Documente Academic
Documente Profesional
Documente Cultură
1
2
I would like to thank Ryan Sandell, Clara Felisari, Brent Vine, Andrew Byrd, Jesse Lundquist,
Tony Yates, and Christina Skelton for their generous help with various stages of the research
and writing of this paper. Remaining infelicities are of course entirely mine.
For a general overview of the literature on initial yod, see Felisari 2008.
Words in Set B (Gk. ) were supposedly rural, as opposed to the high-class words in Set
A (Gk. /h-/). The inclusion of warriors belt (Hom., 17) among the lower-class rural terms was especially problematic.
Earlier solutions involving laryngeals include PIE *h1/2y- > Gk. h- vs. PIE *h3y- > (Lehmann 1955:74ff., Peters 1980:3 and addenda), and *Hy- > vs. *y- > h- (Hammerich 1948:
7, Beekes 1969:958 and 1995:143, Rix 1976:60, Garca Ramn 1999).
Chiara Bozzone
The late development of *y- > h- would be mirrored in the oscillations of Mycenaean spellings
for the sentence-initial element jo-/o- (etymologically built to the stem of the relative pronoun); see Bartonk 2003:139. This alternation, however, has been persuasively treated as
morphological by Probert 2008, while the phonetics remain speculative. As we shall see in
4.3 below, the Mycenaean evidence on this point seems too scanty and problematic to be decisive.
The standard formulation of the law is problematic and has many counterexamples (see Piwowarczyk 2008). In reassessing the evidence, Byrd is able to
identify a more precise environment for the change: only tautosyllabic *-h2/3yclusters reliably undergo laryngeal loss. Within this narrower environment, Byrd
is also able to provide a phonetic justification for the loss of laryngeal. In close
coarticulation with a *y (as would happen in the environment for the law), segments tend to palatalize. Not all segments, however, can equally undergo palatalization: this is true for the PIE laryngeals as well. In this environment, Byrd
explains, we have typological reasons to expect that a glottal segment (*h1)
would behave differently from a pharyngeal (*h2 and *h3). Byrd argues that *h2
and *h3 are deleted before tautosyllabic *y, since the palatalization of a pharyngeal sound is virtually impossible to produce and is not found crosslinguistically.5 By the same account, however, *-h1y- clusters would remain,
since the palatalization of a glottal sound ([h] or []) is not as problematic (the
results of such a palatalization will concern us below in 1.3). Heterosyllabic sequences of *-H.y- would not undergo the change, since the syllable boundary
would prevent close coarticulation.
Byrds solution to Pinaults Law has important implications for our problem:
first, if we extend Byrds account to word-initial tautosyllabic *Hy- sequences,
we have a phonetic motivation for laryngeal deletion before a palatalizing sound
(which could have occurred already in PIE, if it is analogous to Pinaults Law),
but only within the sequences involving *h2 and *h3. Second, if we accept this
extension, we should reconstruct all *Hy- sequences that yield Gk. h- as *h1y-;
*h2/3y- sequences would have lost their laryngeal.
1.3. The development of PIE *h1y- in Greek
One could thus rewrite the rule accounting for Set A above as PIE *h1y- > Gk.
h-. While I will survey evidence in support of this rule in 1.4.2 below, I shall
now focus on the phonological development that would lead from a PIE sequence
*h1y- to Greek aspiration.
The normal way to explain the development of *h1y- into a glottal fricative
[h], as per Rau (2010:1767), is to drop the laryngeal early on and to lenite the
newly word-initial y- into h- at some later stage. There are two problems with this
account:
See the references in Byrd 2011 to the phonetic literature, especially Bessell 1993.
Chiara Bozzone
a. Motivation for the sound change: why would *h1 drop in this environment?
b. Phonetic details of the sound change: why would a [j] lenite into [h]?
1.3.1. motivating the sound change
Let us focus on (a) first. In the consensus view, *h1 has to drop before laryngeal
vocalization;6 otherwise, it would vocalize like every other *HC- sequence.
But if a PIE *h1y- sequence was tolerated so far, and if laryngeals are still present
in Proto-Greek, what triggers the laryngeal loss? There does not seem to be a
good reason for *h1- to drop only initially and only before *y at this stage of
Greek.
If instead of focusing on the laryngeal we focus on the *y, we may find a
likely explanation. In Proto-Greek, *y was gradually being eliminated from a variety of environments, which often resulted in widespread palatalizations of adjacent segments (still ongoing in Mycenaean). In first-millennium Greek, this
resulted in the complete elimination of the phoneme /y/, which can be seen as a
conspiracy or as a snowballing phenomenon. Several second-millennium sound
changes were involved in the process:
1. *y- > an affricate, later spelled (also part of our story; the change is
complete in Mycenaean)
2. Palatalization of stops and resonants before y (still ongoing in Mycenaean)
3. Loss of intervocalic y (still ongoing in Mycenaean)
These changes, as mentioned above, are complete in first-millennium Greek.7
While (1) eliminated word-initial *y-, (2) and (3) took care of word-internal occurrences. In Classical Greek, [j] survived only as an allophone of /i/.
