Friday, 4 May 2012

phoneme and allophone

Here’s another Google n-gram (blog, 13 April), this one for the words phoneme and allophone.

The OED’s first citation for phoneme in its modern sense is dated 1896. As the n-gram shows, though, it was only in the 1950s that this term really took off, reaching a peak around 1970. It was in 1968 that Chomsky and Halle’s The Sound Pattern of English was published, the first widely influential publication to rubbish the whole notion of the phoneme as the basis of phonological analysis (though Jones’s great SOAS rival Firth had attacked it years before). This explains the abrupt decline of the term after that date.

The OED records only one instance of the popular misuse of the term ‘phoneme’ to mean nothing more than ‘speech sound’, often encountered among speech and language therapists, language teachers, and drama teachers. That is from the novelist Kingsley Amis.

The graph shows allophone coming into use around 1940. The OED’s first citation, from Benjamin Lee Whorf, is dated 1938. Daniel Jones seems to have used the term only very rarely, speaking rather of ‘sounds’, the ‘grouping of sounds into phonemes’, and the ‘members’ of a phoneme.

The only case I have found of Jones using the term ‘allophone’ is in The Phoneme: its Nature and Use, §24.

When a phoneme comprises more than one member, it generally happens that one of the sounds seems more important than the other(s). … Such a sound may be termed the “principal member” or “norm” of the phoneme. The other sounds in the same phoneme may be called “subsidiary members” or “subsidiary allophones”.

The American structuralists were always clear that all the members of a phoneme are its allophones: the phoneme comprises its allophones. Phonemes are realized or manifested as their allophones. Nevertheless, students often suppose that the principal member is the phoneme, and that the allophones are what replace it in particular environments. This misunderstanding is encouraged by definitions such as Potter’s (1957):

Robins (for whom I used to write linguistics essays when I was a postgraduate) explains the orthodox position much more clearly.

Although many modern phonologists routinely pooh-pooh the notion of the phoneme, there seems to be no satisfactory replacement term or concept for us to use in its stead.

_ _ _

I shall now be away for three weeks. Next posting: 28 May.

46 comments:

  1. I wonder if we'll see an increase in the use of the term phoneme with the current interest in cognitive phonology and the work of John Taylor, for instance, arguing against the Chomskyan position and for the phoneme to be considered the 'basic category' in phonology?

    ReplyDelete
  2. Mr Wells:

    'students often suppose that the principal member is the phoneme, and that the allophones are what replace it in particular environments. This misunderstanding is encouraged by definitions such as Potter’s (1957): [etc.]'

    I know this problem. I don't teach linguistics but I sometimes use linguistic _Gedankengut_ teaching something else, and am often hard put to explain to students that a phoneme be just an abstract class of allophones, precisely because 'students often suppose...' and so on.

    One explanation thereof I'd like to offer is this: human mind as such (in the middle of the Gaussian curve, that is) hates things abstract. And a phoneme in that correct sense is an abstract entity: a class of allophones in which it is manifested. But the ingenuous mind, which believes that language be made of sounds (well, it is, in a sense) requires something tangible, audible in this case --- and since 'phoneme' is the closest non-ingenuous approximation of the ingenuous 'sound of speech' (is it really?), the yet-ingenuous mind of a student identifies phonemes with sounds, that is, with those sounds which they hear most often, the principal allophones. Especially that subtle differences between different allophones, perceptible to trained linguists, are seldom heard by students.

    So from the philosophical point of view the American structuralist approach is _correttissimo_. (And I agree with Mr. Wells that there is no replacement for 'phoneme'.) From the didactic point of view by contrast --- I dunno; compromises have sometimes to be struck with the ingenuous and abstraction-hating, intensely non-Platonic mind of a 'punter' (as David Crosby would say).

    ReplyDelete
    Replies
    1. Wojciech, at an even lower level of sophistication than punter, the totally uniformed language critic believes that words are composed of letters in speech just as in writing.

      The pre-theoretical concept of letter in speech rests on an assumption (a reasonable on) that alphabetical spelling corresponds to the realities of speech in some way, but also on the assumption (a misguided one — especially if you speak English) that it it's a reasonably close correspondence.

