back to course outline                                  

Phones, phonemes and allophones.

As a starter:

In any one language or dialect there are usually rather more sounds than speakers are aware of. For instance, in many types of Southern British English the GOAT vowel  the words "code" and "cold" are usually different from each other.  Can you hear the difference?

If we swap these vowels around, we can hear that the words do not sound the same - for many people, this pronunciation of "cold" will sound old -fashioned or "posh". However, normally speaking, people don't realise they are using different vowels io these words.  For one thing, we use the same letter - o - to write them.

Similarly, Southern British speakers usually use a different l-sound in "lip" than they do in "pill". Listen now to these two words, first in a normal Southern British accent, and then with the consonants swapped around:

Again, swapping the sounds around makes "lip" sound a bit Scottish or American, and "pill" perhaps a bit Welsh. At all events, it's not a normal Southern British pronunciation.

But as I said, people are not usually aware of these differences. If you look up these words in a good pronunciation dictionary such as Well's's Longman Pronunciation Dictionary you'll find that only is given for both "code" and "cold" and only is given for both "lip" and "pill". The reason for this is that there are hard and fast rules for when each variety is used:

And so when we talk about the sound "l" or "o", we are really talking about a group of sounds which speakers usually "feel" are the "same sound", although they vary according to the sounds which come before them or after them. The point to remember is that, in any one accent or dialect, these variations are strictly according to rule, althouth the speakers themselves are generally quite unaware of them, and produce the right sounds without thinking.

One of the tasks of phonology is top discover these rules.

So, to sum up:



We can start out by thinking of the SET OF PHONEMES for any language as a PRACTICAL, LOGICAL ALPHABET for that language. It's an alphabet with a one-to-one relationship between letters and sounds - no ph for f, no more silent letters such as k in 'knife' and gh in 'thought' and e in' stone'. The words 'phone' and' fun' would both be written f-something -n. 'Write', 'right' and 'rite' would all be written the same: two consonants with a vowel between them. And no letter would have more than one sound - forget c, forget gh. The bad speller's paradise, in fact.

And, crucially, it's an alphabet which shows all the differences between phones (=sounds) that MATTER, and ignores those that don't.

Some differences matter, some don't...???

The difference between t and d is obviously significant in English - 'town' and 'down' are different words. The same goes for 'cod' and 'cad' and 'keyed' and 'code' - they're all k-d with different vowels in the middle.

But some differences between phones are non-significant, and usually we don't think about them, or even notice them. For instance, in most varieties of English, we use a different t-sound in 'top' and 'stop' - the t in 'top' has quite a strong puff of air following it, while the t in 'stop' doesn't. In the south of England, most people have a different o-sound in 'code' and 'cold', and in RP the l in 'lip' is rather different from the l in 'milk'.

These differences are non-significant: they have nothing to do with meaning, but with the structure of the language. If you use the same type of o in 'code' and 'cold', or the 'wrong' type of l in 'lip' and 'milk', you'll sound strange, or old-fashioned, or posh, or uneducated, or foreign, - but you won't be changing meaning of the word. These two types of o or l are different realizations of the 'same' sound.

Our 'logical' phonemic alphabet doesn't need to show the non-significant differences, because they conform to simple rules which the speakers follow unconsciously, and usually without even knowing that the rules exist . For instance, there's a rule which says that the diphthong o is pronounced one way before an l (as in 'cold') and another way elsewhere (in 'code'). Another rule says that l has one form before a vowel ('light l') and another if no vowel follows ('dark l').

These non-significant differences are thus rule-bound; they are predictable according to context.

Phonemes - allophones

  • PHONEMES are the basic sounds - the significant , non-predictable ones.
  • The different ways the phonemes are realised in various positions are called ALLOPHONES - predictable, and non-significant.
  • Same difference!

    Usually, of course, the different ALLOPHONES of the same PHONEME are all similar to each other - they form a FAMILY of sounds. The two o's in 'code' and 'cold' sound very similar - at least to an English speaker. But we mustn't fall into the trap of thinking that ALLOPHONIC difference is small while PHONEMIC difference is large. There is actually no real difference between these differences! We can see this by the fact that the same difference can be allophonic in one language, and phonemic in another.

    Let's look at some examples. English as we know has the sounds s and sh:

    seat/ sheet, massive/machine, basic/nation Obviously these two sounds are SIGNIFICANTLY different from each other in English. Japanese also has s and sh: look at 'Mitsubishi' and ' Subaru'. But to our surprise, we find that Japanese speakers tend to mix up s and sh in English - it's as if they can't hear the difference between them as well as English speakers. And it turns out that the difference in Japanese is ALLOPHONIC rather than PHONEMIC - it is non-significant, predictable, unconscious. In Japanese there is only one phoneme, s, which is realised as sh in front of front vowels, and s elsewhere. So the Japanese word 'shimasu' (do) is phonemically 'simasu'. The distributuion of s and sh in Japanese is entirely predictable by their environment. We say that allophones have complementary distribution (Icelandic 'fyllidreifing'). In English, s and sh are phonemes, and so have contrastive distribution.

    Languages differ as to which differences are significant or not.

    PHONEMES are realised as ALLOPHONES:







    contrastive distribution

    complementary distribution

    broad transcription /.../

    narrow transcription [...]


    We use the method of building MINIMAL PAIRS to discover whether a sound is a significant one or not. So that we can say that in English p and b are phonemes, because they make a distinction for instance between the minimal pairs 'pat' and 'bat'.

    Sometimes it isn't easy to find an actual set of words to make a minimal pair. For instance°we cannot find a a minimal pair for the medial sounds in 'mission' () and vision' (), so we make do with MINIMAL CONTEXTS

    But we cannot make minimal pairs or minimal contexts with predictable sounds - allophones - a new word is not produced, but simply an unusual or "wrong" pronunciation of the old word.

    A phoneme is thus an abstract idea, not a sound. When a phoneme is REALISED (= translated into sound) we use one of its ALLOPHONES.

    back to course outline back to Phonetics back to Pétur Knútsson's home page

    If you have any questions mail me at