Relational-adjective forming suffixes

Of the deverbal suffixes -ful (forgetful), -ant (defiant), -ory (advisory), -ive (restrictive), -ful only attaches to monosyllabic verbs or disyllabic verbs with ultimate stress, which automatically rules out all suffixed verbs, since they are either trisyllabic (-ify) or do not have final (primary) stress (-en, ize, -ate).¹

A similar situation we encounter with -ant, which attaches regularly to verbs with ultimate (primary) stress, especially to Latinate verbs. Suffixed verbs fall out of this domain. There are, however, some counterexamples, which show that sometimes semantic considerations may overrule conflicting morphophonological constraints. That this is only rarely the case, can be seen from the fact that the OED lists only chemical acidifiant, physical calorifiant, ossifiant, personifiant (of which the last two are marked as rare), and cognizant (also with three prefixes), obsolete agonizant, medical obsolete apophlegmatizant, mathematical canonizant, medical cicatrizant, and obsolete symbolizant.

With derivatives of the form V-ory we find lots of counterexamples to Fabb's generalization that deverbal -ory does not attach to already suffixed verbs. In fact, verbs ending in -ate may take -ory as an adjectival suffix productively, cf. acceleratory, calculatory, stipulatory. In addition, there seems to be a phonological constraint at work that only verbs ending in a consonant may take -ory. Hence verbs ending in -ify are not possible input to -ory suffixation. Neither do verbs in -ize take -ory as a suffix.² This leaves us with verbs in -en, which are equally never followed by -ory. Although this failure may be due to undesired effects of a semantic or pragmatic nature³, the Latinate Constraint introduced above can nicely account for this. For -ory, a [+ Latinate] suffix, the constraint demands that it cannot attach to the native suffix -en.

Yet another example of the operation of the Latinate Constraint is the behavior of -ive. This suffix, of which Fabb says that it does not attach to suffixed verbs, only occurs with [+ Latinate] bases, which, furthermore, must feature /d/, /t/, or /s/ as their final consonant. Together, these constraints bring down the number of possible suffixed bases to one, namely those ending in -ate. And, as already mentioned above, these take -ive quite productively, often alongside the rival -ory: accelerative, complicative etc.

In conclusion, we can state that it is primarily the phonological, morphological or semantic properties of the base plus independent constraints on the morphological processes involved that decide on the attachability of any of the adjectival suffixes discussed here. The fact that these adjectival suffixes do not attach to suffixed verbal bases is merely the consequence of the described mechanisms and not a significant structural generalization, the latter being empirically unjustified anyway.

We may now move on to the selectional restrictions of the denominal adjectival suffixes -ful (peaceful), -ous (spacious), -y (hearty), -ly (ghostly), -ish (boyish), -an (reptilian), -ed (moneyed). In general, of the possible suffixed nouns which adjectival forms may be derived from, quite a number take suffixes that are conditioned by base-driven selectional restrictions. Thus -ion takes -al, -ment takes -al/-ary⁴, -ing and -ant are also used as adjectival endings, -ism nouns convert into adjectives by replacement of -ism by -ist or -istic, -ist nouns are homophonous to their corresponding adjectives, or take -ic. In sum, all of the complex nouns just mentioned are more or less systematically excluded from the domains of the adjectival suffixes under discussion, which considerably reduces the number of possible types of bases. Furthermore, a number of nominal suffixes do not seem to lend themselves easily to adjectival derivations, consider -ette- ?, -let- ?, -ling- ?, -ship- ?, -ee- ?, -hood- ?, -ure- ?, most probably for semantic reasons.⁵

In any case, there are numerous counterexamples to Fabb's claim that the adjectival suffixes under discussion do not attach to already suffixed bases. For Ν-ful there is healthful, meaningful, for N-ous we find treacherous, traitorous, harmonious (the number of free forms as bases of -ous is very small anyway, the majority being bound roots). The suffix -y occurs in the following multiply suffixed nouns: Japanesy, healthy, wealthy, ancestory (and many more in -OR AGENT -y), layery (rare), physicky, strengthy (rare/dialectal), weighty.⁶ The most important constraint on -y is perhaps that mostly mono- and disyllabic words take -y (bases in -ese and -or are systematic exceptions), which considerably reduces the number of potential suffixed noun types.

Contrary to Fabb's analysis, denominal -ly seems to attach quite easily to nouns denoting human beings, be they simplex or derived, as exemplified by christianly, teacherly. The same holds for Ν-ish (Romanish, toadyish, undergraduateish). It should also be noted that there is a rival morphological process which is extremely productive and which may hinder the proliferance of synonymous affixed forms, namely -like attachment.

With Ν-ed we also find a number of counterexamples like kingdomed, qualitied, conditioned, whereas for adjectival denominal -an (reptilian) the same arguments hold as with nominal denominal -an (see the discussion above). There is however one set of bases for -an not mentioned so far, namely a semantically and phonologically conditioned set of Latinate animal names (see Marchand 1969:247 for details). Such names are in general not suffixed, at least not according to English morphological processes.

With respect to deadjectival -ly (deadly) Fabb's observation is correct, there are indeed no suffixed bases attested. Does that mean we need a selectional restriction on -ly? Certainly not, because the suffix is completely unproductive in the modern language, which means that the forms involving deadjectival -ly have to be stored individually anyway. This makes a special restriction of the type suggested by Fabb superfluous.

To summarize we have to state that again we have found that numerous counterexamples considerably weaken Fabb's claim, and that, furthermore, base-driven and domain-specific constraints can account for the existing stacking restrictions in a more systematic fashion. Similar arguments apply to the last group of suffixes that supposedly do not attach to suffixed bases.

¹ There are also good reasons to say that V-ful is lexically governed, which excludes all verbs, derived or not, that are not especially marked for this process. In view of the phonological restrictions, we need not decide on this issue.

² This is not a purely phonological constraint since words like revise, advise take -ory as a suffix.

³ It is not clear to me what the meaning of a relational adjective relating to a causative event should be. Incidentally, this semantic argument equally holds for -ize. Note also that for quite a number of (derived and non-derived) verbs no derived adjectives are attested (go, sleep, buy, etc.), which calls for an explanation independent of the morphological make-up of the base verbs.

⁴ It has been argued by Aronoff (1976:53-54) that the attachment of -al to -ment is subject to the negative restriction that the base of the -ment form may not be a word. Although this restriction could be formulated in a positive way, i.e. the base must be a bound root, there are still some counterexamples (Aronoff himself discusses governmental and departmental, the latter of which is not a real counterexample because it is no longer related to the base verb depart). In view of these counterexamples, Giegerich (1998) suggests that -al attaches only to semantically non-transparent -ment forms such as fragment, ornament, department, pigment, experiment, government, compartment but not to transparent forms like deployment, enrollment (Giegerich 1998). Raffelsiefen (1996:205) proposes that the fused suffix -mental only attaches to bases with non-final stress, which correctly rules out * employmental, * deploymental, * enrolImental, but unfortunately incorrectly* employmental, * deploymental, * enrollmental, but unfortunately incorrectly predicts the non-existence of attested forms like fragmental, pigmental. This phenomenon certainly merits further investigation.

⁵ Person nouns in -ee, -let should be semantically, if not phonologically, compatible with -ish and -y, and -ful with abstract nouns in -ship.

⁶ Some of the bases of the derivatives mentioned may not be considered morphologically complex, such as healthy, wealthy, weighty. However, even if the relationship to the related free form is opaque to most native speakers, and the nominal suffixes -th and -t are not productive any more, they can paradigmatically be identified as suffixes.