DISCOURSE PROCESSES, 36(1), 19–46 Copyright © 2003, Lawrence Erlbaum Associates, Inc.
Reading Proverbs in Context: The Role of Explicit Markers Albert N. Katz Department of Psychology University of Western Ontario
Todd R. Ferretti Department of Psychology Wilfrid Laurier University
We examined how explicit markers or “introductory formulae,” which are used for signaling that statements should be interpreted literally or nonliterally, influence the online processing of proverbs. Familiar or unfamiliar proverbial statements were presented in contexts that were biased toward either their literal and nonliteral meanings, and were always preceded immediately by either proverbially speaking, in a manner of speaking, literally speaking, or no marker. The main hypothesis was that the markers, in combination with the contexts, should act as strong constraints on whether people interpret the statements literally or nonliterally. The results demonstrated that each form of marker had a unique effect on the reading of different regions of the proverbial statements, and that they have a stronger influence in reducing ambiguity associated with the meanings of unfamiliar proverbs than with familiar proverbs. This research is the first to systematically examine the role of explicit markers in the moment-by-moment processing of nonliteral language. Results are discussed in relation to existent models of nonliteral language comprehension.
In spoken and written language, expressed meaning (what is said) is often at variance with the meaning intended by the speaker. Examples include metaphor (e.g., “that car is a lemon”), irony (“you sure are my best friend”), idioms (“Jim kicked the bucket last night”) and proverbial language (“the grass is greener on the other side of the fence”). In each case, the intended meaning actually might be the accepted literal sense of each of those statements, such as occurs when a given car is Correspondence and requests for reprints should be sent to Albert N. Katz, Department of Psychology, University of Western, Ontario, London, Ontario, Canada, N6A 5C2. E-mail:
[email protected]
20
KATZ AND FERRETTI
decked out as a fruit in a parade, when someone is describing a good friend, when Jim kicked an object, or when someone was commenting on their neighbors’ wellfertilized lawn. However, in each case the intended meaning might instead be the familiar nonliteral sense, as when one is describing a car that is often in for repairs or someone is describing the fact that Jim had died the night before. Virtually all processing models posit that context plays an important role in recovering intended meaning from expressed meaning. However, the models differ in the role that context plays. At one end of the theoretical spectrum are models that assume an initial obligatory processing of the expressed meaning. In the most traditional models of this type, the literal sense of the expressed meaning is extracted and only if that fails to make sense in context does a later, optional phase come into play in which one attempts to infer a sense that is context compatible (see Glucksberg, 2001 for a review). A more recent variant of this general class of models posits that one is obligated to initially process the salient sense (and not necessarily the literal sense) of the expressed meaning. That is, literal and salient meanings are considered orthogonal to one another, and in many cases the most familiar (or salient usage) for an expression might not be the literal use but a nonliteral use. If the salient sense is the nonliteral usage (as would be the case with the examples given previously), the nonliteral meaning would have processing priority (see Giora, 1999, 2003). Thus, when faced with a proverb, such as it never rains but it pours, a literal-first model would argue that in the comprehension process one initially attempts to understand this sentence as dealing with rain, whereas the graded saliency position would argue that this proverb is so familiar that, regardless of context, the nonliteral meaning (“when bad things happen, they happen in droves”) is automatically activated. Thus, both of these types of models posit that, regardless of context, at the initial moments of processing, literal or salient meaning is aroused. In contrast to such obligatory-first models are direct-access models in which context can bias processing so that only the context-appropriate interpretations are drawn; a phenomenon true for literal and nonliteral language alike. That is, with a sufficiently rich, elaborated, and constrained context, the meaning accessed will be that appropriate to the context. As extended to the literal–nonliteral language distinction, this theoretical approach holds that, at the initial moments of processing, the important distinction is not whether literal or salient meaning is accessed, but rather that the context-appropriate sense is accessed directly during comprehension (for a review of this position see Gibbs, 1994; Giora, 2003). Recent constraint–satisfaction models of language processing (e.g., McRae, Spivey-Knowlton, & Tanenhaus, 1998) would be a variant of this general approach, although this mechanism has seldom been suggested for issues of nonliteral language (but see Katz & Ferretti, 2001). The logic of constraint satisfaction is that the understanding of text involves constructing an interpretation that fits the available information better than it fits with alternative interpretations. As
LINGUISTIC MARKERS IN PROVERBS
21
