A category with contextual variants
Allophones are realizations associated with a phoneme under a particular analysis. In complementary distribution, two sounds occur in different environments rather than competing in the same position. That pattern can support grouping them together, especially when they are phonetically related. It is a hypothesis about a system, not a rule that any two nonoverlapping sounds must automatically be one phoneme.
Prediction needs an environment
An allophonic statement should say where a realization occurs: next to a particular sound, in a stressed syllable, at a word edge, or under another specified condition. Without the environment, “these sounds are the same” is too vague. The sounds may be perceptually and physically different while serving as members of one contrastive category. Speakers need not be consciously aware of the pattern for it to be systematic.
Variation can have several sources
Not every difference is a categorical allophonic rule. Speaking rate, register, individual habits and dialect can affect realizations. A researcher may distinguish contextual allophony from variation within an environment, while noting that the boundary is not always simple. The same phonetic difference can distinguish words in another language. Always keep the variety and the level of transcription visible when comparing analyses across sources.
Allophony links surface differences to a category and an environment; the analysis belongs to a particular system.
Try it for yourself
In The sound switch, select Place only and change voicing. Observe a physical label change without a category change; this grouping is not itself a contextual allophonic rule.
- Essentials of Linguistics, 2nd edition — 4.1 Phonemes and allophones (printed pp. 165–168)Catherine Anderson; Bronwyn Bjorkman; Derek Denis; Julianne Doner; Margaret Grant; Nathan Sanders; Ai Taniguchi · 2023, version 2.2 (February 2023), repository PDFRights & edition ↗
- Essentials of Linguistics, 2nd edition — 4.4 Complementary distribution (printed pp. 175–177)Catherine Anderson; Bronwyn Bjorkman; Derek Denis; Julianne Doner; Margaret Grant; Nathan Sanders; Ai Taniguchi · 2023, version 2.2 (February 2023), repository PDFRights & edition ↗