      I'd reserve punter to denote some sort of consumer of writings on language. This would include language teachers like you and me. At this slightly higher level of sophistication what laymen have the sense to consume is linguistic theory. We consume in the quest of better understanding what we teach and how we might teach it. But we also consume out of sheer curiosity and fascination with the subject matter — as do many amateur enthusiasts, some of them with considerable acumen.

      At this level of sophistication we operate at what I consider a manageable level of abstraction. There are abstract types which we recognise through concrete tokens.

      [OK, it might be better to think of concrete realisations of abstract tokens — but lets leave that to a higher level of sophistication.]

      We're sophisticated enough to see that a spoken word isn't a noise between silences in the way that a written word is a pattern between two white spaces. We take it for granted that a word — as in What does that for mean? Can you repeat that word? etc has characteristics that remain constant between utterances. In this sense word is very much an abstraction. We accept with alacrity the idea that words can be segmented into sounds — yielding the usable model of an infinite set of meaningful symbols (words and word fragments) assembled from a smallish finite set of meaningless symbols. (It's the only model that explains the uniqueness of human language among systems of symbolic communication by animals.) In this sense sound too is very much an abstraction — albeit one that corresponds simply to a token of concrete noise.

      Problems come only when linguists direct us to a higher level of sophistication — and thus of abstraction. Theoreticians expect too much of the phoneme concept. It's not enough for it to be a sound — an abstraction from concrete speech noises — and a building block. Phonemes have to be an identifiable closed set, an exclusive category. But we punters don't care whether, for example, the START, NURSE, NORTH and FORCE sound is one phoneme or two in rhotic and non-rhotic accents. Or whether the happY, lettER and commA, vowels have the same status as the TRAP vowel. Once I discovered lexical sets, I ceased to care about vowel phonemes.

      Well, I might be curious about the theory, but it doesn't actually matter. I've even studied some generative phonology but I'm only really comfortable when thinking of of distinctive features as properties of phonemes — which I suspect is seriously missing the point. I finally abandoned any hope of deep understanding in a class with Gill Brown when she tossed of an aside to the effect that Daniel Jones didn’t have a clue about phoneme theory.

      Delete
  3. I don't know what terminology is used by specialists in artificial speech synthesis and speech recognition, but whenever journalists attempt a popular explanation they seem to use phoneme to mean something concrete and particular. Something like 'realisation of an allophone'.

    ReplyDelete
    Replies
    1. 'phoneme' is a fancy, flashy word, minds of journalists, on the other hand, are rarely firmly set on anything Platonic...

      Delete
  4. "To cut Sinatra off in mid-segment" would have invited misinterpretation.

    ReplyDelete
  5. Maybe part of the problem is that in English there is no word for what language is ultimately segmented into, that was neither very learne`d and susceptible to various misinterpretation or journalistic kidnappery, as is 'phoneme' nor ingenuous like 'sound of speech'. In Polish we have a term that is taught to schoolchildren, 'głoska', like 'kot' (cat) has /k/, /o/, /t/ as 'głoski' --- alongside 'fonem' (which is NOT taught to schoolchildren) and 'dźwięk' ('sound'), which is naive, ingenuous, folksy.

    I sometimes say to my students, look, different 'sounds' can be allophones of the same phoneme in one language but two different phonemes in another, like e.g. the rolled r and the uvular sound: allophones of /r/ in Polish and most (all?) European languages, but different phonemes in Arabic (ra, ghain, respectively). Though not very exact, this helps some to understand better what use this terminology is. From the Platonic point of view the uvular and the apical r's are manifestation, phainomena, of one abstract phoneme /r/, neither of them IS this phoneme.

    ReplyDelete
  6. Could it be because we are told that ἄλλος means 'other'?

    ReplyDelete
  7. Really, though, an allophone is just as abstract as a phoneme: it's a bundle of possible sounds clustered around a prototype, just as a phoneme is. Doubtless no two utterances are ever exactly alike, so what we strictly mean when we say that English /p/ has the conditioned allophones [p] and [pʰ] is that a variety of unvoiced labials with varying voice onset timings are acceptable to anglophones, but that in certain contexts the VOT is quite long, in others, relatively short. By contrast, /b/ has the unconditioned allophones [b] and [p], meaning that the VOT varies from zero to short.

