Blocking

The term 'blocking' has been mainly used to refer to two different types of phenomena, namely the non-existence of a complex form due to the existence of a synonymous competing form (* stealer - thief) on the one hand, and the non-existence of a form due to the existence of a homonymous form (?liver 'someone who lives' - liver 'inner organ') on the other. I will first discuss briefly the latter type and then turn to the more interesting type of synonymy blocking.

Although frequently mentioned in the pertinent literature, homonymy blocking cannot be assigned real significance since in almost all cases cited, the would-be blocked derivative is acceptable if used in an appropriate context. For example, Jespersen (1942:231) tries to explain apparent gaps in the formation of agent nouns with the suffix -er in terms of homonymy blocking, but does not fail to mention counterexamples such as the pun Is life worth living?-It depends on the liver (see also the quotation in the OED's entry liver n 2: "The country for easy livers, The quietest under the sun."). The slight oddness of liver outside appropriate contexts may be explained by the semantics of -er, which is given by Marchand (1969:273) as follows: "Deverbal derivatives (in -er, I. P.) are chiefly agent substantives ... denoting the performer of an action, occasional or habitual". If this characterization is correct, the oddness of liver falls out automatically: live is neither a typical action verb, nor does it denote anything that is performed occasionally or habitually, in any reasonable sense of the definition. Notably, in the quotations the derived form liver receives a more intentional, agentive interpretation than its base word usually has.

Plank (1981:165-173) discusses numerous similar cases from different languages in which homonymy blocking does not provide a satisfactory solution. In essence, it seems that homonymy blocking serves as a pseudo-explanation for facts that appear to be otherwise unaccountable. In a broader perspective, homonymy blocking is only one of the many phenomena that functionalists seek to explain by evoking a universal language principle of ambiguity avoidance. Here, as in other cases, this approach fails to expound why language tolerates innumerable ambiguities, but should avoid this particular one. In summary, homonymy blocking can be disposed of as a relevant morphological mechanism. Let us therefore turn to the more fruitful concept of synonymy blocking.

With regard to German nominalization, Paul (1896:704) already noted that "the formation of nouns in -ung is blocked by the existence of a simpler formation of the function of a nomen actionis" (my translation, I.P.). Perhaps due to its intuitive appeal, the notion of blocking remained in an imprecise and pre-theoretical state until the 1970s and 1980s, when Aronoff (1976), Kiparsky (1982, 1983), van Marie (1985, 1986) and Rainer (1988) tried to formalize this notion.

Elaborating on van Marie's work, Rainer (1988) lays out the weaknesses of Aronoffs and Kiparsky's notions of blocking and develops the most pronounced theory of blocking to date.¹ He distinguishes between two forms of blocking, type-blocking and token-blocking. Type-blocking concerns the interaction of more or less regular rival morphological processes (for example decency vs. ?decentness) whereas token-blocking involves the blocking of potential regular forms by already existing synonymous words, an example of which is the blocking of arrivement by arrival or stealer by thief. The latter type of blocking is the one envisaged by Paul in the above quote. I will first discuss the relatively uncontroversial notion of token blocking and then move on to type-blocking.

Token-blocking occurs under three conditions: synonymy, productivity, and frequency. The condition of synonymy says that an existing word can only block a newly derived one if they are synonymous. Thus, many of the doublets that do occur are not synonymous but convey (at least slightly) different meanings.

Another condition on the occurrence of blocking is that the blocked word must be morphologically well-formed, in other words, it must be a potential word, derived on the basis of a productive rule. Although this condition looks rather trivial it serves to exclude cases of blocking that are only apparent. For example, Wellmann (1975:213) attributes the non-existence of German *Schlagung 'hitting' to the existence of Schlag 'hit (noun)'. As argued convincingly by Rainer (1988:162-163) * Schlagung is ruled out even in contexts where it is not synonymous with Schlag, and in which it should therefore not be blocked. In other words, *Schlagung must be ill-formed due to reasons other than blocking.

Let us turn to the last condition. The crucial insight provided by Rainer is that, contrary to earlier assumptions, not only idiosyncratic or simplex words (like thief) can block productive formations, but stored words in general can do so. As already discussed, the storage of words is largely dependent on their frequency. This leads to the postulation of the frequency condition, which says that in order to be able to block a potential synonymous formation, a word must be sufficiently frequent. This hypothesis is corroborated by Rainer's investigation of a number of rival nominalizing suffixes in Italian and German. The higher the frequency, the higher the blocking force of a word with respect to a synonymous derivative. Therefore, both idiosyncratic words and regular complex words can block, provided that they are stored.