implemented, constraint-satisfaction models assume that different sources of information (i.e., constraints) provide immediate probabilistic support for competing interpretations in parallel over time (in this case, literal and nonliteral interpretations of proverbs). As such, these models do not posit modular obligatory processing priorities, as suggested in the nonliteral language processing literature by obligatory-first models. In general, competition duration (and thus reading time) is itself a function of the strength of the various alternatives: If the constraints all point to the same interpretation, competition is resolved rapidly, whereas settling on an interpretation is delayed as support for different alternatives becomes more equal. Katz and Ferretti (2001) suggested that constraint satisfaction might prove to be a useful way of resolving some of the issues in the time taken to comprehend a statement (such as a proverb) used either literally or in its nonliteral proverbial sense. In the following research, we examine the issue of how strong contextual constraints influence the comprehensions of proverbs. Proverbs provide a unique opportunity to investigate the role of context for several reasons. First, the theoretical contrast between obligatory-first and direct-access models have their counterparts in the proverb processing literature with some (Honeck, 1997) arguing for a variant of literal-first processing (also see Turner & Katz, 1997) and some arguing for directaccess (Gibbs, 1995). Second, proverbs are an ideal linguistic form with which to address the theoretical contrasts because proverbs are based on convention and hence, for familiar proverbs, the nonliteral (proverbial) meaning is the salient one, whereas for unfamiliar or novel proverbs the salient sense is the literal meaning. Because proverbs vary in familiarity, one can readily disentangle the effects of literality from saliency to see which (if either) form of the obligatory-first models better fit the data. Third, because many proverbs are concrete instantiations of a general “truth,” they are also sensible as literally true statements. Thus, for proverbs, unlike some other tropes, one can readily create discourse contexts in which the same proverb is used sensibly as literally true or as proverbially true. For example, when someone states lightning never strikes the same place twice, the literal sense that lightning actually never strikes the same place twice is true, and thus this statement is perfectly sensible when read in a literal context. Alternatively, metaphors such as he broke her heart, or children are precious gems, are not actually true when interpreted literally. People do not literally break hearts, and children are not literally precious gems. As a result, these statements are not as sensible when placed in literally biasing contexts. Thus, one major advantage of studying proverbs is that they permit a strong test in which one can compare the same statement when the context favors the literal or the nonliteral sense of that statement. Despite the obvious advantages of using proverbs to examine the theoretical questions of interest, there is a surprising paucity of studies in which proverbs have been employed, and in the few cases that can be found (e.g., Temple & Honeck, 1999, Turner & Katz, 1997), either offline measures were employed, or,
22
KATZ AND FERRETTI
as has been the case generally for tests of obligatory-first versus direct-access reading studies, researchers have employed an insensitive measure of momentby-moment reading time, usually the time taken to read complete statements (see Gibbs, 1994 for a review of such studies). More recently, however, Katz and Ferretti (2001) investigated whether context plays an immediate role in constraining the meaning of a proverbial statement or whether contextual effects come into play at a later stage of processing, after some initial context-independent processing has occurred. Katz and Ferretti (2001) employed a word-by-word moving window procedure to index the online time course of proverb comprehension. In that research, they crossed whether a given target statement was being used literally or nonliterally (by manipulating the discourse context) and saliency (by choosing target items that were either familiar or unfamiliar proverbs). The results did not consistently support any of the current models of nonliteral language processing. They found no evidence for a literal-first model of nonliteral language comprehension (e.g., as proposed by Grice, 1975). The strongest support for direct-access models was found with familiar proverbs: Regardless of discourse context, familiar proverbs were read equally quickly whether used literally or proverbially. However, in contrast with direct-access predictions, even when we ensured equally supportive discourse contexts, unfamiliar proverbs were, by about the middle words of the proverb, read more slowly than familiar proverbs. Another important theoretical finding was that by about the middle of the target, unfamiliar proverbs were read more slowly when placed in contexts biased toward their nonliteral senses than when placed in contexts biased toward their literal senses. This difference is especially marked at the last word of the target and can still be found at the reading of the first word of the following sentence. In order to account for these findings, Katz and Ferretti (2001) appealed to a constraint-satisfaction theoretical position. As extended to their findings specifically, Katz and Ferretti suggested that among the constraints of importance would be the context itself (the extent to which it supports a literal or a nonliteral reading) and the nature of fixed expressions (the reading of the initial words of a familiar fixed expressions, such as a proverb or idiom or frozen metaphor, might trigger the rest of the expression). Taking just these two constraints into consideration, Katz and Ferretti (2001) could provide an explanation as to why it takes less time in general to settle on a literal over nonliteral interpretation of unfamiliar proverbs and the more rapid reading of familiar over unfamiliar items. Here we extend our previous research on online proverb comprehension by manipulating the strength of contextual constraints present in the discourse. Both Mieder (1982, 1990) and Honeck (1997), in their compendiums of proverb research, indicate that there are various ways in which proverbs are often introduced, what some linguists would label as introductory formulae, or, in other words, explicit markers that are the standard ways of signaling that the receiver will be presented with a proverb. Proverbs are often preceded by explicit markers, because often their
