A trained phonetician - someone trained to distinguish sound types as mechanically and efficiently as a machine - would soon notice that from time to time Welsh speakers produce sounds rather like middle consonant of Spanish caja (and the final consonant of Scottish loch) and unlike anything ever produced by an English speaker, at least while speaking. But the same phonetician would also note that voiceless l-sounds - voiceless laterals, as he or she would say - occur in utterances in both languages. (What is a voiceless lateral? Well, it is the kind of 'l' which you produce in English in the word pleat . Just say the word slowly and listen to yourself.)
OK, you say. Welsh and English share the use of some sounds but not others. This is really hardly surprising, is it? No. It isn't. But what is interesting about Welsh versus English is that although the two languages happen to agree on using essentially the same voiceless laterals, they use them in very different ways.
Let's look at English first and compare the lateral in pleat with the lateral in bleat. If you say the two words to yourself - and pay close attention - you will notice that the two l's are quite different from one another. The first is said to be voiceless (as you already know) and the second is voiced. (What the labels really mean is not important for the moment. All that matters is that you detect the difference.)
Now notice that, however much you repeat these two words, although there will inevitably be minor variations in your pronunciation , the voiceless-voiced difference between the l's will be consistently maintained. Notice too that other words which start off in the same way show the same consistent difference. Try, for example, planned versus bland or plume versus bloom. Even in pairs of words with other consonants preceding the lateral we find the same general pattern repeated. The lateral of clean is voiceless as against the the lateral of glean which is voiced.
It is clear from this, then, that an English speaker makes systematic use of two kinds of lateral. But - and here comes the crucial point - despite an English speaker's ability to articulate voiceless and voiced laterals with perfect control, this difference in sounds is never used in English as the only distinguishing characteristic of any two words. And the reason for this is that the two different laterals can never appear in the same environments. Their distribution is non-overlapping or complementary. The voiceless lateral appears after p and k but never after b and g, while the voiced lateral appears after b and g but never after p and k. They are, therefore, never in a position to be used contrastively. In these circumstances, a phonologist would say that difference between the two sounds is non-distinctive (or non-contrastive) and would treat the two different laterals as no more than positional variants - or allophones - of some rather more abstract element of the English sound system, the /l/ phoneme. (Phoneme symbols are usually written between slashes and allophones between square brackets when it is important - as it is here - to make the distinction clear.)
Now, in Welsh, the situation is very different. Welsh speakers also make systematic use of voiced and voiceless laterals, but in Welsh the two different sounds can be used in identical contexts - they can, in other words, be opposed or contrasted. As a result of this possibility the very same sound difference that we found in English is, in Welsh, capable of supporting a meaning difference between words, even to the extent of being sometimes the only such support. For example, the word pwll (two l's are used in the spelling for a voiceless lateral) is more or less equivalent to English pool but pwl, which differs only in having a final voiced lateral, means fit. Llard means scrap or fragment, but lard means lard. Palu means dig, but pallu means fail. Such minimal pairs obviously provide the most convincing proof that in Welsh - unlike English - the difference between voiced and voiceless laterals is distinctive and we are obliged to say, as a result, that there are two lateral phonemes in the language. From a physical point of view, the sounds in English and Welsh may be by and large the same: their function, however, is quite different.
Treating phonemes as atoms is fair enough when we are directing our attention to sequential elements, but notice that, in talking of the voiced and voiceless lateral phonemes of Welsh, we have, of course, already implicitly assumed that these sequential atomic elements may be 'decomposed' into yet smaller simultaneous components. The next page takes a closer look at these sub-atomic elements.