(25) a. a piece/slice of the action b. a skeleton in the closet/cupboard c. the calm/lull before the storm d. hold a gun/pistol to someone’s head e. throw ones’ hat/cap into the ring Variations of adjectives are less common than those of nouns or verbs, but several examples illustrate that changing adjectives is not disruptive for many phrases’ figurative meanings. Consider the following: (26) a. a bad/rotten apple b. a level/even playing field c. close/near to the bone d. the best/greatest thing since sliced bread Prepositional and adverbial participles may also exhibit little shift in meaning, as seen in the following instances: (27) a. at/in a single sitting b. by/in leaps and bounds c. on/along the right lines d. out of/from thin air e. go round/around in circles Finally, conjunctions may vary as seen in the following instances: (28) a. hit and miss/hit or miss b. when/if push comes to shove c. when/while the cat’s away, the mice will play Not surprisingly, the analyzability of idioms influences people’s intuitions about their lexical flexibility. A series of studies examined the role of semantic analyzability on the lexical flexibility of idioms (Gibbs et al. 1989). Participants were presented with a series of phrases that were left unchanged (e.g., pop the question), had their verbs substituted with relatively synonymous words (e.g., burst the question), had their nouns substituted with synonymous words (e.g., pop the request), or had both their verbs and nouns changed (e.g., burst the request). Ac- companying each phrase was a figurative definition of the unchanged idiom (e.g., propose marriage). The participants’ task was simply to read each phrase and judge its similarity in meaning to the paraphrase. Changing the verbs and nouns of both semantically decomposable and nondecomposable idioms was disruptive to their figurative meanings. However, changing the lexical items in semantically nonde- composable idioms was far more disruptive to these phrases’ figurative interpre- tations than was the case for decomposable idioms. For instance, both noun and verb changes were rated as significantly less acceptable for nondecomposable idi- oms (e.g., punt the pail for kick the bucket) than for decomposable phrases (e.g., burst the request for pop the question). These findings suggest that the semantic 710 raymond w. gibbs, jr. analyzability of idioms provides important constraints for the lexical flexibility of idioms. In addition, changing the individual words in an idiom may not totally dis- rupt that phrase’s figurative meaning, at least in the sense of rendering the phrase literal. But changing some words will alter an idiom’s figurative meaning in slight, but still important, ways (Glucksberg 2001). For example, the idiom break the ice can be altered to form shatter the ice, which now has the meaning of something like ‘to break down an uncomfortable and stiff social situation flamboyantly in one fell swoop!’ (McGlone, Glucksberg, and Cacciari 1994). Shatter the ice is an example, not of lexical flexibility, but of semantic productivity (McGlone, Glucksberg, and Cacciari 1994). Examples of semantically productive idiom variants appear fre- quently in conversation, the media, and literature. People can understand seman- tically productive idiom variants (e.g., Sam didn’t spill a single bean) quite readily, and the more familiar the original idiom, the more comprehensible the variant (McGlone, Glucksberg, and Cacciari 1994). Variant idioms can also be understood as quickly as their literal paraphrases (e.g., Sam didn’t spill a single bean versus Sam didn’t say a single word) (Glucksberg, Glucksberg, and Cacciari 1994). Overall, speakers tend to be significantly more creative in their use of seman- tically analyzable idioms, both in terms of their syntactic productivity and their lexical flexibility. 8. Analyzability in Idiom Comprehension and Learning The analyzability of idioms also plays an important role in their immediate, ‘‘on- line’’ interpretations. Because the individual components in decomposable idioms contribute systematically to the figurative meanings of these phrases, people may process idioms in a compositional manner where the semantic representations of each component is accessed and combined according to the syntactic rules of the language. A series of reading-time studies showed that people took significantly less time to process the decomposable idioms than to read the nondecomposable ex- pressions (Gibbs, Nayak, and Cutting 1989). Both normally and abnormally de- composable phrases took less time to process than their respective literal control phrases, but nondecomposable idioms actually took longer to process than their respective literal controls. These data suggest that people attempt to do some com- positional analysis