LINGUISTIC MARKERS IN PROVERBS
23
literal and nonliteral meanings are sensible in both literal and nonliteral biasing contexts. Thus, explicit markers are helpful for disambiguating the intended meanings of the speaker for the comprehender. The following research examines how and when these markers are used to disambiguate the alternative meanings of proverbs. To our knowledge, this research is the first to study the moment-by-moment processing of explicit markers in conjunction with nonliteral statements. The goal is to examine a range of markers as there is no precedent for how different markers influence online literal and nonliteral ambiguity resolution. Toward this end, we chose to examine the markers proverbially speaking, in a manner of speaking, and literally speaking because they disambiguate the intended meanings of proverbs and are consistent with the types of markers typically used in introducing proverbial (see Mieder, 1982, 1990) and nonliteral usages (Moon, 1998).
EXPERIMENT 1 In this experiment, passages with nonliteral biasing contexts ended with the marker in a manner of speaking just before the proverb was encountered. The logic of this manipulation was simply that the marker should strengthen the constraints that point towards a nonliteral interpretation, but should not point to the type of nonliteral interpretation (i.e., a proverb) required. That is, we asked whether just knowing that a nonliteral fixed expression is coming can constrain the interpretive process. For instance, one could argue that the finding of Katz and Ferretti (2001) of slower reading of unfamiliar proverbs in nonliteral biasing contexts versus literal biasing contexts is due to an initial failure to realize that the sentence being read is of a nonliteral usage. As such, the marker might draw people’s attention to the nonliteral meaning of the sentences earlier, perhaps leading to faster reading of unfamiliar proverbs used nonliterally rather than literally, especially by the end of the sentences. Note that the marker might not provide the same advantage for familiar proverbs placed in nonliteral contexts because the familiar, nonliteral sense of these expressions is well known to the readers. That is, readers know early during the processing of familiar proverbs (i.e., within the first couple of words) what the nonliteral senses of the proverbs are; therefore, there may not be a reading time advantage for these proverbs when preceded by the nonliteral marker. Nonetheless, because the markers preceded the fixed expressions, it was possible that people would pick up the intended nonliteral senses of the expressions more quickly, perhaps leading to faster reading times in contrast to the same statements presented in literal biasing contexts. In contrast to the nonliteral biasing contexts, proverbs presented in literal biasing contexts were always preceded immediately by the marker literally speaking. If the default is to read initially for literal meaning regardless of contextual bias, then one possibility was that the literally speaking marker would be redundant and
24
KATZ AND FERRETTI
have no effect on reading time. However, similar to nonliteral markers, the fact that the markers preceded the literal statements might enable readers to attend to the literal interpretations more quickly, and thus read the statements more quickly. Note that if this were true, then we might expect a bigger advantage of this marker for familiar proverbs because the marker should increase support for the literal, less salient sense of familiar proverbs. Finally, another possibility was that deviation from the default use might be suggested by the literal markers paired with unfamiliar proverbs and nonliteral markers paired with familiar proverbs, because the statements and markers were consistent. That is, the markers might create ambiguity in the meaning of the statements when they were not expected, and hence readers might slow down their reading to consider subtle alternative senses of the statements and preceding contexts. Method Participants. Fifty people participated for course credit, 25 per list. All participants were native English-speaking psychology undergraduates from the University of Western Ontario and had normal or corrected-to-normal visual acuity. Materials. The items used in this experiment were a subset of those used in Turner and Katz (1997) and Katz and Ferretti (2001), and are described most fully in the former. The items consisted of 12 familiar and 12 unfamiliar proverbs and their corresponding literal and nonliteral biasing contexts (see Appendix for a full list of the familiar and unfamiliar proverbs). Thus there were 24 critical passages that were either biased toward the literal or nonliteral meaning of each proverb. The nonliteral contexts that immediately preceded the target proverb always ended with the proverbial marker in a manner of speaking and the literal biasing contexts always ended with the marker literally speaking. Examples of the experimental stimuli are presented later. The first instance is an example of a familiar proverb preceded by a nonliteral biasing context that included a marker also biased to the nonliteral usage, whereas the second is an example of the same proverb in a literal biasing context with the appropriate literal biasing marker. The markers and target sentences are in italics for exposition here; naturally italics were not present in the materials presented during the experiment. Nonliterally biasing context. “What you need is an investment to shelter your profit,” said Ann. “But it’s been a volatile market since the crash,” replied George. “Look, I lost a lot of money last year.” “Don’t worry, you’ll be alright,” she assured him. “In a manner of speaking, lightning never strikes the same place twice.” “How can you be certain?” he asked. “It’s true, the market goes in cycles; it won’t crash again for years.” “Who told you that?” he asked. “I don’t know, I read it somewhere,” she replied.
