When we speak, we produce a continuous stream of sounds. This stream of sounds is then divided into small pieces that are called segments. For a better understanding let’s take as example the word "man". It is made up of three segments: the first segment /m/, the second segment /æ/, and the third segment /n/.
Everything looks very easy but actually it is not. Deciding on the number of segments is not easy and in the past it was always a controversial topic. Let’s take another word, for example the word "mine". The first and the last segment are identical to the first and last segments in the word "man"; but what about the segment(s) in between them? Is /aɪ/, in the middle, one segment or two?
As I have just said, we can divide speech up into segments, and we can find a great variety in the way these segments are made. So, there is an abstract set of units as the basis of our speech. These units are called phonemes and the complete set of these units is called phonemic system of the language. The phonemes themselves are abstract, but there are many different ways we make the sounds that represent these phonemes.
There are cases where it makes little difference which of two possible ways we choose to pronounce a sound. Let’s take as example the word "bad". The /b/ at the beginning of the word "bad" is normally pronounced with no voicing. Sometimes the speaker produces this /b/ with full voicing (maybe in speaking very emphatically). However, the sound is still identified by the listener as being /b/ although we can hear that it is different in some way.
So, in this example, we have two different ways of making a phoneme; one can be substitute for another without changing the meaning of the word. In this example, we have two different realizations of the phoneme "b", one can be substitute for the other one without changing the meaning. The two realizations are said to be in free variation.
We also can find cases in speech where one phoneme can occur where the other one can not. Let’s take as example the word "tea". The /t/ in the word "tea" is aspirated exactly as are all voiceless plosives when they occur before stressed vowels at the beginning of a syllable. In the word "eat", the realization of /t/ is un-aspirated exactly as are all voiceless plosives when they occur at the end of a syllable and are not followed by a vowel.
When two realizations can only occur in a particular place (as is shown above) we say that the realizations are in complementary distribution. Definitely, the aspirated and the un-aspirated /t/ will both be identified by the listener as /t/ but the aspirated realization will never be found in the place where the un-aspirated realization is appropriate (or vice-versa).
Sometimes, these different realizations of phonemes are called allophones. So, we have taken a look at the aspirated and un-aspirated allophones of the phoneme "t". When symbols (that represent sounds) are written, usually the allophones are not indicated.