Recognition of allomorphs

Rule R 2 is concerned with allomorphy. Allomorphy is variations in the form of a morpheme. These variations clearly must not be so extreme that a native speaker fails to identify the phonemes with the morpheme they claim to represent. If a native speaker does not recognize the variations as instances of the same morpheme, then he recognizes them as two different morphemes and the matter is no longer a question of allomorphy.

The central problem is how to define the permissible limits of allomorphic variation, i. e. how much divergence, and of what kinds, before the identity is broken and a new morpheme has to be established. Allomorphy is usually limited to variations within the phonemic shape of a morpheme. Occasionally (as here) graphemic variation is considered; semantic variation much less often. All three variations can be subsumed under the term allomorphy, always provided a recognizable degree of identity is maintained.

These examples illustrate the range of the phenomenon.

Phonemic variation only:

worth: worthy (shows [th]: [ th ] shift only).

Graphemic variation only:

bone: bony (shows variation E: Ø only); panic: panicky (shows C: CK shift).

Semantic variation only:

accident: accidental (shows semantic shift: accident = “a sudden event … producing an unfortunate result”; accidental = “occurring sometimes [my italics] with unfortunate results”).

There can naturally be combinationsof phonemic, graphemic andsemantic variation.

Examplesare:

(with both phonemic and graphemic variation)

face: facial

(shows CE: CI shift and [s]: [sh] shift)

abdomen: abdominal

(shows E: I shift, [ә ]: [i] shift, [ә ]: [ä ] shift).

climax: climactic

(shows X: CT shift and [ks]: [kt] shift).

Recognition rule R 2 provides umbrella coverage for a wide range of allomorphic variation. (A complete list of variations permitted under this rule will be found in Appendix 1). If our concern were to propound a complete system for morpheme identification then R 2 would need to be reconsidered more stringently from that point of view. For our present purposes — to obtain reliable segmentation of our four suffixes — R 2 is fully adequate.

It is worth noticing that although R 2 appears to be generous its effect is never to force upon us segmentations to which intuition or good sense objects. On the contrary it prevents our segmenting in a number of cases where intuition or good sense tell us we could be right to do so. Scholastic and parochial remain unsegmentable since the patterns school: scholastic and parish: parochial are notechoed in another Webster or Brown form.

In other words, we follow our stated principle of restricting our material to the most clearcut instances. There is no implication that the forms excluded should be excluded in general, under other circumstances, or for other purposes.