    This should be no surprise. We do not think of dog as an abstract term, yet individual dogs vary far more than individual performances of a phoneme. It's not even clear that there is a prototypical dog, as there is a prototypical speech sound for a phoneme — rather, different people have different dog prototypes based on their different experiences in life. (Note that "prototypical" does not imply "fuzzy"; an animal is either a dog or not a dog: there are no phrases like "technically a dog" or "almost-dogs".)

    ReplyDelete
    Replies
    1. 'an allophone is just as abstract as a phoneme: '

      yes, but the differences are barely recognisable, we are talking about Perceian types not tokens here, and it is obviously true that no two tokens are absolutely alike. But that makes also the recognition of allophones difficult, if they are not very unlike acoustically.

      From a philosophical point of view, 'dog' is an abstract term indeed. What it signifies is a canine nature, which is abstract. 'Fido', by contrast, is a concrete term. 'Fido happens to be a dog'='Fido instantiates, is an individual instance of, canine nature'.

      Delete
    2. Very interesting, Cowan.

      I agree that both "allophone" and "phoneme" are abstract concepts, because any allophonic transcription (no matter how narrow is intended to be) is just an attempt at decoding the complexity of an utterance. And the only "concrete representation" of an utterance can be a spectogram or a recording itself. And neither the spectogram nor the recording are absolutely identical to the original utterance, but as the range of frequency they capture is much greater than the human ear (if the recording equipment is good, of course) they are rather faithful descriptions.

      However, I still believe the "phoneme" is more abstract a concept than an "allophone" in that the latter is intended to describe physical/phonetic segments while the former is intended to describe language/phonological segments. Allophones are more tangible in this respect.

      Yet I think you've made a very interesting point.

      Delete
    3. Very interesting, I agree! Another claimed difference between phonemes and allophones is that the former but not the latter change meaning. (Reese Heitner entertainingly documents futile attempts to rid the phoneme of meaning in 'An odd couple: Chomsky and Quine on reducing the phoneme', Language Sciences 27, 2005.) Inspired by John Cowan's argument that allophones are also abstract objects (cf. Linda Wetzel's 'Types and Tokens', MIT Press, 2009), one might insist that allophones, too, are necessarily defined by meaning --how else could we determine that they are non-contrastive?

      Delete
    4. Isn't it in principle possible that phonemes could actually be concrete entities realized somehow inside people's brains? Or has that idea been completely dropped by today's linguists?

      Delete
    5. So either [b] or [p] can be used in bat, i.e., they're in free variation? I didn't realize that, but it makes sense because pat has [pʰ], so there is still a clear contrast between the two.

      Delete
    6. On the contrary vp, most linguists assume as you say that phonemes have concrete psychological or neural reality. Platonists like Jerry Katz who argued that language deals in abstract objects are almost extinct! Of course, the phoneme is considered extinct by many linguists too. As Elan Dresher bemoans in the new Blackwell Companion to Phonology: "The phoneme, to all appearances, no longer holds a central place in phonological theory. Two recent and voluminous handbooks devoted to phonology, edited by Goldsmith (1995) [2d edn. 2011 --DF] and by de Lacy (2007), have no chapter on the phoneme. It is barely mentioned in the indexes. ... it is not much talked about, and when it is, it is more often to dispute its existence than to affirm it" (p. 241). Like others here (Wojciech, Cowan, ...) I suspect that rumours of the deaths of phonemes and platonists may be premature if not exaggerated ;)

      Delete
    7. John Cowan

      Really, though, an allophone is just as abstract as a phoneme: it's a bundle of possible sounds clustered around a prototype, just as a phoneme is.

      I agree that an allophone is an entirely theoretical concept. But is it as theoretically sound as the phoneme concept?

