Minimal contrastive pair
The concept of the minimal contrastive pair is central to the scientific study of language, particularly that of sound.
Very logically, it describes two things (a 'pair') which can be told apart ('contrastive') in such a way that the difference is the smallest ('minimal') needed to tell two items of the language apart.
For example the stress difference between 'REcord' and 'reCORD' is enough to tell native speakers that the first is a noun and the second a verb. Hence these two stress patterns form a minimal contrastive pair - they are the only audible difference between the two words.
Most of the consonants used in English can be grouped into pairs by their voicing. For each pair of voiced and unvoiced consonants, it is possible to find a pair of words whose difference is only marked by the difference between the members of that pair. 'bit', for example, is only distinguished from 'pit' by the voicing of the consonant made with both lips: 'b-' is voiced, and 'p-' is unvoiced. But the two words are markedly different! Similarly, the first consonants of 'got' and 'cot' give us the only difference between the two words, and the only difference between them is their voicing. 'bed' and 'bet' are distinguished in speech by the voicing of their final consonants. In writing, several spelling differences help us to distinguish between 'face' and 'phase'; but in speech, it is the voicing of the sibilants '-s-' and '-z-' that provides the minimally contrastive pairing.
(Minimal contrastive pairs in which a voiced consonant in the first word is replaced by the corresponding unvoiced consonant in the second (e.g.. ‘bed’ and ‘bet’, ‘goat’ and ‘coat’, ‘view’ and ‘few’, ‘found’ and ‘fount’) involve, when spoken, exactly the same movements of the lips. Consequently they will look alike to a lip-reader and cannot be told apart by a deaf person lip-reading. Such pairs of words are known as homophenes.)
In research in linguistics, much time and effort is expended in finding examples of minimally contrastive pairs by which to prove the significance of a language feature that has been noticed.