Minimal pair in general

In phonology, a minimal pair is a pair of significative units – commonly, words – which differ only in one phonological segment (phone). Engl. bit vs. bet are an example.

Minimal pairs play an important methodological role in determining whether a certain phone found in recordings of a language represents a phoneme of its own, or rather, whether two phones represent two distinct phonemes or are just variants. The presentation of a minimal pair counts as a demonstration that the two phones by which the two significative units differ

By this argument, the [ɪ] and [ɛ] contained in the example pair represent distinct phonemes, viz. /i/ and /e/.

Generally, if two sounds can at all constitute an opposition in a language, they do so in more than one minimal pair. This is just a consequence of an economic deployment of system resources. The higher the number of such minimal pairs, the higher the functional load of the opposition in question. Occasionally, only one minimal pair can be found for a given pair of phones. For instance, [h] and [x] (the latter being an allophone of the phoneme /ç/) contrast in one minimal of the German language, viz. Chauke [member of some West-Germanic tribe] - Hauke [male given name]. Both of these words are marginal to the lexicon. Then the question arises whether such evidence is sufficient to establish two phonemes. After all, these two particular phones are mostly in complementary distribution: [h] occurs exclusively in the syllable onset, while [x] (like [ç]) occurs almost only in non-initial position, and in initial position almost only in loans.

Near minimal pair

If one desires to make progress in the analysis in such a methodological situation, a possible solution is to loosen the requirements. The device frequently invoked is the near minimal pair. A German example of a near minimal pair opposing [ç] (the other allophone of /ç/) to [h] is Chinin 'quinine' - hinan ‘upwards’. As may be seen, the two words differ not only in the sound in question, but also in the vowel of the second syllable.

It is clear that a near minimal pair does not demonstrate that only just the sound in question is sufficient to distinguish a meaning and therefore must be distinctive. The other difference occurring in the near minimal pair is equally responsible for the difference in meaning. In such a situation, it is important that the phonetic difference at stake cannot possibly be conditioned by the other difference in its context. For the example in question, this means that the alternation of [ç] ~ [h] in Chinin vs. hinan cannot be conditioned by the contrast between /iː/ and /a/ in the following syllable. If this were the case, then [ç] ~ [h] would be (conditioned) allophones of one phoneme, and the argument based on a near minimal pair would be spurious. In fact, the exact opposite of what one wanted to demonstrate would be shown.

In the case at hand, such a conditioning can indeed be excluded. Not only is no phonological rule known by which /iː/ vs. /a/ would condition an alternation of [ç] ~ [h] in the onset of the preceding syllable. What is more, vowels do not, in general, condition consonantal allophony in the onset of the preceding syllable. Thus, a conditioned allophony can safely be excluded for this near minimal pair; and /ç/ ~ /h/ come out as German phonemes, even though their contrast carries an extremely low functional load.

An example of a phonetic difference for which there is not a single minimal pair in the language is presented by the pair [ð] vs. [θ] in English. What we have are exclusively near minimal pairs like that - thatch, these - thesis, this - thistle, thou - thousand and thy - thymus (the enumeration is complete). In comparing these pairs, one notes two kinds of differences between their members:

  1. The member of the pair containing [θ] is longer by a few phonemes than the member containing [ð].
  2. The member of the pair containing [ð] is a monomorphematic and monosyllabic grammatical formative, while the member containing [θ] is a lexical word.

While no phonological mechanism of English is known which would condition an alternation of these two phones at the onset of a word simply because one is longer than the other, there might well be a condition restricting initial [ð] to grammatical formatives. Thus, the decision whether these two phones are considered distinct phonemes of English depends on whether one accepts such a morphological condition as #b for a complementary distribution of phones.

At the same time, it is clear that the concept of near minimal pair must be handled very restrictively. As a matter of fact, the argumentation based on minimal pairs is commonly necessary at an early stage of the investigation of a language. In such a situation, one cannot be sure that one knows all the phonological rules that can condition an alternation. Consequently, if for a given difference only near minimal pairs can be found, doubts remain. By the same token, the additional difference in a near minimal pair must indeed be minimal. It makes no sense to consider though and thought a near minimal pair of English, since apart from the final consonant (which presumably makes no relevant difference), the vowel immediately following the fricatives in question differs, too. A too liberal use of the device of ‘near minimal pair’ jeopardizes the analytic methodology.

Optimal minimal pair

On the other hand, the condition on minimal pairs can also be straightened. An optimal minimal pair is one in which the two significative units compared differ only in the value of one phonological feature. For instance, pet vs. bet differ only in the voicing of the initial bilabial stop; all its other features are the same. This is, thus, an optimal minimal pair. On the other hand, pet vs. let is a minimal pair, too, but a suboptimal one. Such a pair can prove that [p] and [l] must represent distinct phonemes. But it does not exclude the possibility that [p] is a conditioned variant of [b]. This is only excluded by the optimal minimal pair adduced.

The net balance is: The phonological contrasts of a language are best demonstrated and illustrated by optimal minimal pairs. Failing these, minimal pairs in general do the service to a large extent. In the absence of minimal pairs, near minimal pairs can be resorted to. However, their controlled use presupposes a degree of knowledge of the phonological system which is generally beyond the endeavour of discovering the phoneme system of the language.