LINGUISTIC MARKERS IN PROVERBS
25
Literally biasing context. “Let’s take shelter from the rain under this broken tree,” said Ann. “But it’s dangerous to hide under a tree during a storm,” replied George. “Look, the tree has been hit once already.” “Don’t worry, we’ll be alright,” she assured him. “Literally speaking, lightning never strikes the same place twice.” “How can you be certain?” he asked. “It’s true, once the energy dissipates; it takes a while to rebuild.” “Who told you that?” he asked. “I don’t know, I read it somewhere,” she replied. Because it is important that the contexts be similar in all conditions, the target items were rated in context on a number of dimensions by Turner (1989). The dimensions and the corresponding ratings are listed in Table 1.
TABLE 1 Mean Ratings for Quality Dimensions Familiar Literal Target Type Dimension Familiarity Proverb Paraphrase Ease of comprehension Proverb Paraphrase Appropriateness Proverb Paraphrase Plausibility Proverb Paraphrase Similarity Proverb Paraphrase Preference Proverb Paraphrase Humor Proverb Paraphrase Creativity Proverb Paraphrase
Unfamiliar
Nonliteral
Literal
Nonliteral
M
SD
M
SD
M
SD
M
SD
2.3 2.3
.97 .97
2.2 2.2
.84 .84
3.9 3.9
.97 .97
4.0 4.0
1.1 1.1
2.0 1.7
.83 .79
2.1 1.8
.90 .76
2.8 2.1
.93 .95
3.1 1.9
1.1 .81
3.1 2.3
1.1 .96
2.6 2.4
.61 .93
3.2 2.4
.93 .85
3.3 2.3
.98 .83
2.8 2.5
.77 1.2
2.5 2.5
.78 1.1
2.5 2.7
1.0 1.2
3.7 2.4
1.3 .92
4.2 5.3
1.2 1.1
4.7 5.6
1.3 1.1
5.6 5.6
1.0 1.1
4.7 5.6
1.3 .98
4.2 5.3
1.2 1.1
4.7 5.6
1.3 1.1
5.6 5.6
1.0 1.1
4.7 5.6
1.3 .98
4.2 5.3
1.2 1.1
4.7 5.6
1.3 1.1
5.6 5.6
1.0 1.1
4.7 5.6
1.3 .98
3.3 5.1
1.1 .89
3.2 4.8
1.1 .84
4.2 4.6
.76 1.0
2.6 4.8
1.1 .83
Note. The ratings are based on a 1 to 7 scale. Lower numbers indicate a higher score (e.g., very familiar) and higher numbers indicate a low score (very unfamiliar).
