I was wondering how do you write these rules in features using the minimum number of features to describe the segment(s) targeted by the rule in the input. I am still super new to linguistics! Thank you!

/ t̪ , n̪ / → [ ʈ , ɳ ] / [ ʈ , ʂ , ʐ , ɳ , ɭ ] __

/ k / → [ c ] / ___ [ i , e ]

/ u / → [ i ] / [ c , ɲ , j ] ___

  • 1
    Apart from the fact that we don't give homework answers, this is an opinion-based question – it depends on your opinion of the best set of features. Every author has their own feature system, some have two or three. Also, "targeted" in the context of rules is a technical term – thing on the left of the arrow. Is that what you mean?
    – user6726
    Commented Oct 22, 2020 at 0:47
  • I can assure you that this wasn't homework, I am not a linguistics student yet but I like to learn with some help :) I just wanted to see the possible answers but thanks anyway I guess!
    – ddd
    Commented Oct 22, 2020 at 1:00
  • And I don't think its opinion based when these segments can definitely obtain general features to describe. This is a rule going under change.
    – ddd
    Commented Oct 22, 2020 at 1:03
  • I’m voting to close this question because it does not benefit visitors other than the OP.
    – Nardog
    Commented Oct 22, 2020 at 2:44
  • It could be a nice discussion but sure go ahead.
    – ddd
    Commented Oct 22, 2020 at 3:09

1 Answer 1


The best answer to this question is to explain why it can't be answered, as is. There is no single theory of features for rule-writing – I don't think anyone has done a thorough survey covering the last 70 years, but it probably is in the hundreds of systems, just in terms of names of features alone. Some theories have features with plus-and-minus values; some have just names like "front", "round", without minus values (the theory of privative features). In addition, every theory of features can be paired with some theory of rules, and there are dozens of them. These are the theory issues that make this an unanswerable question. Then there is a very practical question, that the list of segment has to be paired with a list of all segments that exist in the language. For example, in the case of the first rule, the triggering class of segments is [ʈ, ʂ, ʐ, ɳ, ɭ]. Let's use the SPE features: these segments are [+coronal,–anterior]. But that is not necessarily sufficient. How do we exclude [ɖ] or [ʈʷ] from the trigger class? Or, do we have to? Do the segments [ɖ, ʈʷ] exist in the language and do they have to be excluded? Given a list of phonemes in the language, we can form a meaningful question: what features identify {α} and exclude {β,γ...}. You have to know what must be excluded.

As an example of the theory-dependent nature of the question, for the last rule, the difference between the input and output vowel is that the input is (using SPE features) [+round,–back] and the output is [–round,–back]. Does that mean that both features have to be specified as changing in the rule? Not necessarily. It is possible that, if the language only has the vowels [i e a o u] that there is a redundancy relationship where non-low front vowels are non-round and non-low back vowels are round (or you can state the relationship s.t. rounding implies fronting). That's a difference in theories – do such relations have to be explicitly encoded in rules, or is there behind-the-scenes machinery that guarantees those relations (and how does it do it)?

There is also the unfortunate use of the word "target" in the question. Elements of rules have conventional names: the thing that undergoes a rule is the "target", the thing that it becomes is the "structural change" and the context where it happens is the "trigger". But, the question might be using "target" in the sense "is identified", i.e. any class mentioned in the rule.

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.