Could you please help me revise the phonological rule in Turkish below?

Laxing rule

It means that "/e/ becomes [ɛ] when followed by a non-dorsal sonorant in the same syllable and if /e/ occurs at the beginning of a word, then it becomes [ɛ] when preceding /ɾ/ in the same syllable." However, when the feature [+tap] is activated, the features [+sonorant] and [-dorsal] become redundant since the feature [+tap] is sufficient to represent /ɾ/ in the language. Therefore, I want to express this situation in just one rule without any redundant features. I also thought about writing [+sonorant, -dorsal] and <[+tap]> features in curly brackets, but then would it be possible to know if <#> is still connected to <[+tap]> rather than [+sonorant, -dorsal] because curly brackets simply denote the possibility of both feature sets. Or are angled brackets enough in this case to show that # only goes together with [+tap] not [+sonorant, -dorsal]?

Maybe I can use indices to show such a relationship, but I'm not sure if there exists such a convention.

I appreciate your time and thank you so much in advance!

  • Based on your interaction with @user6726, I feel the need to ask: Are you asking a question about theory or notation? As far as I can tell it's a bit of both. I'm actually a Turkish speaker myself, so I could help you by giving an answer on how Turkish academic tradition handles such "phonological rules", if you think that might be of any help. Mar 2, 2020 at 9:08
  • Oh, really? I'm a Turkish speaker as well. I'd be more than happy to receive some support on the issue. If you are available, I'm very much interested in a lengthy conversation about the topic. Mar 2, 2020 at 11:18

1 Answer 1


The rule is a bit of a theoretical anomaly in mixing SPE formalism with apparent autosegmental assumptions (features being "activated" with implications that features may be underspecified, plus autosegmental ~ feature-geometric features), so you need to accompany the rule with a summary of the phonological theory that you are assuming. In the context of SPE rule theory, what your rule says, applied to Turkish, is that all lingual and labial consonants, and the vowels /e ø/, become lax in a context. If you want to specifically restrict this to /e/, you have to add more features to say "non-round vowels".

The context says (roughly) "before {n,l,r,j,m} within the same syllable", but the angled brackets complicates the rule. The formal definition of angled brackets in SPE is a bit of a mess, but the standard interpretation of the notation is that "A → B /[a,<y>] __<z> Q" means that "A → B /[a,+y] __z Q" and "A → B /[a,–y] __ Q". This means that z can intervene between A and Q if the preceding segment is featurally [a,+y] (whatever a is, and it must be specifically plus y), or else if A is immediately before Q and is preceded by a segment that is [a,–y]. The effect of putting [+tap] in angled brackets is to say that {n,l,r,j,m} trigger this laxing rule very generally, but r triggers it only in the specific context that the vowel is word-initial (for example erdogan and not pervez). I cannot tell if that is what you wanted to say. In general, the correct way to write a rule is to say in plain language what you want the rule to do, then find the matching formal expression. I can't figure out what you say the rule is supposed to do.

Your first sentence suggests that you did mean this, so I'd independently like to see proof that this is a factually correct description of the language (a personal curiosity thing). Your concern with "redundant" features is not without merit, but in a vacuum it can't be addressed. Who cares if a feature is redundant? (I do, actually, though technically I'd say the problem is with unmotivated features). My theory of features evaluates features "as they exist in a grammar", so if X is unnecessary, it is unnecessary, and does not have a special status depending on whether some other features are assigned in the derivation. You apparently have a different theory of features. A satisfying answer to your question requires that you set forth your theory of features, a bit much for here, but I'm sympathetic to the problem.

It is not clear what the facts of the language are, but you seem to assume that there is a single rule that derives [ɛ]. The single-rule assumption is often made in phonological analysis, so we don't say that there are is a rule in German devoicing final /g/, and a separate rule devoicing final /b/, and a third one devoicing /d/. However, we also don't say that there is a single rule of vowel harmony in Mongolian, or a single rule of vowel harmony in Turkish, or a single rule of gradation in Finnish, because in the context of "the" theory, it's impossible to state these processes as a single rule. Any theory limits what can be expressed as a single rule: except, it is not clear what constitutes a formally unstateable mapping in SPE theory. An alternative that you need to consider, if there is good evidence that /r/ behaves differently from other consonants which still do the same thing, is that there are two rules.

  • 1
    First of all, thank you so much for your response and enthusiasm about my problem. I use the feature theory adopted in Bruce Hayes' Introductory Phonology. He states that [tense] and [back] are features used only for vowels. These features are only applied to consonants if there is any secondary articulation such as velarization, palatalization, etc. Therefore, [-high, - back] features only refer to /e/ in the phonemic inventory of Turkish. [+sonorant, -dorsal] sounds are /m n ɾ l/ in Turkish. However, /e/ does not become [ɛ] at the beginning of a word except when followed by /ɾ/. Mar 1, 2020 at 15:55
  • Accordingly, /e/ becomes [ɛ] in erdoğan, but this alternation does not occur in emsal, entari or elma. The laxing rule should apply to pervez as well because /ɾ/ is [+sonorant, -dorsal]. Additionally, considering [tense], [back], [high] [low] and [front] features, /l/ is not [-tense, -back, -high, -low, -front], rather [0tense, 0back, 0back, 0high, 0low, 0front]. This use of features really reasonates with me since not all sounds have to possess a kind of feature to define them. If you're still interested in the topic, I'll be happy to have a lengthy discussion with you! Mar 1, 2020 at 16:05
  • 1
    The problem is that we don't know what features specifications you're assuming, indeed what you take the phoneme inventory to be ([ɛ] is an allophone), plus we don't know what the facts are (treat this as an elementary allophony problem to solve, give /e/ in the context A_B for all possible values of A and B (including Ø=ğ). Lowering before r is not difficult to state; adding a word-initial condition is not hard. Why do we have to go further than that? I realize now that you're also assuming a single rule, see my revision.
    – user6726
    Mar 1, 2020 at 16:18
  • 1
    I simply want to cover all the alternations in a single rule; however, it seems that providing two or three rules for laxing would be more beneficial and understandable. If you are interested, I can give you the phoneme inventory of Turkish that I use so that we can meet on common ground. Mar 1, 2020 at 17:15
  • 1
    The first segment in a word is preceded by a [-segment] unit (actually, -segment,+WB]). The second or subsequent segment is not preceded by a [-segment] unit. That is the featural approach. Alternatively, the disjunction {[+syl],[-syl]}, but only if you assume boundaries are not specified either way for phonetic features.
    – user6726
    Mar 2, 2020 at 15:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.