I'm a high school student who will be going to college to study linguistics next fall. I'm already knowledgeable about some areas, but I'm currently trying to expand my knowledge in phonology. I have learned about both phoneme and distinctive feature theories. I think I understand their principal ideas, but I'm still unclear as to the justification for distinctive features.

I don't know a lot about semantics, but I'm aware of two basic theories of lexical meaning. One sees meaning as determined by the presence or absence of certain semantic characteristics associated with a word, but this one was rejected a long time ago for a prototype-based theory which sees meaning as a general categorization of this related by their varying degrees of similarity to a prototypical example. I'm sure semantics is a lot more complicated than that, but I don't see why phonology can't be viewed in a similar way. Distinctive feature theories are like the first semantic theory because they think speakers distinguish segments by the presence or absence of relevant features. It seems like there could be a phoneme theory that works like prototype theory, i.e., speakers distinguish segments by their degree of similarity to a certain auditory or articulatory standard for each phoneme.

I feel like the justification for phonemes is very obvious: speakers can tell when a segment is one phoneme or another. But distinctive features seem quite a bit more arbitrary. They assume that a particular characteristic is the specific thing allowing people to make the distinction between two segments, and they often seem to assume that many other distinctions in the language are made due to the exact same characteristic.

For example, the SPE theory [+/- anterior] seems particularly absurd: Doesn't [f] differ from [s] in the exact same way it differs from [ʃ], and how is having a constriction at or in front of the alveolar ridge relevant those distinctions?

Phonemes also seem less arbitrary since there is less room for uncertainty about what the phonemes of a particular language are. Different people will come to generally similar conclusions as to what the phonemes of a particular language are, but the features they include or the number they use could probably differ quite a bit. I mean, you could create completely random features like [+/- my friend sounds silly when she makes this sound]. Without the theory being 100% based in psychological evidence, I don't see any reason to unequivocally claim that one distinctive feature model is more valid than another (except maybe for reasons for computational efficiency, which doesn't have to exist a priori). That makes the justification for their existence at all seem dubious.

What is the justification for distinctive feature theories? I'm especially interested in psychological or cognitive justifications since those would surely be the most important, right? Content I've seen so far just says things like "there are some psychological justifications" without elaborating at all and switches to talking about other things like natural classes and allophonic rules.

  • You seem to have all the answers, so what is your question?
    – user6726
    Commented Oct 24, 2023 at 15:09
  • 4
    @user6726 The question is "What is the justification for distinctive feature theories?" I only discussed ways in which the theories seem to possibly NOT be justified. As I mentioned, I'm most interested in psychological or cognitive justifications, of which I don't currently know any.
    – Graham H.
    Commented Oct 24, 2023 at 15:39
  • 1
    The justification for phonemes (which I agree are useful) is actually the alphabet. Phonemic theory was originated by Bible translators who wanted to print in all languages. See SIL and Pike's book Phonemics. A Technique For Reducing Languages To Writing
    – jlawler
    Commented Oct 25, 2023 at 17:25
  • 1
    Phonemes are to words as atoms are to matter. They are the smallest differentiating unit in a word but have no intrinsic meaning. Bear in mind that language was spoken before it was written.
    – Lambie
    Commented Oct 27, 2023 at 15:28
  • 1
    see 2 Phonetic and psycholinguistic evidence in Mielke 2008 The Emergence of Distinctive Features and 9.2. Psychological Reality of Distinctive Features in Atkinson et al. 2014 Foundations of General Linguistics
    – Alex B.
    Commented Oct 27, 2023 at 16:20

2 Answers 2


I have come to believe that the one most important thing to understand about linguistic theories is that our explanations are only ever models—models that can be more or less helpful for a given purpose, that can make less or more correct predictions when applied to new situations. And that's it; no 'truth' or 'mental reality' to see here.

One commenter said that the origin of phonemes lies with the alphabet, and this suspicion is not new; linguists have long accused other linguists of having come up with phoneme theory with the sole intention to be able to type exotic languages more conveniently on their mechanical typewriters. And that's OK—as long as one refrains to assert that "now that I've establish how to efficiently and exhaustively notate the utterances of language X with only k symbols, I KNOW HOW speakers of language X think about their language, slash: how their neurons are wired". That's ridiculous, and isn't SPE full of such 'mentalist' claims? SPE has been refuted on many levels; what it sparked though is what is real and valuable about SPE.

With that in mind, I don't find your doubts unreasonable; rather, I'd ask: Por que no los dos? You mention 'similarity to a prototype' by which distinctive features have 'long been refuted'; well, no? those two things are hardly contradicting each other, are they, and anyway: In my model, Earth is flat and I can draw a straight line from Hamburg to New York, and that's how I'll ELI5 my vacation plans; for the captain of that aircraft, a sphere with geodesics will be more appropriate. Earth can't both be a flat plane and a sphere, but when we're talking in models, the flat map is more practical for some purposes and gets a lot of work done without some of the complexities of a sphere (3D space for example). In this example, we can readily tell which model is a closer fit to reality, but in the case of natural languages, we can not.