26
KATZ AND FERRETTI
As can be seen, except for the manipulated variable of proverb familiarity, the items were perceived as very comparable to one another on a range of dimensions, including ease of comprehension and plausibility, especially when one compares the ratings given the familiar proverbs in literal and nonliteral contexts and the ratings given the unfamiliar proverbs in literal and nonliteral contexts. Thus, on face, the contexts would appear to be equally supportive of the literal and nonliteral (proverbial) uses of targets and thus equally likely to directly engage the contextappropriate meaning. The 24 proverbs and their 2 contexts were divided into 2 lists. Each list contained 12 familiar and 12 unfamiliar proverbs. Moreover, half of the familiar and unfamiliar proverbs were preceded by literally biasing contexts, and the other half of the proverbs were preceded by nonliteral biasing contexts. Across the lists, each proverb appeared with both types of contexts, and within each list no proverb appeared more than once. In addition, 72 paragraphs were created that served as filler trials. These trials were similar to the experimental trials in length and narrative form. However, none of these paragraphs included a proverb or any other form of nonliteral language. Finally, 10 practice paragraphs were created to give participants a chance to get used to the moving window format. These trials resembled the filler trials and also did not include any proverbial statements. Procedure. The paragraphs were displayed, one at a time, on a 17-inch Apple monitor controlled by a Macintosh G3. They were presented using PsyScope (Cohen, MacWhinney, Flatt, & Provost, 1993) in a one-word-at-a-time self-paced moving window format. Thus, paragraphs were initially presented on the screen with each nonspace character replaced by a dash. Participants pressed a button to reveal the first word of the paragraph. Each subsequent button press revealed the next word and replaced the previous word with dashes. Participants read each paragraph in this manner and then answered a yes–no comprehension question. These questions were simple queries about a fact in the passage and were presented to ensure that our participants were actually reading the passage. Testing sessions began with 10 practice items. Participants then read the remaining 96 experimental trials, taking a break after 40 trials. Participants were instructed to read at a pace that resembled how they would typically read a magazine or newspaper. Each session lasted approximately 40 min. Reading latencies for each word were recorded by the computer and were measured as the time interval between successive button presses. Design. The analyses were conducted on seven critical regions. We examined the word that preceded the target because that word was the last word of the marker and we were interested in examining how the marker influences reading times. The target itself was examined in five regions: the first word, the second word, the second to last word, the last word of the target plus an average reading
LINGUISTIC MARKERS IN PROVERBS
27
time for the middle words. To examine whether there are “spill over” effects after the target was read, indicating a continued difficulty in comprehension, we also assessed the reading time to the first word of the sentence that followed the target. These regions were chosen because past research (e.g., Katz & Ferretti, 2001; Pexman, Ferretti, & Katz, 2000) with nonliteral language has shown that most of the differences in reading time occur at the end of the trope, with some less marked effects at the beginning. This procedure also allowed us to do analyses across proverbs of differing lengths. The main analyses were a series of 2 (context: literal biasing vs. proverb biasing) × 2 (proverb type: familiar or unfamiliar) analyses of variance (ANOVAs). One such analysis was conducted for each of the seven regions discussed above, followed by planned comparisons to examine the difference between familiar and unfamiliar proverbs in the two contexts. Analyses were performed using participants (F1) and items (F2) as random variables. Context was within participants and item factors, whereas familiarity was within participants and between items. Any reading latency that was greater than three standard deviations from the mean was removed from the analysis. Items in which the participants incorrectly answered the comprehension question were removed from the analysis (less than 1% of trials). For all inferential statistics, p < .05 unless otherwise noted. The same 7 regions and analytic strategy were also employed in Experiments 2 and 3. Results and Discussion The mean reading times for familiar and unfamiliar proverbs placed in literal and nonliteral biasing contexts are illustrated in Figure 1. As can be seen there is (a) initially faster reading of both unfamiliar and familiar proverbs when placed in the nonliterally biased context; (b) slower reading throughout of unfamiliar proverbs, compared to familiar proverbs; (c) a large difference in reading times, especially at the final word of the proverbs and spilling over to the next sentence, in which unfamiliar proverbs used nonliterally are read much more slowly than the same items used literally; and (d) familiar proverbs were read similarly regardless of contextual bias. The region-by-region analyses largely confirm these observations. Word before proverb. The two-way interaction between target type and context was significant, F1(1, 49) = 6.20, F2(1, 22) = 5.84. This interaction occurred because people read the last word before the proverb (i.e., the word speaking) at the same rate for familiar and unfamiliar proverbs when the context was figuratively biasing, both Fs < 1. However, people read the last word of literal biasing contexts more quickly for familiar than unfamiliar proverbs, F1(1, 49) = 9.21, F2(1, 22) = 7.56. Planned comparisons also revealed that the word before familiar proverbs was read more quickly when preceded by figurative biasing contexts than by literal
28
KATZ AND FERRETTI
FIGURE 1 Experiment 1 reading times.
