Wrapping up: remaining issues

We are confident that the discussion presented above accurately represents the facts of velar palatalization as a regular phenomenon in Turkish, and as we can back it up using examples from the speech of the speaker who recorded our data. If we look hard enough, we will find that

Velars in word initial "clusters"

While dorsals occurring as the second C in a word-final consonant cluster are colored by the preceding vowel (e.g. sirc, fark in (14)), the mirror image situation is not evident when a dorsal occurs as the first C in a word-initial cluster. Native Turkish words do not have initial C clusters, but Turkish has borrowed a number of foreign words with initial sequences consisting of a stop followed by a sonorant consonant. Several examples are given in (26). Our Turkish consultant, Dr. Erdener, reports that initial consonant sequences in words such as these can be pronounced as clusters, as in (26a), but that the alternative pronunciations in (26b), in which the consonants of the borrowed sequence are pronounced with an intervening high vowel (generally [ɨ], resulting in disyllabic forms) is more natural. The examples in (13) show that only velars (no palatals) occur in initial clusters, regardless of the backness of the following vowel. For this reason, initial "clusters" cannot be treated as the "mirror image" of final clusters, even when the consonants of the clusters are identical (compare borrowed [kr] and native [rk]).


a. Initial Cluster b. Epenthetic V
gri gɨri 'grey'
kral kɨral 'king'
grup gɨrup 'group'
krem kɨrem 'cream'

In sum, there is some question as to whether word-initial consonant sequences such as those in (13) should be considered clusters at all, since they are generally separated by a vowel on the surface (even though it is not represented orthographically). Since this vowel is the back unround vowel [ɨ], the fact that only velars appear in this environment is consistent with the distribution of velar stops as stated so far (see the discussion associated with figure (2)). For these reasons, we will not treat cases such as those in (13) as exceptions.

Palatals in velar environments

In this module, we have argued that:

  • Velar and palatal stops are in complementary distribution
  • The palatal set are derived allophones of velar phonemes

This analysis predicts that:

  • Velar allophones should never appear in palatal environments
  • Palatal allophones should never appear in velar environments

The generalization that velars do not surface in palatal environments seems to be right - as we expect, if palatals are derived from velars in syllables containing front vowels. However, it turns out that the generalization that palatals never occur in velar environments isn't quite surface true. Clements and Sezer (1982) provide minimal and near minimal pairs showing that both palatals and velars can occur in a syllable containing the vowel [a]. Examples appear in (2).


kar 'snow' car 'profit'
gaz 'gas' ɟavur 'infidel'

Normally, when we find two sounds in a single environment, we analyze them as belonging to phonemes that contrast in the language. Since velars are excluded from palatal environments, we still need the rule in (25) that derives palatal allophones of the velar phonemes. In this scenario, palatal allophones have two sources: underlying velars and underlying palatals. We might be forced to conclude that velars and palatals are not in complementary distribution, as we argued above, but are rather in neutralization distribution. The "phoneme-to-allophone map" under this alternative scenario is shown in (28).

(28) /k/ /c/ /g/ /ɟ/
[k] [c] [g] [ɟ]

However, before embracing the alternative scenario in (28), consider that things are not as straightforward as they seem. It seems that syllables with [a] is the only environment in which both palatals and velars surface, and it seems that there are not many examples of palatals occurring in the velar context. Here are some things for you to think about:

  • Is the existence of a few pairs such as those in (27) sufficient to justify assigning phonemic status to palatal stops?
  • If Turkish has palatal stop phonemes, would we (or would we not) expect to find them in all environments in which velars can surface?
  • What are the "pros" and "cons" of the complementary distribution analysis we presented earlier?
  • What are the "pros" and "cons" of the neutralization alternative sketched here?

For a more extensive discussion of the disharmonic cases of palatal stops, as illustrated in (27), see Clements and Sezer (1982).

prev | top