Lastly, when we talk about the 'perceived similarity' between sound classes in human languages, it's no different from similarity perceived between any two old things: If you give me an apple A, a banana B and a citron C and ask me to 'group them by similarity', then (AB) vs (C) is an obvious good choice (you most often do not want to chew on a lemon although some do). Re-thinking it tho, (AC) vs (B) is also 'correct' when we consider their geometric shapes. So, do I have to use 'taste' or 'shape' to build my groups? How about 'color' or 'price'? What about 'articulatory' vs 'perceptive' features, which one is the way to go? Just so happens you can do useful stuff with either of those, so why throw those babies out with the bathwater?

And you're right, it might just be that speakers do form categories that are highly idiosyncratic like "that strange hiss I hear from that other person". Sure, why not. Hardly universal but maybe correct for a specific case.

But, more interestingly, whenever we assemble things into classes as opposed to other classes, then inevitably there must be 'distinctive features' involved, because the membership in a class is a binary attribute dictated by logic alone (and let's not forget that some distinctions are not binary). Importantly, we can say that some binary distinctions are at play in language because we observe so-called 'natural classes' (and sometimes not-so natural classes) that put speech sounds / phonemes into groups that do or don't participate in a given phonetic process, or are allowed at certain phonotactic positions, and so on.

This was a great advance at the time when it became obvious that oftentimes we can talk about 'the sonorants', 'the liquids', 'the non-rhotic liquids', 'the obstruents' as sub-sets in the inventory of a certain language that all participate in a certain procedural or positional regularity in a given language or sometimes across many languages. These classes establish features! you see (and maybe that word 'distinctive' is the real culprit here; I've long be suspicious about the fixation that some phonologists suffer from when it comes to determine the one, true, minimal, logically inevitable number of features and phonemes in languages. It's good to do it for fun or practice or to devise a parsimonious orthography, not such a good thing to get hooked on for good).

When pressed about the 'reality' of phonemes I'll have to admit that according to my own limited observation, even educated and multilingual speakers of Mandarin Chinese, for which I delivered a partial phonological description at the time, had apparently very little appreciation for the concept, although they were taught Zhuyinfuhao (a phonemic orthography) as kids. Also, when talking to speakers of, say, English or German, the orthography of those languages is approximately how deep one will get, including any number of hallucinations of sound differences that provably do not exist (caused by semantic or spelling differences). What does that tell me? I have no clue.

Lastly, may I be so bold as to suggest a bonmot: "a phoneme doesn't have to be letter-sized". A phoneme may be a tonal melody across a syllable. 'ts' in the onset may behave a bit like 't+s' in some cases or like an inseparable affricate in others. Por que no los dos!

  • 1
    Since commas often designate precisely "a tonal melody across a syllable", should one call a comma a phoneme?
    – jlawler
    Commented Oct 27, 2023 at 14:56
  • @jlawler not the comma itself of course; also phrasal intonations are above the level of the word, thus not considered in classical phonemic theory. Other than that, why not? The prototypical question intonation is shared across many European languages and beyond where the pitch rises to the end, and a few others that I feel are 'categorial' (i.e. perceived and produced as one particular sort of expressivity). But in English "Ma—", "Ma?", "Ma...", "Ma!" in high, rising, deep, falling tone is same meaning modified; in Chinese these are 4 unrelated words (mother, hemp, horse, scold) Commented Oct 28, 2023 at 5:36
  • Many structuralist accounts of English phonology distinguished 4 pitch levels, with contours like comma, yes/no question, and full stop. In fact, the distinction was made between US English, with 4 levels, and Mexican Spanish, with 2. This was used to explain why American English speakers (even speaking Spanish) tend to sound excited and anxious to Spanish speakers, because their extreme intonations are rare and expressive in Spanish, but normal in English.
    – jlawler
    Commented Oct 28, 2023 at 16:08

The trick lies in the disconnect between phones and phonemes. Phones have some physical reality—we can measure acoustic properties of the sound, or articulatory properties of the vocal tract, and dissect them under a metaphorical microscope—and phonemes have some psychological reality, as we can see through categorical perception studies.

But now we need to explain how the two are connected! For example, in historical linguistics, sound changes seem to apply to phonemes, not phones. We might say, for example, that the non-emphatic stop phonemes turned into fricatives in a certain context in the history of Aramaic. But how can a phoneme be a non-emphatic stop? Our usual definition of "stop" involves closure of the vocal tract, but that's not phonology, that's phonetics!

One solution is to say that these properties are actually part of the phoneme, in an abstracted, phonological way. That is, whether or not you fully block the vocal tract is a phonetic thing, but phonemes can also be labelled +stop or -stop. (Or -cont or +cont, if that's how you want to analyze it.) Now we can say this sound change applied to phonemes that are +stop and -emph, without having to directly invoke things like voice onset time and the closure of the vocal tract.

But now, if we don't have physical sound waves and articulations to measure, what can we really say about these properties? What is the nature of ±stop, if it's not in the physical articulation of the vocal tract? Well, we can't say much…except that they contrast with each other. Hence "distinctive" features: they're defined by opposition. In the pure, abstract realm of phonology, the most certain thing we can say about +stop is that it contrasts with -stop.

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.