Journal of Psycholinguistic Research. Vol. 28, No, 1, 1999
Proverb Comprehension: The Primacy of Literal Meaning Jon G. Temple1,2 and Richard P. Honeck1
According to the multistage model of figurative language understanding, literal meanings must be developed before figurative meanings. Although the model implies that figurative understanding should take longer than literal understanding, Kemper (1981) reported that figuratively biased proverbs were processed more quickly than literalized proverbs. By contrast, in the present study the results from six experiments yielded the opposite conclusion. These results support the multistage model and the conceptual base theory of proverb comprehension (Honeck, 1997; Honeck & Temple, 1994; Honeck, Voegtle, Dorfmueller, & Hoffman, 1980) which incorporates it. Discrepancies between studies that have examined the multistage model may crucially depend on methodological factors such as type of experimental design, materials, and, as apparent in the present case, the task and dependent measure used.
Proverbs present an interesting theoretical challenge. This is because their use and comprehension entail a variety of issues in the areas of categorization, mental representation, pragmatics, attention, and more (Honeck, 1997). The present study focused on one issue, namely whether proverbs are understood in a serial or in a more direct, nonserial way. That is, does comprehension of a proverb's figurative meaning first require the development of a literal meaning for the proverb? This issue applies to essentially all forms of indirect language (i.e., figuration, indirect speech acts, etc.) and by convention it is called the "multistage" issue. We will return to this issue momentarily, but first some general considerations of proverb comprehension are in order. 1 2
University of Cincinnati, Cincinnati, Ohio 45221-376. Address all correspondence to Jon G. Temple, Ph.D., Mail Drop 252, IBM Global Services, 150 Kettletown Road, Southbury, Connecticut 06488. 41 0090-6905/99/0100-0041 $ 16.00/0 © 1999 Plenum Publishing Corporation
42
Temple and Honeck
There are two general kinds of situations in which proverbs are used and understood—irrelevant-context and relevant-context situations (Honeck & Temple, 1994). Our description of these situations will assume the viability of a multistage processing model. In an irrelevant-context situation, a proverb is uttered in a nonsupportive context that does not contain a topic for the proverb. That is, the context is irrelevant for the construction of either a figurative or literal meaning for the proverb. This situation occurs, for example, when people are given intelligence tests such as the Wechsler Adult Intelligence Scale (Matarazzo, 1972) or Gorham's Proverb Test (Gorham, 1963) and are asked to interpret some proverbs. On these tests someone might be asked to interpret a proverb such as, "A used key is always bright." A more extreme example of an irrelevant-context situation would occur if we were approached by a stranger in an elevator, and the stranger uttered the "key" proverb. Presentation of proverbs in irrelevant-context situations presents an interesting set of problems for would-be interpreters. Most important, they must recognize that a literal meaning is inappropriate, and that they must construct a newer, more appropriate meaning. In some irrelevant-context situations (e.g., the elevator example) they must even decide whether it is worth trying to understand the proverb. From a theoretical and practical standpoint, there is nothing wrong with presenting proverbs sans context, although one must be aware of the cognitive challenges that it presents. Despite its widespread use in psychiatric testing, for example, serious questions have been raised about its validity (Andreason, 1997; Gibbs, 1995; and Honeck, 1997, discuss these questions) The second, and more common, situation in which proverbs are used is a relevant-context situation. In this situation the larger context and the topic constitute important relevant information for constructing a figurative or literal proverb meaning. For example, suppose that a high school student is complaining to a parent about a paper he or she has to write for a history class. The crux of the student's complaint is that he or she fails to see the necessity of having to continually write papers. In response to this the parent says, "The used key is always bright." In uttering this statement the parent has suggested that practice keeps one's writing skills at a high level and that this is to be valued. By implication, the student has been exhorted to work on the paper because there is a larger payoff. This form of proverb use is common but simple compared to proverb performances that can occur in some African societies in which proverbs are more highly valued than in Western cultures (Finnegan, 1970; Monye, 1996; Penfield & Duru, 1988). Indeed, the appropriate use of proverbs in some cultures may require the satisfaction of a large number of pragmatic felicity conditions (Briggs, 1985;
Comprehension of Proverbs
43
Gokhan, 1992) and may occur as a complex but integral part of everyday social interactions (Monye, 1996). The interpreter's task in a relevant-context situation involves a number of components. The interpreter must understand the topic and the proverb, and effect a mapping or correspondence between them. The interpreter must also get the underlying pragmatic point of the proverb's utterance. The theoretical problem is that the topic and the proverb do not match on a literal level. In the high school student example, the paper writing issue (i.e., the topic) and the "key" proverb do not comport. The writing of papers has no obvious connection with keys, brightness, etc. The student must nevertheless come to understand that the "key" utterance means basically that things that are frequently used can improve in function. This (figurative) meaning must then be mapped to his or her own writing skills, and to the value of improving them. It is important to note that the mapping process is fundamental to what we mean by proverb comprehension in this paper. Moreover, there seems to be no way to validate such comprehension without requiring a mapping of some kind. The relative comprehension of a proverb can only be assessed by examining how proverbs are related to interpretations, instances (topics), or, in general, some other input. This is the case because proverb figurative meaning is an unobserved, hypothetical construct whose nature can only be inferred from the various ways in which this meaning can be expressed (Honeck, 1997). As applied to the "key" example, the stages issue amounts to asking about the relative serialness of the mental processes that occur in proverb comprehension. In short, could the proverb be understood in a direct manner, without first constructing meanings about keys, key use, and brightness, or are these sort of meanings prerequisite? A conventional position on the issue has been distilled from the writings of Clark and Lucy (1975), Grice (1975), and Searle (1979). The multistage model presupposes Grice's cooperative principle, that a speaker intends to convey truthful, relevant information when speaking to others. When confronted with an utterance, a listener tries to infer the speaker's intended meaning. In Stage 1 of the model, the listener does this by first computing the literal meaning of the utterance. In Stage 2, given contextual constraints, such as knowledge of the immediate situation and other background knowledge, the listener decides whether this literal rendering is appropriate. If the literal meaning is deemed appropriate, then processing stops here. However, if it is found to be inappropriate or defective, processing continues into Stage 3, where the listener will attempt to construct a new meaning for the utterance which will make it consistent with the context.