when understanding idiomatic phrases. When an idiom is de- composable, readers can assign independent meanings to its individual parts and will quickly recognize how these meaningful parts combine to form the overall figurative interpretation of the phrase. idioms and formulaic language 711 Finally, children’s comprehension of idioms depends on their intuitions about the internal semantics of these figurative phrases (Gibbs 1987, 1991; Nippold and Rudzinski 1993; Levorato and Cacciari 1999). Children’s learning of idioms is generally thought to depend on their associating a given sequence of words with arbitrary figurative meanings (e.g., kick the bucket means ‘to die’). However, the evidence shows that children attempt to do some compositional analysis when understanding idiomatic expressions (Gibbs 1991). Younger children (kindergart- ners and first graders) understood decomposable idioms much better than they did nondecomposable phrases. Older children (third and fourth graders) understood both kinds of idioms equally well in supporting contexts but were better at inter- preting decomposable idioms than they were at understanding nondecomposable idioms without contextual information. Children did not understand idioms with well-formed literal meanings any better than they did ill-formed idioms. Conse- quently, it is unlikely that young children find analyzable or decomposable idioms easier to comprehend simply because these phrases possess well-formed literal meanings. Instead, the younger children found it easier to assign figurative mean- ings to the parts of decomposable idioms and did not simply analyze each expres- sion according to its literal interpretation. There is now a vast body of work from Cognitive Linguistics and psychology showing that the traditional view of idioms as being fixed expressions is quite wrong. A comprehensive theory of idiomaticity must acknowledge the complexity of these phrasal patterns, paying detailed atten- tion to their various lexical, semantic, syntactic, and pragmatic properties, and not assume that idioms are fixed, dead expressions. 9. Idiom Schemas The extensive evidence reviewed in this chapter against the idea that idioms and related speech formulas are noncompositional or fixed expressions has not ad- dressed an important issue. Any argument about the variability of idioms presup- poses that the variant forms of an individual expression are to be considered as var- iations rather than as separate expressions with coincidentally the same meaning and with some lexis in common. Consider the idiom pairs in (29) and (30) (Moon 1998): (29) a. champ at the bit b. chafe at the bit (30) a. hit the roof b. hit the ceiling Do these pairs reflect single idioms with minor variations or completely indepen- dent idioms? This question is relevant to explaining many American/British idioms pairings such as (31) and (32): 712 raymond w. gibbs, jr. (31) a. the shoe is on the other foot b. the boot is on the other foot (32) a. blow off steam b. let off steam Most linguists and lexicographers view the examples in each pair as variations from some core phrase. But may they actually be completely different idioms in the same way that the words gasoline/petrol or apartment/flat are closely related but discrete lexical items? Moon (1998) argues that pairs like the above represent idiomatic ex- pressions within a single idiom scheme. Under this view, variation in idiom ex- pressions reflects further evidence of instability rather than pointing to the idea that each phrase is actually an entirely different idiom. Although linguists have often proposed grammatical constraints on idioms, few scholars have pursued the idea that different conceptual schemes govern var- iation among these phrases. But how do we identify the canonical form of an idiom? Consider the phrases in (33): (33) a. have an axe to grind b. have no axe to grind c. with an axe to grind d. without an axe to grind e. with no axe to grind The phrases here either represent a variable idiom cluster, where there are sev- eral possible related forms, or a frozen, unvarying idiom nucleus (i.e., axe to grind) that collocates with preceding have/with/without and a/no. There may be advan- tages to seeing the above phrases as a frozen nucleus with a collocating structure, but the core axe to grind is not by itself a meaningful unit. In fact, many idioms do not have fixed forms. Even the classical example kick the bucket, which is often cited in favor of the idea that idioms are ‘‘fixed’’ expressions or ‘‘frozen’’ phrases, can meaningfully be used in various forms such as kick it, kick off,orkick it off . Moreover, many traditional proverbs and sayings are truncated from their canon- ical or earliest forms to create lower-level grammatical units. Consider the fol- lowing