English Word Formation

I. Aronoff, Selkirk and Lieber on the structure of the English lexicon and the nature and use of word-formation rules.

In the wake of Chomsky's "Remarks on Nominalization" (1970), in which Chomsky makes a strict distinction between syntax and derivational morphology, Aronoff (1976) proposes a word based theory of the lexicon. This lexicon is a separate component of the grammar in which derivational word formation processes are dealt with. This hypothesis is called strong lexicalist hypothesis. Aronoff suggests that inflection and compounding are not taking place in the lexicon but in the syntax (Spencer 1991: 82). As to the nature of the listed lexical items Aronoff does not go along with Halle who in 1973 assumed that the lexicon is made up of three lists: a list of morphemes, a list of actual words, and a list of words that are regularly formed but are non-existent (McCarthy 1992: 25). In his theory Aronoff reduces the three lists to one single list, stating that it could only be words that are listed in the lexicon, not morphemes. A reason for this assumption is that morphemes, other than words, are not persistent in meaning and sometimes they do not seem to have any meaning at all. A good example for meaningless morphemes are the so-called cranberry morphemes. If you analyse the last unit of the words (1) cranberry, boysenberry, huckleberry as the meaningful morpheme #berry# you will be left with the morphemes (2) #cran#, #boysen#, #huckle# that are meaningless in isolation (Aronoff 1976: 10). As far as persistence of meaning is concerned, Aronoff points out that there are elements that can be isolated and therefore recognised as morphemes but that do not seem to have any constant meaning. In the cases of (3 ) receive, perceive, deceive, conceive for example the meaning of the morpheme #ceive#, if there is any at all, changes depending what prefix is attached to it. Since the morphemes in (2) do not constitute a Saussurean minimal sign and the morphemes in (3) do not have a constant meaning, they cannot be in the lexicon because Aronoff's lexicon is a system put together from a list of minimal signs plus a set of word formation rules (from now on WFRs) that operate over those minimal signs. Aronoff therefore adopts a hypothesis in which the listed minimal meaningful sign is the word and not the morpheme (Aronoff 1976: 18). According to his theory all lexical items have to belong to one of the major lexical categories. The same has to be true for the output of WFRs.

Aronoff does not want to list any possible or actual word because WFRs seem to overgenerate. Therefore he proposes certain restrictions as to whether a word gets a lexical entry or not. His main criteria for the listing of a word is the question of arbitrariness. "[All] and only those words which are exceptional, i.e. arbitrary in at least one of their various features, will be entered in the lexicon" (Aronoff 1976: 43). The word transmission for example will get a lexical entry due to the fact that it has an unpredictable meaning ('gearbox of a car') next to its expected meaning, whereas emission with only one straightforward meaning will not be listed (McCarthy 1992: 28f). Aronoff sees idiosyncrasy as a direct result of the persistent meaning of lexical items. If a word does not have a lexical entry there is no expectation that the word will take on idiosyncratic meanings or, in other words, there will be no semantic drift Aronoff points out that this shows the effect lexical listing can have on semantic coherence (Aronoff 1976: 43).

As far as the production of new words is concerned Aronoff points out that a new word can only be formed "by applying a regular rule to a single already existing word" (Aronoff 1976: 21). The new word has to be member of some major lexical category, "the exact category being determined by the WFR which produces the word" (Aronoff 1976: 49). The problem Aronoff faces here is that there are in fact affixes that are attached to roots that are not words. In order to overcome this problem Aronoff introduces a set of allomorphy and truncation rules, that patch up "the phonological form of words which have been produced by WFRs" and that work as adjustment rules (Spencer 1991: 83). Also there is the possibility of back formation. If you take the words aggression and aggressive and separate the morphemes #ion# and #ive# you will be left with the particle #agress# which does not surface as a lexem. Aronoff derives the word aggressive from aggression by a rule of truncation that eliminates #ion# in aggression in order to turn it into a possible base for the derivation of aggressive. This explains how to get the form aggressive but it leaves the reader to wonder why suddenly #ive# is attached to a noun even though normally its an affix that attaches to verbs only. Aronoff's WFRs do not necessarily function synthetically even though there are rules that are productive. Generally speaking WFRs work analytically, i.e. they analyse existing words into their component morphemes (Spencer 1991: 85). Aronoff's theory is what McCarthy (1992) refers to as associative. In an associative theory, "meaning and morphological structure are seen as intimately associated. [...] [The] mechanisms which account for a word's structure must also determine (or at any rate suggest) what its meaning should be" (McCarthy 1992: 32).