I believe that the Proto-Greek treatment of *Hy- (and likely *Hi-) belongs to
this larger pattern. Specifically, I argue that the loss of /y/ resulted in the palatalization of the adjacent laryngeal in the sequence*h1y-. This palatalized version of
*h1-, not *y- on its own, is what yields Greek aspiration.
6
7
By which I mean loss of the laryngeal and concomitant phonologization of the suitably colored
phonetic schwa that accompanied it. See n.15 below.
Hajnal (2006:128) conjectures that the /y/ phoneme was already lost in late Mycenaean times.
Fig. 1. The Weakening Hierarchy (after Hock 1991:83).
The commonly adduced parallel of word-internal *-s- > -h- > -- behaves differently. Upon loss, internal *-h- < *-s- leaves traces in the form
of:
Word-internal *-y-, upon loss, does not leave any such traces.8
If word-internal *-y- did not in fact become a glottal fricative, the parallel
with *y- falls away. And while word-initial *y- could undergo fortition and yield
8
For apparent Mycenaean evidence of *-y- > -h- see 4.3 below.
Chiara Bozzone
Gk. 9 swarm (of bees) < PIE *h1yeh1- throw, send. The initial
laryngeal in the root is guaranteed by metrical considerations (cf. Peters
1976 on the scansion of ).
Gk. holy < PIE *h1yeh2- worship (with Lubotskys Law) or
*h1ya- (LIV). The -i- vocalism of the reduplicant in the Sanskrit perfect
j supports the laryngeal. The form must be from *Hi-Hih2-, since a
root *yeh2- would have yielded *ya-ih2- > **yeje (LIV s.v.).10 Note also Riekens derivation of HLuv. i-zi-ya- do, make < *Hig-y/- (2007),
which supports initial *h1-, since initial *h2- and *h3- would not be lost in
Anatolian before syllabic i, as in Hitt. imma- image, copy < PIE
*h2im- and ia- hitch-pole < PIE *h2/3ih1/3- (Craig Melchert, p.c.).
Gk. battle < PIE *h1yewd- move. Several traces of the laryngeal (in non-initial position) are found in Indo-Iranian: Ved. yyudhir
bellicose (RV 10.149.4a), 3pl.perf. yyudhur (RV 3),11 amitr-ydhfighting enemies (RV 3.29.15a), YAv. asp-iiaoa- fighting on a
horse, fr-iiaoa- fighting in front (PN). See also Ittzs 2012:93 and
n.27.
1.4.2. supporting evidence for Rule [2] (PIE *h2/3y- > PIE *y- > Gk. )
There also exist proposed reconstructions with *h2/3y- that support Rule [2]
above. Of these, one has clear evidence for *h3:
Gk. west wind (and perhaps darkness) < PIE *h3yebenter. Compare futuo < *h3e-h3ib- (Cheung 2007:175), Skt.
ybhati futuit.
Two more have some evidence for a laryngeal, which we could write as *h2/3:
9
10
11
12
Gk. yoke < PIE *h2/3yewg- yoke. Support for the laryngeal:
Ved. yunak (3sg.act.impf.), yukta (3sg.mid.aor).12
Chiara Bozzone
Gk. grains < PIE *h2/3yew- grain?. Support for the laryngeal:
Ved. syva- having good pastures and derivatives, e.g., yavasawithout pastures (probably the name of a king); see the discussion in
Ittzs 2012:989 and n.42.
The above etymologies were originally used to argue for PIE *Hy- > Gk. (see
the references in n.3), but they can now be incorporated into my account.
1.4.3. counterevidence to Rule [2]
To my knowledge, the only reconstruction that is incompatible with Rule [2]
above is Weisss (1994) etymology of healthy, beneficial (from the zero
grade of *h2y-u time of living and the root *gih3- live), i.e., < PIE
*h2yu-gih3-s. Weiss compares Cypriote Gk. u-wa-i-se /za-ne forever and ever, Lat. igis everflowing, Av. yauua living forever, Goth.
ajukds eternity.
Weiss himself does not present this equation as conclusive, though the cognate set is striking. One may perhaps object only in terms of semantics, since
healthy, beneficial (in the Greek form) does not seem to contain the notion ever (precisely the *h2yu- component, which seems to work adverbially in all of
the examples above) that is prominent in the rest of the set.13 Thus the older etymology, with the root for good (PIE *h1su-) in the first member and loss of the
laryngeal in the compound14 remains as an equally viable option, at least if one
wants to accept Rule [2] above. It is interesting to note that both etymologies
were proposed in 1892 by Saussure, who favored the *h1su- solution. Semantically, having good health seems unobjectionable.
13
14
phenomenon he dubs embarrassing if we reconstruct the root without laryngeal. (In his survey, brevis in longo is just as frequent before yodh- as it is before yug-.)