biasing contexts, F1(1, 49) = 5.04, F2(1, 22) = 19.82. Moreover, unfamiliar proverbs were read more quickly when preceded by figurative biasing contexts than by literal biasing contexts, F1(1, 49) = 33.26, F2(1, 22) = 61.71. There was a main effect of context in which the word before the proverb was read more quickly when preceded by figurative than by literal contexts, F1(1, 49) = 37.44, F2(1, 22) = 75.05. There was also evidence of an effect of familiarity, in which familiar proverbs were read more quickly than unfamiliar proverbs, but this effect only approached significance in the subject analysis, F1(1, 49) = 3.63, p < .07, F2(1, 22) = 1.78, p > .19. First word of target. The two-way interaction between target type and context, was nonsignificant, both Fs < 1. There is a suggestion that familiar proverbs were read marginally more quickly when preceded by literal biasing contexts than by nonliteral biasing contexts, but this effect was only apparent in the item analysis, F1 < 1, F2(1, 22) = 3.40, p < .08. However, unfamiliar proverbs were read more quickly when preceded by nonliteral biasing contexts than by literal biasing contexts, F1(1, 49) = 4.91, F2(1, 22) = 4.55. There was also a main effect of context in which the first word was read more quickly when preceded by nonliteral
LINGUISTIC MARKERS IN PROVERBS
29
than by literal contexts, F1(1, 49) = 8.72, F2(1, 22) = 7.66. There was no effect of familiarity, F1(1, 49) = 1.72, p > .19, F2 < 1. Second word of target. The two-way interaction between target type and context, was nonsignificant, both Fs < 1. Neither familiar proverbs (both Fs < 1) nor unfamiliar proverbs, F1 < 1, F2(1, 22) = 2.09, p > .16, were read significantly different when preceded by literal biasing contexts and nonliteral biasing contexts. There was no overall effect of contextual bias, both F < 1, but there was a marginal effect of familiarity in the subject analysis, F1(1, 49) = 3.42, p = .07, F2(1, 22) = 1.72, p > .20. That is, there is a hint in the data that unfamiliar proverbs were being read more slowly than familiar proverbs by the second word of the target, in line with the findings of the earlier studies. Middle region of target. The two-way interaction between target type and context was nonsignificant, both Fs < 1. Planned comparisons revealed that familiar proverbs were read at the same rate whether preceded by literal biasing or nonliterally biasing contexts, both Fs < 1. Unfamiliar proverbs were also read equally quickly when preceded by both types of contexts, both Fs < 1. Overall, the first words from proverbs were read at a similar rate in the two contexts, both Fs < 1. However, familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 49) = 24.56, F2(1, 22) = 9.67. Second last word of proverb. The two-way interaction between context and familiarity was, now, marginally significant, F1(1, 49) = 3.63, p < .07, F2 < 1. Planned comparisons revealed that familiar proverbs were read at the same rate in the two types of contexts, both Fs < 1. However, unfamiliar proverbs were read marginally more quickly when they were preceded by literal biasing contexts than by nonliterally biasing contexts, but this effect was only evident in the subject data, F1(1, 39) = 3.19, F2 < 1. The main effect of context was nonsignificant, both Fs < 1. Familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 39) = 29.83, F2(1, 22) = 26.12. Last word of proverb. The two-way interaction between context and familiarity was significant in the subject analysis only, F1(1, 49) = 6.33, F2 < 1. Planned comparisons again revealed that familiar proverbs were read similarly in the two types of contexts, both Fs < 1. Unfamiliar proverbs were read more quickly when they were preceded by literal biasing contexts than by nonliterally biasing contexts, F1(1, 49) = 10.01, F2 < 1. Proverbs presented in literally biasing contexts were read more quickly than when they were preceded by nonliterally biasing contexts, but this effect only reached significance in the subject analysis, F1(1, 49) = 6.30, F2 < 1. Familiar proverbs were still read more quickly than unfamiliar proverbs, F1(1, 49) = 42.15, F2(1, 22) = 21.34.