44
Temple and Honeck
For figurative language, this new meaning is the figurative meaning, involving all three (or more) stages of processing. The empirical implications of the multistage model are straightforward. Assuming that each stage adds to mental processing time, the model predicts that people should take longer to understand utterances in a figurative than a literal fashion. Although a small literature has produced findings that are consistent with the multistage model (e.g., Clark & Lucy, 1979; Janus & Bever, 1985), the bulk of the research has led to the conclusion that figurative utterances need not take longer to understand than literal utterances (Clark, 1979; Gibbs, 1979, 1986; Gibbs, Nayak, & Cutting, 1989; Glass, 1983; Pollio, Fabrizi, Sills, & Smith, 1984). These results are usually interpreted in one of two ways. When an utterance is more rapidly understood in a figurative fashion than in a literal one, this is typically taken as support for a direct access model (Gibbs, 1980, 1983, 1986; Kemper, 1981; Schweigert & Moates, 1988) in which figurative meanings are automatically selected from memory without first generating a literal meaning for an utterance. On the other hand, some researchers have tested the null hypothesis, interpreting a finding of no difference between figurative and literal reading times as supporting a parallel model, whereby literal and figurative meanings are generated simultaneously and independently (Estill & Kemper, 1982; Inhoff, Lima, & Carroll, 1984; Ortony, Schallert, Reynolds, & Antos, 1978). As a result of this groundswell against the multistage model, many in the area are convinced that the multistage model is dead (Temple & Honeck, 1991) Although the majority of studies on this issue have used idioms, metaphors, and indirect requests, a smaller literature exists using proverbs. The present study focused on proverbs. Proverbs can have the important characteristic, not present for metaphors, similes, and many indirect speech acts, that their figurative and literal senses are distinct but related in a principled way. That is, the literal and figurative meaning are qualitatively different, yet figurative meaning can be constructed using literal meaning and other comprehension strategies. For example, the proverb, The best fish are found near the bottom, can be construed as a handy fishing tip or, more figuratively, as asserting that the most valuable things are often the hardest to obtain. Moreover, unlike idioms, novel proverbs can be created anew, permitting tighter control over potential confounds, such as familiarity. For 16 years, the only study published and cited vis-a-vis proverbs and the multistage model was Kemper's (1981). For this reason, we will present and critique this study in some detail. Kemper conducted four experiments. In Experiment 1, participants read short or long paragraphs that were either figuratively or literally related to a proverb. Participants were required to judge whether a proverb that followed the paragraph was "appropriate,"
Comprehension of Proverbs
45
that is, whether it was, "originally the final concluding sentence of the preceding paragraph" (Kemper, 1981, p. 184). Kemper found that reaction times (RTs) were faster in the figurative than in the literal condition, although this was primarily due to smaller RTs in the figurative condition using long contexts and larger RTs in the literal condition with short contexts. This general finding is inconsistent with the multistage model. An initial problem here is that it is not clear what "appropriateness" of the context-proverb relationship would mean to participants. They could have construed it to mean aptness, plausibleness, stylishness, or naturalness, all of which involve metalinguistic skills and go beyond mere comprehension. Kemper (1981) was similarly concerned that the appropriateness judgments may have led participants to adopt an unnatural or artificial processing strategy. Therefore, in Experiment 2, participants' time to merely read (only appropriate) proverbs was measured. This reduced response times in the literal condition from M = 4.63 s (Experiment 1) to M = 2.37 s, and in the figurative condition from M = 3.66 s (Experiment 1) to M = 2.12 s. Nevertheless, the 0.25 s difference was statistically significant. However, the reduction in RTs suggests that the appropriateness judgments were more complex than reading time and may have tapped factors that were tangential to comprehension and therefore for the multistage model. A related point is that, in the examples of the materials Kemper (1981) provided for Experiment 1, it is clear that some transitions between the paragraph and proverb were awkward, particularly in the literal condition. However, participants would not necessarily have considered these awkward literal trials to be inappropriate, because there also were trials where the proverb and the context were clearly unrelated. Instead, the awkward transitions may have led participants to treat the literal trials as near misses, requiring additional processing time. In this regard, the RT was M = 5.15 s in the literal condition using short contexts in Experiment 1, which seems unduly long. It is unclear why it would take participants this long to judge the appropriateness of, for example, "Diamonds come in small packages," unless this statement was somehow appropriate but awkwardly related to its context. Even in the reading-time task in Experiment 2, awkwardness would have made it more difficult for participants to relate the proverbs to their contexts. Another possible confound in the materials was familiarity. Kemper (1981) reported her proverbs as being unfamiliar to participants. However, they were really only moderately unfamiliar, having received a mean rating of 3.3 on a 5-point scale where 5 = very familiar. More recent research has indicated that familiar proverbs are understood in a nonliteral fashion more quickly than unfamiliar proverbs (Case, 1991; Turner & Katz, 1990). This
46
Temple and Honeck
result is consistent with the multistage model, which implies that the figurative construal of familiar utterances will take less time, since familiarity may facilitate both the literal and figurative processing of an utterance. Kemper (1981) used within-subject designs in the first two experiments. However, this may have produced a figurative processing set, thereby facilitating RT in the figurative condition and artificially inflating it in the literal condition. In Experiment 3, Kemper therefore used a design in which half of the participants received all literal contexts first, followed by all figurative contexts, while the other half received the opposite order. Once again, RT was slower in the literal bias condition, due primarily to smaller RTs in the figurative condition using long contexts and larger RTs in the literal condition with short contexts. Unfortunately, in statistical analyses for this experiment the data were collapsed across both orders, thereby rendering the data uninterpretable regarding the effect of processing mode. Moreover, participants received the same proverbs in their respective literal and figurative conditions, thereby further negating these potentially important results. Experiments 1 to 3 examined how proverbs were processed in relation to a larger context. In Experiment 4, proverbs were preceded by a singleword biasing cue. The results revealed that appropriateness-based RTs were now faster in the literal than the figurative condition. This result is consistent with the results of Clark and Lucy (1975) for indirect requests, and Ortony et al. (1978) for metaphors presented in short contexts. In sum, serious questions can be raised about the results reported by Kemper (1981). Particularly troublesome are the problems regarding the appropriateness judgments and the fit between the contexts and the proverbs. The reading-time measure used in Experiment 2 softened these problems but this measure provided no guarantees regarding a participant's level of understanding of a proverb or of its connection to its context. The attempt to clarify the role of figurative processing sets was unsuccessful. In our view, no clear conclusion can be drawn about the kind or accuracy of comprehension shown by Kemper's participants. It follows that her study may have a limited bearing on the multistage model. Kemper (1981) concluded that her results supported an expectancybased account rather than an inferential account. She stated that "inferences from the literal meaning of the proverb to their figurative meaning seem to be required only when subjects cannot use contextual information to generate expectations about what the speaker is likely to say'' (Kemper, 1981, p. 197). That is, background, contextual and mutually shared information lead to strong expectations about meanings, so that a particular utterance, such as a proverb, can be processed directly. For idioms and indirect requests that are highly conventionalized, and where nonliteral meanings are generally prestored, the notion of direct access
Comprehension of Proverbs
47