Weiss (1994:151) justifies the difference in semantics by noting that in Greek as opposed to
Avestan and possibly also Latin, the first member of the compound has not been replaced by
the dative, or directive, which we have hypothesized above was the original locus of the meaning eternity. Yet Weisss analysis of the Germanic form (1489) does not mention such a
replacement, and the meaning eternity is still present; this suggests that the replacement in
Avestan could be just a surface renewal of the first member of the compound, which preserved
its original meaning for life, for eternity. Nor is the proposed meaning for the Greek compound (having a life which is with vitality (to it)) as clear as one would like. One may conclude that the cognate set without Gk. is phonologically more regular and semantically
more cohesive, and remains an important discovery on Weisss part.
If one is willing to accept the loss of laryngeal in compound, then Weisss etymology itself
becomes problematic, for without the initial laryngeal we would expect ** instead.
2. Gk. horse and initial *HiThis discussion also bears on the reconstruction of the notoriously problematic
Greek word for horse. In treating the etymology of , de Vaan (2009:203)
comments: the only element still defying an explanation is the initial h-, which
is not found in the PN -. Ruijgh (1995:355) explains it from analogy
with chariot. De Vaan explains the i-vocalism in the root by schwa
epenthesis,15 after Vine (1999), a phenomenon situated in Pre-Greek. The zero
grade in Greek is explained as deriving from a Late PIE (= Core PIE) thematization of the original u-stem noun (which may be preserved in Hittite as the form
underlying the paradigm of ANE.KUR.RA-u), based on the stem-form of the
genitive singular.
In this section, I focus on explaining the initial aspiration. (Discussion of the
derivational history of the word is beyond the scope of this paper.)
2.1. A new rule for *h1iInterestingly, the aspiration in the root for horse in Greek correlates systematically with:
16
17
10
Chiara Bozzone
(though not identical) to the cases of *h1y- above, and could be used to explain
the aspiration. Specifically: this environment likewise involves a sequence of *h1
plus a palatalizing segment (but in this case *i rather than *y). Postulating a glide
between the laryngeal and the vowel (i.e., *h1i-) would make the parallel even
more precise. If we accept this parallel, we could then explain the initial aspiration as lautgesetzlich rather than analogical.
The following development can therefore be posited:
1. Late PIE: *h1w-o2. > Pre-Greek schwa epenthesis (before laryngeal vocalization): *h12.wo- > *h1i.wo3. > Proto-Greek *h1 palatalization: *hi.kkwo- > Gk. /hppos/
If one accepts this etymology, one can then formulate the following rule:
Rule [4]: Pre-Greek *h1i- > Gk. hi- (likely via the intermediate stage ProtoGreek *hi-).
This is entirely parallel to Rule [3] (in 1.4), i.e., PIE *h1y- > Gk. h-.
2.1.1. evidence for Rule [4]
The best supporting evidence for Rule [4] proposed above is found in the following item:
Gk. send < PIE *h1yeh1- throw, send: PIE *h1i-h1yeh1-mi > ProtoGreek *h-(y)-mi > . The initial aspiration is otherwise unexplained
in the reduplicated forms (it would have to be analogical to the aorist). In
my account, the aspiration would be lautgesetzlich.
There is one piece of possible additional supporting evidence for Rule [4], which
has various proposed etymologies:
Gk. (Old Attic h-, cf. 2.2.1 below) oven < ? PIE *h1eptake. A possible derivation would be: PIE *h1p-n- > Pre-Greek *h12.pno- > *h1i.pno- > Gk. /hipns/; but the semantics are not very satisfactory. I shall discuss this word below in 2.2.1.
There are three more pieces of evidence that are compatible with Rule [4]
above, where Grassmanns Law arguably hides the expected aspiration. Of these
forms, two have independent evidence for *h1:
Gk. be! < PIE *h1es- be, i.e., PIE *h1s-di > Pre-Gk. *h12s-di >
Proto-Gk. *his-ti > *his-ti > Grassmann, whence Gk. /isti/. The rest of
11
Gk. fish, i.e., PIE *h1duH- > Pre-Gk. *h12duH- > Proto-Gk.
*hik-t- > *hik-t- > Grassmann, whence Gk. /ikts/. The laryngeal
must have been present for the schwa secundum to be inserted; without
laryngeal, we would expect .
Finally, there is one apparent piece of counterevidence against Rule [4], for
which paradigmatic leveling can nevertheless be reasonably invoked:
Gk. , , , zero-grade forms of go < PIE *h1ey-go. The expected aspiration in **hmen, **hte, **hsi could have been lost on the
model of the singular (/emi, e, esi/; and so also in modal forms).
From this survey of the evidence, we can conclude that Rule [4] is justified.
2.2. What about *h2i- and *h3i-?
Rule [4] above makes the treatment of *h1i- parallel to the treatment of *h1y-, in
that both sequences develop aspiration in first-millennium Greek. What about *h2
and *h3 in these same environments? In Rule [2], I proposed that *h2/3 are lost
before *y by the time of Proto-Greek (because, following Byrd 2011, pharyngeals
cannot be palatalized). In this section, I propose that in Proto-Greek, *h2/3 are lost
before *i as well, for the same phonetic reason.