      Yes there are some EITHER-OR cases like when we bend the paper with pʰɪn and not with spɪn, or where the distinction is so well recognised as to attract labels like clear/dark l. But against this there are variants along a scale — e.g. slightly glottalise, more glottalised, even more glottalise and so on. Or would you not speak of allophones in such cases?

      Yes, allophone is an extremely useful pre-theoretical concept. But does it have anything to contribute to phoneme theory?

      Delete
    8. there are no phrases like "technically a dog" or "almost-dogs"

      Are dingoes dogs or wolves?

      Delete
    9. I can't resist a Cowan-style essentialist reply here (http://home.ccil.org/~cowan/essential.html): "Dingoes are essentially dogs that have re-evolved into wolves." (retrieverman.wordpress.com, 2 Jun 2011); sorry this one isn't funny

      Delete
    10. Cowan is notorious for his various 'essentially' statements, all of which are funny but not all of which are very accurate. Dingoes, as I established this morning by looking at them in the Zoological Garden in Gdańsk, are not very lupine, rather vaguely jackal-like, if anything.
      Wolves and dogs are technically the same species, canis lupus. But dingoes seem to be closer to dogs than to their wild ancestors, wolves.

      Delete
  8. I guess it's sorta similar to chemistry then:

    An element comprises its isotopes.

    Cl-35 isn't more chlorine-ish than Cl-37, there just happens to be 3 times as much of it.

    ReplyDelete
  9. I don't think it's really like chemistry. Any molecule you examine will have a discrete number of atoms for each constituent element. And it can be checked against a known inventory of thus-defined isotopes.

    If you examine a segment of sound, it won't necessarily display discrete numbers of anything. And there's no guaranteed of a definitive inventory of allophones against which to check it.

    ReplyDelete
  10. I have to give a thumbs up to Sili and thumbs down to you here, David. But I am maybe wrong on this and you got the point.

    However, if i look at the situation as a reader, not as an orthodox or a hardline linguist, the elaborations are analogously sounds similar segments to me—environmentally bipolar (or having their biplicity), predictable variants in their neutralization, and in complementary distribution (an explicit phonemic inventory against which to check).

    I agree with John and Leni in the sense to narrow things down, but a further complication here is a question like--what is then wrong with ‘cats’ and ‘dogs’ examples?

    That is, if we are to assume that ‘g’ in ‘dogs’ is as a rare voiced stop but as a variant slake-voiced /ɡ/or half-voiced, then we have to take also the assumption that the plural ‘s’ cannot usually be as the allomorphy of the /z/ in complementary distribution for /s/ in ‘cat’ but the plural ‘s’ in ‘dogs’ as a predictable variant allophone of the phoneme /s/ for the voiced 'g' in 'dogs' and voiceless 'g' in 'dogs' as for their complementary distribution.

    ReplyDelete
    Replies
    1. Nevill

      Please correct me if I've misunderstood but I take you to suggest that there are a discrete number of ways of pronouncing the /g/ phoneme in dogs.

      Well, I'll allow there's variation, but i fail to see why we should recognise only two variants — or any other discrete number.

      Incidentally, you bring out what is lost when we think in terms of phonemes. The final consonants in dogs and cats are clearly related. The spelling brings out the similarities but phonemic transcription brings out the difference. Morphophonemic analysis offers a solution with the notation {s} but life's too short to run both phonemic and morphophonemic transcription in tandem every time.

      Delete
  11. Ad vp

    'Isn't it in principle possible that phonemes could actually be concrete entities realized somehow inside people's brains? '

    this is a naive psychologism for one, a doctrine dead for 100 years, now rescitated (?) But even if it had been so, those 'entities realised somehow inside people's brains' would have come, them too, in types (=universals) and tokens (particulars), there is no escape from a (moderate) realism about universals, I am afraid. Various nominalisms have been popular in the English-speaking world for over 600 years, yet they are all wrong.

    ReplyDelete
  12. How can we doubt that there is something concrete within the brain, set of circuit-like configurations that convert the infant's perception and production from a wide undifferentiated spectrum of sound to a repertoire of discrete sounds which can be used in analysing and synthesising the language around them. Other parts of the brain produce and perceive tokens; the phonological part identifies the types.