can be defensible. However, it is less defensible with proverbs—a proverb need not have a prestored meaning to be understood. Moreover, Kemper's (1981) findings of faster figurative than literal RTs, upon which her directaccess model was based, are anomalous and unreplicated. In our view, the theoretical problem is that the direct-access position places essentially all of the basis for understanding in the context and none whatsoever in the proverb. That is, people are presumed to already have an understanding of some topic, with the proverb serving as a mere redundant confirmation of it. This amounts to saying that the proverb requires no understanding because the required, particular understanding has already been derived, the proverb itself being of little conceptual value. Thus, a radical contextualist, directaccess model is uninformative about the role of a proverb in relevant-context situations. Moreover, this model leaves unexplained how people can map topics to proverbs despite clear differences in their literal content. Apparently, this model assumes that a proverb's figurative meaning is prestored along with whatever literal meaning it might have. A particularly acute problem with a contextualist, direct-access model is that it should predict no difference in the time it takes to process literal and figurative statements. If context can so overdetermine the meaning of figurative statements, and reduce the time for their comprehension, why should it not do the same for literal statements? Thus, the longer RTs in Kemper's literal conditions are an anomaly, unless, as we have stated, it is assumed that all proverbs, unfamiliar and familiar, have prestored meanings. Of course, this is a dubious assumption. Kemper's (1981) results and conclusion are inconsistent with the conceptual base theory of proverb comprehension (Honeck, 1997; Honeck & Temple, 1994; Honeck, Voegtle, Dorfmueller, & Hoffman, 1980). Briefly, according to this theory, full proverb comprehension entails five phases— literal meaning construction, recognition that this meaning is inadequate, literal transformation of the literal meaning, construction of a figurative meaning, and instantiation of this meaning. Like the multistage model, the conceptual base theory makes figurative comprehension the result of serial processes by which the literal meaning of a proverb is transformed to help construct a nonliteral meaning. The concept of literal meaning employed in the conceptual base theory stems from Dascal's (1987, 1989) suggestion that early conceptualizations have been too simplistic in assuming that compositionality is the sole determinant of literal meaning, the view that one should be able to determine a sentence's literal meaning completely from its component morphemes and from the rules of syntactic composition in the absence of any contextual information (Searle, 1979). Dascal has argued that conventionality is also an important aspect of literal meaning. There are many composite expressions, such as frozen idioms and dead metaphors,
48
Temple and Honeck
that are learned by rote, and have higher-order literal meanings that cannot be completely determined through a compositional analysis. Moreover, conventionality factors into the compositional literal meaning, since all word assignments and pragmatic rules are established through social consensus. Per Dascal's (1987) scheme for characterizing literal meaning, both conventionality and compositionality must be viewed as determinants of literal meaning. Considered in this manner, literal meaning is based on the default values of the words and their syntactic combination, activated background knowledge, lexicalized phrasal constituents, and their conventional usages. Using this construal of literal meaning, the conceptual base theory has generated results that are consistent with the multistage model, but we will defer discussion of these and related data until the General Discussion. No study has been published that has challenged Kemper's (1981) claims. Since we believe that there is strong reason to do so, we have taken up this challenge. The core of our approach was to develop a different kind of task and measure—a mapping task involving a choice-RT measure. For this technique, there are two contexts and a proverb. Both contexts are related in either a figurative or a literal way to the proverb, but only one context bears the correct relationship to the proverb. Participants first read the two contexts and then the proverb. They must then choose the context which is more meaningfully related to the proverb, and their latency in choosing is measured. This task and measure have several advantages that together make it more valid than the appropriateness judgment task coupled with either the reading-time or RT measure used by Kemper (1981). First, because the relationship between the context and the proverb is defined (as either literal or figurative), a more valid inference can be made about the kind of meanings participants use in mapping the proverb to the context. Second, because the task requires only a mapping (i.e., categorization) operation, and not a metalinguistic judgment such as appropriateness, it provides a purer measure of proverb comprehension. Third, because participants' accuracy is measured, it can be determined whether they understood the mapping in the correct way. Fourth, the measure is online, in the sense that it solicits a response immediately upon an understanding of how the proverb maps to context. That is, context processing is complete and offline, whereas proverb processing is measured theoretically when the requisite literal or figurative (and literal) meanings have developed. Finally, the measure has a degree of ecological validity because proverb understanding is measured in relation to a topic, which mimics the relevant-context situation for proverb use. While the present technique involves two potential topics, this undoubtedly reflects what happens in everyday situations in which the proverb user selects out one topic from among several in a discourse to comment on (Briggs, 1985;
Comprehension of Proverbs
49
Penfield & Dura, 1988). Contexts are multidimensional, involving many kinds of common ground information. Proverbs categorize one of these elements and typically make a pragmatic point about it. Two blocks of three experiments apiece were conducted, for a total of six experiments. The first block played the critical role in testing the multistage model. The second block served largely as methodological checks on the stimulus materials used in the first block.
EXPERIMENTS 1, 2, AND 3 The first three experiments were conducted to determine whether it takes longer to process a proverb in a literal or figurative fashion. Experiment 1 used a between-subjects design, Experiment 2 used a within-subjects design, and Experiment 3 replicated Experiment 1 but for a different methodology for presenting stimuli. Experiment 1: Method Participants Thirty-two students in introductory psychology courses at the University of Cincinnati served as participants. Eight men and eight women served in the figurative condition, while six men and 10 women served in the literal condition. Their participation satisfied a course requirement or earned them extra credit in the course, as for all the experiments reported. Materials There were 22 proverb families. Two were used in the practice trials and 20 in the test trials. A family consisted of a proverb, a pair of twosentence contexts that were literally related to the proverb, and a pair of two-sentence contexts that were figuratively related to the proverb. This is illustrated in Table I for two proverb families. Proverbs. The proverbs were syntactically and semantically nondeviant. Thus, they could be used in conjunction with literal contexts without providing obvious cues for a figurative interpretation. In addition, the proverbs had nonoverlapping literal and figurative meanings. Most of the proverbs had been used in prior research (Honeck, Case, Temple, & Firment, 1990; Honeck, Kibler, & Sugar, 1985). An independent group of 40 participants used a 5-point scale to rate the familiarity (5 = high familiarity) of the literal form of the proverbs. The mean rating was 2.79 with a range of 1.57 to 4.62. A second independent group of 40 participants used a 5-point scale
Temple and Honeck
50
Table I. Example of the Materials for Two Proverb Families "Cow" family Proverb New proverb Context: literal, relevant Context: literal, irrelevant Context: figurative, relevant Context: figurative, irrelevant
The cow gives good milk but kicks over the pail. The ocean delights the sailor then drowns him. The dairy farmer was complaining about his most productive cow. When it was being milked it just couldn't stand still. The girl who lived in the country had a favorite cow named Elsie. Elsie provided lots of fun for the girl. The stockbroker made lots of money for her clients. Then, suddenly, every stock she recommended lost big money. The student became upset when her college applications were turned down. She quickly applied to other colleges. "Fish" family
Proverb New proverb Context: literal, relevant Context: literal, irrelevant Context: figurative, relevant Context: figurative, irrelevant
The best fish swim near the bottom. The tastiest coffee beans grow in the highest branches. Bill went fishing in the old pond. He discovered that the more he lowered the bait in the water, the bigger the fish he caught. The owner of the seafood store sold every sort of fish. It cost him more to do it, but his customers were more satisfied. The student was looking for a cheap computer. She looked all over and finally found a bargain in an offbeat magazine. For years, maintenance of the building had been neglected, and now it was condemned. The building manager just shook his head.
to rate how easily the proverbs were comprehended (5 = very easy to comprehend). The mean rating was 3.72 with a range of 2.75 to 4.59. In addition, an independent group of 40 participants used a 5-point scale to rate the imagery aroused by the literal form of the proverbs (5 = high imagery). The mean rating was 3.43 with a range of 2.09 to 4.19. Contexts. For each proverb, a pair of two-sentence contexts was written such that each was related to the proverb in a literal way. One context, the relevant (correct) context, was designed to confirm or otherwise to be consistent with the proverb considered as a literal proposition. The irrelevant (incorrect) context was designed neither to confirm nor to disconfirm the proverb, but to comment in a neutral way on the proverb's topic. Both literal contexts within a family shared either one or two content words with the proverb. Content word overlap served as a cue that a literal interpretation was called for. The overlap also prevented the participants from using a simple word matching strategy for choosing the correct context item. The
Comprehension of Proverbs
51
