The phoneme

Children do not learn the rules of spoken language by explicit instruction, but rather by a combination of copying what they hear, and building up mental generalizations based on their experiences. How much they are helped in this by some internal structure in the brain dedicated to language acquisition, which linguists call a Language Acquisition Device or Language Faculty, is still a matter of debate.

Nonetheless, aspects of spoken language show very strong similarities to the types of patterns outlined above for writing. Again, some differences between units matter, because replacing one with another will cause a different meaning to be conveyed in the language in question: replace the initial sound [k] in call with [t], and you have tall, an entirely different English word. Correspondingly, English speakers perceive [k] and [t] as entirely separate sounds, and find them rather easy to distinguish.

In other cases, two sounds which phoneticians can equally easily tell apart will be regarded as the same by native speakers. For instance, say the phrase kitchen cupboard to yourself, and think about the first sounds of the two words. Despite the difference in spelling, native speakers will tend to think of those initial consonants as the same – both are [k]s. However, if you say the phrase several times, slowly, and think uncharacteristically carefully about whether your articulators are doing the same at the beginning of both words, you will find that there is a discernible difference. For the first sound in kitchen, your tongue will be raised towards the roof of your mouth, further forward than for the beginning of cupboard; and for kitchen, your lips will be spread apart a little more too, while for cupboard your mouth will be more open. Unless you are from Australia or New Zealand, this difference is even clearer from the phrase car keys, this time with the first word having the initial sound produced further back in the mouth, and the second further forward.

In IPA terms, these can be transcribed as [k], the cupboard sound, and [c], the kitchen one. However, in English [k] and [c] do not signal different meanings as [k] and [t] do in call versus tall; instead, we can always predict that [k] will appear before one set of vowels, which we call back vowels, like the of cupboard or the [ɑ:] a Southern British English speaker has in car, while [c] appears before front vowels, like the [I] of kitchen or the [i:] in Southern British English keys. Typically, speakers control predictable differences of this type automatically and subconsciously, and sometimes resist any suggestion that the sounds involved, like [k] and [c] in English, are different at all, requiring uncharacteristically close and persistent listening to tell the two apart. The difference between [k] and [c] in English is redundant; in phonological terms, this means the difference arises automatically in different contexts, but does not convey any new information.

Returning to our orthographic analogy, recall that every instance of a hand-written a or A will be different from every other instance, even produced by the same person. In just the same way, the same speaker producing the same words (say, multiple repetitions of kitchen cupboard ) will produce minutely different instances of [k] and [c]. However, a hierarchical organization of these variants can be made: in terms of spelling, we can characterize variants as belonging to the lower-case or capital set, and those in turn as realizations of the abstract grapheme .

The subclasses have a consistent and predictable distribution, with upper-case at the beginnings of proper nouns and sentences, and lower case everywhere else: we can say that this distribution is rule-governed. Similarly again, we can classify all the variants we hear as belonging to either fronter [c] or backer [k], although we are not, at least without a little phonetic consciousness-raising, aware of that difference in the way we are with a and A; presumably the fact that we learn writing later, and with more explicit instruction, accounts for our higher level of awareness here.

In turn, [c] and [k], which native speakers regard as the same, are realizations of an abstract unit we call the phoneme (where the ending -eme, as in grapheme, means ‘some abstract unit’). Phonemes appear between slash brackets, and are conventionally represented by IPA symbols, in this case /k/. As with graphemes, we could in principle use an abstract symbol for this abstract unit, say /§/, or /❂/, or give it a number or a name: but again, it is convenient and clear to use the same symbol as one of its realizations. Those realizations, here [k] and [c], are allophones of the phoneme /k/.

To qualify as allophones of the same phoneme, two (or more) phones, that is sounds, must meet two criteria. First, their distribution must be predictable: we must be able to specify where one will turn up, and where the other; and those sets of contexts must not overlap. If this is true, the two phones are said to be in complementary distribution. Second, if one phone is exceptionally substituted for the other in the same context, that substitution must not correspond to a meaning difference. Even if you say kitchen cupboard with the [k] first and the [c] second (and that won’t be easy, because you have been doing the opposite as long as you have been speaking English – it will be even harder than trying to write at your normal speed while substituting small a for capital A and vice versa), another English speaker will only notice that there is something vaguely odd about your speech, if that. She may think you have an unfamiliar accent; but crucially, she will understand that you mean ‘kitchen cupboard’, and not something else. This would not be so where a realization of one phoneme is replaced by a realization of another: if the [k] allophone of /k/ is replaced by the [t] allophone of /t/, then tall will be understood instead of call.

Finally, just as the orthographic rules can vary between languages and across time, so no two languages or periods will have exactly the same phonology. Although in English [k] and [c] are allophones of the same phoneme, and are regarded as the same sound, in Hungarian they are different phonemes. We can test for this by looking for minimal pairs: that is, pairs of words differing in meaning, where the only difference in sound is that one has one of the two phones at issue where the other has the other (think of tall and call). In Hungarian, we find minimal pairs like kuka [kuka] ‘dustbin’ and kutya [kuca] ‘dog’. It follows that [k] and [c] are not in complementary but in contrastive distribution; that interchanging them does make a meaning difference between words; and hence that [k] and [c] belong to different phonemes, /k/ and /c/ respectively, in Hungarian. Unsurprisingly, speakers of Hungarian find the difference between [k] and [c] glaringly obvious, and would be extremely surprised to find that English speakers typically lump them together as the same sound.

As for differences between periods of the same language, it is straightforward to demonstrate that Modern English [f ] and [v] contrast, or are in complementary distribution, since minimal pairs like fat [f ] versus vat [v], leaf versus leave, or safer versus saver are easy to come by. The phoneme system of Modern English therefore contains both /f/ and /v/. However, the situation was very different in Old English, as the examples in (3) show.

Instead of minimal pairs, we find predictable, complementary distribution, with [v] appearing medially, between vowels, and [f ] in other positions. Consequently, [f ] and [v] can be analyzed as allophones of one phoneme, which we might call /f/: Old English speakers would have regarded [f ] and [v] as the same, just as Modern English speakers think of [k] and [c] as the same sound. Later in the history of English, many words like very, virtue and veal were borrowed from French, bringing with them initial [v], which had not previously been found in English. The distribution of [f ] and [v] therefore ceased to be complementary, since both could appear in word-initial position, creating minimal pairs like very and ferry, or veal and feel. In consequence, [v] stopped being an allophone of /f/, and became a phoneme in its own right, producing the opposition of /f/ (realized as [f ]) and /v/ (realized as [v]) we find today.