Table of Contents:
Know hope: Metaphor, optimal innovation and pleasure
Introduction: Processing models
What are the processes involved in language comprehension? Do literal and nonliteral instances of language use require different interpretation mechanisms or do they follow the same processing routes? To illustrate the question, consider the following example:
(1) On Wednesday, February 20th, 2002, Yinnon Hiller, aged 20 and Amir Malenky, aged 18, will appear before the Israeli High Court of Justice... Both ask to be released from military service on the grounds of their pacifist beliefs... Both conscientious objectors ask to perform alternative civilian service instead of military service, but the Israeli army so far refused to allow them this option.
New Profile - Movement for the Civil-ization of Israeli Society supports Yinnon Hiller and Amir Malenky in their struggle. We believe that the State of Israel should recognize that an individual can participate in society in ways other than bearing arms. http://www.gush-shalom.org/archives/ forum_eng.html (19.2.02)
How do we make sense of these utterances? What makes us interpret the last word (arms) or collocation (bearing arms) nonliterally (as ‘weapons’) rather than literally (as ‘hands’)? Is it contextual information biased toward the military sense of arms that makes us select this appropriate sense? Is it the salience or dominance of the military sense of the word that makes that meaning available swiftly? What about the meaning of struggle, then? Given the contextual bias toward the military interpretation, would its literal (‘battle’) meaning be primed during its processing on account of its contextual relatedness?
Whether it is context or the lexicon that affects our understanding primarily has been an enduring debate in linguistics and psycholinguistics for over three decades or so (Gibbs 1994; Glucksberg 2001; Giora 1997, 2002, 2003). Indeed, various models of figurative language have come up with different proposals as to how we make sense of literal and nonliteral utterances.
The Standard Pragmatic Model (Grice 1975; Searle 1979) assumes the temporal priority of the literal interpretation of utterances. In this view, literal meanings are obligatory - they are automatic and immune to contextual information. Nonliteral meanings, on the other hand, are derivative and optional. They are induced only when a literal interpretation fails to resonate with contextual information. In this view, then, it is the literal meaning (‘hand’) of arms that should be accessed first and adjusted to contextual information only as an aftermath; similarly, it is the literal (‘battle’) meaning of struggle that should be induced first and revisited later. The Standard Pragmatic Model thus assumes different processing routes for literal and nonliteral language uses, regardless of strength and bias of context. While both literal and nonliteral utterances are being processed literally initially, only nonliteral language is expected to involve an additional phase of adjustment to contextual information. According to the Standard Pragmatic Model, then, comprehending the figurative utterances in (1) should incur some integration difficulty compared to their interpretation in a literally biasing context.
Unlike the Standard Pragmatic Model, the Direct Access View (or its more recent version entitled the Constraint Satisfaction Model) does not assume that lexical processes are immune to contextual information. Rather, context interacts with lexical processes very early on and if it is sufficiently constraining, it should result in selecting the contextually appropriate interpretation exclusively or at least initially. Such a view disputes the temporal priority of literal meanings (Glucksberg 1998,1995, 2001; Glucksberg, Gildea and Bookin 1982; Keysar 1989, 1994). Instead, in realistic, social contexts, comprehenders should be able to understand the figurative interpretations of metaphors, irony/sarcasm, idioms, proverbs and indirect speech acts directly without having to first analyze and reject their literal interpretations (Ferretti, Schwint and Katz 2007; Gibbs, 1994, 2001; 2002; Katz and Ferretti 2001, 2003). The Direct Access View thus assumes no processing differences for literal and nonliteral language, provided prior context is supportive and specific enough. Rather, both types of language should be comprehended directly, without involving an inappropriate interpretation first. According to this view, then, in the strongly biasing context of (1), the utterance including struggle and arms should be interpreted metaphorically; no integration difficulties are anticipated compared to their interpretation in a literally biasing context (Ortony et al. 1978).
Following Fodor’s modular assumptions (1983), the Graded Salience Hypothesis (Giora 1997, 1999, 2003; Giora and Fein, 1999a, b; Giora, Fein and Schwartz 1998; Peleg, Giora and Fein 2001, 2004; Peleg and Eviatar 2008) assumes that comprehension involves two distinct mechanisms that run in parallel, without interacting initially. One is bottom-up, sensitive only to domain specific (here) linguistic stimuli; another is top-down involving inferential and integrative processes, susceptible to both linguistic and nonlinguistic information. Diverging from the classical modular view (Swinney 1979), however, the Graded Salience Hypothesis further assumes that bottom-up processes are ordered: Salient responses/mean- ings are accessed faster than less-salient ones (Duffy, Morris and Rayner 1988; Rayner et al. 1994).
To be salient, a meaning has to be coded in the mental lexicon and be foremost on our mind due to e.g., conventionality, frequency, familiarity, or prototypicality. Coded meanings, low on these parameters, are less-salient and slower to reach sufficient levels of activation than salient meanings. According to this view, then, coded meanings would be accessed automatically upon encounter, regardless of contextual information or authorial intent. Meanings not coded in the mental lexicon, although nonsalient, may be made available via the contextual, predictive mechanism.