An important issue in Aronoff's theory is the productivity of a WFR. He states that if the output of a WFR allows for phonological idiosyncrasies or if it is lexically governed or if it does show semantic incoherence then the productivity of the very rule is reduced. Aronoff proves this by setting up an experiment to investigate the productivity of two WFRs that derive Nouns out Adjectives. He applies the ness-rule and the ity-rule to a constant set of adjectives ending in the affix #ous. He observes that two phonological idiosyncrasies go along with the #ity words. #ity words are always stressed on the second last syllable and in some words #ity causes #ous -bases to undergo truncation which eliminates the #ous morpheme. Examples for these cases would be cúrious but curiósity and various but variety. The fact that it is totally unpredictable whether a word does truncate or not means that the #ity words need to be lexically marked to indicate possible truncation. As far as transparentness of the outcome is concerned, the #ness rule seems to be totally transparent whereas the words that are derived by the #ity rule have in some cases idiosyncratic meanings. The productiveness of the #ity rule seems to be reduced on various occasions whereas #ness rule produces a set of totally transparent derivates. The conclusion Aronoff draws from this experiment is that fully productive WFRs do not need to be treated in the lexicon since their outcome is to no degree arbitrary. The blocking phenomenon also shows that words derived by fully productive WFRs need not to be listed in the lexicon. Aronoff introduces the notion of blocking of possible lexems because the lexicon has a low tolerance for synonymy. If there is an already existing non-derivational noun corresponding to, for example, the adjective glorious, the lexically governed output of the #ity rule is blocked. Since the word glory is already existing there is no room for * gloriosity. Gloriousness though is a possible derivation from glorious and this is due to the fact that gloriousness is generated by a fully productive WFR whose output is not listed in lexicon because of its transparency. Since the output of these WFRs is not listed it cannot possibly be blocked (Spencer 1991:88f).

Selkirk (1982) suggests a different theory of the lexicon. Like Aronoff she locates so-called rules of word structure (Aronoff's WFRs) in the lexicon even though she admits that they might as well be seen as part of the syntactic component of the grammar. Her theory can also be referred to as a lexicalist theory. Since she focuses on the nature of word structure and on the rule system generating it, the question about where to locate the word structure rules is not of primary interest to her though (Selkirk 1982: 10). An important distinction to Aronoff's theory is that for her "word structure has the same general formal properties as syntactic structure and [...] that it is generated by the same sort of rule system" (Selkirk 1982: 2). Since she regards morphological and syntactical structures as formally related, the morphological structure, as the phrasal structure for sentences, conforms to "certain restrictive patterns, the characterisation of which requires the x-bar theory of categories" (Selkirk 1982: 6). Each morphological category has to be understood as a binary structure, consisting of a level specification category and a feature specification category. The features which are relevant to word syntax (i.e. morphology or word structure), can be divided in two classes. "Syntactic category features" such as [+-Noun], [+-Verb], etc... are assigned to the first class, so-called "diacritic features" belong to the second class of features. This second class includes those features "relevant to particulars of inflectional and derivational morphology" (Selkirk 1982: 7). Inflectional features include for example tense, gender, person and number, derivational features would for example be [+-latinate]. Selkirk, other than Aronoff, argues that inflection as well as derivation and compounding all have to be regarded as word syntactical processes. Derivational and inflectional morphology are according to Selkirk "not strictly disjoint": In her system an inflectionally marked element may serve as the base for a derivational process (Selkirk 1982:8).

[...]