
Stevens, A. 1968. Madurese phonology and morphology. New Haven: American
Oriental Society.
Szemerenyi, 0.1999. Introduction to Indo-European linguistics. Oxford: Claren-
don Press. ·
Vennemann, T. 1989. Phonological and morphological consequences of the 'glot-
talic theory'. In: T. Vennemann (ed.) The new sound of Indo-European: essays
in phonological reconstruction. Berlin: Mouton de Gruyter.

DepartmentofGermanics C h a r l e s M. B a r r a c k
University of Washington
Seattle, WA98195-3130
USA
cbarrack@u.washington.edu

Brought to you by | Cornell University Library


Authenticated
Download Date | 10/17/16 4:53 PM

You might also like