Productivity in shape: formal generality and regularity

Earlier, we have observed that some processes of inflection and derivation are more widely used than others. For example, among ways of forming abstract nouns from adjectives, -ness (as in greyness, happiness, richness) is more widely used than -ity(as in sensitivity, purity) or -th (as in depth, length). I will use -ness, -ity and -th to tease apart different ways in which a process can be ‘productive’.

The suffix -ness is formally general in the sense that, when attached to most adjectives, it yields an abstract noun which is either in common use (greyness, richness etc.) or would not need to be listed as a lexical item because its existence is predictable, given the existence of the adjective. Thus, once one has learned the existence and meaning of the adjective dioecious, one does not have to learn separately the existence of a noun dioeciousness. (Dioeciousness thus resembles the adverb dioeciously) The suffix -ness is also formally regular, in the sense that one can specify what sort of structure an adjective must have in order to be a possible base for it – namely, any structure whatever. That is, whatever adjective -ness is attached to, the result sounds like a possible noun, even though it may not be one that is conventionally used (e.g. sensitiveness, pureness, longness). If native English speakers hear a non-English-speaker use the word longness instead of length, they will almost certainly be able to understand what the speaker means, even if longness is not a word that they themselves would use.

By contrast, both -ity and -th are much less general. With most adjectives, the result of attaching either of these is something that is not only not an actual noun but also not a possible noun. For example, *greyth and *richity sound not merely unconventional but positively un-English; by contrast with longness, they are not words that we would understand without effort in the unlikely event of our hearing them used. But this does not mean that both these suffixes are equally irregular. In fact, -ity is formally quite regular, in the sense that possible bases for it are easy to specify: adjectives in -ive (selective, passive), -able or -ible (capable, visible), -al (local, partial),-ar (insular, polar),-ic (electric, eccentric),-id (liquid, timid) and -ous (viscous, various). Formally irregular are the relatively few nouns in -ity formed from adjectives outside this range, e.g. dense, immense, pure, rare. (Compare dense with tense: they look alike, but they form their abstract nouns density and tension in different ways.) Also somewhat capricious is the behavior of adjectives in -ous, some of which preserve this suffix in the allomorph -os-, e.g. viscosity, curiosity, while others lose it, e.g. ferocity, variety related to ferocious, various – an idiosyncrasy already noted By contrast, -th is formally quite irregular, in that the adjectives that choose it share no common structural characteristic beyond the fact that they are monosyllabic (deep, wide, broad, long, strong) – a characteristic that they share with hundreds of other adjectives, however.

The behavior of -ness and -ity shows that regularity does not imply generality. Even with the bases where -ity is regular, it is by no means totally general. It is easy to think of adjectives which on formal grounds are suitable bases for a noun in -ity but for which no such corresponding noun is in common use. Examples are offensive, aggressive, social, chemical, lunar, nuclear, strategic, allergic, languid, horrid, gracious, devious. I say ‘not in common use’ rather than ‘never used’, because a noun such as offensivity, sociality or languidity does not sound wrong in the way that *richity or *greenth does. A check in a large dictionary may reveal that some of these nouns have indeed been used. The important point, however, is that a noun in -ity does not exist automatically just through the existence of a suitable base adjective, as with dioeciousness and dioecious. The suffix -ity has more gaps in its distribution, even in the domain where it is regular, than the suffix -ness has. This kind of gappiness is particularly characteristic of suffixes borrowed directly or indirectly from Latin, rather than inherited from Proto-Germanic.

The kinds of formal regularity that we have discussed so far have involved characteristics of the base that are either purely syntactic (for example, the bases to which -ness attaches are adjectives) or partly morphological (for example, the bases to which -ity attaches are adjectives that contain certain suffixes). But formal regularity can involve phonology too. The noun-forming suffix –al, can be attached only to bases whose final syllable is stressed. Thus the actual nouns survival, proposal, referral and committal are all formally regular, but the hypothetical nouns *edital , *punishal and *reckonal are non-existent not merely by accident but because they are formally irregular, violating the final-stress requirement. (Only one noun exists that violates this requirement, namely burial.) Does it follow, then, that any verb with final stress can be the base for a noun in -al ?, namely the phonological requirement that verb-forming suffix -en can attach only to mono-syllabic bases that end in plosives (as in redden, thicken, dampen) or fricatives (as in stiffen, lengthen).We left unanswered the question whether all such adjectives are bases for existing verbs with -en, or whether there are hypothetical verbs that do not exist even though they comply with the phonological requirement. It is in fact quite easy to find relevant examples. The verb meaning ‘make wet’ that corresponds to the adjective wet is not ‘wetten’, as one might expect, but simply wet; and there is no ‘limpen’ corresponding to limp (meaning ‘flabby’), nor ‘badden’ corresponding to bad. Similarly, despite the existence of reversal based on reverse, there is no ‘conversal ’ based on converse; and, despite the existence of arrival, revival and survival, there is no ‘derival ’ based on derive. So -al suffixation and -en suffixation, although they both exhibit formal regularity of a phonological kind, are both less than totally general.

If a derivational process can be formally regular without being highly general, it is natural to ask whether the reverse situation can obtain: can a process be general without being formally regular? This would be the situation of a process that is used in the formation of relatively many lexemes, but so randomly that one cannot discern any formal or structural characteristics shared by the bases that undergo it. Imagine, for example, that the adverb-forming -ly suffix could be attached not only to adjectives but also to nouns and verbs, so as to form numerous adverbs such as ‘invently’ (meaning ‘inventively’) and ‘gloomly’ (meaning ‘gloomily’) – but that the existence of such noun-derived and verb derived adverbs (as well as of adjective-derived ones) is haphazard and unpredictable, so there happens to exist no word ‘selectly’ (meaning ‘selectively’), nor ‘cheerly’ (meaning ‘cheerily’). It is hard to find any example in English of a derivational process so haphazard as that. But this is not surprising, because it is hard to imagine how a collection of words with just these properties would come into existence. Unless a process is relatively regular, few new words are likely to be created by means of it, or to become established in general usage once they have been introduced – so, if -ly suffixation were as irregular as we are assuming, the class of words exhibiting it would never be likely to be numerous. We can therefore take it that in practice, although not by definition, formal generality presupposes formal regularity, but not vice versa.