Formal analysis: take 1

Now that we have a generalization we can work with (at least for now), the next step is to provide a formal analysis (a rule or rules). Before we can do this, it will be necessary to make an analytical decision that we have been deferring. We decided earlier that because we have two sets of sounds, palatals and velars, that are in complementary distribution, the best phonemic analysis is one that posits one set as underlying (phonemic, or basic) and the other as derived. Either palatals are basic and velars are derived, as in (6a). Alternatively, velars could be basic, in which case the palatals are derived, as in (6b). We commented earlier that since the distribution of both palatals and velars was fully predictable from their context (and neither set was more restricted than the other), that there was no solid basis for privileging one set over the other as candidates for phonemic status. This is still the case: we have seen a range of data that have provided useful information, but we still face the which-is-phonemic predicament we were in near the beginning of the discussion.

In cases where system-internal evidence does not clearly indicate which of two sets of sounds in complementary distribution has phonemic status, it is reasonable to make an inference based on external information, for example typological information, that contributes to our understanding of how sound systems work in a broad sample of languages. Two relevant typological considerations are summarized in (16).

(16) Typological markedness

Sound inventories Velar Cs but no palatal Cs More common
Palatal Cs but no velar Cs Less common (marked)
Sound processes Velar fronting (palatalization) More common
Palatal backing Less common (marked)

The two characteristics in (16) are linked: many sound inventories with both velars and palatals have both due to synchronically productive or diachronic processes of palatalization. These typological considerations weigh in favour of the phonemic analysis in (17). This analysis assumes that palatals are derived from underlying velars in Turkish.

(17) /k/ /g/
[k] [c] [g] [ɟ]

Having decided in favour of (17), our analysis now requires a formally stated rule that palatalizes velars in the context of front vowels. Intuitively, the process seems quite straightforward: dorsal-vowel agreement is a form of assimilation in which the [-back] feature of a front vowel is transferred to a dorsal. However, it turns out that producing a rule-based account is not as simple as it might initially appear to be. Because the palatalization process is bidirectional, our formal analysis will require more than one rule:

  • A rule of progressive palatalization, in which a velar assimilates to the backness of an immediately following vowel.
  • A rule of regressive palatalization, in which a velar assimilates to the backness of a preceding vowel, and which applies in multiple environments.

The rule of progressive assimilation can be stated in formal notation as Prevocalic Velar Fronting in (18).

(18) Prevocalic Velar Fronting

a. /velar stop/ → [palatal stop] /___ [front vowel]
b. In phonological features:
-sonorant → [-back] ___ [+syll, -back]

Prevocalic Velar Fronting accounts for the data in (2) (e.g. cel 'bald'); in (9a) (e.g. harecet 'movement'); in (10b) (e.g. haɲɟi 'which'); in (11) (e.g. acis 'echo'); and the form istimlaaci 'expropriation, acc.' in (12), in which an underlying velar assimilates to an immediately following front vowel [i e y ø], regardless of the pre-velar environment.

Regressive assimilation can be formally stated as Postvocalic Velar Fronting in (19). Regressive palatalization occurs in four contexts, with and without an intervening C, and in either word final position, or before a consonant.

(19) Postvocalic Velar Fronting i and ii

a. /velar/ → [palatal] / i. [front vowel] ___ # e.g. istimlak, (12)
ii. [front vowel] ___ C e.g. icram, (10a)
iii. [front vowel] C[+son] ___ # e.g. fark, (14)
iv. [front vowel] C[+son] ___ C e.g. reɲcler
b. In phonological features:
-sonorant → [-back] i. [+syll, -back] ___ #
+back ii. [+syll, -back] ___ C
+high iii. [+syll, -back] C[+son] ___ #
iv. [+syll, -back] C[+son] ___ C

By convention, the notation we use to write rules such as those in (15) and (11) often "collapses" double environments like (i) and (ii), representing them as a single environment that allows for an optional consonant, [+syll,-back,-round] (C) ___ , by including the intervening consonant in parentheses to indicate that it is an optional feature. However, this notational convention, though convenient, obscures the fact that an environment so stated represents distinct rules.

Encoding the final word boundary in (i) and (iii), and the preconsonantal environment in (ii) and (iv) are ways of specifying positions in which no vowel immediately follows the target dorsal consonant. This seems like a very roundabout way of encoding this restriction.

prev | top | next