M is for Minimal pairs

8 10 2017

The story of the Australian pig farmer whose livestock were decimated by floods has been circulating on the Internet recently. A reporter misheard him say that ‘Thirty thousand pigs were floating down the river’, and reported it as such. In fact, what he had said was: ‘Thirty sows and pigs…’.  A nice example of how a minimal pair mistake can cause problems even among native speakers.

Just to remind you, here’s how minimal pairs are defined in The New A-Z of ELT:

A minimal pair is a pair of words which differ in meaning when only one sound (one phoneme) is changed. Pair and bear are minimal pairs, since their difference in meaning depends on the different pronunciation of their first sound: p versus b. However, pair and pear are not minimal pairs, since, although they differ in meaning, they are pronounced the same. Minimal pairs are widely used in pronunciation teaching to help learners discriminate between sound contrasts, particularly those that don’t exist in their L1, for the purposes of both recognition and production.

On the MA course I teach for The New School, I set the students a task in which they describe how they might exploit this kind of minimal pairs activity (from Baker 2006):

ship or sheep 2006

Here’s my feedback on the task:

As I suggest, such activities may have limited usefulness. Indeed, does anyone still do them?


Baker, A. (2006) Ship or sheep? (2nd edn). Cambridge: Cambridge University Press.


P is for Phoneme

17 03 2013

aeIs the phoneme dead?

We’ve been doing a unit on phonology, and my doubts about the phoneme are partly a reflection of my students’ own difficulties with the concept.  Not surprisingly, I’ve been having to tease out the difference between phonemic symbols and phonetic symbols, and even between phonology and phonics.

But all the time I’ve been dreading the day when someone challenges this definition (from An A to Z):

‘A phoneme is one of the distinctive sounds of a particular language. That is to say, it is not any sound, but it is a sound that, to speakers of the language, cannot be replaced with another sound without causing a change in meaning’.

The definition has an authoritative ring to it, not least because it simply re-states what by many is considered a founding principle of all linguistics. Listen to Jakobson (1990: 230) who practically bellows the fact: ’The linguistic value … of any phoneme in any language whatever, is only its power to distinguish the word containing this phoneme from any words which, similar in all other respects, contain some other phoneme’ (emphasis in original).

dHow is it, then, that we regularly teach that the ‘s’ at the end of cats is a different phoneme than the ‘s’ at the end of dogs?  If different phonemes flag different meanings, what change of meaning is represented in the difference between /s/ and /z/? Or, for that matter, between final /t/ and final /d/, as in chased and killed?   If there is no difference in meaning (since /s/ and /z/ both index plurality, and /t/ and /d/ both index past tense), aren’t they simply different ways of pronouncing the same phoneme?

Phonemes, after all, are not phones, i.e. sounds. Acoustically speaking there are many different ways – even for a single speaker – of realizing a specific phoneme. This is why Daniel Jones (1950: 7) defined phonemes as ‘small families of sounds, each family consisting of an important sound of the language together with other related sounds’ (my emphasis). These related sounds are the different allophones of the phoneme.

Hence the analogy with chess pieces: the way individual chess pieces are designed will vary from set to set, but they will always bear certain family resemblances, bishops all having mitres, and knights having horse heads, etc. More important than their form (and one reason that this analogy seems to work so well),  is the relationship that they have with one another, including the ‘rules’ that constrain the way that they may behave. Bishops can’t do what knights do, nor go where knights go, and vice versa.

Phonemes – like chess pieces – are defined in relation to one another. As Bloomfield (1935: 81) put it, ‘the phoneme is kept distinct from all other phonemes of its language. Thus, we speak the vowel of a word like pen in a great many ways, but not in any way that belongs to the vowel of pin, and not in any way that belongs to the vowel of pan: the three types are kept rigidly apart.’

ngIn fact, a purely structuralist argument would say it’s not actually about meaning at all, it’s about ‘complementary distribution’, or, as Jones (1950: 132) puts it (also bellowing): ‘NO ONE MEMBER EVER OCCURS IN A  WORD IN THE SAME PHONETIC CONTEXT AS ANY OTHER MEMBER’.  That is to say, the /s/ at the end of cats and the /z/ at the end of dogs never occur where the other occurs, and vice versa. But is this true? What happens to the /z/ at the end of dogs in the sentence: The dogs seem restless? Hasn’t it become /s/?