    Theory calls out for an inventory of types and process of assigning tokens. Because the job was first tackled by phoneticians, the tokens were such speech sounds as could be precisely observed and calibrated and the tokens were idealisations — total abstractions but defined in essentially the same manner as the concrete observed tokens.

    But why was the job grabbed phoneticians? Because they were there. And why were they there? Because of a long tradition of thinking in terms of vowels and consonants, albeit with a somewhat muddle-headed grasp of the difference between speech and writing. And yet, and yet…. the alphabet was a brilliant invention. It clarified valid but ill-defined insights into what words were and how they were were built. Phonetics stood on the shoulders of alphabeticisation and saw more far, more clearly.

    The phoneme as defined by phoneticians is a wonderful heuristic. It allows us to think discuss at the type level in terms that are intuitive and graspable at the token level. Phoneme, allophone, phone — however abstract, however concrete — they address what is accessible directly or indirectly to the conscious mind.

    But suppose it's not like that. Suppose that the observed tokens and the idealised types operate at the same level and speed as grammar rules. Suppose, that is, that the reality behind phonemes etc is inaccessible to the conscious mind. It's conceivable that, for example, generative phonology is the reality and phonemes are a convenient fiction. Certainly, the brain is fast enough to operate at the level of bundles of features.

    Not the conscious brain, though. Let's assume for a moment that distinctive features are the real types in principle discoverable by brain science, although not with the technology of the present or, indeed, of ages to come. But even if scientists in the far future identify all the synapse circuitry, we'll still — as now — need a concept that the conscious brain can operate with.

    We need the concept phoneme in the way we need the concept gene. But I don't believe we impose a structuralist template on genes. We don't read of a closed set comprising all the possible genes. Rather we read of the discovery of sequences within DNA that are inherited intact and are associated with some (also inherited) biological trait.

    To be fair, phonemes are more of a category and more of a closed set. But is it really necessary to decree that there can be no marginal phonemes, no ambiguity in recognising allophones and assign them to phonemes? I'm sorely tempted to view phonemes and allophones as atheoretical conveniences.
    • Phoneme: the fictional object allegedly 'signified' by a symbol suitable for transcribing speech in such a way as to eliminate individual speakers' different pronunciation of the same lexis
    • Allophone: the fictional object allegedly 'signified' by a symbol suitable for transcribing speech in such a way as to identify individual speakers' differences

    To exaggerate a little, an allophone is a pronunciation that a phonetician has a symbol for. Probably my reluctance to recognise a large number of allophones is due a poor ear and a dislike of exotic symbols.

    ReplyDelete
    Replies
    1. David,

      I am not sure if you were talking to me or not, but in any event: an allophone and its perception and the physiological basis of its perception are three different things. Just like a cat, its perception by Mr X and its physiological basis are three different things.

      In this sense, neither a cat nor an allophone nor a phoneme is anything inside anyone's brain. That was vp's psychologism.

      Now it is of course possible, but I'd leave this to brain-scientists, that in Mr X's brain there are some permanent structures, after a time, responsible for his perception of the given allophone or of a cat. It would remain to be seen if this structure is typically the same in other people's brains, or if it necessarily the way it is in Mr X's brain.

      Delete
    2. An allophone is only like a cat if allophones exist.

      We don't talk of cats out of convenience, but It's at least arguable that convenience is the only excuse for taking of allophones.

      Suppose that a phoneme is more like an illusion — not the sort you see in a fevered or drugged state, but the sort that your brain creates as a sort of correction to a confusing input. For example, there's a famous one that may not work without full control of the typesetting:

      PARIS IN THE

      THE SPRING

      With the right justification and spacing, nobody notices the repeated THE.

      If this is really the case — as I believe it might be — then the phoneme exists only in the brain.

      Delete
    3. I'd say that the 'brain' (mind) simply disregards, makes abstraction from, the differences between allophones and concentrates on what is common to them, and in so doing perceives a phoneme, e. g. the phoneme 'd', whether in 'fiddle' or 'didn't, or 'draw' or 'bade' or 'had it'. This perception is not a purely passive copy of what is being heard, it's sense-data reformed and reorganised, but not an illusion (as if though all those phones above, conventionally spelt 'd', had nothing in common.