figurative contexts did not share any content words with the proverbs. Obviously, if there was literal content word overlap this would have prompted participants to search for a literal interpretation when, in fact, the correct choice was figuratively related to the proverb. Such overlap would therefore have artificially inflated RT under the figurative condition. The relevant figurative contexts were written to be excellent instances of the figurative meaning of the proverbs. The irrelevant figurative contexts were written to be unrelated either literally or figuratively to the proverbs. In fact, they were constructed to be excellent instances of the figurative meaning of proverbs that were not used in this study, and therefore they had the same qualities as the correct instances, except for relevance. In order to check the experimenter's construction of the contexts, an independent group of 30 participants judged which of the two literal contexts and which of the two figurative contexts "made the most sense" with respect to the proverb. The participants chose the relevant literal context 91.1% of the time, and the relevant figurative context 93.7% of the time. Thus, without the pressure of a RT task, the experimenter-derived correct contexts were judged as correct over 90% of the time, and equally so for the two conditions. In the RT portion of the experiment, the participants read a pair of contexts before seeing a proverb. To rule out the possibility that the contexts would be differentially available in short-term memory, the paired contexts were equated as much as possible on three variables—number of syllables, comprehensibility, and concreteness (imagery value) The two contexts within a bias condition varied by no more than one syllable. The mean number of syllables for the literal contexts was 32.38, while the mean was 32.45 for the figurative contexts. An independent group of 25 participants judged which of the two literal contexts and which of the two figurative contexts was the "hardest to understand." These judgments were made without the presence of the proverbs. The relevant literal contexts were judged as harder to understand 57% of the time and the irrelevant literal contexts 43% of the time. The relevant figurative contexts were judged harder to understand 44.4% of the time and the irrelevant figurative contexts 55.6% of the time. These values suggest that, if anything, it would be harder to select the correct choice in the literal than the figurative condition. In this case, the RT data would be biased against finding support for the multistage model. However, overall, only 5 times out of 40 was one context selected more often than 69.6%, which was the lower bound for above chance performance, z = 1.95, p < .05. Two of the cases were incorrect choices, and three were correct choices. Thus, the contexts by themselves were approximately equally easy to understand. An independent group of 24 participants rated the mental imagery aroused by each of the contexts using a 5-point scale on which 5 = high
52
Temple and Honeck
imagery. These ratings were made without the proverbs being present. The mean ratings of 3.21 for the relevant literal contexts and 3.02 for the irrelevant literal contexts did not differ significantly, t(19) = 1.35, p > .05. The mean ratings of 2.66 for the relevant figurative contexts and 2.89 for the irrelevant figurative contexts also did not differ significantly, t(19) = 1.68, p > .05. Procedure The participants sat alone in a quiet room in front of a computer and a monitor. The experimenter showed them how to place their hands on the keyboard—the forefinger of the left hand for the T (top context) key and the forefinger of the right hand for the B (bottom context) key. The participants were first given two practice trials that were unique to their condition, with appropriate feedback, followed by two more practice trials with the same materials in which the experimenter explained why the correct choice was correct and if an incorrect choice was made, why it was incorrect. No responses were recorded during the practice trials. A trial consisted of the following. A pair of contexts appeared on the monitor screen for 35s, one in the upper area of the screen and the other below it. Pilot work indicated that this was ample time to read the two contexts and to remember their respective positions on the monitor screen. An audible beep was then sounded which acted as a get ready signal, followed 3 s later by a proverb. The participant was instructed to hit the key that corresponded to the correct context. The correct context was defined as the one that ' 'makes the best fit'' with the proverb. Instructions emphasized quick but accurate responses. Once the participant responded the screen went blank for 2 s and then the next trial began. In the instructions, the contexts were called "situations" and the proverbs were called "sentences." The word proverb was not used in the instructions. Also, the participants were told that the top situation would be correct half of the time and the bottom situation the other half, but that there was no pattern as to whether top or bottom was correct. The practice trials were followed by 20 test trials. The latter proceeded exactly as the practice trials, except that no feedback as to correctness was expected or provided. Response times were recorded to the nearest millisecond. The correctness of the response was also recorded. An IBM-compatible computer managed verbal and auditory stimulus presentation, intertrial interval, RT recording, and response accuracy recording. Design A simple mixed design was used. The between-subjects factor was bias, either literal or figurative. The within-subjects factor was the 20 proverbs.
Comprehension of Proverbs
53
The correct response was the top context on 10 trials and the bottom context on 10 trials. Two pseudorandom orders of top-bottom were used such that the correct response could appear in the top or bottom position no more than twice in a row. This was done in order to disrupt participants' attempts to look for patterns, something that pilot work indicated they were very likely to do. Also, two random orders of families within a condition were used. These four combinations of orders were approximately balanced across participants. Results and Discussion Correct Responses
The percentage of correct responses in the literal condition (M = 92.50) did not differ from that in the figurative condition (M = 91.25), t(30) < 1, p > .05. This high level of accuracy in the RT portion of the study is consistent with performance on the judgment task. Reaction Time
Performance
A one between-subjects (bias: literal vs. figurative), one within-subjects (proverb) analysis of variance was computed on the median RTs for all responses. In keeping with Clark's (1973) recommendation, proverb was treated as a random factor.3 The analysis revealed that RTs were longer in the figurative bias condition (M — 3.68 s, SD = 2.14) than in the literal bias condition (M = 2.33 s, SD = 1.29), F(l, 39) = 9.64, p < .01. In absolute terms, in 19 cases the mean figurative RT was longer than the mean literal RT, while in one case it was equal. There was also a significant effect due to proverb, ,F(19, 570) = 4.75, p < .01, and to the interaction between proverb and bias condition, F(19, 570) = 4.31, p < .01. The above overall analysis included both correct and incorrect responses. Incorrect responses were included for two reasons. First, most commonly accepted procedures for compensating for missing data can artificially 3
Following Clark's (1973) recommendation, "quasi-F-ratios" (F) were computed for the bias factor instead of the more common practice of using separate F1 and F2 tests. Results computed using F have the advantage that they will generalize across both materials and subjects. Moreover, F will not be significant unless both F1 and F2 are also significant. In comparing the two approaches, Clark (1973) concluded that "F is probably the safest test to use in most instances" (p. 339). We have computed F only for bias because this factor lacks the error term necessary for computing an F-ratio, a consequence of treating proverb as a random factor. F-ratios were computed for all other factors, as the analysis of variance procedure already provided the correct error terms for these factors.
54
Temple and Honeck
stabilize the data. Second, accuracy was relatively high. Nevertheless, the RTs for the correct responses were collapsed across proverb and analyzed separately to determine if they were consistent with the results for the overall analysis. For correct responses, the mean of the median RTs was longer in the figurative condition (M = 3.34 s, SD = 1.20) than in the literal condition (M = 2.15 s, SD = .54), t(30) = -3.72, p < .01, a result consistent with the overall analysis. For the sake of completeness, the RTs for the incorrect responses were also analyzed. For incorrect responses, the median RT was again longer in the figurative condition (M = 5.83 s, SD = 3.7) than in the literal condition (M = 3.35 s, SD = 2.5), but this difference was not significant, t(22) = -1.95, p > .05. Because in further experiments, the separate analyses of RTs for correct and incorrect responses almost always produced results identical to those for the combined analysis, these analyses will not be reported henceforth. The overall results are consistent with the notion that figurative processing takes longer than literal processing, as predicted by the multistage model. Multiple Regression A stepwise multiple regression was performed with RT as the criterion variable, and with bias condition, ratings of proverb familiarity, ratings of proverb comprehensibility, and ratings of proverb imagery as predictor variables. Predictor variables were retained only if they were significant at alpha = .15. The analysis revealed a significant R2 = .55, F(2, 37) = 22.854, p < .01, adjusted R2 = .53, in which bias condition and proverb familiarity ratings were the only contributing variables. The partial r was .71 for bias condition and — .23 for familiarity rating. Thus, bias condition was the dominant and almost exclusive contributor to RT in this experiment. These results provide converging evidence that figurative processing was the primary contributor to RT, which is consistent with the multistage model. Experiment 2: Method Experiment 2 used a within-subjects design, in which the participants received both literal and figurative biasing contexts across trials. However, this procedure may lead to the formation of a figurative processing set, in which literal RTs would become elevated as participants searched for more possibilities than normally required for the literal condition. Participants Eight men and eight women from the introductory psychology courses at the University of Cincinnati served as participants.