Ah, yes, you say – but sounds in connected speech are influenced by their environment, blending with or accommodating to the sounds around them. The true test for a phoneme is if it distinguishes isolated words, like pin and pen – those infamous minimal pairs. But when are words ever isolated? When does the phonetic environment not have an effect?  And isn’t the voiced /z/ at the end of dogs, and the unvoiced /s/ at the end of cats also an effect of the phonetic environment? That is to say, where does connected speech start becoming connected if not at the juxtaposition of two sounds?

It gets even trickier when we consider weak forms. There are at least two different ways of saying can, as in I can dance: I /kæn/ dance, or I /kən/ dance. Both are possible, even where the stress remains on dance. The latter is simply more reduced. But the meaning is unchanged. [kæn] and [kən] are not minimal pairs. They are different phonetic realizations of the same word (hence the square brackets). Phonetic. Not phonemic. Shouldn’t, therefore, they both be transcribed as /kæn/?

In researching this, I’ve encountered a lot of debate as to whether the concept of the phoneme has any currency at all any more. As one scholar puts it, ‘the phoneme, to all appearances, no longer holds a central place in phonological theory’ (Dresher 2011: 241). The problem seems to boil down to one of identification: is the phoneme a physical thing that can be objectively described, or is it psychological – a mental representation independent of the nature of the acoustic signal?

eThe answer to the first question (is it physical?) seems to be no, there are no ‘distinctive features’ or family resemblances (such as voicing or lip-rounding) that unequivocally categorize sounds as belonging to one phoneme family and not another.

On the other hand, there is some evidence, including neurological, that the phoneme does have a psychological reality, and that speakers of languages that share the same sounds will perceive these sounds differently, according to whether they flag meaning differences or not. (This is analogous to the idea that if your language does not distinguish between blue and green, you will see both blue and green as being shades of the same colour).  This, in turn, is consistent with Jakobson’s claim that ‘if we compare any two particular languages, we will see that from an acoustic and motor point of view their sounds could be identical, while the way they are grouped into phonemes is different’ (p. 223).

It’s not for nothing, therefore, that the concept of the phoneme has given us the very valuable distinction between emic and etic, i.e. the perspective of the insider vs that of the outsider. Phonemes capture something that we, the insiders, intuit about language, even if their objective reality is elusive. We know that pronunciation impacts on meaning, even if we don’t quite know how.

Perhaps Jakobson (op. cit. 230) had good reason to claim, therefore, that ‘the phoneme functions, ergo it exists’.


Bloomfield, L. (1935) Language, London: George Allen & Unwin.

Dresher, E. (2011) ‘The Phoneme’, in van Oostendorp, M., Ewen, C.J., Hume, E., & Rice, K. (eds) The Blackwell Companion to Phonology, Oxford: Blackwell, available here

Jakobson, R. (1990) On Language, edited by Waugh, L.R. & Monville-Burston, M., Cambridge, Mass: Harvard University Press.

Jones, D. (1950) The Phoneme: Its nature and use, Cambridge: W. Heffer & Sons.

Illustrations from the very clever phonemic chart that comes with English File (Oxenden, C. and Seligson, P., 1996, Oxford University Press).

P is for Pronunciation

1 08 2010

Read my lips

I’ve just completed a nine-hour block of sessions on phonology on the MA TESOL course that I’m teaching at the New School. Apart from the inevitable (and sometimes intractable) problems involved in reconfiguring my knowledge of phonology so as to accommodate North American accents, the question that simply will not go away is this: Can pronunciation be taught?