      Such perception is also holistic in the sense that it is structured to some extent in the light of former perceptions, e. g. you perceive two different phonemes in 'caddy' and 'city' (if you are attuned to American English) even though what attacks your ears is exactly the same. This is analogous to seeing a cat for what it is even though it at a given moment looks like something else, e. g. a weasel or such.

      I don't like this 'in the brain' talk among other things because it suggests that we should look into brains in order to hear phonemes. When a cat is seen but there is not cat to be seen it is better to say, methinks, that there is an illusion of a cat (more or less vivid) than to say that a cat is in the brain (can it possibly damage the brain with his claws?)

      Delete
    4. Wojciech

      I don't like this 'in the brain' talk among other things because it suggests that we should look into brains in order to hear phonemes.

      I don't see where else we should look. A serviceable description of a phoneme may be expressed in acoustic terms, but it can only be approximate — close but approximate. Phonemes are learnt in the brain, recognised in the brain, produced in the brain. If the brain can't define it, then it isn't a phoneme.

      As for the cat, if you hallucinate that you see one, it follows that there is a cat-recognition template in your brain. And that's what a phoneme is — a speech-sound recognition template.

      There is a difference. If there's a flaw in your cat-recognition template, it doesn't affect too much your interactions with small domestic mammals. But a flaw in your speech-sound recognition template is liable to degrade your ability to converse with people.

      Delete
  13. I find this all puzzling. Of course ɑː is the most frequent, and I'd say the only possible phone in careful and correct speech, in BATH words. For example. And so is əː (I'm ditching the ɜː notation, it makes no sense, schwa can be stressed in proper English) in NURSE words. That is why these symbols were chosen - otherwise we'd have all sorts of silliness and would use Geordie vowels to describe proper RP sounds or something. These symbols simply correspond to the sounds of educated speech and thus the phonemes represents the most frequent and desirable phones.

    ReplyDelete
    Replies
    1. I am not a linguist but I suspect they'd tell you that your ɑː is a bit different in your baths from what it is in your afters or your asks or your dances or your fathers or such, and again a bit different from one bath to another, from one ask to another and so on. The nearly infinite variability of allophones-types (first set of contrasts) and allophones-tokens (second set). Depending on how good your ear is, you will hear these differences or not, most often not, I suspect. Most of the time they will be just ɑː's to you, and that is why it is good to have all of this conceptual grid (phoneme, allophone, and so on) and the IPA alphabet. Take this from a consumer of phonetic wisdom, not from one of its manufacturers.

      Delete
  14. Not so much, Duchess. John uses /e/ on this blog and elsewhere for the DRESS vowel, which is far closer to [ɛ] than to [e]. However, it's easier to type [e], and it's theoretically entirely in order, so he uses it.

    And of course some of us use /æ/ and /ɚ/ in our BATH and NURSE words, no matter how excruciatingly correctly pronounced.

    ReplyDelete
    Replies
    1. Duchesse, Wojciech, John Cowan

      I was strangely unengaged — to the point of bafflement when I read this exchange. It's suddenly occurred to me why that was.

      I spent my professional life teaching non-native speakers to communicate in acceptable spoken English. For me that entailed a target phonology more or less equivalent to RP — broadly similar to my own accent and a huge number of middle-class British accents which John W. might class as 'near RP' or 'modified RP' and a great many regional accents that speakers have modified in the direction of RP.

      For teachers like me, what we find in the standard resources (Jones, Gimpson etc and their successors) is a model based on RP. If students are directed to emulate what Jones etc describe, they almost certainly won't sound like RP speakers. (OK, I've met some exceptions.) But hopefully they will sound 'acceptable' with a foreign variant of that broad RP-like accent. (Yes, of course it's different for American teachers)

      So I really, really don't care what phone students use in BATH words. All I want is that they use either a TRAP vowel or a PALM vowel. In the broad accent there's no room for three target vowels, although specific accents might call for a for specific native-speaker vowels.