30
KATZ AND FERRETTI
Word after target. The two-way interaction between context and familiarity was significant in the subject analysis and marginal in the item analysis, F1(1, 49) = 5.86, F2(1, 22) = 2.95, p < .10. Planned comparisons again revealed that familiar proverbs were read similarly in the two types of contexts, F1 < 1, F2(1, 22) = 1.57, p > .22. Unfamiliar proverbs were read more quickly when they were preceded by literal biasing contexts than by nonliterally biasing contexts, but this effect was only significant in the subject analysis, F1(1, 49) = 7.10, F2(1, 22) = 1.34, p > .25. The main effect of context was nonsignificant, F1(1, 49) = 1.93, p > .17, F2 < 1. Familiar proverbs were still read more quickly than unfamiliar proverbs, F1(1, 49) = 9.81, F2(1, 22) = 20.54. Taken together these data are very similar to that found in Katz and Ferretti (2001), when no markers at all were employed, indicating that Katz and Ferretti’s finding of more reading difficulty associated with unfamiliar proverbs placed in literal versus nonliteral contexts cannot simply be attributed to the fact that readers are unaware that a nonliteral interpretation is in order. That is, even when explicitly forewarned that a nonliteral reading is in order our participants had no integrated meaning of the critical sentence into the proceeding context by the end of the sentence and even beyond, into the next sentence. The only detectable influence of the markers was that the nonliteral maker, in a manner of speaking, was clearly easier to integrate in the discourse contexts than the literal marker, literally speaking, leading to faster reading times at the marker and the beginning of the proverbs. Thus, the results were somewhat unexpected because the explicit markers employed had no additional measurable influence on the reading times associated with the end of the proverbs than the previously used contexts by Katz and Ferretti (2001). However, there may be a couple of reasons for why the markers did not influence reading times at the end of the proverbs. First, the use of explicit markers for both the nonliteral and literal biasing contexts may have simply strengthened both contexts relative to each other and, therefore, perhaps it is not surprising that we obtained similar results as in Katz and Ferretti (2001). Second, one could reasonably argue that a stronger test would be to use explicit markers that not only inform the reader that the upcoming text is being used nonliterally, but also provides more specific information about the nonliteral nature of that text. Experiments 2 and 3 addressed these issues in two important ways. First, in Experiment 2 we substituted the phrase proverbially speaking for the phrase in a manner of speaking. Thus, for nonliteral contexts the reader will be informed by the nonliteral marker that the upcoming sentence is being used nonliterally as a proverb. In Experiment 3, we again strengthened the nonliteral contexts by employing the marker proverbially speaking, but did not include a corresponding marker for the literal biasing contexts. Thus, Experiment 3 indexed the strength of the proverbial marker in comparison to a well-established literal baseline (i.e., Katz & Ferretti, 2001). Conversely, comparing the results of Experiments 2 and 3
LINGUISTIC MARKERS IN PROVERBS
31
also provided insight into the strength of the literal marker, because the only difference between the two experiments was the absence of the literal marker.
EXPERIMENT 2 The main finding from Experiment 1 was that an underspecified marker (in a manner of speaking) produced very little effects on the reading times of proverbs in context, relative to the situation in which a literal marker was employed and relative to situations when no markers are given (cf. Katz & Ferretti, 2001). In Experiment 2, we placed the marker proverbially speaking just before a proverb is to be encountered for proverbial biasing contexts. The logic of this manipulation was simply that it should strengthen the constraints that point toward a nonliteral interpretation and, more specifically, should point to the type of nonliteral interpretation (i.e., a proverb) that is upcoming. Moreover, because this experiment also employed the marker literally speaking for literal biasing contexts, we expected that familiar and unfamiliar proverbs placed in literal contexts should be read in a similar manner to Experiment 1. Note, however, that there should be differences between the two contextual conditions at the end of the proverbs if the more specific nonliteral marker proverbially speaking strengthened the nonliteral contexts relative to the more general nonliteral marker used in Experiment 1. Method Participants. Forty people participated for course credit, 20 per list. All participants were native English-speaking psychology undergraduates from the University of Western Ontario who had normal or corrected-to-normal visual acuity. Materials. The items used were identical to those employed in Experiment 1 with one exception: The nonliteral biasing contexts always ended with the phrase proverbially speaking,….” Note that the literal contexts ended, as they had in Experiment 1, with the proverbial marker literally speaking,.…” Procedure and design. Experiment 1.
