Where might I find ready-made contrastive feature hierarchy trees for these languages? In the case that they aren’t available anywhere online, I may need to make my own, in which case I’m looking for instructions on how to make contrastive feature hierarchies, and where to find the required phonological information on any of these languages to put them together. I am neither an expert on this topic or a student who is meant to be studying it, so this is purely a question from a hobbyist.


It depends on what you mean by "contrastive feature hierarchy". This particular combination of words does refer to an existing theory, promulgated by Elan Dresher, but there are similar approaches, going back to Halle's work on Russian in the 1950s. The word "hierarchy" is typically used to limit the discussion to theories where the features have some "order", and "contrastive" typically refers to only those features needed to represent minimal pair, excluding any features that are introduced to account for allophonic variants of sounds. There are also number-of-feature requirements in some theories, for example it might be stipulated that the central goal of analysis is to reduce the number of specifications required for the phonemes of a language, or to reduce the number of features used. And finally, there is a difference between theories that assume a universal (small) inventory of phonetically-defined features – the SPE theory, for example, versus Radical Substance Free Phonology, where features are not pre-defined phonetically and instead are induced based on phonological behavior (e.g. [m,p,b] act as a class in triggering a rule, therefore they have a feature in common).

The book The constrastive hierarchy in phonology (B. Elan Dresher) is a good explication of one particular theory. The basic idea of the theory is that you start with an ordering of features (that's what you need in advance – or, that's what the child has to discover). Let's say that you have [syllabic] as feature 1: then you divide the inventory into the +syllabic vs -syllabic subsets. Then you move on to the next feature, if there are any undivided subsets with more than one member, separating into the +X and -X subsets. At some point you have have just the set {ʃ} – now you have a complete specification of that segment, and you don't assign any other features to it.

An idea underlying this theory is that by putting a feature high in the hierarchy, it becomes "more phonologically available", so that if the distinction between "vowel" vs "consonant" is rampant in the rule system, you would make sure that [syllabic] is high in the hierarchy. If some segments, such as sonorants, are excluded from a class that they are phonetically in (such as voicing), this would result from the ordering of [voice] in the hierarchy, relative to sonorant – once [r l m n] are distinguished with [sonorant], you don't further distinguish them with voicing, since voicing is not "contrastive" under that ordering.

There are about 20 features in most phonetically-driven theories of features, which means that there are over a million orderings of the features, and a million feature analyses possible. The main issue with hierarchical approaches is figuring out which order in the "correct" one. There are variants of the hierarchical approach which allow even more possibilities (in case the ordering can be different within a sub-tree). IMO the best evidence comes from phonological activity, which is what Dresher's approach focuses on – if it acts like a voiced consonant, it is voiced. In the compression-driven approach (minimize the number of pluses and minuses), there isn't a huge savings achieved by one hierarchy over another, so it hardly matters what the hierarchy is.

For Celtic, specifically Brythonic, I recommend Pavel Iosad's dissertation. He actually draws trees.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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