Indeed, the observation by Peters (1980:113ff.) that laryngeals fail to vocalize before *i ( vs. , cf. 1.1 above) can be explained under this proposalexcept that I would restrict this mechanism to *h2 and *h3, since Peterss
evidence is only for *h2i-, while initial sequences with *h1i- are covered by Rule
[4] above. I thus propose the following rule for *h2/3i-:
Rule [5]: Pre-Greek *h2/3i- > Proto-Gk. *i-.
Under this proposal, the same symmetry holds between *h1y- and *h1i- sequences
as between *h2/3y- and *h2/3i- sequences in Greek: the first group develops aspiration, the second loses the laryngeal.
12
Chiara Bozzone
Gk. cheerful, bright < PIE *h2eyd- kindle, i.e., PIE *h2id- >
Proto-Greek *it-. Compare the full grade in clear sky.
Cypriot Gk. wounded (ICS 217.3, spelled i-ki-ma-me-no-se)
< PIE *h2ey- pierce. Compare the full grade in spear point,
spear (1.3.2).
Gk. sharp; epithet of wind in Homer. This adjective
does not have an established etymology and is normally translated as favorable; in the epics, it is routinely said of a wind sent by a god to help
in navigation (as in Il. 1.479). Because folk etymology associated the adjective with the verb come, Herodian (On the Prosody of the Odyssey, , 7) insists on its lack of initial aspiration:
. The connection with PIE *h2ey- pierce seems
convincing both semantically (a sharp wind) and phonologically, and
would support Rule [5].
Gk. footstep, track (of an animal) < PIE *h3eyg- go forth. The
connection to the PIE root *h3eyg- is normally presented as doubtful; yet
there is no counterevidence, and the semantics are convincing. Note that,
although *h3 is suggested by LIV, *h1 is also possible if an o-grade is involved. In this case, we could reconstruct the root as *h1eyg-, and Gk.
could be evidence for Rule [4] instead (with Grassmanns Law).
13
Gk. take, seize < PIE *h2ey- id.. The present stem shows the
full grade instead of the expected zero grade, which should have come
out as **; the parallel with Ved. inti drive, send seems particularly telling. The verbal noun *-, taken as the basis for ask,
also does not show the expected zero grade.
go, walk < PIE *h3eyg- go forth. The expected zero-grade
form should have been ** (compare , 2.2.1 above). The full
grade was probably restored on the basis of pres. be gone.
swell < PIE *h2eyd- . The zero grade *h2id-nw/nu- should have
given ** (- verbs normally involve late creations). The -nupresent is supported by Arm. aytnowm swell, also with secondary full
grade of the root (LIV s.v.). In Greek, a non-analogical full grade would
have given **-. Here the present most likely served as a model.
will bear < PIE *h3eyt-s- fetch. The expected zero grade is not
visible in the verbal adjective bearable. The *h3 is supported by
14
Chiara Bozzone
Lat. tor (Tichy 2004) and the Luwian verb izza(i)- fetch (Melchert
2007).
One could take the evidence above to argue for PIE *h2/3i- > Gk. a/oi- instead
(that is, one could say that these are zero grades, and that the laryngeals vocalize
before *i). There are two problems with this interpretation:
PIE *h1y- > Gk. hPIE *h2/3y- > Proto-Gk. *y- > Gk.
PIE (or Pre-Greek) *h1i- > Gk. hiPIE (or Pre-Greek) *h2/3i- > Gk. i-
in terms of manner of articulation, [h] would pattern better with the other
laryngeals, which are also fricatives;
15
I believe that *h1 [h] itself was lost in Proto-Greek, but that its palatalized variant,
phonemicized as /h/, remained much longer and eventually evolved (in initial
position) into Greek aspiration.
In phonetic terms, it is likely that a palatalized glottal fricative /h/ would be
realized as [] (palatal fricative). By the first millennium (but not so in Mycenaean), this sound had merged completely with Gk. h-, which in turn was the result
of the lenition of PIE *s. To understand the merger, I shall now turn to the phonetics of PIE *s.
4.2. The phonetics of PIE */s/
We know that PIE *s was a voiceless sibilant. In terms of place of articulation,
Vijnas (2010) has made an interesting case for PIE *s being post-alveolar,
namely:
+continuant
+coronal
[] =
+strident
spread glottis
I believe that the evolution of PIE *h1y- in Greek can offer support to this view.
Following Vijnas 2010, the lenition of this inherited [] in Greek would have
yielded something more constricted than plain [h]. One can write this sound as:
+continuant
(same as [], but without the stridency).
[] = +coronal
spread glottis
If we compare the result of our palatalized *h1 and the result of lenited PIE *s,
they seem indeed very close in terms of features (shared features are italicized):
+continuant
+dorsal
+coronal
[] =
+high
+front
spread glottis
+continuant
[] = +coronal
spread glottis
These two sounds are somewhat distant from Greek aspiration as we usually conceive of it:
16
Chiara Bozzone
+continuant
dorsal
[h] =
coronal
+spread glottis
The [spread glottis] and [+CORONAL] features, resulting in a more turbulent
sound, seem to be what unites our candidates and opposes them to plain [h].