As a teacher, I have to confess that I can’t recall any enduring effects for teaching pronunciation in class – but then, I very seldom addressed it in any kind of segregated, pre-emptive fashion. Most of my ‘teaching’ of pronunciation was reactive –  a case of responding to learners’ mispronunciations with either real or feigned incomprehension. There are only two pron-focused lessons that I can remember feeling good about: one was where I used an inductive approach to guide a group of fairly advanced learners to work out the rules (or, better, tendencies) of word stress in polysyllabic words (the students seemed generally impressed that the system was not as arbitrary as it had appeared), and another where I used a banal dialogue that happened to be in the students’ workbook to highlight the different spellings of the /ay/ phoneme – a lesson that was more about spelling than pronunciation, really – but, again, one that helped dispel the myth that there are zero sound-spelling relationships in English.

As a second language learner, any attempts to improve my pronunciation have fallen (almost literally) on deaf ears. I remember being told by a well-intentioned Spanish teacher: “Your problem is that you use the English ‘t’ sound instead of the Spanish one”. To which I replied, “No, the ‘t’ sound is the very least of my problems! My problem is that I don’t know the endings of the verbs, that I don’t have an extensive vocabulary, that I can’t produce more than two words at a time. … and so on”. That is to say, in the greater scheme of things, the phonetic rendering of a single consonant sound was not going to help me become a proficient speaker of Spanish. Nor was it something I would be able to focus any attention on, when my attention was so totally absorbed with simply getting the right words out in the right order. And nor, at the end of the day, would I ever be able to rid myself of my wretched English accent, however hard I tried (assuming, of course, I wanted to).

Hence, I’m fairly sceptical about the value of teaching pronunciation, and I suspect that most of the exercises and activities that belong to the canonical pron-teaching repertoire probably have only incidental learning benefits.  A minimal pairs exercise (of the ship vs sheep type) might teach some useful vocabulary; a jazz chant might reinforce a frequently used chunk. But neither is likely to improve a learner’s pronunciation. Certain learners (a small minority, I suspect) with good ears and a real motivation to “sound like a native speaker” might just squeeze some benefit out of a pron lesson, but for the majority it will probably just wash right over them.

In An A-Z of ELT, I hint obliquely at these doubts – doubts which I claim are justified by research studies. What studies?

Well, here’s one for starters. In an early attempt to tease out the factors that predicted good pronunciation, Suter (1976) co-opted a panel of non-specialist informants to assess the pronunciation of 61 English learners from a range of language backgrounds and with different histories of exposure and instruction. Twelve biographical factors were found to correlate with good pronunciation, and, in a subsequent re-analysis of the data (Purcell and Suter 1980), these were reduced to just four. These four predictors of acceptable pronunciation were (in degree of importance):

  • the learner’s first language (i.e., all things being equal, a speaker of, say, Swedish is more likely to pronounce English better than a speaker of, say, Vietnamese)
  • aptitude for oral mimcry (i.e. ‘having a good ear’)
  • length of residency in an English-speaking environment
  • strength of  concern for pronunciation accuracy

Significantly, none of the above factors is really within the teacher’s control (although the last – the motivtaional one – could arguably be nurtured by the teacher). Nevertheless, the learners’ histories of instruction seemed not to have impacted in any significant way on the accuracy of their pronunciation. The researchers commented: “One of the most obvious [implications of the study] relates to the fact that teachers and classrooms seem to have had remarkably little to do with how well our students pronounced English”.

Now, is this bad news (we can’t do much to help our learners achieve acceptable standards of pronunciation)? Or is it good news (we don’t have to teach pronunciation, and can spend the time saved on more important stuff)?


Purcell, E.T., and Suter, R.W. 1980. Predictors of Pronunciation Accuracy: a Re-examination. Language Learning, 30, 271-287.

Suter, R.W. 1976. Predictors of Pronunciation Accuracy in Second Language Learning. Language Learning, 26: 233-253.