Phonology: case studies
By Scott Myers and Megan Crowhurst
Department of Linguistics at the University of Texas
The other CL phenomena
The examples in (9) show that deletion and CL are not observed following a back vowel, (9a), or before a nonsonorant consonant, (9b).
| (9) | No yod deletion |
| Formal | Informal | ||||
| a) | [+syll,+back] ___ | kaj.mak | *ka:.mak | 'dairy cream' | |
| kuj.ruk | *ku:.ruk | 'tall' | |||
| b) | ___ [-son] | mej.dan | *me:.dan | 'square' | |
In another set of forms, exemplified in (10), yod is deleted in the same contexts as in (8). However, in (10), vowel lengthening doesn't occur.
| (10) | Yod deletion without CL |
| Formal | Informal | ||||
| a) | i.ji | i.i | 'good' | ||
| b) | u.ju | *u.u | 'sleep' | ||
| bɨ.jɨk | *bɨ.ɨk | 'moustache' | |||
Yod deletion in general is described by the rule in (11), where "X" refers to any sonorant segment, either a consonant or a vowel.
| (11) | Yod Deletion /j/ → Ø / V[-back] ___ X[+son] |
The rule in (11) describes yod deletion in all the cases in which it occurs, both those in (8), and those in (10). However, (11) does not provide for vowel lengthening. Compensatory lengthening is restricted to a subset of the set of cases in which yod is deleted. Yod can be deleted before any sonorant, but CL occurs only the cases in which yod precedes a sonorant consonant. Thus, we need the rule in (12), which stipulates the more stringent conditions required to account for the CL cases.
| (12) | Compensatory Lengthening with yod /V[-back] j/ → V:[-back]Ø / ___ C[+son] |
Notice that the rules in (11) and (12) must apply in a specific order in order for the right results to be generated. The derivation in (13) shows that both yod deletion and CL are correctly derived if the lengthening rule (12) applies first, followed by the deletion-only rule in (11).
| (13) | A derivation that produces the attested results |
| Phonemic representation | / sejret / | / iji / |
| Syllable structure | sej.ret / | i.ji |
| Rule (12) | se:.ret | n/a |
| Rule (11) | n/a | i.i |
| Surface representation | [ se:.ret ] | [ i.i ] |
The counter-derivation in (14) shows that if the rules (12) and (11) apply in the opposite order, an unattested outcome is predicted for the CL form.
| (14) | The counter-derivation falsely generates *e.lyl with short [e] |
| Phonemic representation | / sejret / | / iji / |
| Syllable structure | sej.ret / | i.ji |
| Rule (11) | se.ret | i.i |
| Rule (12) | n/a | n/a |
| Surface representation | *[ se.ret ] | [ i.i ] |
If the deletion-only rule were to apply before the lengthening rule, then yod deletion would result in both cases, but vowel lengthening is suppressed in the forms where it should apply (yielding, for example *e.lyl instead of the attested form e:.lyl). The lengthening rule, (12), must apply first because rule (11), if applied first, would rob (12) of its environment. Below, we will see that a similar situation occurs in the case of /v/ deletion.
/v/ deletion
In the third CL phenomenon in Turkish, vowel lengthening accompanies the deletion of the fricative /v/. In casual speech, /v/ is deleted when it is followed by a labial segment (consonant or vowel), or when preceded by a labial vowel. (Labial vowels are [+round] vowels.) Sezer (1985) provides verb paradigms that show the roots øv- 'praise', ov- 'rub', sav- 'rebuff', and sev- 'love' before the C-initial Infinitive and past tense suffixes, and before the V-initial aorist suffix, so that we can observe alternations between root allomorphs with and without the final /v/. Before the inifinitive suffix, which begins with the labial consonant [m], all roots can occur without /v/ (whether or not the preceding vowel is round). Neither the aorist nor the past tense suffix begins with a labial C or V. In these cases, /v/ is deleted only in the roots øv- and ov-, in which the vowel before /v/ is labial. However, /v/ deletion is not possible in the roots sav- and sev-, in either the past tense or aorist, in which /v/ is neither preceded nor followed by a labial segment (C or V).
| (15) | /v/ deletion |
| Infinitive | Aorist | Past | |||||||||||||
| øv.mek | ~ | ø:.mek | ø.ver | ~ | ø.er | øv.dy | ~ | ø:.dy | 'praise' | ||||||
| ov.mak | ~ | o:.mak4 | o.var | ~ | o.ar | ov.du | ~ | o:.du | 'rub' | ||||||
| sav.mak | ~ | sa:.mak | sa.var | ~ | *sa.ar | sav.dɨ | ~ | *sa:.dɨ | 'rebuff' | ||||||
As the examples in (15) show, the conditions on /v/ deletion are the same regardless of /v/'s position within the syllable. These examples also show that when /v/ occurs syllable finally (with a vowel preceding, but not following), /v/ deletion is accompanied by vowel lengthening (e.g. sev.mek ~ se:.mek). When /v/ deletion occurs in syllable initial position (prevocalically), lengthening is not observed (e.g. o.var ~ o.ar). Additional examples showing that lengthening doesn't accompany syllable-initial /v/ deletion aappear in (16).
| (16) | intervocalic /v/ deletion |
| da.vul | ~ | da.ul | 'drum' | ||
| du.var | ~ | du.ar | 'wall |
The deletion of /v/ in both lengthening and nonlengthening cases could be accounted for by the general rule in (17), which states that /v/ is deleted postvocalically, if either the preceding vowel or following segment (whether a consonant or a vowel) is labial.
| (17) | /v/ Deletion / v / → Ø / V ___ X Conditions: Either V or X is labial; X may be C or V. |
An account of vowel lengthening in the CL cases involving /v/ requires the rule stated in (18). Notice again the similarity between the CL rule in (18) and the non-lengthening rule in (17): /v/ deletion occurs in essentially the same segmental environments in both cases, but vowel lengthening can only occur when the segment following /v/ is a consonant.
| (18) | Compensatory Lengthening with /v/ /V v/ → V:Ø / ___ C Conditions: V (preceding /v/) is labial, or C is labial. |
Notice that the rules in (14) and (18) must apply in a specific order in order for the right results to be produced. The derivation in (19) shows that both /v/ deletion and CL are correctly derived if the lengthening rule (18) applies first, followed by the deletion-only rule in (17).
| (19) | A derivation that produces the attested results |
| Phonemic representation | / øvmek / | / øver / |
| Syllable structure | øv.mek | ø.ver |
| Rule (18) | ø:.mek | n/a |
| Rule (17) | n/a | ø.er |
| Surface representation | [ ø:.mek ] | [ ø.er ] |
If the deletion-only rule in (17) were to apply fbefore the lengthening rule, then /v/ deletion results in both cases, but vowel lengthening would be suppressed in the infinitive form, yielding unattested outcomes like *o.mak.
The account we have developed works on a descriptive level, in that it correctly derives the forms that actually occur. However, some important concerns arise. In particular, our analysis uses three separate rules to account for similar CL phenomena, whereas a more general account would be preferable. We turn to this and other issues below, before moving on to construct a general analysis for compensatory lengthening ion Turkish.