Are "distinctive features" PROPERTIES of "phonemes"?

[zaba]

Is it correct to think of distinctive features as properties of phonemes?
If not, how would you consider the relationship between DFs and phonemes?

[Daniel]

Phonemes are mentally contrastive categories. They are realized through assigning features (that is, sound) to them then through phonological processes (leading to allophonic variation).
So at a philosophical level your question needs to specify things like exactly what you mean by "relationship" and "X as Y". But in short, sure, you can think of phonemes as having features. Certainly that's what the features are there for-- to explain the contrasts of phonemes and (to some extent) deal with allophonic variation.

[zaba]

Thanks.

Certainly that's what the features are there for-- to explain the contrasts of phonemes and (to some extent) deal with allophonic variation.

So in the "famous" intro textbook example of non-contrastive p vs ph in English but contrastive p vs ph in hindi, how does distinctive feature theory help?

[Daniel]

To get into the technical details, that's a very complex question. For one thing, even if valid, are the features language-specific or cross-linguistically valid? Can we say that English and Hindi have the same voicing feature but only Hindi uses the aspiration feature? (Abstracting from what they might be called specifically in a given theory.)

The point is that the phonemes contrast in one language in a way they don't in another.

From the perspective of a student, operating under the understanding that the features and phonemes are essentially the same thing and that you can reasonably compare languages this way is both useful and relevant to what you'll read about. From a deeper theoretical perspective there are various things to work out that are still uncertain. What "is a feature"? Is it a gesture? A mental construct? A cross-linguistic property of Universal Grammar? Is it even a thing, more than just a convenient notation system?

It's hard to answer these questions in the abstract. (Also, as I've probably mentioned to you, phonology is not my area, so for a state-of-the-art answer on some of this you should check with a phonologist. I believe what I'm saying is in essence reasonable, and I know the situation is very complex, but check on any details you need to know more about.)

Also, importantly, the answers to these questions depend very much on your theory (and which particular version of it your textbook uses). So while nothing is certain, if your textbook says something, that might be on the exam

[zaba]

So while nothing is certain, if your textbook says something, that might be on the exam

I guess you know me enough to consider that I don't really care so much about exams, I'm really interesting primarily in understanding this stuff because the most interesting things are just glossed over as if doctrine sometimes.

From a deeper theoretical perspective there are various things to work out that are still uncertain. What "is a feature"? Is it a gesture? A mental construct? A cross-linguistic property of Universal Grammar? Is it even a thing, more than just a convenient notation system?

I mean, there is something that [voice] has in common across languages (and something which it may or may not depending). From my perspective we're trying to get to the "atoms" of phonological structure here, right? The atom itself is also a theoretical postulate, which doesn't exist in the way we think of it.  Anyway, "features" are not gestures. Gestures are gestures. They may involve some gesture, but the correspondence isn't 1-to-1. And let's forget about UG for now. 

But you raise a good question: Are features mental constructs?  Do they have the same "kind" of reality as a phoneme? How am I to think about this?

[jkpate]

It depends on what you want to model. If you want to model language as its own thing, then features should be something that lets you provide a more concise and complete description of alternations and contrasts. If you want to model language as a mental process, then you might want features to be something else. For example, if you want to work out a motor theory of speech perception, then features might very well be gestures. The basic idea would be that, as a listener, you're applying Bayes' rule:

[latex]P(\mbox{word}~|~\mbox{acoustics} ) \propto P( \mbox{acoustics}~|~\mbox{word} ) P( \mbox{word} )[/latex]

On the right-hand side, the P( acoustics | word ) term is a probability distribution over acoustics given a word, which, because listeners are usually talkers, listeners can compute because they know how to pronounce each word. So the listener manages to determine the word based on the acoustics (the term on the left-hand side) because they know the gestures involved in producing the word and the overall probability of each word.