N. V. Tatsenko introduction to theoretical phonetics of english

phoneme allophone
t
IK
] [t]

^hDp StDp
l ^K
Figure 3.1 - A Phoneme and an Allophone
The phoneme is a functional unit. Function is usually understood to mean discriminatory function, that is, the role of the various components of the phonetic system of the language in distinguishing one morpheme from another, one word from another or also one utterance from another.
The opposition of phonemes in the same phonetic environment differentiates the meaning of morphemes and words, e. g. said - says, sleeper -sleepy, bath - path, light­like. Sometimes the opposition of phonemes serves to distinguish the meaning of the whole phrases, e. g. He was heard badly - He was hurt badly. Thus we may say that the phoneme can fulfil the distinctive function.
The phoneme is material, real and objective. That means that it is realized in speech of all English-speaking people in the form of speech sounds, its allophones. The sets of speech sounds, that is the allophones belonging to the same phoneme are not identical in their articulatory content though there remains some phonetic similarity between them [Vrabel 2009, p. 26].
As a first example, let us consider the English phoneme [d], which when not affected by the articulation of the preceding or following sounds is a plosive, fore-lingual apical, alveolar, lenis stop. This is how it sounds in isolation or in such words as door, darn, down, etc., when it retains its typical articulatory
characteristics. In this case the consonant [d] is called the principal allophone. Principal or typical allophones do not undergo any distinguishable changes in the chain of speech.
At the same time there are quite predictable changes in the articulation of allophones that occur under the influence of the neighbouring sounds in different phonetic situations. Such allophones are called subsidiary. For example:
[d] is slightly palatalized before front vowels and the sonorant [j], e. g. deal, day, did, didyou.
[d] is pronounced without any plosion before another stop, e. g. bedtime, bad pain, good dog; it is pronounced with the nasal plosion before the nasal sonorants [n] and [m], e.g. sudden, admit, could not, could meet; the plosion is lateral before the lateral sonorant [l], e.g. middle, badly, bad light.
Followed by [r] the consonant [d] becomes post-alveolar, e. g. dry, dream; followed by the interdental [9], [6] it becomes dental, e.g. breadth, lead the way, good thing.
When [d] is followed by the labial [w] it becomes labialized, e.g. dweller.
Allophones of each phoneme possess a bundle of distinctive features that make this phoneme functionally different from all other phonemes of the language concerned. This functionally relevant bundle of articulatory features is called the invariant of the phoneme. The articulatory features which form the invariant of the phoneme are called distinctive or relevant. To extract a relevant feature of the phoneme we have to oppose it to some other phoneme in the same phonetic context. If the opposed sounds differ in one articulatory feature and this difference brings about changes in the meaning of the words, the contrasting features are called relevant. For example, the words port and court differ in one consonant only, that is the word port has the initial consonant [p], and the word court begins with [k]. Both sounds are occlusive and fortis, the only difference being that [p] is labial and [k] is backlingual.
47
Therefore, it is possible to say that labial and backlingual articulations are relevant in the system of English consonants.
The articulatory features which do not serve to distinguish meaning are called nondistinctive, irrelevant or redundant; for instance, it is impossible in English to oppose an aspirated [p] to a non-aspirated one in the same phonetic context to distinguish meanings. That is why aspiration is a non­distinctive feature of English consonants.
If an allophone of some phoneme is replaced by an allophone of a different phoneme the mistake is called phonological, because the meaning of the word is inevitably affected, e. g.: beat - bit.
If an allophone of the phoneme is replaced by another allophone of the same phoneme the mistake is called phonetic. It happens when the invariant of the phoneme is not modified and consequently the meaning of the word is not affected [Vrabel 2009, p. 27].
Allophones often show up when people have different accents. One good example is the word "butter". Some native speakers will say [Ьлгэ]. Others will say [bvto]. You can see here that [t] and [r] are allophones of the same phoneme. Whatever way you say it, the meaning of the word does not change. It's still the yellow stuff made from milk that you put on bread.
Thus, allophones are the linguistically non-significant variants of each phoneme. In other words, a phoneme may be realised by more than one speech sound and the selection of each variant is usually conditioned by the phonetic environment of the phoneme. Occasionally allophone selection is not conditioned but may vary from person to person and occasion to occasion (ie. free variation).

2. 
   Notation
   

The abstractional and material aspects of the phoneme have given rise to the appearance of transcription. Transcription is a set of symbols representing speech sounds. The symbolization of sounds naturally differs according to whether the aim is to indicate the phoneme, i. e. a functional unit as a whole, or to reflect the modifications of its allophones as well. The International Phonetic Association (IPA) has given accepted values to an inventory of symbols, mainly alphabetic but with additions. The first type of notation, the broad or phonemic transcription, provides special symbols for all the phonemes of a language. The second type, the narrow or allophonic transcription, suggests special symbols including some information about articulatory activity of particular allophonic features [Vrabel 2009, p. 27].
The broad transcription is mainly used for practical expedience; the narrow type serves the purposes of research work. The first type was introduced by Daniel Jones. He realised the difference in quality as well as in quantity between the vowel sounds in the words sit and seat, pot and port, pull and pool. According to Daniel Jones' notation, English vowels are denoted like this: [i] - [i:], [e] - [ж], [л] - [a:], [э] - [э:], [u] - [u:], [э] - [з:]. This way of notation disguises the qualitative difference between the vowels [i] and [i:], [э] and [э:], [u] and [u:], [э] and [з:] though nowadays most phoneticians agree that vowel length is not a distinctive feature of the vowel, but is rather dependent upon the phonetic context, that is it is definitely redundant. For example, in such word pairs as hit - heat, cock - cork, pull - pool the opposed vowels are approximately of the same length, the only difference between them lies in their quality which is therefore relevant [ibid].
The other type of broad transcription causes no phonological misunderstanding providing special symbols for all vowel
phonemes: [i], [i:], [e], [x], [a:], [л], [v], [э:], [u], [u:], [э], [з:]. The narrow or phonetic transcription incorporates as much more phonetic information as the phonetician desires, or as he can distinguish. It provides special symbols to denote not only the phoneme as a language unit but also its allophonic modifications. The symbol [h], for instance, indicates aspirated articulation: [kheit] - [skeit]. The following symbols of phonemes are used in the modern transcription: