I'm currently enrolled in an introductory linguistics course at my school, and I am having a hard time pinning down what phonemes are. My text defines phonemes as:

"A distinctive structural element in the sound system of a language. A phoneme is an abstract element that can have alternative manifestations in different phonological environments."

I studied the chapter, and formulated my own definition so that I could check my understanding by asking my professor. Mine read:

"A set of phonetically similar sounds that can be used interchangeably within a language without affecting meaning; a set of allophones."

I was told that I am mistaken. According to her, I am equating the phoneme with the allophones, and that phonemes are the 'idea' of the sound we are saying. However, there is no 'idea horse'or 'idea garment' that is not derived from concrete examples, and I do not understand how that can be the case for a sound. Furthermore, how can we say that phonemes "have" allophones, if a phoneme is not a set of allophones? Or that [p] is an allophone of /p/?

Please don't misunderstand. I am not intentionally arguing with my instructor, and I suspect that I am really mistaken. However, I would very much like to know why and how I am mistaken, as well as learn to understand what a phoneme is.

  • Your professor is right. Usually a phoneme is understood to be an abstract, underlying representation (or base form) "which is altered in particular contexts by the rules of the phonology" (Hayes 2011). Abstract phonemes are realized phonetically in its allophones, or in other words, each phoneme has a set of allophones. That is why you cannot even pronounce a phoneme, you can only pronounce its allophone with the largest distribution (aka an elsewhere allophone). – Alex B. Feb 11 '16 at 4:38

There has been a conflict among linguists since, well, forever, about whether phonemes are analytic devices created by linguists, or whether phonemes actually exist, in some sense, and are discovered by linguists. Whichever side you choose, you'll be in good company.

Personally, I take the realistic side, which I take it is the side argued by Edward Sapir in The Psychological Reality of Phonemes. A version of that view is that of David Stampe, in various publications, and in the unpublished paper "Yes, Virginia ..." (... there is a phoneme).

Your idea about it appears to be a version of the analytic view, similar to that of Daniel Jones'. And if you take the opposing psychological view, it is indeed an error. But some would agree with you.

The division of opinion in linguistics is reminiscent of the controversy in mathematics on whether mathematical objects are created by mathematicians or discovered by them. The latter position is called Platonism.

  • 1
    "Forever" is a long time. Actually the word "phoneme" was coined in 1873. – fdb Feb 12 '16 at 0:10

This is one of those questions that has no right answer, because it depends on how you define the concept. So one way to look at it is that your instructor is right, by definition.

You answer closely approximates the most rigorous definition of the concept, employed by taxonomic structuralists (who, b.t.w. fell into disfavor in the 60's). That is, they viewed a phoneme as being the union of a set of phones (in saying specifically "of a single phoneme", we call them "allophones"). They viewed phonemes, morphophonemes, morphemes and so on as concepts for organizing the facts. There are a number of principles at play in deciding how to organize phones into phonemes. The basic idea is that two phones cannot be subsumed under the same phoneme unless the contexts where they appear is complementary. Given the words [pʰɪt] and [bɪt], since these are distinct words, we know that [pʰ] and [b] cannot be members of a single phoneme (they are members of different phonemes). Given words like [pʰɪt] and [spɪt], it is possible that [pʰ] and [p] are members of the same phoneme (allophones of a single phoneme).

Under that view, or honestly any view, "changes meaning" is not the right way of looking at the pit/bit difference. It's not that changing p to b changes meaning, since then you'd expect there to be some change in meaning if you replace p with b in play. Instead, you don't get a word at all. Different words are made up of strings of phonemes, so by selecting a different phoneme, it is possible that you will identify a different word. And different words tend to have different meanings.

There are other views of the popular relationship "phoneme", some of which depend on having the "phoneme" and the "allophone" be made of the same "stuff", typically some kind of definition in terms of phonetic features. This view is wide-spread amongst practicioners of generative phonology (not exclusive to them, though). Thus the "phoneme" /p/ would be a voiceless unaspirated bilabial stop, and its two main allophones are a voiceless unaspirated bilabial stop [p] and a voiceless aspirated bilabial stop [pʰ] -- note the the phoneme and one of the allophones are the same thing. In that view, the aspirated allophone is the result of applying a rule.

Now to focus on this "idea of a sound" matter. In the structuralist view, allophones are not physical sounds, indeed they had no way of dealing with physical sound, other than to listen and transcribe, assuming some reference sounds. Structuralist phones are abstractions, although they are not "ideas" since they were behaviorists. They refer to classes of behavior – in fact, the behavior of the linguist who is exposed to language stimulus and responds with transcriptions. Though they tended to think of themselves as outside the equation, thus ideally they are mindless automata that transcribe the language behavior of speakers. In the generative view, segments (phones / allophones) are not physical sounds, they are mental representations of sounds. (Specifically, they are the conjunction of a set of features, which are the "intent" to articulate in a particular way). This view is a consequence of the rationalist philosophy adopted by generative grammar: we are supposedly modeling mental states that cause speaker behavior, not modeling the behavior itself.

As I hope to have shown you (superficially and incompletely), the ontology of "phoneme" is very complex (historically) and generally is not at all well handled at the introductory level. Generally, we are happy if students can grind out marginally correct answers to questions like "are p and b allophones in English".

  • 1
    If you think replacing p with b in play makes no change in meaning, you must think play and blay mean the same thing. – Greg Lee Feb 11 '16 at 14:53
  • 1
    You are wrong about what I must think. I think that the meaning relationship between [plɛɪ] and [pʰlɛɪ] is the same as the meaning relationship between [blɛɪ] and [pʰlɛɪ]. – user6726 Feb 11 '16 at 16:41

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