cases: (34) a. a bird in the hand (is worth two in the bush) b. birds of a feather (flock together) c. don’t count your chickens (before they’re hatched) d. make hay (while the sun shines) In addition to idiomatic and proverbial phrases like the above, a quick look at any idiom dictionary reveals that many idiom schemes share a prominent verb or noun. Consider the word hit in the following clusters of expressions: (35) a. hit the desk b. hit the hay c. hit the sack idioms and formulaic language 713 (36) a. hit the beach b. hit the road c. hit the surf (37) a. hit the bottle b. hit the plum wine c. hit the sauce Some linguists claim that these phrases are not individual idioms (i.e., that hit is polysemous), in the way that these phrases tend to be listed in idiom dictionaries. Instead, the above phrases are part of a broader pattern (Ruhl 1989). For example, hit is essentially monosemic, and there are principles of metonymy and analogy governing the production of these different phrases. Yet Moon (1998) suggests that rather than seeing hit as monosemic, there are broader patterns of idiom schemes governing the production of related idiomatic phrases. Consider the following re- lated phrases: (38) a. shake in one’s shoes b. quake in one’s shoes c. shake in one’s boot d. quake in one’s boot e. quiver in one’s boots f. quake in one’s Doc Martens The main verb in each phrase means ‘shake’, and this is associated with nouns meaning ‘footwear’ to connote fear and apprehension. Any words that convey sim- ilar meanings as do these verbs and nouns will make equally appropriate idioms. Another set of expressions reflective of an idiom scheme is the following: (39) a. scare the life out of someone b. scare the shit out of someone c. scare someone shitless d. scare the pants off someone e. frighten the life out of someone f. be frightened out of one’s mind g. be scared out of one’s wit Once again, these formulaic phrases have several significant conceptual parts, such as scaring someone which results in the loss of something important (e.g., one’s life, one’s mind, one’s wit, one’s pants). As long as this conceptual scheme is main- tained, a linguistic expression will convey idiomatic/formulaic meaning. One more set of synonymous variations where the idiom schema is relatively clear is seen in the following list of idiomatic phrases: (40) a. down the chute b. down the drain c. down the pan d. down the plughole 714 raymondw.gibbs,jr. e. down the toilet f. down the tubes/tube Idiom schemes like the above have some reference, a metaphor (or metonymy) or cognate words, in common, but without (necessarily) any fixed structure or specific words. The concept of idiom schemes explains several facts about the analyzability and variability in idioms and related speech formulas. As Moon (1998: 163)comments: Idiom schemas represent concepts embedded in the culture and associated with particular lexicalizations. They are characterized by an underlying concept (the relationship between tenor and vehicle) and an overlying preferred lexical reali- zation, usually with connotated evaluation. The exact form of words may vary or be exploited but is still tied to the underlying concept which provides the driving or motivating force in these idiom phrases. There has been little empirical work attempting to identify major idiom schemes or looking at people’s possible use of such schemes in idiom production and un- derstanding. But quite notably, cognitive linguists have tried to explain similar for- mulaic patterns in an approach known as Construction Grammar (Fillmore, Kay, and O’Connor 1988; Goldberg 1995; Kay and Fillmore 1999). Construction grammar is a monostratal, unification-based syntactic theory that represents form-meaning complexes as construction templates, each with a specific set of morphosyntactic, semantic, and pragmatic principles, which may be com- bined with other templates to form more complex structures. Under this approach, grammatical constructions are not reduced to simple rewrite rules, but have spe- cific semantic and pragmatic properties that must be captured (see also Croft, this volume, chapter 18) Construction grammar has important implications for the study of idioma- ticity in language, especially in providing a unified account of grammar and what have traditionally been viewed as peripheral aspects of language. For instance, in their analysis of the phrasal construction let alone, Fillmore, Kay, and O’Connor (1988, 534) argue that ‘‘it appears to us that the machinery needed for describing the so-called minor or peripheral constructions of the sort which has occupied us here will have to be powerful enough to be generalized to more familiar structures, in particular those represented by individual phrase structure rules.’’ For example, consider the following two sets of expressions: (41) a. Mary won’t eat shrimp, let alone squid. b. I barely got up to eat lunch, let alone cook breakfast. c. I was too young to serve in World War Two, let alone World War One. (42) a. What is the scratch doing on the table? b. What do you think your name is doing on my book? c. What is it doing raining? The phrase let alone allows speakers to simultaneously address a previously posed proposition and to redirect the listener to a new proposition that should be more informative. Constructions of the What’s X doing Y? type express both a request or idioms and formulaic language 715 a demand for an explanation and the speaker’s belief in the incongruity of the scene proposition. The constructional structure of these clauses is dictated by the or- dinary core constructions that license familiar subject-predicate structures, verb phrases, and inverted clauses. ‘‘In grammar, the investigation of the idiomatic and of the general are the same; the study of the periphery is the study of the core—and vice versa’’ (Kay and Fillmore 1999: 30). Work on formulaic phrases in construction grammar has not yet made contact with the broader empirical study of idioms in linguistics and psychology. But it seems evident that principles of idiomaticity may be governed by various con- structions or idiom schemes that have more in common with more productive aspects of language than has been assumed in the past. One important possibility to explore is the idea, inherent in Construction Grammar, that construction templates are related through inheritance hierarchies, containing more and less general pat- terns. Thus, idiom constructions, or idiom schemes, may be organized in higher- order structures that reflect something of how people conceptualize related and less related idiomatic phrases. One additional observation, which may relate to broader proposals on idiom schemas/constructions, is that few idioms have literally animate NPs, and there are few idiomatic phrases of the form V þ Goal, such as throw NP to the wolves (Nunberg, Sag, and Wasow 1994). For example, English has several dozen idioms of the form hit þ NP, as mentioned earlier, such as hit the bottle, hit the road, and hit the sack, where the objects denoted are inanimate. This contrasts with nonidio- matic phrases with transitive hit which often include objects with animate themes. What accounts for this statistical tendency? Nunberg, Sag, and Wasow ( 1994) argue that the rarity of some idiom chunks arises from two general facts about meaning transfer in figurative uses of language. First, abstract situations are usually described in concrete terms. Second, inanimate concepts are mapped onto animate ones, more so than the opposite. Because ani- mates are necessarily concrete, literally animate NPs in idiom phrases are not used to denote abstract entities. For this reason, literally animate NPs are rare in idio- matic expressions. Agents and Goals are also rare to find in idioms because they too are typically animate. This analysis is also supported by analysis of proverbs involving NP arguments with animate literal meanings and inanimate idiomatic readings. Thus, the phrase look a gift horse in the mouth maps the animate horse onto an inanimate entity that has been freely given. This mapping illustrates the general tendency of meaning transfer where concrete things and situations (e.g., the body, spatial relations) serve as the source domains for more abstract target domains (e.g., social interactions, causal relations, etc). Under this view, it would be quite unusual to find idioms or proverbs such as divulge the information to mean something like ‘spill the soup’, as in The waiter divulged the information all over my new suit (Nunberg, Sag, and Wasow 1994). It appears, then, that there are important conceptual constraints on permis- sible idioms and proverbs. Idiom schemas are likely constituted, at least in part, by meaning transfer processes that are grounded in nonlinguistic patterns of thought. 716 raymond w. gibbs, jr. 10. The Conceptual Basis for Idiomaticity The important empirical demonstrations on idiom analyzability suggest that the internal semantics of idioms might be correlated in systematic ways with the con- cepts to which idioms refer. One possibility is that the figurative meanings of idioms might very well be motivated by people’s conceptual knowledge that is consti- tuted by metaphor (and to a lesser extent metonymy) (Gibbs 1994;Ko ¨ vecses and Szabo ´ 1996). In fact, people make sense of many idioms because they tacitly rec- ognize the metaphorical mapping of information from two domains that give rise to idioms in the first place. Cognitive linguistic research supports this view (Lakoff 1987;Ko ¨ vecses 2000). For example, the idiom John