This leads us to another aspect of token-blocking, not explicitly discussed in Rainer, namely the problem of really synonymous doublets, which may indeed occasionally occur. Plank (1981:181-182) already notes that blocking of a newly derived form does not occur in those cases where the speaker does not activate the already existing alternative form. This is frequently attested in child language (e.g. Clark 1981), where the lexicalized alternatives are not yet present, and in adult speech errors, i.e. when lexical access fails. For obvious reasons, the likelihood of failing to activate a stored form is negatively correlated to the frequency of the form to be accessed. By making frequency and storage the decisive factor for token-blocking, the theory can naturally account for the occasional occurrence even of synonymous doublets.

In the light of these considerations, token-blocking is not "a prophylactic measure to avoid undesired synonymy", as claimed by Plank (1981:182, my translation, I.P.), but the effect of word storage and word processing mechanisms.

We may now move on to the notion of type-blocking, which is less well understood. According to van Marie's Domain Hypothesis, rival suffixes are organized in such a way that each suffix can be applied to a certain domain. In many cases one can distinguish between affixes with an unrestricted domain, the so-called general case (e.g. -ness suffixation, which may apply to practically any adjective), and affixes with restricted do mains, the so-called special cases (for example -ity suffixation). The latter are characterized by the fact that certain constraints limit the applicability of the suffixes to a lexically, phonologically, morphologically, semantically or otherwise governed set of bases. Van Marie further distinguishes two kinds of special cases, systematic and unsystematic ones. A systematic special case is definable by type, i.e. there is a rule that governs the distribution of the affix in question. The unsystematic special case is characterized by the fact that it is lexically governed, i.e. definable only by token. According to the domain hypothesis, the domains of special cases and general cases interact in ways that allow predictions about possible and impossible derivatives. One such prediction is that there is a tendency for the domain of the general case to be systematically curtailed by the domains of the special cases. The pre-emption of the application of the general case by the systematic special case is labeled type-blocking by Rainer (1988). Rainer also observes that not only the general case may be curtailed by the systematic special case, but that rival systematic special cases may also preclude each other. For instance, it seems that the domain of the German nominal suffix -heit is a special case which is curtailed by the more restricted domain of the special case -ität, which speaks for the extension of the mechanism of type-blocking even to rival special cases.

Type-blocking may occur under almost the same conditions as token-blocking, except that frequency does not play a role, for the obvious reason that storage of individual words is not crucial here. Thus only the conditions of synonymy and productivity hold for type-blocking.

For an evaluation of the domain hypothesis and Rainer's blocking theory we will consider briefly an example from English derivational morphology, -ness suffixation and its rivals. Aronoff (1976:53) regards formations involving nominal -ness as ill-formed in all those cases where the base adjective ends in -ate, -ent or -ant, hence the contrast between decency and ?decentness. This would be a nice case of type-blocking, with the systematic special case -cy (decency) precluding the general case -ness. There are, however, three problems with this kind of analysis. The first one is that, on closer inspection, -ness and its putative rivals -ity or -cy are not really syn onymous (cf. Riddle 1985), so that blocking could - if at all - only occur in those cases where the meaning differences would be neutralized. Blocking effects could therefore only be demonstrated in discourse contexts that specify synonymous and non-synonymous interpretations. Riddle's study provides such contexts, thereby showing that rival formations are often possible.

This brings us to the second problem, which concerns the status of forms like ?decentness, for which it remains to be shown that they are indeed morphologically ill-formed. The occurrence of many doublets (see Riddle 1985, or many entries in the OED) rather indicates that the domain of the general case -ness is not systematically curtailed by -ity or -cy. The final problem with putative cases of type-blocking is to distinguish them from token-blocking. Thus, the avoidance of ?decentness could equally well be a case of token-blocking, since one can assume that, for many speakers, the word decency is part of their lexicon, and is therefore capable of token blocking.

Especially in view of the empirical problems, van Marie (1985:238ff) has conceded that the Domain Hypothesis is too strong, and should therefore be used chiefly as a heuristic mechanism, i.e. as a "tool with which we can subject the paradigmatic dimension of morphological structure to a further and more penetrating investigation" (van Marie 1985:239). Rainer comes to a similar conclusion when he states that "The main task for future research in type-blocking will be to try to find out to what extent it is predictable which one of two word-formation rules with rival systematically restricted domains will supersede the other" (1988:179-180).

We will present an in-depth analysis of an interesting test case for van Marie's and Rainer's type-blocking theories, the rival verb-deriving processes in English, i.e. the affixes -ize, -ate, -ify, -en, eN-, be and conversion. It will turn out that the data cannot be adequately handled in terms of type-blocking, and that the distribution of the different processes is not governed by primarily paradigmatic forces, as claimed by van Marie. To the contrary, I will demonstrate that the applicability of a rule to a given base can be most naturally accounted for by referring to the individual syntagmatic properties and restrictions of each process.

¹ For our purposes it is not necessary to repeat here Rainer's (1988) detailed review of Aronoffs and Kiparsky's theories of blocking, in which the serious drawbacks and inconsistencies of these approaches are pointed out. The interested reader is referred to the first section of Rainer's article.