The problem of unintentionality
The first problem concerns Schultink's notion of unintentionality. He claims that one should distinguish between unintentional expansions of the lexical stock by the unconscious implementation of word formation rules on the one hand, and the intentional creation of new words in order to produce a special effect on the other. The former is often referred to as morphological productivity, the latter as morphological creativity. Creatively formed neologisms are usually perceived by other speakers as somehow remarkable and have often a humorous or repulsive effect, whereas truly productive neologisms are typically not noticed as such by the speakers. This distinction between intentional and unintentional creations is still considered crucial by more recent authors who claim that coinages on the basis of productive patterns go unnoticed, whereas those on the basis of unproductive patterns always draw attention to themselves (e.g. Baayen and Lieber 1991:808, Lieber 1992:3). Thus, if the unproductive de-adjectival nominalizing English suffix -th (as in depth) were attached to an adjective not belonging to the established set of bases that take this suffix (such as broad, deep, long etc.), this would certainly not go unnoticed (cf. *steepthl*stepth), whereas new words in -ness are typically not recognized as new formations. These considerations are corroborated by the findings in Aronoff (1983) which we already discussed above: with truly productive formations speakers often cannot tell whether the form is a neologism or not.

There are, however, two serious counterarguments against the necessity of the 'unintentional' criterion. First of all, the notion of unintentionality is notoriously vague. Some speakers have a higher level of awareness of the manipulation of linguistic signs than others, i.e. what goes unnoticed by one speaker may strike the next as unusual. It is hard to see how this criterion could be operationalized.

But, for the sake of the argument, let us leave aside these doubts and take some notion of 'unintentional' as given. Even under this assumption the criterion is not a necessary one, because even productive rules can be applied intentionally. For example, many of the -ize neologisms listed in the OED below can be assumed to be intentional creations, because they are very often coined to designate new concepts in science, which is often done purposefully and with some consideration. Nevertheless, these words should be regarded as productive formations and not incidents of morphological creativity in the above sense because they are in perfect harmony with the semantic and phonological specifications of -ize derivatives.¹ The same reasoning would hold for the word crucialness, which I consciously and intentionally invented a minute ago, and which is nevertheless undoubtedly a productively formed derivative. Some readers would have noticed the novel character of this word in a text, others would have not.

Given these considerations, I do not take unintentionality to be a necessary characteristic of productive processes, although it may be pertinent in many instances of productive formations.

The second problem concerns the criterion of uncountability. Schultink assumes that productive rules give rise to an in principle uncountable number of forms whereas unproductive rules can only yield a fixed, countable, and presumably small number of derivatives (cf. also Lieber 1992:3-4). Again, Schultink's points can be illustrated with the two de-adjectival nominalizing suffixes -th and -ness in English. The unproductive suffix -th only attaches to a limited number of bases,² whereas -ness can practically occur with any adjective of the English language. Since the number of adjectives is, in principle, infinite, so is the number of possible -ness derivatives. With such examples, uncountability seems a clear criterion to distinguish productive and unproductive processes. There are, however, some less clear cases conceivable. Imagine a productive rule that is restricted in such a way that there is a definite limit to the number of potential bases, and therefore to the number of possible derivatives. For example, a suffix may attach productively to words of a certain prosodie make-up, and the number of such words is limited (e.g. because the stress rules of the language no longer permit words of the required make-up). Or, consider the case of -ment which is productive with bases that contain the prefixes en or em- (henceforth eN-).³ Since this prefix is no longer productive, the number of possible -ment forms on the bases of these forms is also limited. If the criterion of uncountabillity is upheld, the attachment of -ment to such bases must be considered unproductive. This result is, however, undesired since, in view of the fact that all bases of the pertinent kind can potentially take -ment, one would want to classify this process as fully productive.

Scrutinizing the hedge "in principle" in the above definition may shed some more light on this problem. Given a rule like -ness suffixation it seems that all adjectives may undergo this rule.⁴ Although the number of adjectives an individual speaker knows at a given point in time may be countable, it is in principle unlimited because it is an open class to which new items can easily be added, for example by derivation or borrowing. If the class of possible input words is open, so must be the class of possible derivatives of productive rules. This is the essence of Schultink's reasoning. But what about apparently closed classes of possible bases? According to Schultink, word formation rules making reference to closed classes would be unproductive by definition. Under an alternative view, such a rule would be nevertheless considered productive if all members of the class, and hypothetical new members, could undergo the rule, and provided that the class can be defined intensionally. The latter is possible with the class of eN- prefixed verbs, but crucially not with the bases of processes that are lexically governed, such as -th nominalization of adjectives.

Let us look again at our example of eN- prefixed verbs and -ment nominalizations. If the rule were really productive, we would predict that, if there were new possible base verbs in eN-, which is of course highly un likely due to the rule's non-productivity, these base words could take -ment as their nominalizing suffix. Fortunately, this prediction can be tested since, in spite of the general non-productivity of eN- there are six 20th century neologisms with this prefix attested in the OED, all of them apparently creative or analogical formations. For two of these, endistance and embrittle, nominalizations in -ment are also attested, and it seems that the other four derivatives {encode, envision, emplane, enhat) could also lend themselves to -ment suffixation. What this shows is that the rule of -ment attachment is productive with an intensionally defined class of base words inspite of the fact that this class cannot increase indefinitely. However, in principle an uncountable number of productively formed -ment derivatives could be created. Seen from this angle, the criterion of uncountability becomes entirely vacuous, because uncountability in principle follows auto matically from the general potential to create words on the basis of existing patterns. If a rule has this property, the number of possible derivatives is always, in principle, uncountable, but crucially only in principle. To summarize, Schultink's criterion of uncountability can be disposed of because this property falls out automatically from other aspects of the definition of productivity.

¹ Neuhaus (1971:158) evokes a similar argument against intentionality with regard to the rise of the suffix -ite in the 19th century which "was used systematically in the expansion of scientific terminology, for example in mineralogy and chemistry. Hence, productivity can be the result of planning and need not follow a general tendency in a more or less unconscious fashion." (my translation, I. P.)
² See Bauer (1988:59-60) for an exhaustive list of lexemes involving the suffix -th.
³ See Abstract-noun forming suffixes for a thorough discussion of the interaction of -ment and eN-.
⁴ To my knowledge, no one has ever proposed a rule-specific restriction on -ness that goes beyond the specification that the base words are mostly ad jectives. The majority of reseachers stress the fact that, in the words of T. Williams, it "is found in abundance in all linguistic environments" (1965:283).