These shared features could explain two things about these two segments:
Their gradual merger can be observed in the spellings from Mycenaean to alphabetic Greek. I shall come back to the phonetics of Gk. /h/ in 4.4 below.
4.3. PIE *h1y- and lenited *s in the Mycenaean writing system
As far as one can tell, [] and [] were still partially distinct in the Mycenaean
writing system. The areas of overlap in the spelling are due both to the limitations
of the script and the closeness of the sounds. In Linear B:
[] had a distinct spelling only in the sequence /ha/, often spelled with
the sign a2. /ho/ was spelled like the plain vowel /o/, with the sign o,
and similarly for /he/ via e;
[] in the sequence /ho/ seemingly had two alternative spellings, jo and
o. For the sequence /ha/, the spelling seems to be both ja and a2
(and perhaps a as well). A possible example for /he/ (a-pe-e-ke [PY An
724], if /apehke/ to ) is spelled with e.
Yet the evidence for this last point is slim and debated; we have two sources for
the writing of initial h- [] in Mycenaean:
the sentence-initial particle /yo-/, which has two distinct spellings, o and
jo;
the obscure forms ja-ke-te-re (PY Mn 11.2) and a2-ke-te-re (KN V 118),
often taken as alternative spellings of a-ke-te-re (PY Jn 832, 2) workers of raw materials (see Skelton 2013 for a thorough discussion).
If taken together, these two sources indicate a spelling alternation in writing h- <
PIE *h1y-: one spelling involving aspiration (o and a2), and another involving
the etymological yod (jo and ja). This alternation has usually been taken as
diachronic in nature: the spellings with yod would be older (or etymological),
17
while the spellings with aspiration (whether notated or not) would show the later
development of *y- > h-.Yet the connections are fragile.
Probert (2008) has argued that the sentence-initial particles written as jo and
o have distinct syntactic roles: jo appears to be nominative (perhaps masculine
plural), and o accusative (perhaps neuter singular). The spellings are then used
conventionally, and the phonetics involved remain obscure (both particles should
start with the same sound).
Skelton (2013) has argued that ja-ke-te-re and a2-ke-te-re (< PIE *Hyeh2kheal, cure; *h1yeh2k- in my theory) belong with Gk. cure (cf. Old Ionic
H) and should not be connected to a-ke-te-re (which belongs with
Gk. make, produce); this excludes the use of a for writing /ha-/. Furthermore, the chronological sequence for the spellings is the opposite of what the
story above would predict: the spelling with aspiration (a2-ke-te-re) dates from
about 1350 BCE (moreover from Knossos, which may be psilotic), while the etymological spelling with ja (ja-ke-te-re) dates from 1200 BCE (Skelton 2013).
This complicates the account rather than clarifying it.
At the present state of our knowledge, one must admit that the Mycenaean
evidence is not clear enough to guide our discussion. One can, however, try to
reconstruct a scenario that would justify the usages found in our corpus. In order
to do so, I shall make use of the reconstructed phonetic features of each phoneme,
and see how they map onto the graphemes available in Linear B. Phonetically, I
assume that we are dealing with:
+continuant
+dorsal
+coronal
+high
+front
spread glottis
[] =
+continuant
[] = +coronal
spread glottis
+sonorant
+continuant
18
Chiara Bozzone
sonorant continuant
+
+
+
dorsal
+
coronal
+
+
+
+
The only features that are present in [] and absent in [] are [+high, +front,
+dorsal]: these are features that the spellings with jo and ja could capture (in
that these features are contained in the glide [j]), and the spellings o, a, and
a2 could not. We would then predict that jo and ja were used for writing []
and not []. This is indeed what we find.
Now imagine that the scribes were trying to map the phonological features
onto their spelling as closely as possible; the following scenario emerges:
Writing Myc. h- < PIE *h1y-:
18
It is possible, as Probert suggests (2008:162), that the proximity of i (at the end of nom.pl.
masc. /hoi/) increased the chances that the features [+high, +front] would be perceived as salient enough to notate. This could explain why nom.pl.masc. was spelled jo, while nom/acc.
sg.neut. was spelled o.
19
This is indeed what we find in the tablets. To sum up, we can formulate the following principles:
1. the j series is used to spell /h/, and not plain /h/;
2. the plain vowel signs o and e are used to spell both /h/ and /h/; a is
used to spell /ha/, but not /ha/;
3. a2 is used to spell /ha/ and /ha/.
We can conclude that Mycenaean spelling keeps /h/ and /h/ distinct, in that the
former can be notated by the j series, while the latter never is, and that the sign
a is used to write /ha/ but not /ha/.
In word-internal position, however, there are some famous exceptions to
these conclusions (and to point (1) specifically), to which we now turn.