spilled the beans maps our knowledge of someone tipping over a container of beans onto a person revealing some previously hidden secret. English speakers understand spill the beans to mean ‘reveal the secret’ be- cause there are underlying conceptual metaphors, such as the mind is a con- tainer and ideas are physical entities that structure their conceptions of minds, secrets, and disclosure (Lakoff and Johnson 1980). Although the existence of these conceptual metaphors does not predict that certain idioms or conventional expressions must appear in the language, the presence of these independent con- ceptual metaphors by which we make sense of experience provides a partial mo- tivation for why specific phrases (e.g., spill the beans) are used to refer to particular events (e.g., the revealing of secrets). A good deal of empirical work in psycholinguistics has investigated the met- aphorical motivation for idiomatic meaning. For instance, various psycholinguistic evidence supports the idea that metaphors such as anger is heated fluid in a container are really conceptual and not, more simply, generalizations of linguistic meaning. Studies have looked at people’s mental imagery for idioms and proverbs (Gibbs and O’Brien 1990; Gibbs, Strom, and Spivey-Knowlton 1997), people’s context-sensitive use of idioms (Nayak and Gibbs 1990) and euphemistic phrases (Pfaff, Gibbs, and Johnson 1997), people’s folk understanding of how the source domains in conceptual metaphors constrain what idioms mean (Gibbs 1992), people’s use of conceptual metaphors in organizing information in text processing (Allbritton, McKoon, and Gerrig 1995), and people’s use of conceptual metaphors in drawing inferences when reading poetic metaphors (Gibbs and Nascimento 1996). These psycholinguistic findings support the hypothesis that different kinds of metaphorical thought partly explain why many metaphors and idioms have the meanings they do for contemporary speakers. Let me briefly describe two sets of studies that illustrate the importance of conceptual metaphor in idiom understanding. Emphasized in this work is the idea that many conceptual metaphors are grounded in recurring patterns of embod- ied experience. Thus, the source domains that are mapped onto different target idioms and formulaic language 717 domains in idioms are themselves ‘‘image-schematic’’ structures (Johnson 1987; Gibbs and Colston 1995). One set of psycholinguistic studies examined how people’s intuitions of the bodily experience of containment, as well as several other image schemas which serve as the source domains for several important conceptual metaphors, underlie speakers’ use and understanding of idioms. These studies were designed to show that the specific entailments of idioms reflect the source-to-target domain map- pings of their underlying conceptual metaphors (Gibbs 1992). Most importantly, these metaphorical mappings preserve the cognitive topology of these embodied, image-schematic source domains. Participants in a first study were questioned about their understanding of events corresponding to particular bodily experiences that were viewed as motivating specific source domains in conceptual metaphors (e.g., the experience of one’s body as a container filled with fluid). For instance, participants were asked to imagine the embodied experience of a sealed container filled with fluid and then were asked something about causation (e.g., What would cause the container to explode?), in- tentionality (e.g., Does the container explode on purpose, or does it explode through no volition of its own?), and manner (e.g., Does the explosion of the container occur in a gentle or a violent manner?). Overall, the participants were remarkably consistent in their responses to the various questions. To give one example, people responded that the cause of a sealed container exploding its contents out is the internal pressure caused by the increase in the heat of the fluid inside the container. They also reported that this explosion is unintentional because containers and fluid have no intentional agency and that the explosion occurs in a violent manner. These brief responses provide a rough image-schematic profile of people’s understanding of a particular source domain concept (i.e., ‘heated fluid in the bodily container’). These different image-schematic profiles about certain abstract concepts al- lowed me to predict something about people’s understanding of idioms. For in- stance, people’s understanding of anger should partly be structured by their folk concept for heated fluid in the bodily container as described above. Several studies showed this to be true (Gibbs 1992). Not surprisingly, when people understand anger idioms, such as blow your stack, flip your lid,orhit the ceiling, they inferred that the cause