Does knowing separately phonetic features X,Y,... imply that a speaker will know a phoneme characterized solely and completely by X,Y,...?

Question

In The Language Instinct, Steven Pinker claims that when babies learn to talk, if they have learned in certain phonemes a set of features, then they automatically learn other phonemes characterized solely by these features, even though these sounds involve a novel combination of such features. For example, if they know the feature "voiceless fricative" because they can do \f\ and "alveolar articulation" because they can do \t\ , then they automatically will be able to produce \s\ .
He is an expert in language acquisition, so I guess I probably did not understand something, but are there not obvious counterexamples? Even though the idea seems pretty intuitive, consider for example Arabic, which lacks \p\ even though it has the bilabial \b\ and many voiceless phonemes. What is missing in the picture of this apparent contradiction?

Answer 1

If you cite page numbers in the book which gave you that impression, we could give you a more detailed analysis of what he said. Instead, I'll have to proceed on the basis of what I know that he knows and believes.

First, his theoretical premise and the ideological point of the whole book is that children already know all of the building blocks of language from birth, they just don't know the specific arrangements in the local language. (This is a point that I strongly disagree with, but I can at least say what the position is). All language sounds are defined in terms of a fixed set of universal phonological features, which are phonetically defined. If a language has [θ] then the child will encounter [θ], and automatically knows the features that constitute [θ] because there is a build-in detection system for saying whether a sound is [+continuant] or [-continuant], [+coronal] or [-coronal], and so on.

He uses idiosyncratic features of his own invention, mainly because most of his reading audience isn't interested in learning the rather complicated feature system that was actually assumed, plus at the time he was writing the book, there was no agreement in what the features are which is deadly if you are trying to convey the impression that there is a "scientific consensus". There could be a feature voiceless, and a feature fricative, and a feature alveolar, where [s] in English is adequately defined by the conjunction of those three features. That is not sufficient for Tigrinya which also has ejective [s'], but wih one more feature you can now characterize the lingual fricatives of Tigrinya.

Since the features are built in, you don't learn the features (and there is no feature "voiceless fricative", there are separate, combinable features). What you learn is that certain combinations exist in the language, and others do not. A child learning Arabic would learn that there is [f, b, m] but not [p, v] (let's assume a dialect that has no actual [p]).

One problem with this idealized view of features in relation to sounds is that if you know that English, French and Navaho all have the phoneme /t/, you still can't pronounce /t/ correctly in the language. This is way beyond the scope of what Pinker has anything to say about, but there is still a question and controversy over how these abstract phonemes are mapped to physical productions, so that you will get the correct degrees of aspiration in /t/ for French (none), English (some) and Navaho (lots). Within a language, there is a certain degree of phonetic coherence to each essential property of phonemes, for instance the Arabic emphatics have a core phonetic similarity – they are not wildly different from each other in their production. There is variation in French in terms of the amount of aspiration that voiceless stops have, but they are fairly similar and do not expand to include the values that you find in English and Navaho. Perhaps that is the kind of "then you automatically know" that he is referring to. You do have to learn the "physical value" of /t/, which has a universal definition in terms of phonetically-defined phonological features.

Answer 2

For another obvious counterexample, basically all native English-speakers can produce voiced sounds, velar sounds, and fricatives, but most cannot produce a voiced velar fricative (which is not part of the English inventory).