The net result of this technique is the extraction of 2-length and longer components from all sequences that can be matched against non-occurring sequences (a sequence being an environment and the phoneme that occurs in it); e.g. from English /rtr/ matched against /trt/. These components do part of the work of identifying and phonetically describing the phoneme over which they extend (e.g. the ‘gives the aspiration component of Greek ɸ), so that only a residue of the original phoneme is required to accompany them (in this case /p/ to accompany ‘: /’p/ = ɸ). This residue in turn can designate another phoneme which occurs without the component (/p/ = π).
- Harris (1944), Pag. 198