The results of these different psycholinguistic studies demonstrate that the
syntactic versatility, lexical flexibility, and semantic productivity of idioms are
not arbitrary phenomena, perhaps due to historical reasons, but can be at least
partially explained in terms of an idiom’s semantic analyzability.
The analyzability of idioms also plays an important role in their immediate
online interpretations. Because the individual components in analyzable idioms
(e.g.,lay down the law) systematically contribute to the figurative meanings of
these phrases, people process idioms in a compositional manner in which the
meanings of the components are accessed and combined according to the syn-
tactic rules of the language (Peterson & Burgess 1993). On the other hand, a
strict compositional analysis of semantically unanalyzable idioms (e.g.,kick the
bucket) provides little information about the figurative meanings of these
expressions. Understanding unanalyzable idioms requires that people first do
some sort of analysis where the individual words are examined to see if they
have independent meanings that contribute to the meaningful interpretation of
the idiom as a whole. Once this process fails to produce an acceptable inter-
pretation in a context, then people probably retrieve the conventional, figura-
tive meanings of these phrases from their mental lexicons.
Support for this idea about idiom comprehension comes from reading-time
studies that showed that people took significantly less time to process decom-
posable or analyzable idioms than to read unanalyzable expressions (Gibbs,
Nayak, & Cutting 1989). These data suggest that people normally attempt to do
some compositional analysis when understanding all types of idiomatic phrases.
This does not mean, however, that people automatically compute the literal,
context-free interpretations of idioms (Gibbs 1980, 1985, 1986). The results of
one study, for instance, showed that literally ill-formed idioms (e.g.,pop the
question) are understood just as quickly as are well-formed phrases (e.g.,kick the
question) (Gibbs, Nayak, & Cutting 1989). Thus, people do not appear to be bi-
ased toward processing the putative literal meanings of idioms. Rather, some
compositional process attempts to assign some context-sensitive meanings to
the individual components in idioms during understanding. Children actually
experience greater difficulty learning the meanings of semantically unanalyz-
able idioms precisely because these phrases’ nonliteral interpretations cannot
be determined through analyses of their individual parts (Gibbs 1987, 1991;
Nippold & Martin 1989). These data show that idiom learning does not occur in
a rote manner but develops in stages as children acquire linguistic and meta-
linguistic skills (Cacciari & Levorato 1989; Levorato 1993; Levorato & Cacciari
1992).
Once again, the traditional view that idioms are dead metaphors with non-
compositional meaning cannot account for any of these empirical findings.
Many linguistic studies indicate that analyzability is an important concept in
understanding the linguistic behavior of many other formulaic expressions,
including verb-particle constructions (Bolinger 1971; Lindner 1981) and bino-
mial expressions (Lambrecht 1984). One future challenge will be to see whether
these observations and empirical findings on idiom analyzability extend to
languages other than English and to linguistic constructions that are not nor-
mally classified as idiomatic (cf. Coulmas 1981).
738 Raymond W. Gibbs Jr.