4.3.1. cases of j to spell /h/
The first possible exceptions are two forms with variant spellings that could be
taken to indicate the usage of j to spell word-internal /h/. Importantly (as we
shall see), these are both forms where the aspiration follows the high vowel /i/.
These forms are:
Both forms show a spelling variant without j, and a spelling variant with j
bridging a vowel hiatus. Following Jimnez Delgado (2008:84, 86), I believe that
the variants spell the presence and absence of aspiration, respectively, and that
the j is motivated by the rules of Mycenaean spelling and has no phonemic reality (and thus no etymological connection to /h/).
The glides w and j are regularly used in Mycenaean spelling to bridge hiatus between high vowels and vowels that follow (see Meiner 2007 for a recent
survey of the phenomenon). In these environments, j and w can actually be
said to signal the phonemic hiatus: they indicate that no segment occurs between
the two vowels; the scribes just write the phonetic glide between these vowels.
Conversely, the absence of j and w signals the absence of phonemic hiatus,
and the presence of another segment between the vowels; this segment is /h/,
which is written by the vowel signs e, o, a2, or a.19
19
See Jimnez Delgado (2008:73 with n.2), who calls this phenomenon hiato grfico.
20
Chiara Bozzone
20
As many have pointed out, the morphology of the standard proposal is problematic, in that (a)
- verbal adjectives are not clearly found before 5th-century Greek (see Hutton 1993:127
n.61), (b) they normally employ zero grade of the root, and (c) they are usually thought to belong with the Sanskrit gerundive suffix -tavya-, which is incompatible with the Mycenaean
forms.
21
vowels in general (/i/ and /e/), and not just after /i/.21 Under this assumption, all of
the forms could be accounted for as follows:
We would thus be looking at the neut.gen.sg. and at the neut.nom./acc.pl. of an sstem noun such as KeTes- < PIE *KeT-es- (a solution already explored by Hutton 1993 just for qe-te-a2), perhaps an s-stem to the PIE root *ged- want, desire, which would have given Gk.**, next to desire. This solution
may be less exciting than the traditional one (and may require some additional
work in the interpretation of the texts), but is more straightforward, both morphologically and phonologically.
For our purposes, it is crucial to note that qe-te-a2, in the traditional interpretation, is the only example in which etymological -y- would be unambiguously
spelled as /h/ in Mycenaean, and therefore the only direct evidence of the supposed change *-y- > -h-. Given how problematic the etymology is, we cannot ask
this form to bear much weight.22
One final piece of indirect evidence in support of *-y- > -h- is the form po-nike-a (KN Se 880.2, normally taken as the fem.nom.sg. of adj. purple).
If we take the hiato grfico at face value, we should read /poinke(h)/, with /h/
presumably coming from the suffixal *-y- in *poink-ey. Had there been no /h/
between the vowels (that is, had the *-y- disappeared without trace), we would
have expected the spelling po-ni-ke-ja /poinke/ instead. po-ni-ke-ja, however,
would also have been the spelling for the form with the suffixal /y/ preserved,
namely /poinkey/; we actually find this spelling in KN Ln 1568, where it most
likely designates the profession: po-ni-ke-ja /poinkeia/ worker in purple
(Killen 2004:230). Such ambiguity may have led scribes wanting to mark the
21
22
Phonetically, that a glide would be present is easy to imagine; unfortunately, this form is precisely the example that is normally used to establish this ruleand even more unfortunately,
such a spelling is not otherwise found in the gen.sing. of s-stems (Hutton 1990:130 n.76).
Moreover, if additional secure forms showing a2 as a spelling for /ya/ were to surface, one
could not exclude the possibility that the overlap between a2 and ja for writing word-initial
/ha/ mentioned above could have led some speakers to use a2 as an allograph for ja even
when the phonetics did not support the equation (that is, when writing simple /ya/). At present,
however, I am not aware of any such additional forms.
22
Chiara Bozzone
+continuant
+dorsal
[] = +low
+back
spread glottis
This fact has been noticed several times (Allen 1987:534, Woodard 2010:31),
but has not been explained on a linguistic basis.
Interestingly, the Phoenician sound [] in particular shares the features
[+dorsal, spread glottis] with [], and [spread glottis] with [], above, in contradistinction to a glottal fricative [h], which is [dorsal, +spread glottis]; this
could provide a satisfactory explanation as to why it would have been chosen
over [h] to represent Greek aspiration. The most likely explanation for the merger
is for [] to have lost the [+high, +front] features, which would bring it closer to
both [] and [].
All in all, the above considerations suggest that we should imagine Greek aspiration as a more turbulent sound than [h], and that /h/ [] and /h/ [] converged
into a single phoneme only in post-Mycenaean times.
23
The fact that the scribe did not explicitly mark the aspiration between the vowels with a2 is
not in itself probative, given the near-absence of a2 at Knossos.
23
References
Allen, W. Sydney. 1987. Vox Graeca: The Pronunciation of Classical Greek. Cambridge:
Cambridge University Press.
Bakker, Egbert (ed.). 2010. A Companion to the Ancient Greek Language. Oxford:
Wiley-Blackwell.