Comprehension of Proverbs
55
Materials The materials were the same as those used in Experiment 1, except that two additional families were constructed for the practice trials. Procedure The procedure was identical to that in Experiment 1, except that participants were given two practice trials twice on each kind of bias condition. Design A completely within-subjects design was used in which all participants were repeated across 10 proverbs in the literal bias condition and 10 different proverbs in the figurative bias condition. Across participants, each proverb appeared equally often in each bias condition. There were two orders of proverbs, two orders of correct choice (top or bottom), and two orders of bias. These eight combinations were approximately balanced across participants. Results and Discussion Correct Responses The percentage of correct responses was 90.00 in the literal condition and 90.65 in the figurative condition, a nonsignificant difference, t(15) < 1, p > .05. This high level of accuracy was consistent with that in Experiment 1 and confirmed the experimenter's labeling of choices as either correct or incorrect. Reaction Time Performance A within-subjects analysis of variance on median RTs for all responses revealed that it took longer to respond in the figurative condition (M = 3.90 s, SD = 1.71) than in the literal condition (M = 3.47 s, SD = 1.61), F(l. 6) = 7.40, p < .05. The effect due to proverb was not significant, F(9, 135) = 1.24, p > .05, nor was the interaction between bias condition and proverb, F(9, 135) < 1 , p > .05. These results are consistent with the notion that figurative processing takes longer than literal processing, as predicted by the multistage model. However, it is clear that the literal RTs in Experiment 2 (M = 3.47 s) were considerably larger than in Experiment 1 (M = 2.33 s). For figurative RTs, there was a much smaller difference between Experiment 2 (M = 3.90 s) and Experiment 1 (M = 3.68 s). This is consistent with the possibility
Temple and Honeck
56
that a figurative processing set may have elevated literal RTs in Experiment 2. Multiple-Regression Analysis A stepwise multiple regression analysis was performed in which RT was the criterion variable and bias condition and the three proverb ratings were predictor variables. Using an alpha = .15 cutoff point, only familiarity rating and bias condition were contributors to an R2 = .34, F(2, 37) = 9.30, p < .01, adjusted R2 = .30. Proverb familiarity contributed more heavily to RT than bias condition as indicated by partial rs of —.50 and .28, respectively. These contributions are opposite to those in Experiment 1. However, these results were not unexpected. If a figurative processing set was operating, then the literal RTs would become inflated as participants attempted to generate more possibilities than were necessary for the literal bias condition, thereby lowering the contribution for bias. Under these pressures, familiarity might contribute more to RT, because familiarity can facilitate the processing of a proverb. This is consistent with the multistage model, which implies that the figurative construal of familiar utterances will take less time, since familiarity facilitates the literal processing of an utterance. Experiment 3: Method
In both Experiments 1 and 2, the contexts remained on the screen when the proverb appeared. Participants were explicitly instructed not to re-read the contexts once the proverb appeared, but to use them simply as visual reminders of their respective positions (on the top or bottom). This manipulation was intended to reduce the memory load of the participants. However, during Experiments 1 and 2, it was observed during practice trials that sometimes participants did in fact reread the contexts, thereby introducing an undesired source of variance into the results. To tackle this problem, Experiment 1 was replicated with an important difference: The contexts were removed from the screen when the proverb appeared, thereby making it impossible for the contexts to be reread. Participants Twelve men and 12 women from the introductory psychology courses at the University of Cincinnati served as participants. Materials The materials were the same as those used in Experiment 1.
Comprehension of Proverbs
57
Procedure The procedure was identical to that in Experiment 1 except that, 3 s after the get ready signal, the contexts were completely removed from the screen, and the proverb was presented by itself. Design As in Experiment 1, a simple mixed design was used. The betweensubjects factor was figurative bias, either literal or figurative. The withinsubjects factor was the 20 proverbs. The same four combinations of orders used in Experiment 1 were approximately balanced across participants. Results and Discussion Correct Responses The percentage of correct responses in the literal condition (M = 95.0) did not differ from that in the figurative condition (M = 92.1), t(22) = -1.54, p > .05. This high level of accuracy in the RT portion of the study is consistent with performance on the judgment task and with that in Experiments 1 and 2. Reaction Time Performance A one between-subjects (bias: literal vs. figurative), one within-subjects (proverb) analysis of variance was computed on median RTs for all responses. The analysis revealed that RTs were longer in the figurative bias condition (M = 4.20 s, SD = 1.96) than in the literal bias condition (M = 2.47 s, SD = 1.27), F(l, 27) = 17.32, p < .01. In absolute terms, in 19 cases the mean figurative RT was longer than the mean literal RT, while in one case it was equal. There was also a significant effect due to proverb, F(19, 418) = 7.15, p < .01, and to the interaction between proverb and bias condition, F(19, 418) = 2.72, p < .01. As in Experiments 1 and 2, these results are once again consistent with the notion that figurative processing takes longer than literal processing, as predicted by the multistage model. Multiple Regression A stepwise multiple regression was performed with RT as the criterion variable, and with bias condition, ratings of proverb familiarity, ratings of proverb comprehensibility, and ratings of proverb imagery as predictor variables. Predictor variables were retained only if they were significant at alpha
58
Temple and Honeck
= .15. The analysis revealed a significant R2 = .69, F(2, 37) = 40.65, p < .01, adjusted R2 = .67, in which bias condition and proverb familiarity ratings were the only contributing variables. The partial r was .75 for bias condition and —.35 for familiarity rating. As in Experiment 1, and unlike Experiment 2, bias condition was the dominant and almost exclusive contributor to RT in this experiment, providing converging evidence that figurative processing was the primary contributor to RT, a result consistent with the multistage model. Discussion of Experiments 1, 2, and 3 Experiments 1 to 3 provide strong support for the multistage model. In each experiment, the reaction times for the figurative condition were significantly longer than for the literal condition. Moreover, in absolute terms, a consistent pattern of larger RTs in the figurative condition for 19 out of 20 proverbs, and no difference between conditions for one proverb, directly contradicting Kemper's (1981) finding of longer literal RTs in every case. This pattern of longer figurative RTs, including virtually every proverb in the study, held for both familiar and unfamiliar proverbs. Although stepwise regressions indicated proverb familiarity was an important contributor to RT, as predicted by the multistage model, processing type (literal vs. figurative) was the most important contributor to RT in the two experiments utilizing between-subjects designs. These results also provide a long-needed demonstration of figurative processing sets operating in within-subjects designs. Experiment 2 used a within-subjects design. As expected, there was a smaller difference in RTs between literal and figurative conditions for Experiment 2, a difference attributable to its larger literal RTs (M = 3.47 s), versus the smaller literal RTs found in both Experiment 1 (M = 2.33 s) and Experiment 3 (M = 2.47 s), which used between-subjects designs. This finding of inflated literal RTs in a within-subjects design is consistent with the notion that within-subjects designs can engender a figurative processing set, since other than its withinsubjects design, Experiment 2 was identical to Experiment 1. Additional evidence for a figurative processing set is suggested by the stepwise regressions for Experiment 2, in which the respective contributions of processing type and familiarity to RT were reversed relative to the two experiments using between-subjects designs. In a situation where a figurative processing set is operating, there is a confounding in processing modes that may prime participants to interpret both literal and figurative inputs in a figurative fashion, even when it makes no sense to do so. Consequently, a participant might rely more on proverb familiarity since this facilitates the (literal) processing of a proverb.
Comprehension of Proverbs
59
EXPERIMENTS 4, 5, AND 6 If the literal materials were easier to understand and process than the figurative materials, irrespective of the biasing condition, this could cast doubt on the conclusions for Experiments 1 to 3. Of course, ratings data indicated that the literal contexts were not easier to understand than the figurative contexts. However, it is possible that the use of the contexts in the RT task aroused either different degrees of understanding or some other processing factor that could differentially affect RT. We therefore decided to use a variation of the original RT task to get at this possibility. For the next three experiments, this involved the presentation of new proverbs along with the old literal contexts. The RTs obtained were then compared with those obtained for the old figurative contexts, which were used either in conjunction with their old paired proverb (Experiment 4) or with the new proverbs (Experiments 5 and 6) The new proverbs were designed to arouse figurative meanings that were very similar to those aroused by the old proverbs. For example, the proverb Bees have honey in their mouths and stingers in their tails has a similar meaning to the new proverb The friendliest cats can have the sharpest claws. However, while the bees proverb is literally related to the following context, the cats proverb is figuratively related to it: "Her dad told her that even though bees make honey, she should stay away from the hive near the fence. But the girl was curious and wandered in that direction anyway." The old irrelevant literal contexts remained irrelevant, but irrelevant in a nonliteral way. Indeed, all contexts were now related in a nonliteral way to the proverbs. Note that this methodology involved testing the null hypothesis. That is, if the old literal contexts, when figuratively biased, did not produce different RTs than the old figurative contexts, then it could not be claimed that the literal contexts were easier to process. Nor for that matter could this claim be made if the literal contexts produced slower RTs. Either result would be consistent with the notion that the literal contexts were not easier to process than the figurative contexts, and therefore, simply for this reason, could not have produced the faster literal RTs in the first experiments. Experiment 4: Method For this experiment, the old literal contexts were paired with new proverbs whose figurative meanings were highly similar to those of the old proverbs. The old figurative contexts were presented along with their old proverb mates.
60
Temple and Honeck
Participants Thirty-four students in the introductory psychology courses at the University of Cincinnati served as participants. Nine men and eight women were in the new figurative condition and 15 men and 2 women in the old figurative condition. Materials The same families were used as those in Experiment 1, except that, in the new figurative condition, new proverbs were used. These new proverbs were written to be very similar in figurative meaning to their mates in Experiment 1, To verify this, an independent group of 25 participants used a 9-point scale to rate the paired proverbs for similarity of figurative meaning (9 = high similarity). The mean rating was 7.18 with a range from 6.08 to 8.12. The materials for the old relevant and old irrelevant figurative contexts were the same as in Experiment 1, using the original proverbs. An independent group of 40 participants used a 5-point scale to rate the familiarity (5 = high familiarity) of the literal form of the new proverbs. The mean rating was 2.52 with a range of 1.60 to 4.12. An independent group of 40 participants used a 5-point scale to rate how easily the new proverbs were comprehended (5 = very easy to comprehend). The mean rating was 3.83 with a range of 3.22 to 4.22. In addition, an independent group of 40 participants used a 5-point scale to rate the imagery aroused by the literal form of the new proverbs (5 = high imagery). The mean rating was 3.25 with a range of 2.53 to 3.98. Procedure In all respects, the procedure was identical to that used in Experiment 1. Of course, since there were only figurative relationships between the contexts and the proverbs, only the instructions for this condition were used. Design A mixed design was used, with figurative condition, new figurative versus old figurative, as the between-subjects factor and proverb as the within-subjects factor. As in Experiment 1, the correct context appeared on top half of the time and on the bottom the other half, with two random orders of proverb families and two pseudorandom orders of top-bottom correctness. These four combinations were approximately balanced across participants.
Comprehension of Proverbs
61
Results and Discussion Correct Responses The percentage of correct responses was 87.94 in the new figurative condition and 90.60 in the old figurative condition, a nonsignificant difference, t(32) = 1.08, p > .05. The high accuracy level corroborated the designation of the contexts as either relevant (correct) or irrelevant (incorrect) Reaction Time Performance A one between-subjects (bias: new figurative vs. old figurative), one within-subjects (proverb) analysis of variance was computed on median RTs for all responses. This analysis indicated that there was no difference between the new (M = 4.63 s, SD = 2.81) and the old (M = 4.34 s, SD = 2.42) figurative conditions, F'(l, 39) < 1, p > .05. Also, there was a significant effect due to proverb, F (19, 608) = 4.38, p < .01, and to the interaction between bias condition and proverb, F(19, 608) = 2.82, p < .01. Thus far, these results are consistent with the notion that the two contexts were equally difficult to understand. Item Analysis Recall that semantically similar proverb pairs were used, with one proverb used in the old figurative condition and its mate in the new figurative condition. It was of some interest therefore to determine whether the proverb pairs produced similar data. To test this, a simple effects procedure was used in which cell-unique error terms were computed. This analysis indicated no difference in RTs in 15 cases, all Fs(1, 32) < 2.44,p > .05, and a difference in five cases, all Fs(1, 32) > 4.38, p < .05, and in only two of these significant cases were RTs for the new figurative condition less than for the old figurative condition. These results are consistent with the notion that the two contexts were roughly of equal comprehensibility. Experiment 5: Method One problem with Experiment 4 was that the new and old figurative contexts were paired with different proverbs. So, although the two proverbs were figuratively similar, the literal information in the proverbs was quite different. The present experiment used the same (new) proverb to figuratively bias the two contexts, therefore permitting a cleaner comparison between them.
Temple and Honeck
62
Participants Twenty-four men and 24 women from the introductory psychology courses at the University of Cincinnati served as participants. Materials The materials were the same as those used in Experiment 4, except that the new proverbs were used in both the new figurative and old figurative conditions. Procedure The procedure was identical to that in Experiment 4. Design A mixed design was used. Results and Discussion Correct Responses The percentage of correct responses was 88.54 in the new figurative condition and 87.50 in the old figurative condition, a nonsignificant difference, t(46) < 1,p > .05. Reaction Time Performance A one between-subjects (bias: new figurative vs. old figurative), one within-subjects (proverb) analysis of variance was computed on median RTs for all responses. This analysis indicated that there was no difference between the old (M = 4.69 s, SD = 2.35) and the new (M = 3.98 s, SD = 1.96) figurative conditions, F(1, 58) = 2.78, p > .05. Also, there was a significant effect due to proverb, F(19, 874) = 7.37, p < .01, and to the interaction between bias condition and proverb, F(19, 874) = 4.67, p < .01. Consistent with Experiment 4, these results support the notion that the two contexts were equally difficult to understand. Experiment 6: Method As for Experiment 3, there was some concern about whether the presence of the contexts on the screen after the proverb was presented might introduce an unwanted source of variance, and thereby obscure the "true" nature of the materials. The present experiment therefore replicated Exper-
Comprehension of Proverbs
63
iment 5, but using the context removal methodology developed in Experiment 3. Participants Twenty-eight men and 28 women from the introductory psychology courses at the University of Cincinnati served as participants. Materials The materials were the same as those used in Experiment 5, using the new proverbs for both new figurative and old figurative conditions. Procedure The procedure was identical to that in Experiment 4, except that, 3 s following the get ready signal, the contexts were completely removed from the screen, and the proverb was presented by itself. Design A mixed design was used. Results and Discussion Correct Responses The percentage of correct responses was 88.40 in the new figurative condition and 87.14 in the old figurative condition, a nonsignificant difference, t(54) < 1, p > .05. This high accuracy level is consistent with that in the other experiments. Reaction Time Performance A one between-subjects (bias: new figurative vs. old figurative), one within-subjects (proverb) analysis of variance was computed on median RTs for all responses. This analysis indicated that RTs were longer in the new figurative condition (M = 4.39 s, SD = 2.12) than in the old figurative condition (M = 3.62 s, SD = 1.68), F(1, 65) = 4.76, p < .05. Also, there was a significant effect due to proverb, F(19, 1026) = 8.56, p < .01, and to the interaction between bias condition and proverb, F(19, 1026) = 5.50, p < .01. These results suggest that the literal contexts used in Experiments 1 to 3 were somehow more difficult to understand than the figurative contexts used in those experiments. Therefore, it is clear that faster literal RTs in
64
Temple and Honeck
Experiments 1 to 3 were not the consequence of the literal contexts being intrinsically easier to understand in the RT task situation. Discussion of Experiments 4, 5, and 6 Experiments 4 to 6 strongly suggest that the results of Experiments 1 to 3 were due to the type of processing manipulated, not to inherent biases in the materials. In both Experiments 4 and 5, there was a failure to reject the null hypothesis of no difference between the two conditions, permitting a provisional, if imperfect, acceptance of the null hypothesis, since it was tested under conditions of unbiased replication and adequate power (Cook, Gruder, Hennigan, & Flay, 1979; Greenwald, 1975). Moreover, the difficulty-of-understanding ratings for the contexts provide converging evidence that the two types of contexts were equally comprehensible when judged sans proverbs in a non-RT situation. On the other hand, in Experiment 6 the RTs for the new figurative condition were longer than for the old figurative condition, suggesting that the literal contexts in Experiments 1 to 3 were, if anything, harder to comprehend and process than the figurative contexts. Therefore, it is possible that Experiments 1 to 3 may have been biased against finding faster literal RTs and therefore against the multistage model. These several results effectively counter any suggestion that the findings of slower figurative RTs in Experiments 1 to 3 were driven by the greater difficulty of understanding the figurative contexts, as opposed to the mapping operations required to connect the proverb with the figurative contexts.
GENERAL DISCUSSION In sum, contrary to Kemper (1981), the results strongly indicate that proverbs take longer to process figuratively than literally. This was demonstrated with a between-subjects design, a within-subjects design, and when memorial access to the proverbs' contexts was controlled. Two additional experiments indicated that the literal and figurative sets of materials were roughly equal in difficulty, a finding consistent with the difficulty-of-understanding ratings. Another experiment showed that the literal materials may actually have been harder to understand than the figurative materials, in which case the literal RT data of Experiments 1 to 3 may have been unduly inflated. In addition, the two literal contexts were equated for lexical overlap with the proverb, thereby precluding the participants from using a simple word-matching strategy in the literal condition. In general, the numerous ratings and experimental checks on the materials indicate that the results cannot easily be attributed to some confounding in these materials.
Comprehension of Proverbs
65
Stepwise regressions indicated that, when a between-subjects design was used, processing mode was the most important contributor to RT. That is, processing a proverb figuratively or literally contributed more to RT than simply whether the proverb was familiar. This is consistent with the view that the figurative meanings were computed, not simply retrieved from memory. Familiarity was the second most important contributor to RT. For the within-subjects design, the primary contributor to RT was familiarity, followed by processing mode as the next most important. Note here that a range of proverb familiarity values was used. Our regression results are consistent with experimental evidence that familiarity can speed proverb processing (Case, 1991; Turner & Katz, 1990). In none of our three experiments, however, did the imagery or comprehensibility of the proverb uniquely contribute to RT. Our slower figurative RT results are consistent with those of several prior studies on proverbs. Research that has been conducted on the conceptual base theory of proverb comprehension has used mainly unfamiliar proverbs. In this case, figurative meanings could not have been prestored. Yet, using judgment and memory tasks, a large number of studies have revealed that people are able to reliably connect proverbs with verbal and visual interpretations, and with verbal and visual examples. Even young children can connect totally novel proverbs with visual examples of the figurative meanings of the proverb (for review of this body of work, see Honeck, 1997; Honeck & Temple, 1994; Honeck et al., 1980). Moreover, Nippold and Haq (1996) found that children in Grades 8 and 11 could better select an interpretation for concrete than abstract proverbs. The authors contended that this reflected an ability to decode the literal meanings of the proverb's words, rather than the retrieval of a prestored meaning. In general, it is hard to imagine how people could make these connections without having first developed literal meanings for the proverbs. Turner and Katz (1992a, 1992b) provided additional evidence for this premise. First, they required participants to read proverbs in either literally or figuratively biasing contexts. Contrary to Kemper's (1981) findings, they found that literalized proverbs were read more quickly. They also found that familiar proverbs were read faster than unfamiliar proverbs, replicating Case (1991) and Turner and Katz (1990). Similarly, they found that proverbs that were presented in literalizing contexts were recalled better when prompted by literal than figurative cues, while proverbs set in figurative contexts were prompted equally well by literal and figurative cues. Clearly, literal meaning is playing an important role in these data. In general, the overall pattern of results with proverbs does not support an expectancy-based, radical contextualist, direct-access model. Instead, they support a more modest contextualist view in which the literal meanings of
66
Temple and Honeck
proverbs must first be developed to help construct a figurative meaning which can, in turn, be used for mapping purposes. Context is important in this scenario, but it is not the sole factor. This is the position entailed by the conceptual base theory of proverb comprehension (Honeck & Temple, 1994; Honeck, et al., 1980), which incorporates the basic tenets of the multistage model. In this theory, literal meaning is considered a sort of conventional meaning, based on the default values of the words and their syntactic combination, activated background knowledge (inferences), and any phrasal constituents retrieved from the lexicon with their own default (conventional) usages (Dascal, 1989). The resulting literal meaning is a miniature mental model. For example, for the proverb, Bees have honey in their mouths and stingers in their tails, a hearer must construct a mental model about bees and things that bees do. Familiarity can play a vital role, since it represents prior cognitive effort in deriving a literal meaning, thereby reducing the overall amount of online processing required later on. Consequently, the literal meaning can be quite complex, but past cognitive work can facilitate future comprehension. If this literal model maps satisfactorily onto the context, then processing stops, and the hearer concludes that the literal meaning is in fact the intended meaning. However, if the literal meaning does not map well onto the context, then the hearer must use the literal meaning and the context to compute an alternative meaning. The hearer must generate inferences that can tie the various elements of the discourse context together in order to decipher the speaker's intended meaning. This process can involve, for instance, an analogizing process by which the hearer attempts to map referents in the context to those in the figurative utterance. If the figurative meaning is fairly familiar, this may greatly facilitate, but not eliminate, the mapping process. Although computing a nonliteral meaning can be quite difficult, it is not necessarily harder than computing the literal meaning, since literal meanings can be complex and figurative meanings can be simple. Rather, figurative meanings take longer to construct primarily because literal meaning must be constructed first, thus reflecting an additional stage of processing. Overall, perhaps the best way to describe the figurative comprehension of a proverb is that it involves problem solving, entailing understanding and integration of the proverb topic, discourse context, figurative meaning, and speakers' pragmatic points. More automatic processing, such as lexical computation, is also involved. Perhaps the most important contribution of this study is methodological. Whether the multistage model is confirmed or not depends crucially on the task, dependent measure, experimental design and materials used. We have detailed the difference in these respects between Kemper's (1981) study and our own. There is additional evidence for considering these factors, however.
Comprehension of Proverbs
67
For example, Honeck et al. (1990) found that literalized proverbs were more rapidly read than figuratively biased proverbs, but in subsequent pilot work, they also found no difference in reading latencies. For metaphor processing, Ortony et al. (1978) found similar null results using reading-time as the dependent variable. But when eye regressions (Inhoff et al., 1984) or a more sensitive phrase-by-phrase reading time measure (Janus & Bever, 1985) were used, evidence consistent with the multistage model was found for metaphor processing. Clearly, the global reading-time measure suffers from insensitivity to online processing mode and complexity. In contrast, for proverb processing at least, the choice RT measure in conjunction with the mapping task is more sensitive to these factors. This study also implicates the important role of experimental design. Experiment 2 indicated that within-subjects designs foster a figurative processing set that can inflate literal RTs. Thus, these designs can reduce the sensitivity of the dependent variable and lead to Type II errors. Inasmuch as within-subject designs are common in the area, this is an important consideration. Finally, two potential issues can be raised regarding the present study. First, like Kemper's (1981) study, this study has not directly shown that literal meaning was in fact used to construct nonliteral meaning. Rather, it was shown that it takes less time to understand that a proverb, such as The used key is always bright, is about keys, brightness, and a general key-using schema, than about frequently used instruments retaining their functional value. There is accumulating evidence that such literal meanings are used in processing unfamiliar metaphors (e.g., Blasko & Conine, 1993) and even idioms (e.g., Cacciari & Glucksberg, 1991; McGlone, Glucksberg, & Cacciari, 1994). In any event, it remains a future task to demonstrate the online contribution of literal proverb meanings to figurative ones. Another issue is whether the multistage model applies only to proverbs, and not to other forms of figurative language. Though this argument has superficial plausibility, it is too soon to be certain. However, given the accumulating evidence that literal meanings may be important in all forms of figurative language, and the fact that, generally speaking, highly conventional, relatively uncomplicated materials have been used in research on figurative language, this model or some version of it may be tenable. Of course, the other factors that we have emphasized will also play a role, that is, dependent measure, task, and type of design. In our view, it is unlikely that once all these factors are taken into account that proverbs will turn out to be special vis-a-vis the multistage model. REFERENCES Andreason, N. (1977). Reliability and validity of proverb interpretation to assess mental status. Comprehensive Psychiatry, 18, 465—472.
68
Temple and Honeck
Blasko, D. G., & Conine, C. M. (1993). Effects of familiarity and aptness on metaphor processing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 295-308. Briggs, C. L. (1985). The pragmatics of proverb performances in New Mexican Spanish. American Anthropologist, 87, 793—810. Cacciari, C., & Glucksberg, S. (1991). Understanding idiomatic expressions: The contribution of word meanings. In G. B. Simpson (Ed.), Understanding word and sentence. Amsterdam, The Netherlands: North-Holland Press. Case, T. (1991). The role of literal and figurative familiarity in proverb comprehension. Unpublished doctoral dissertation, University of Cincinnati. Clark, H. H. (1973). The language-as-fixed-effect fallacy: A critique of language statistics in psychological research. Journal of Verbal Learning and Verbal Behavior, 12, 335-359. Clark, H. H. (1979). Responding to indirect speech acts. Cognitive Psychology, 11, 430477. Clark, H. H., & Lucy, P. (1975). Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 14, 56-72. Cook, T. D., Gruder, C. L., Hennigan, K. M., & Flay, B. R. (1979). History of the sleeper effect: Some logical pitfalls in accepting the null hypothesis. Psychological Bulletin, 86, 662-679. Dascal, M. (1987). Defending literal meaning. Cognitive Science, 13, 259-281. Dascal, M. (1989). On the roles of context and literal meaning in understanding. Cognitive Science, 13, 253-258. Estill, R. B., & Kemper, S. (1982). Interpreting idioms. Journal of Psycholinguistic Research, 11, 559-568. Finnegan, R. (1970). Oral literature in Africa. Oxford, England: Clarendon Press. Gibbs, R. W. (1979). Contextual effects in understanding indirect requests. Discourse Processes, 2, 1-10. Gibbs, R. W. (1980). Spilling the beans on understanding and memory for idioms in conversation. Memory & Cognition, 8, 449-456. Gibbs, R. W. (1983). Do people always process the literal meanings of indirect requests? Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 524— 533. Gibbs, R. W. (1986). Skating on thin ice: Literal meaning and understanding idioms in conversation. Discourse Processes, 9, 17—30. Gibbs, R. W. (1994). The poetics of mind: Figurative thought, language, and understanding. New York: Cambridge University Press. Gibbs, R. W. (1995). What proverb understanding reveals about how people think. Psychological Bulletin, 118, 133-154. Gibbs, R. W., Nayak, N., & Cutting, C. (1989). How to kick the bucket and not decompose: Analyzability and idiom processing. Journal of Memory and Language, 28, 576-593. Glass, A. L. (1983). The comprehension of idioms. Journal of Psycholinguistic Research, 12, 429-442. Gokhan, A. G. (1992). What have the ancestors said: An ethnography of speaking proverbs in a Turkish community (Izmir). Unpublished doctoral dissertation, University of Pittsburgh. Gorham, D. R. (1963). Additional norms and scoring suggestions for the Proverbs Test. Psychological Reports, 13, 487-492.
Comprehension of Proverbs
69
Greenwald, A. G. (1975). Consequences of prejudice against the null hypothesis. Psychological Bulletin, 82, 1-20. Grice, H. P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and semantics 3: Speech acts (pp. 41-58). New York: Academic Press. Honeck, R. P. (1997). A proverb in mind: The cognitive science of proverbial wit and wisdom. Mahwah, NJ: Erlbaum. Honeck, R. P., Case, T. S., Temple, J. G., & Firment, M. (1990, April). Nonliteral processing, proverbs, and the multi-stage model. Paper presented at the eighty-second meeting of the Southern Society for Philosophy and Psychology, Louisville, KY. Honeck, R. P., Kibler, C., & Sugar, J. (1985). The conceptual base view of categorization. Journal of Psycholinguistic Research, 14, 155—174. Honeck, R. P., & Temple, J. G. (1994). Proverbs: The extended conceptual base and the great chain metaphor theories. Metaphor and Symbolic Activity, 9, 85-112. Honeck, R. P., Voegtle, K., Dorfmueller, M. A., & Hoffman, R. R. (1980). Proverbs, meaning, and group structure. In R. P. Honeck & R. R. Hoffman (Eds.), Cognition and figurative language (pp. 127-161). Hillsdale, NJ: Erlbaum. Inhoff, A. L., Lima, S. D., & Carroll, P. J. (1984). Contextual effects on metaphor comprehension in reading. Memory & Cognition, 12, 558-567. Janus, R. A., & Bever, T. G. (1985). Processing of metaphoric language: An investigation of the three-stage model of metaphor comprehension. Journal of Psycholinguistic Research, 14, 473-487. Kemper, S. (1981). Comprehension and the interpretation of proverbs. Journal of Psycholinguistic Research, 10, 179-198. Matarazzo, J. (1972). Wechsler's measurement and appraisal of adult intelligence (5th ed.). Baltimore: Williams & Wilkins. McGlone, M. S., Glucksberg, S., & Cacciari, C. (1994). Semantic productivity and idiom comprehension. Discourse Processes, 17, 167-190. Monye, A. K. (1996). Proverbs in African orature: The Aniocha-Igbo experience. New York: University Press of America. Nippold, M. A., & Haq, F. S. (1996). Proverb comprehension in youth: The role of concreteness and familiarity. Journal of Speech and Hearing Research, 39, 166176. Ortony, A., Schallert, I., Reynolds, R., & Antos, S. (1978). Interpreting metaphors and idioms: Some effects of context on comprehension. Journal of Verbal Learning and Verbal Behavior, 17, 465-477. Penfield, J., & Duru, M. (1988). Proverbs: Metaphors that teach. Anthropological Quarterly, 61, 119-128. Pollio, H. R., Fabrizi, M. S., Sills, A., & Smith, M. K. (1984). Need metaphoric comprehension take longer than literal comprehension? Journal of Psycholinguistic Research, 13, 195-214. Schweigert, W. A., & Moates, D. R. (1988). Familiar idiom comprehension. Journal of Psycholinguistic Research, 17, 281-296. Searle, J. (1979). Metaphor. In A. Ortony (Ed.), Metaphor and thought (pp. 92-123). Cambridge, England: Cambridge University Press. Temple, J. G., & Honeck, R. P. (1991, March). The standard pragmatic model is still alive. Paper presented at the eighty-third meeting of the Southern Society for Philosophy and Psychology, Atlanta. Turner, N. E., & Katz, A. N. (1990, November). Context effects in comprehending familiar and unfamiliar proverbs. Paper presented at the thirty-first annual meeting of the Psychonomic society, New Orleans, Louisiana.
70
Temple and Honeck
Turner, N. E., & Katz, A. N. (1992a, June). Contextual appropriate and inappropriate cues for proverbs. Paper presented at the annual meeting of the Canadian Psychological Association, Quebec City, Quebec, Canada. Turner, N. E., & Katz, A. N. (1992b, June). The processing of proverbs: Evidence for the primacy of literal meaning. Paper presented at the annual meeting of the Canadian Society of Brain Behavior and Cognitive Science, Laval, Quebec, Canada.