Group 2: Suffixes that attach outside one other suffix (6 of 43)

Fabb claims that the following six suffixes only attach to one other suffix, noun-forming -ary (attaches to -ion, as in revolutionary), adjective-forming -ary (with the same property, cf. revolutionary), denominal -er (attaches again only to -ion, as in vacationer), -ic (only occurs after -ist, as in modernistic), -(at)ory (only after -ify, as in modificatory), and deadjectival -y (only after -ent, as in residency).

As was the case with group 1 suffixes, Fabb's observations are empirically and theoretically flawed. With -ary, combinations involving -ate and -ment are also attested, as in commendatory, complementary, sacramentary, sedimentary, supplementary, which enlarges the number of possible suffixes preceding -ary to at least three. The same is true for adjective forming -ary, since most of the above forms are also used as adjectives.

Denominal -er (vacationer) attaches to quite a number of suffixes as the following examples show: adventurer, allegorister (rare), annoyancer, conveyancer, aphorismer (obsolete), assurancer, astromancer, astrologer¹, baggager, bondager (Scottish), complimenter², concordancer, con jecturer. Thus, -er may also be preceded at least by -ure, -ist, -ance, -ment, -age, and -ar.

Concerning -ic, Fabb's observation is empirically correct, counterexam ples are extremely rare (although geological agglomeratic may be one). The question, however, remains whether we are faced with a base-driven or with an affix-driven restriction. It seems that the constraint should be formulated as a base-driven requirement, i.e. as a property of -ist rather than as a property of -ic. If we say that -ist takes only -ic as an adjectival suffix, we explain the occurrence of not only -ist-ic, but also the obvious ungrammaticality of *-ist-al, *-ist-ive, *-ist-ent and the like. Again we see the empirical and theoretical advantage of a base-driven restriction over a suffix-driven one, since both legal and illegal combinations can be predicted.

Strangely enough, -(at)ory, Fabb's suffix [38], is also mentioned as a member of the group of suffixes that Fabb claims does not appear outside any suffixes. I have already pointed out that verbs in -ate productively take -ory as an adjectival suffix, and have argued for a base-driven constraint. On the same grounds we can state another base-driven restriction, namely that verbs in -ify regularly take -atory as an adjectival suffix (accompanied by an alternation of -ify into -ific, which is already familiar from -ion suffixation).

Fabb's claim that deadjectival abstract noun-forming -y only attaches to one adjectival suffix is not quite correct, since at least adjectives in -ate also take -y productively as the noun-forming suffix (accompanied by a change of the final plosive into a homorganic fricative). Consider intimacy, privacy, literacy, degeneracy, (con)federacy, accuracy, adequacy. Hence, a base-driven constraint is to be favored along the same lines as argued repetitively above.

To summarize our review of the suffixes that are claimed to attach only outside one other suffix, we can say that, with the exception of -ic, all of them attach to at least two other suffixes. Furthermore, the base-driven selectional restrictions suggested here to replace Fabb's restrictions have been shown to be superior in their predictive power. This group of suffixes does not form a homogeneous class at all. In fact, it seems strange to posit a class of suffixes on the basis of the number of suffixes to which they attach. Since very specific idiosyncratic selectional restrictions have to be posited anyway, nothing is gained for a theory of derivational morphology by generalizing over the number of such idiosyncrasies.

¹ The forms attested in the OED suggest that the combination -log-er has competed with -log-ist for quite some time, with the latter having now superseded the former in most of the cases.

² The primary stress on the first syllable suggests a denominal origin since -er is a stress-neutral suffix.