The velar-palatal alternation


On the sound inventory page for Turkish, the inventory of consonants shows two velar stop phonemes, /k g/. What this inventory does not show, however, is that in addition to the sounds [k g], Turkish also has two palatal stops [c ɟ], at least on the surface. Velar and palatal stops are articulatorily quite similar: together, they form the natural class of dorsal stops, so named because the primary articulator for these stops is the dorsum, or back of the tongue. The chart in (1) provides phonological feature specifications for the obstruents (stops, affricates, and fricatives) of Turkish. As (1) shows, only the feature [back] distinguishes velars and palatals, with the velars representing [+back] and the palatals, [-back].

(1) Phonological features for the obstruents of Turkish.

Articulator LAB COR DOR
p f t s ʃ c k
b v d z ʒ ɟ g
[anterior] + + - _
[distributed] + - - - + +
[back] - +

Notes: All obstruents are [-sonorant]. LAB, COR, DOR represent Labial, Coronal, and Dorsal articulator nodes, respectively.

Given the way we set up the phoneme inventory initially, it will be clear that we assume only velar allophones. Still, when a student of linguistics encounters a language whose sound inventory has both velar stops [k g] and palatal stops [c ɟ] on the surface, how would she decide whether the velar-palatal place difference is contrastive, or whether one set, velars or palatal stops, is derived from the other in a special context? If we find two sounds such as the bilabial nasal [m] and the voiceless alveolar stop [t] in a language, we can safely assume that they belong to distinct sound categories (phonemes), because [m] and [t] differ in many ways (voicing, manner of articulation, place of articulation), and it is hard to imagine one being derived from the other as the result of a simple, natural sound process. We can't make this assumption as easily for the set [k c g ɟ], because palatals and velars form what the late descriptive linguist Kenneth Pike once called "suspicious pairs". "Suspicious pairs" are pairs of sounds that should be given special attention in working out the phonemic inventory of a language because they are phonetically similar, and/or because they commonly interact in phonological processes found in human languages. Grouping the dorsals by voicing, we may view the voiceless set [k c] as a suspicious pair; the voiced dorsals [g ɟ] would likewise be a suspicious pair. We are suspicious when we find both dorsals and velars in the sound inventory of a language, because sound processes that derive one from the other (usually palatals from velars) are typologically very common, although the conditioning factors can be quite different from language to language.

Sound processes deriving palatals or alveopalatals from velars have been found in suchdiverse languages as Bantu varieties (e.g. Kinyarwanda and Lamba); English; historically, in French; dialects of Brazilian Portuguese; and Tupi-Guarani languages such as Guarayu; among many others.

When we look closely at Turkish, we find that the distribution of the velars and palatals is not fully independent. In the rest of this section, we develop an argument that the velar - palatal difference is not contrastive, and that palatals are derived allophones of the velar stops in aparticular context. Developing an analysis, as with any problem solving activity, is a process, and in this section we develop our argument in such a way as to reflect this process. The steps we take when solving a phonological problem can be described as follows:

  • As a first pass, we work out the distribution of our targets (the sounds we are analysing) in a preliminary set of data and provide a generalization that fits the facts.
  • Are there questions about how the process works that can't be answered due to limitations of the data considered so far? If so, then we consider additional data that bears on these questions.
  • Information provided by additional data might require us to revise our generalization.
  • Once we have a generalization that seems adequate, we write a rule that accounts for the generalization.
  • We evaluate the simplicity of the analysis. If the analysis seems overly complicated, we may be missing something.
  • Consider ways of simplifying the analysis.

In developing our discussion along these lines, our aim is to provide you with a "road map" that you can use in solving other phonological problems.

prev | top | next