      For me, that's of no interest whatsoever — as long as I'm wearing my teacher's hat, that is. There are two target vowels, therefore there are two phonemes.

      The PALM vowel is the same phoneme whether pronounced as ɑ: or a — or any modification. And the TRAP vowel has remained the same phoneme even as speakers from the Queen on down the social scale have moved from æ as described by Jones. When I started teaching, we got by with one symbol with or without a length mark:
      TRAP: a or ɑ
      PALM : a: or ɑ:
      In the structuralist climate of the time nothing mattered but difference. There was a slogan in French — possibly from De Saussure. And people quoted the analogy of chess pieces, which could be of any shape and could be made of any material — short of ice.

      An outfit I worked for in Italy placed enormous importance on the symbol æ, for purely pedagogical reasons: to steer learners to a higher-prestige pronunciation that didn't employ Italian a. Pedagogic but snobbish.

      The reason for avoiding the symbol ə: is similarly non-phonetic. In English, vowels frequently reduce to ə in the absence of stress. This is an association which teachers want to develop in learners. Many of us were taught the non-phonetic reason for the use of the term schwa in Hebrew grammar.

      Besides, the vowel in commA and (for non-rhotics) lettER might well be closer to ʌ in accents that are not very unlike RP. And to cap it all, there can hardly be any non-phonetician non-rhotic native speakers who perceive any similarity between the commA/lettER vowel and the NURSE vowel.

      All that matters — with my teaching hat on — is that there are two target vowels, and therefore two phonemes, even in some rhotic accents. Any two symbols would serve to represent them — except that the lettER/commA vowel is associated with teaching items such as ðæt~ðət, and has a distinct name.

      Similarly with DRESS. There's only one E phoneme — as long as we perceive the FACE sound as a diphthong. Of course DRESS and FACE must be distinguished in notation but an added i or ɪ or j will do the trick. It's not the end of the world if foreign learners use something close to e for the DRESS vowel provided that they have a distinctive pronunciation for FACE.

      Delete
    2. What matters to the teacher is not the symbolism but the repertoire. And this repertoire is still defined as the structuralists decreed: by difference. That's why, Wojciech, I find your recurring problem with ʌ so very very simple. Difficult yes, but simple.

      The spoken English I teach (used to) has a distinct vowel phoneme for STRUT words. I came from a region where this was not so, but my mother insisted on a total distinction between STRUT and FOOT. Since my mother's prejudices are still held by some British speakers, I would still teach the STRUT vowel as a distinct phoneme — except for students living in certain parts of England. Any symbol would do but u is needed for GOOSE words and a would introduce needless confusion. The symbol ʌ has the considerable merit of dissociation — it doesn't invite confusion with BOOK, GOOSE, TRAP, PALM or BATH. So even if some native speakers' STRUT vowels are closer to a that's no reason for not using the strikingly unusual symbol ʌ.

      Delete
  15. As as teacher of English, I have always thought of a phoneme as to do with understanding and allophones as to do with sounding like a native speaker.
    I would see a phoneme as an interaction between knowledge in a listener's brain and the acoustic signal. It is perceptual but needs something physical.

    An allophone, as is indicated by the square brackets, is a physical phenomenon independent of the listener. It is an abstraction in the same way as when we measure the length of something as one centimetre, this is an approximation to an ideal version of the centimetre but it is a different order of thing to a phoneme.

    Did there not use to be an argument that because /h/ and /ng/ has complementary distributions there were really just one phoneme? Is this accepted anywhere now?

    ReplyDelete
    Replies
    1. I believe it was never anything but a joke: a satire on the overly broad defiition of phoneme, as if phonetic similarity counted for nothing at all.

      Delete
  16. Richard

    I like your observation of the phoneme

    It is perceptual but needs something physical.

    with the caveat which you may not accept that the something physical is a selection from the acoustic signal, and interpretation which may well not be shared by a speaker of another language.

    I'm not so happy with your take on the allophone

    An allophone, as is indicated by the square brackets, is a physical phenomenon independent of the listener. It is an abstraction in the same way as when we measure the length of something as one centimetre,

    I think it's more like measuring as one thickness of a pen or one flower stem. It's a graspable unit of perception but not a discrete, constant, objective unit.

