Degrees of allomorphy

Allomorphy can be subdivided into several types. It would be quite possible to formulate the allomorphic rules so that variations explainable in consistent morphological, phonological or morpho-phonological terms (for instance race: racy, anger: angry or the shift of stress before (IC) in humanist: humanistic)wereplaced in a category of their own and covered by their own rules. Rule R 2 should then be rephrased to limit its coverage to variations (such as symptom: simptomatic, produce: product) which are not explainable in such terms.

However, there are reasons for thinking this is unnecessary. One is that R 2 provides umbrella coverage for graphological and other such variation already, so to exempt these forms would be superfluous. Secondly, none of the variations quoted is 100% consistent. Beside race: racy we find lace: lacey; beside anger: angry wefind bluster: blustery; beside humanist: humanistic we find Arab: Arabic.

It may also be objected that to allow symptom: symptomat- as allomorphs dangerously facilitates the process of segmentation. The suggestion is that while symptomatic can now be segmented by S 1 — on the grounds that it consists of (SYMPTOM) + (IC) — it might well be difficult to segment by S 2. (This argument is not in fact applicable to symptom: symptomatic since we can easily segment also by S 2 on the pattern:

symptomat-: anatom- + -ize: -ic.

And the same is true of the great majority of these forms. S 2 is applicable in almost every case. But there are nevertheless cases where this objection holds good.)

We see no reason to doubt the rightness of this R 2 at least so far as it serves our present purpose of suffix-segmentation. This means that we shall indeed be willing to recognize symptomat- (in symptomatic) as an allomorphic instance of (SYMPTOM), paralleled by, for example:

(PARADIGM) recognized in paradigmat- + (IC)

(PROBLEM) recognized in problemat- + (IC)

(ERR) recognized in errat- + (IC)

(IDIOM) recognized in idiomat- + (IC)

(SYSTEM) recognized in systemat- + (IC).

It is not true that rule R 2 in any way “corrupts” our segmentation procedures or the statistics which arise from them. We should, that is, have been forced to find some other way to make the symptomat- + (IC) segmentation in any case. When we examine the considerable pattern strength represented in Webster and Brown it becomes clear that this -at- is a feature of our language that has to be dealt with under any circumstances. It is no rare occurrence but a recurring unit, albeit of uncertain function. We shall (if we reject the stem-allomorphy proposal offered above) have to classify -atic either as an allomorphic variant of (IC), or as a new morpheme (ATIC), (an uneconomical suggestion), or in some other way.