Group 3: Suffixes that attach freely outside other suffixes (3 of 43)

According to Fabb, -able, deverbal -er and deadjectival -ness are subject to no selectional restrictions other than those involving part of speech. While at first glance this analysis seems valid, Fabb's conclusion that "About these, nothing need be said" (1988: 535), is premature. Since none of these suffixes can be blindly attached to all verbs or all adjectives, it is necessary to posit additional restrictions in order to rule out apparently impossible derivatives. With -er impossible formations like *resembler come to mind, -able does not attach to verbs ending in a postconsonantal liquid (e.g. *doublable, *saddlable, *wriggleable, Szymanek 1985:102), and even -ness is subject to complex phonological constraints that trigger schwa epenthesis in certain cases (e.g. mark[ə]dness, well-form[ə]dness, see Wi ese 1996b for a detailed analysis).

Fabb adds deadjectival -en to his list of freely attaching suffixes, since it may appear outside -th )lengthen). He concludes that -en is subject to two selectional restrictions, one involving part of speech, the other involving only monosyllables. Unfortunately, Fabb does not discuss -en in more detail, because the former constraint is problematic. In Fabb's article -en is introduced as a suffix taking adjectives as a base, whereas undoubtedly bases ending in -th are nouns. This would call for a constraint of the kind that -en attaches to monosyllabic adjectives and nouns, which is, however, much too weak in view of the fact that the vast majority of monosyllabic nouns do not take -en as a suffix. In fact, only those nouns that end in the nominalizing suffix -th (and nominalizing -t, as in height) take -en (almost) regularly as a verbal suffix (but note that the forms breadthen, depthen (rare) are marginal at best), whereas other denominal forms are entirely lexically governed {threaten, earthen, frighten, hasten, hearten (obsolete), heaten, kidden (obsolete/rare), nighten (obsolete/rare), shapen (rare), soapen, waxen (obsolete/rare)).

In sum, there are indeed some restrictions operating on the would-be freely attaching suffixes, and these restrictions are much more complex than Fabb would have it. Furthermore they do not differ in kind from the ones observed earlier, in that they involve morphological, semantic and phonological properties of the base or the derivative.