Indeed, when specific enough, contextual information may affect comprehension immediately. A highly predictive context would yield meanings on its own accord very early on. However, it would not interact with lexical access and would therefore not block coded but inappropriate responses upon encounter of the lexical stimulus (Giora 2003; Peleg et al. 2001; Peleg and Eviatar 2008).
Given that both the literal and nonliteral meanings of arms and struggle are salient, the Graded Salience Hypothesis predicts that the metaphorical utterances in (1) should incur no integration difficulties compared to their interpretation in a literally biasing context. Since both meanings should be accessed automatically in both types of context, the contextually appropriate meaning is made available swiftly and effects seamless integration processes.
In sum, whereas the Standard Pragmatic Model assumes that nonliteral language should cohere less smoothly with prior context than literal language, the Graded Salience Hypothesis and Direct Access View have different predictions. Both theories assume equivalent processes for figurative and nonfigurative language, though apparently for different reasons. The Direct Access View attributes to a constraining context the role of neutralizing the differences found between literal and nonliteral language embedded in poorly informative contexts so that when context is sufficiently strong and supportive, literal and nonliteral interpretations cohere as smoothly (Ortony et al. 1978; but see Janus and Bever 1985; Peleg et al. 2004 and Giora et al. 2007 for a critique). Particularly, nonliteral language would be tapped directly without having to involve an analysis of the literal interpretation first (Gibbs 2002). The Graded Salience Hypothesis discards the literal-nonliteral distinction altogether and replaces it with the salient-nonsalient continuum (Giora 1997, 2002, 2003). Diverging from the Direct access View, it thus anticipates processing difficulties for language uses whose less or nonsalient meanings or interpretations are invited. Given that salient meanings get accessed automatically upon encounter of the relevant stimulus, when contextually incompatible, they would result in extra adjustment processes resulting in less-salient or innovative meanings and interpretations. When such interpretations are invited, processing would be more effort consuming, at least, locally, compared to when salient meanings are invited, regardless of literality or figurativeness.
Given the Graded Salience Hypothesis, then, the idiom He is singing a different tune in (2c, taken from Gibbs, 1980), or the fixed expression black on white in (3c), or the idiom you don’t know your right from left in (4c) should cohere more smoothly with prior context (e.g., take shorter to read) following (2a, 3a, 4a) than following (2b, 3b, 4b). Whereas the contexts in (2a, 3a, 4a) invite the salient (figurative) meaning of the idioms and the salient (literal) meaning of the fixed expression, the contexts in (2b, 3b, 4b) invite their low-salience (literal and figurative) interpretations. According to the Graded Salience Hypothesis, reading (2c, 3c, 4c) following (2b, 3b, 4b), then, should involve accessing (and reinterpreting) the salient meaning of the expressions in spite of their contextual incompatibility. Such predictions, however, are not invited by either of the alternative views. Both theories predict equal reading times for (2c, 3c, 4c) in all types of context (either 2a, 3a, 4a or 2b, 3b, 4b). According to the Standard Pragmatic Model, both interpretations of (4c) involve a figurative comprehension phase; hence no processing differences are anticipated. Since, however, there are literality differences involved in idioms, processing difficulties would be predicted only for the figurative reinterpretation. According to the Direct Access View, both interpretations of (2c, 3c, 4c) are invited by similarly strong and supportive contexts; hence no processing differences are anticipated. Findings, however, support the Graded Salience Hypothesis. They show that low- salience interpretations took longer to read than salient alternatives, regardless of figurativeness (Brisard et al. 2001; Frisson and Pickering 2007; Gibbs 1980; Giora and Fein 1999a, b; Giora et al. (2009); Giora et al. 2007; but see Ortony et al. 1978):
(2) a. On TV there was a program discussing Carter’s first year in office.
One reporter talked about the military budget. “In the campaign Carter promised to cut that budget.” “But now that he is the president,”
b. Nick and Sue were listening to Jackson Browne on the radio. “All Jackson Browne songs sound alike.” Sue Said. “Now isn’t that the same song we heard him do on TV recently.”
“No.” Nick replied;
c. "He is singing a different tune.”
(3) a. I want your promise documented
b. This cheese cake with chocolate coating is exactly what you wanted:
c. black on white
(4) a. The Comprehensive Lexicon will teach you whatever you are interested in
b. Buy The Comprehensive Guide for the Political Factions in Israel
c. so that you won’t feel you don’t know your right from left.
Why would speakers make use of utterances that might endanger coherence (even if only momentarily) and involve complex processes when less costly utterances are at hand? The explanation we intend to put forward and test here concerns speakers’ pursuit of aesthetic effects. Speakers resort to innovativeness because they wish to disturb without repelling, to attract rather than detract listeners’ attention. Apparently, the text in (1) does not resort to any aesthetic device in spite of its occasional figurative language. We claim here that it takes innovativeness rather than figurativeness to affect pleasure.