of anger is internal pressure, that the expression of anger is unin- tentional, and that it is done is an abrupt violent manner. People do not draw these same inferences about causation, intentionality, and manner when comprehending literal paraphrases of idioms, such as get very angry. More interesting, though, is that people’s intuitions about various source do- mains map onto their conceptualizations of different target domains in very pre- dictable ways. For instance, several later experiments showed that people find idioms to be more appropriate and easier to understand when they are seen in discourse contexts that are consistent with the various entailments of these phrases. Thus, people find it easy to process the idiomatic phrase blow your stack when this 718 raymond w. gibbs, jr. was read in a context that accurately described the cause of the person’s anger as being due to internal pressure, where the expression of anger was unintentional and violent (all entailments that are consistent with the entailments of the source- to-target domain mappings of heated fluid in a container onto anger). But readers took significantly longer to read blow your stack when any of these entailments were contradicted in the preceding story context. These psycholinguistic findings show how people’s metaphorical concepts underlie their understanding of what idioms mean in written texts. Moreover, they provide significant experimental evidence that people’s intuitions about their em- bodied experiences can predict something about their use and understanding of idioms, expressions that are partly motivated by bodily based conceptual metaphors. A different series of experiments demonstrates that people appear to compute or access metaphorical representations during their immediate understanding of idioms like blew his stack (Gibbs et al. 1997). In these studies, participants read stories ending with idioms and then quickly gave lexical decision responses to visually presented letter-strings that reflected either something about the concep- tual metaphors underlying these idioms (e.g., ‘heat’ for anger is heated fluid in a container having just read John blew his stack) or letter-strings that were un- related to these conceptual metaphors (e.g., ‘lead’). There were two important findings from this study. First, people were faster to make these lexical decision responses to the related metaphor targets (e.g., ‘heat’) after having just read idioms than they were after having read either literal para- phrases of idioms (e.g., John got very angry) or control phrases (e.g., phrases still appropriate to the context such as John saw many dents). Second, people were faster in recognizing related metaphorical targets than unrelated ones after having read idioms, but not after literal paraphrases or control phrases. This pattern of results suggests that people are immediately computing or accessing at least something related to the conceptual metaphor anger is heated fluid in a container when they read idioms. In another experiment, participants were faster to make lexical decision responses to metaphor targets (e.g., ‘heat’) after having read an idiom motivated by a similar conceptual metaphor (e.g., John blew his stack) than after an idiom with roughly the same figurative meaning but motivated by a different con- ceptual metaphor (e.g., John bit her head off, which is motivated by the conceptual metaphor anger is animal behavior). Experimental findings like these suggest that people compute or access the relevant conceptual metaphor for an idiom during some aspect of their processing of these phrases. Not all, psychologists in particular, agree with this conclusion (Glucksberg and Keysar 1990; Kreuz and Graesser 1991; Glucksberg, Brown, and McGlone 1993; Stock, Slack, and Ortony 1993; Keysar and Bly 1995; McGlone 1996). Much of the debate over whether metaphors of thought influence verbal metaphor understanding centers on the most appropriate methodology for examining lin- guistic understanding. One criticism about some of the above empirical research is that asking people about their intuitions as to why figurative expressions mean idioms and formulaic language 719 . verb phrases, and inverted clauses. ‘‘In grammar, the investigation of the idiomatic and of the general are the same; the study of the periphery is the study of the core—and vice versa’’ (Kay and Fillmore. idioms. Another set of expressions reflective of an idiom scheme is the following: (39) a. scare the life out of someone b. scare the shit out of someone c. scare someone shitless d. scare the pants off. raymond w. gibbs, jr. (31) a. the shoe is on the other foot b. the boot is on the other foot (32) a. blow off steam b. let off steam Most linguists and lexicographers view the examples in each pair