Bartonk, Antonin. 2003. Handbuch des mykenischen Griechisch. Heidelberg: Winter.
Beekes, Robert S. P. 1969. The Development of the Proto-Indo-European Laryngeals in
Greek. The Hague: Mouton.
. 1995. Indo-European Linguistics: An Introduction. Amsterdam: Benjamins.
. 2010. Etymological Dictionary of Greek. Leiden and Boston: Brill.
Bessell, Nicola. 1993. Towards a Phonetic and Phonological Typology of Post-Velar Articulation. Ph.D. diss., University of British Columbia.
Billigmeier, Jon-Christian. 1976. The Origin of the Dual Reflex of Initial Consonantal
Indo-European *y in Greek. Journal of Indo-European Studies 4.22131.
Brixhe, Claude. 1996. Phontique et phonologie du grec ancien. Louvain-la-Neuve:
Peeters.
Byrd, Andrew. 2011. A Fresh Look at Pinaults Law. Paper presented at the 23rd Annual
UCLA Indo-European Conference.
Cheung, Johnny. 2007. Etymological Dictionary of the Iranian Verb. Leiden: Brill.
Colvin, Stephen. 2006. Autosegmental Phonology and Word-Internal -h- in Mycenaean
Greek. Glotta 82.3654.
Felisari, Clara. 2008. Levoluzione in greco del fonema indoeuropeo jod allinizio di parola. B.A. thesis, Universit Cattolica del Sacro Cuore, Milan.
Garca Ramn, Jos Luis. 1999. Griechisch : , vedisch ytrRcher und die Vertretung von *i - im Griechischen. In Heiner Eichner and Hans
Christian Luschtzky (eds.), Compositiones Indogermanicae in memoriam Jochem
Schindler, 7796. Prague: Enigma.
. 2006. La fragmentacin dialectal griega: Limitaciones, posibilidades y falsos
problemas. Incontri Linguistici 29.6182.
Greppin, John A. C. 1978. On Greek Zeta. Journal of Indo-European Studies 6.1412.
Hackett, Jo Ann. 2004. Phoenician and Punic. In Roger D. Woodard (ed.), The Cambridge Encyclopedia of the Worlds Ancient Languages, 36675. Cambridge: Cambridge University Press.
Hajnal, Ivo. 2006. Grammatik des mykenischen Griechisch. http://www.uibk.ac.at/
sprachen-literaturen/sprawi/mykgr.html.
Hammerich, L. L. 1948. Laryngeal Before Sonant. Copenhagen: Munksgaard.
Hamp, Eric P. 1997. Indo-European Initial Yod in Greek. In Alexander Lubotsky (ed.),
Sound Law and Analogy: Papers in Honor of Robert S. P. Beekes on the Occasion of
his 60th Birthday, 914. Amsterdam: Rodopi.
24
Chiara Bozzone
Hutton, William. 1993 [19901]. The Meaning of qe-te-o in Linear B. Minos 256.105
31.
Hock, Hans Henrich. 1991. Principles of Historical Linguistics2. Berlin: de Gruyter.
Ittzs, Mt. 2012. Initial y in the Rigveda. In Roman Suka and Ondej efik (eds.), The
Sound of Indo-European 2: Papers on Indo-European Phonetics, Phonemics and
Morphophonemics, 86115. Munich: LINCOM.
Ivanov, Vyacheslav V. 1979. Razyskanija v oblasti anatolijskogo jazykoznanija, 10. Xett.
appina plamja oaga : dr.-isl. ofn oag. timologija 1977.145.
Jimnez Delgado, Jos Miguel. 2008. La situacin de *H en griego micnico. Kadmos
47.7390.
Killen, John. 2004. Names in -e and -e-u in Mycenaean Greek. In John H. W. Penney
(ed.), Indo-European Perspectives: Studies in Honor of Anna Morpurgo Davies,
21735. Oxford: Oxford University Press.
Kloekhorst, Alwin. Forthcoming. Proto-Indo-European Thorn-Clusters. Historische
Sprachforschung.
Kmmel, Martin. 2000. Das Perfekt im Indoiranischen. Eine Untersuchung der Form und
Funktion einer ererbten Kategorie des Verbums und ihrer Entwicklung in den altindoiranischen Sprachen. Wiesbaden: Reichert.
Lehmann, Winfred P. 1955 [1969]. Proto-Indo-European Phonology. Austin: University
of Texas Press and Linguistic Society of America.
Lejeune, Michel. 1964. Sur quelques termes du vocabulaire conomique mycnien. In
Emmett L. Bennett Jr. (ed.), Mycenaean Studies. Proceedings of the Third International Colloquium for Mycenaean Studies Held at Wingspread, 48 September
1961, 77109. Madison: University of Wisconsin Press.
LIV = Rix, Helmut, et al. 2001. Lexikon der indogermanischen Verben2. Wiesbaden:
Reichert.
LGPN = Various. 19872010. Lexicon of Greek Personal Names. Oxford: Oxford University Press.