    For language teachers — though not for phoneticians and phonologists — allophones are units in differing in ways that are not obvious to the native speaker — although native speakers can be trained to perceive the differences.

    If you and I choose to teach the pronunciation of English /l/ to suitably advanced and suitably motivated students, we may well choose to teach [l] and [ɫ] as two teaching items. For me, this no longer has any theoretical implications. It's just a practical convenience.

    That complementary distribution paradox belongs to the field of phonology, which I think long since turned away from that sort of abstraction to the consideration of acoustic and/or articulatory features.

    ReplyDelete
  17. Well, since we're talking about the philosophical categories 'concrete' (which is *not* the same thing as 'sensual') and 'abstract':

    It's the phoneme which is concrete, as it is to be thought as the concretion of its heterogeneous allophones. It would be thus an example of an 'concrete universal'. On the other hand, it is the phone which is abstract, as it is in its very concept an abstraction from concrete sensual phenomena, i. e. identifying two concrete speech phenomena as utterances of the same phone is based on abstracting the 'this is it' of the phon from co-articulation, particularities of articulation/intonation, base frequency etc. etc..

    I'd be very interested to learn more about the critique of 'phoneme' among linguists. If there's no 'satisfactory replacement term', then this implies that there's a need for one and that the concept is still 'active'. Is Chomsky and Halle’s "The Sound Pattern of English" still the canonical text? Or is there something which might be more up to date? (I wouldn't mind, if there would be something on the internet, since I have no easy access to university resources.)

    ReplyDelete
    Replies
    1. Cited-above Elan Dresher's chapter from the new Companion to Phonology (Blackwell, 2011) is a freebie.

      Delete
    2. Ad Utis

      'Well, since we're talking about the philosophical categories 'concrete' (which is *not* the same thing as 'sensual') and 'abstract':

      It's the phoneme which is concrete, as it is to be thought as the concretion of its heterogeneous allophones.'

      You are using the word 'concrete' in a Hegelian sense; have you, if I may ask, any Hegelian sympathies or inclinations, or does his language feel german to you?

      Delete
    3. sorry, I meant 'germane' of course, I meant do you feel drawn to this language, be it German or English (in translation)?

      Delete
    4. It's true. I didn't think about it, but I have indeed a Hegelian background. As for whether this language is 'germane' -- well, I'm using it. Habitually even. In my understanding the apparent peculiarities of this kind of philosophical terminology stem from a clarification of what appears only as implications in more traditional philosophical language. Your mileage may vary.

      @Darin Flynn: Thank you very much!

      Delete
    5. Well, good old George William Frederic was not such as madman as some Anglo-Saxon philosophers make him out to be.

      Cheers

      Delete
  18. Interestingly, Amis's son Martin also misuses the term phoneme, in much the same way.

    ReplyDelete
  19. These are always a problem in this area, and many of you are not making things easier for us.

    And if the core questing and answer cannot narrow down their peripheries meaningfully, the redundancies are left for further complications than their answers can be drown from their respective semantic features. So the question is then where linguists and students of linguistics can go for an answer by conventional means of their times and research studies if driven conclusions are far away from such analyzes for correctness. In facebooks? In blogs?

    If it is fine among linguists that 'phoneme' is as the primary anatomical-landmark sound-features of the vocal tract that has underlying orthographic representation (though not really in the sense ‘contrastive’ or ‘predictable’, but as the concept of abstractions from concrete sound phenomena) and an allophone as a secondary anatomical-landmark sound-features, the most striking is the take that what constitutes an allophone depends on the linier order of a target phonetic feature and the conditioned tone in question--like why not [l] and [ɫ] phenomenon if so. An answer to this is, yes, but how correctly is our problem since the dark ‘l’ also has the autosegmental realization as an allophone.

    ReplyDelete
  20. Indeed, if there is no problem with dark-l-only accents like mine and many Americans', why should there be a problem with light-l-only accents?

    ReplyDelete