Person-noun-forming suffixes

Let us start with the denominal suffixes -an (librarian), and -ist (,methodist). The distribution of -ist 'adherent of a doctrine/attitude' is identical to that of the corresponding -ism derivatives denoting the respective doctrine or attitude.¹ Given our analysis of -ism, it does not come as a surprise that we find numerous examples of already suffixed nouns that take -ist as a suffix, cf. e.g. abortionist, expansionist, consumerist, conventionalist.

The suffix -an is more problematic. First of all, it is not quite clear which suffix(es) Fabb means by -an. While Marchand (1969), for example, treats -ian, -an, -oan, -onian as allomorphs, Fabb does not say anything about -ian, -onian or -oan. Fabb's example librarian itself may be interpreted as involving a bound root libr-² and two suffixes, namely -ary ('a place for', OED) and -an. Under the assumption that library is morphologically complex Fabb's claim is empirically inadequate.

There are, however, three possible counterarguments that might be adduced in order to save the generalization that -an does not attach to already suffixed nouns. First, one may assume that words like library are not morphologically complex, i.e. they do not end in a suffix, and therefore allow the attachment of -an. While for library this may still be a plausible analysis, for words like salutatorian it is definitely not. Second, one might say that the example librarian is badly chosen and that we are in fact not dealing with the suffix -an, but with a compound affix -arian. If we follow this argument (which is itself shaky, given the compositional meaning), we have to turn to those cases of -an suffixation which definitely do not involve preceding -ary. Such cases seem to be restricted to either proper nouns and place names (Pennsylvanian, Chomskyan, Andersonian, sometimes accompanied by phonological adaptation) or nouns ending in -ia or -y (academian). Items in the first group, proper names and place names, usually do not involve any suffixes. It is this morphological property of the members of the - probably most productive - semantically defined domain of -(i)an suffixation that is responsible for the impression that -(i)an does not attach to suffixed stems. The second group of stems, those ending in -y, can indeed be argued to be morphologically complex, refuting the claim that -an does not attach to already suffixed nouns. The paradigm barbaric, barbarous, barbarian, barbary for example, strongly suggests that barbary consists of a stem barbar- and an affix -y.

In summary, no matter how we analyze denominal -an derivatives, it is unnecessary to posit an idiosyncratic selectional restriction which forbids the suffix to attach to already suffixed bases.

Let us turn to the deverbal person noun forming affix -ant (defendant), which denotes a personal or material agent. Again, possible verbal bases involve those ending in -ify, -ize, -ate, and -en. A look at Lehnert (1971) and the OED shows that, almost without exception (see below), these verbs are subject to the domain of agentive noun forming -er/-or.³ The rival suffix -ant has a somewhat peculiar distribution, since its attachment is partly lexically governed (i.e. unproductive) and partly rule-governed and productive. In the semantically distinguishable domains of medical/pharmaceutic/chemo-technical and legal/corporate jargon, -ant can be used productively to form words denoting substances and persons, respectively, as evidenced by the following examples disinfectant, repellent, consultant, accountant, defendant, to mention only a few.

The two rival domains of agent nominalizations -er and -ant are potentially in conflict, whenever base forms may conform to both domains. This is necessarily the case where the semantic specification (which is part of the definition of the domain of -ant nominalizations) clashes with the general applicability -er. For instance, verbs ending in -ize denoting an event of the, say chemical or medical realm, are potentially subject to two nominalization rules. Hence it does not come as a surprise that we find tetanizant 'an agent or substance that causes tetanus' (OED), where -ant supersedes -er.

These facts could possibly be interpreted as supporting van Marie's Domain Hypothesis. Since -er is the general case, its domain of application may be systematically curtailed by the domains of the special cases, e.g. -ant. This means that all verbs form agent nouns by default through -er/-or attachment, unless otherwise specified. Thus, apart from lexically marked verbs such as defend, only verbal stems denoting concepts of the chemo-technical and legal/corporate fields may take -ant. The fact that derived verbs generally form agent nouns with -er indicates that they do not carry the diacritic marking for -ant. It is only by way of their belonging to the semantically defined domain of this suffix that derived verbs may become subject to -ant agentive nominalizations. However, contrary to the prediction of the domain hypothesis, they do not do so obligatorily, but rather tend to choose the general case -erl-or. Thus, for the combination -ize-ant, the above-cited tetanizant seems to be the only unquestionable example, whereas a much larger number of derivatives are attested with -er even in the semantic domain of the special case (e.g. atomizer, carbonizer, crystallizer). The same holds for the combination -ate-ant, which is attested only once (inflatant, rarel), while ate-or is rather common (e.g. calibrator, chlorinator, defibrillator). In essence, the special case does not pre-empt the general case, hence type-blocking cannot explain the distribution of -ant and -er/-or.

¹ There are a number of -ist and -ism forms that do not have a counterpart in -ism and -ist, respectively. These cases are, however, confined to semantically distinct forms which simply denote agents (cf. linguist, theorist) or (medical) conditions like mongolism, metabolism (see Giegerich 1998 for discussion).

² Support for the existence of this bound root may come from the analogical coining libricide 'book slaughter' (rare1).

³ I consider -er and -or allographs of one suffix.