Malzahn, Melanie. 20012. Zum participium necessitatis des Griechischen.
Studia Minora Facultatis Philosophicae Universitatis Brunensis 67.13541.
Masson, Olivier. 1983. Les inscriptions chypriotes syllabiques: Recueil critique et comment. Paris: de Boccard.
Meiner, Torsten. 2007. Notes on Mycenaean Spelling. The Cambridge Classical Journal
53.96111.
Melchert, H. Craig. 2007. Luvian Evidence for *H3eit- take along; fetch. UCLA IndoEuropean Studies Bulletin 12.13.
Morpurgo Davies, Anna. 2012. Phonetic Laws, Language Diffusion, and Drift: The Loss
of Sibilants in the Greek Dialects of the First Millennium BC. In Philomen Probert
and Andreas Willi (eds.), Laws and Rules in Indo-European, 10224. Oxford: Oxford University Press.
25
Oldenberg, Hermann. 1909. Rgveda. Texkritische und exegetische Noten. Erstes bis sechstes Buch. Berlin: Weidmann.
Peters, Martin. 1976. Attisch . Die Sprache 22.15761.
. 1980. Untersuchungen zur Vertretung der indogermanischen Laryngale im
Griechischen. Vienna: Verlag der sterreichischen Akademie der Wissenschaften.
Pike, Moss. 2006. Mycenaean qe-te-jo and the Greek Verbal Adjective in -. Paper
presented at the 25th East Coast Indo-European Conference, Columbus, OH.
Pinault, Georges-Jean. 1982. A Neglected Phonetic Law: The Reduction of the IndoEuropean Laryngeals in Internal Syllables before Yod. In A. Ahlqvist (ed.), Papers
from the 5th International Conference on Historical Linguistics, 26572. Amsterdam: Benjamins.
Piwowarczyk, Dariusz. 2008. The Treatment of Laryngeals before Yod in Indo-European
Verbs. M.A. thesis, Leiden University.
Probert, Philomen. 2008. Mycenaean o- Is Accusative; jo- Is Nominative. Glotta 84.126
68.
Rau, Jeremy. 2010. Greek and Proto-Indo-European. In Bakker 2010, 17188.
Rieken, Elisabeth. 2007. Hieroglyphen-luwisch i-zi-ia-: Ein Beitrag zur Rekonstruktion
der urindogermanischen Kulturgeschichte. In Wolfgang Hock and Michael MeierBrgger (eds.), Dar Slovesny: Festschrift fr Christoph Koch zum 65. Geburtstag,
26375. Munich: Sagner.
Rix, Helmut. 1976. Historische Grammatik des Griechischen. Darmstadt: Wissenschaftliche Buchgesellschaft.
Ruijgh, Cornelius. 1995. Observations sur les voyelles dappui en proto-indo-europen et
en grec ancien. In Wojciech Smoczyski (ed.), Kuryowicz Memorial Volume, Part I,
34556. Krakow: University of Krakow.
Sapir, Edward. 1938. Glottalized Continuants in Navaho, Nootka, and Kwakiutl (with a
Note on Indo-European). Language 14(4).24874.
de Saussure, Ferdinand. 1892. Varia: . Mmoires de la Socit Linguistique de Paris
7.8990.
Sihler, Andrew L. 1995. New Comparative Grammar of Greek and Latin. Oxford: Oxford
University Press.
Skelton, Christina. 2013. Thoughts on the Initial Aspiration of HAKESANDRO. In Jenny
Strauss Clay, Irad Malkin, and Yannis Tzifopoulos (eds.), Proceedings of the Conference: Graph in Late Geometric and Protoarchaic Methone, Macedonia (ca. 700
BCE). Berlin: de Gruyter.
Tichy, Eva. 2004. Gr. , Lat. t und die Mittelzeile der Duenos-Inschrift. Glotta
78.179202.
de Vaan, Michiel. 2009. The Derivational History of Greek and . Journal of
Indo-European Studies 37.198213.
Vijnas, Aurelijus. 2010. The Indo-European Sibilant */s/. Historische Sprachforschung
123.4055.
26
Chiara Bozzone
Vine, Brent. 1999. Greek root and Schwa Secundum. In Vyacheslav Ivanov and
Brent Vine (eds.), UCLA Indo-European Studies I, 530. Los Angeles: UCLA Program in Indo-European Studies.
Weiss, Michael. 1994. Life Everlasting: Latin igis everflowing, Greek
healthy, Gothic ajukds eternity, and Avestan yauua- living forever. Mnchener Studien zur Sprachwissenschaft 55.13156.
Willi, Andreas. 2001. Lateinisch iubere, griechisch euths und ein indogermanisches
Rechtskonzept. Historische Sprachforschung 114.11746.
Woodard, Roger D. 2010. Phoinikia Grammata: An Alphabet for the Greek Language.
In Bakker 2010, 2546.
Wyatt, William F. 1968. Early Greek /y/. Glotta 46.22937.
. 1976. Early Greek /y/ and Grassmanns Law. Glotta 54.111.