There are multiple models of distinctive features. Wikipedia distinguishes between three main approaches:


Jakobson and colleagues defined them in acoustic terms,[11]


Ladefoged's system[12] is a purely articulatory system apart from the use of the acoustic term 'sibilant'.

** Acoustic & Articulatory **

Chomsky and Halle used a predominantly articulatory basis, though retaining some acoustic features

In what way are these different models of distinctive features more or less adequate in terms of language description? At least from my perspective, most schools of linguistics have converged on Chomsky & Halle's mixed approach. Does this come at a theoretical price?



The main division is between substance-based theories (as you listed) and substance-free theories. Examples of the latter are Foley's theory, and Radical Substance Free Phonology (and to a lesser extent, the Parallel Structures Model). The latter class of theories does not require that features satisfy some phonetic predefinition, where the former theories do.

The SPE position is spelled out on p. 299: "We shall describe the articulatory of every feature... We shall speak of the acoustical and perceptual correlates only occasionally, not because we regard these aspects as either less interesting or less important, but rather because such discussion would make this section, which is itself a digression from the main theme of our book, much too long". Since SPE is not considered to be a major break with the Jakobson and Halle approach, it would be somewhat redundant to repeat those acoustic characteristics -- but also illuminating, since diffuse, compact and grave do not translate directly into SPE theory.

One failure of the wiki classification is that it omits perceptual and aerodynamic definitions of features. The feature [high] is a typical example of an articulatory feature, since it is defined in terms of a specific articulatory movement. [soronant] on the other hand is more of an aerodynamic feature, defined in terms of an aerodynamic result (spontaneous voicing) that must result -- there are many articulatory paths to that result. Heightened subglottal pressure is another. Distributed is also a hybrid articulatory-aerodymanic feature, describing the shape of the resonator but not what you articulate to achieve that end. Then, of course, strident is a strictly acoustic feature.

For observational adequacy, SPE theory requires a set of features with definitions, plus multiple kinds of coefficients (u, m, +, - and integers). Because SPE is committed to feature matrices being phonetic descriptions of segments, and because for example high and mid vowels across languages are only roughly comparable, they require an auxiliary mechanism for describing the pronunciation difference between [i] in German vs. English. Integer coefficients are necessary in order to satisfy their desideratum of phoneticism. (They reject the notion of "phonetic implementation" and only allow truly automatic properties into a "phonetics").

However, the SPE theory of phonetic implementation via integer-manipulation turned out to be an uncashable check, and the result was that actual phonetic facts were ignored -- or, for some people, would lead to massive upsets in the analysis. The best-known example of that is the question of [voice] in English, owing to the fact that voiced obstruents are not consistently produced with focal fold vibration, as ideally required by the theory. Therefore b and p would be distinguished on the basis of aspiration, for some people -- but that precludes there being a difference between aspirated and unaspirated p, which exists at least on the surface in English. And since b can be actually is produced with vocal fold vibrations in some instances, this leads to the quandry that an instance of b might be [+voice] or [-spread], depending on how it actually ended up being produced.

The main price that has to be paid comes from the premise that features are phonetic descriptions, not phonological ones. Given that premise, a theory which strictly limits definitions to one kind of phonetic fact is more restrictive than one that allows features to have multiple (potentially inconsistent) definitions. A mixed theory is better able to handle conflicts such as the mixed behavior of l as [+continuant] vs. [-continuant]; the consequence of that flexibility is that a smaller set of features is required (whereas in a strict-definition theory, more features would be required, in order to achieve the observed range of phonological class behaviors). Given the background assumption (the requirement to phonetically define features), there is no net price associated with mixed modes of definition, but it does require that you have a language-specific phonetic implementation component.

| improve this answer | |

I don't think the SPE system is mixed -- it seems purely articulatory, to me. There is some mention of acoustics, here and there, but this is supplementary information, or to help readers relate the SPE system to earlier work.

I don't see any "theoretical price". If it were to turn out that some features are truly acoustic, then to some extent, the SPE system would be wrong, but that's a theoretical bonus. We want theories to demonstrate that they have some content by being vulnerable to potential counterevidence.

| improve this answer | |
  • For example, strident is acoustic, right? – Teusz Feb 1 '15 at 13:52
  • I guess strident is acoustic, so maybe I'm wrong about this. – Greg Lee Feb 1 '15 at 15:41

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.