The procedure and design were identical to
Results and Discussion The mean reading times for familiar and unfamiliar proverbs placed in literal and nonliteral biasing contexts are illustrated in Figure 2. The differences from the results found in Experiment 1 are quite clear. First, markers that cued the reader as to the type of upcoming nonliteral statements led to slower reading (as compared
32
KATZ AND FERRETTI
FIGURE 2 Experiment 2 reading times.
to the literal biasing contexts) in the initial regions for both familiar and unfamiliar proverbs used nonliterally. Second, the slower reading time following unfamiliar proverbs used nonliterally (rather than literally) in Experiment 1 was eliminated by the beginning of the second sentence in Experiment 2, indicating that readers in this experiment found it easier to integrate the unfamiliar proverbs placed in the nonliteral contexts. Third, the literal biasing marker had a similar influence on reading times as in Experiment 1, with one exception: For the first time, we found some evidence (in the subject analysis) that familiar proverbs are integrated more easily in literal than nonliteral biasing contexts. Word before proverb. There was a robust effect of context in which the word before the proverb was read more quickly when preceded by literal than by nonliteral contexts, F1(1, 39) = 31.94, F2(1, 22) = 18.34. That is, knowing that the upcoming target was to be interpreted nonliterally slowed processing considerably. No other effect reached significance. The two-way interaction between target type and context was not significant, both Fs < 1. Familiar proverbs were read similarly when preceded by literal biasing contexts and nonliteral biasing contexts, both Fs < 1. Similarly, unfamiliar proverbs were read similarly when
LINGUISTIC MARKERS IN PROVERBS
33
preceded by literal biasing contexts and nonliteral biasing contexts, both Fs < 1. There was no effect of familiarity, both Fs < 1. First word of target. The pattern of results for the first word of the proverbs was very similar to the previous word region. There was no two-way interaction between target type and context, both Fs < 1. There was a main effect of context in which the first word was read more quickly when preceded by literal contexts, F1(1, 39) = 10.73, F2(1, 22) = 4.24. This effect was found in the planned comparisons as well. Both familiar and unfamiliar proverbs were read more quickly when preceded by literal biasing contexts than by nonliteral biasing contexts, but this effect was only significant in the subject analyses, F1(1, 39) = 7.98, F2(1, 22) = 2.03, p > .16 for the familiar proverbs and F1(1, 39) = 4.87, F2(1, 22) = 2.23, p > .14 for the unfamiliar proverbs. There was no effect of familiarity, both Fs < 1. Second word of target. The two-way interaction between target type and context was again nonsignificant, both Fs < 1. Overall, the first words of the proverbs were read more quickly when they were preceded by literal than by nonliteral biasing contexts, F1(1, 39) = 14.05, F2(1, 22) = 5.72. Planned comparisons revealed that familiar proverbs were read more quickly when they were preceded by literal biasing contexts than by nonliterally biasing contexts, F1(1, 39) = 17.12, F2(1, 22) = 4.00, p < .06. Unfamiliar proverbs were also read more quickly when they were preceded by literally biasing contexts than by nonliterally biasing contexts, but this effect did not reach significance in the item analysis, F1(1, 39) = 8.73, F2(1, 22) = 1.82, p > .19. We now begin to see the emergence of familiarity effects. Familiar proverbs were read more quickly than unfamiliar proverbs, but this effect only reached significance in the subject analysis, F1(1, 39) = 4.37, F2(1, 22) = 1.64, p > .21. Middle region of target. The effects were quite similar to that observed in the earlier regions. The two-way interaction between target type and context was again nonsignificant, both Fs < 1 and, as in the previous cases, the words were read more quickly when they were preceded by literal than by nonliteral biasing contexts, F1(1, 39) = 12.82, F2(1, 22) = 5.67. Planned comparisons confirmed this finding for both familiar and unfamiliar proverbs but, in both cases, the effect emerged more strongly in the subject analyses [for familiar proverbs, F1(1, 39) = 6.76, F2(1, 22) = 3.14, p = .09, and, for unfamiliar proverbs, F1(1, 39) = 6.96, F2(1, 22) = 2.62, p > .11]. By this region, the familiarity effect was quite large: Familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 39) = 58.92, F2(1, 22) = 5.55. Second last word of proverb. The main effect of context was now nonsignificant, F1(1, 39) = 2.59, p < .12, F2 < 1, as was the two-way interaction
34
KATZ AND FERRETTI
between context and familiarity, both Fs < 1. Familiar proverbs were still read more quickly than unfamiliar proverbs, F1(1, 39) = 55.24, F2(1, 22) = 10.79. The planned comparisons revealed that familiar proverbs were now being read at the same rate whether used literally or nonliterally, both Fs < 1. However, unfamiliar proverbs were read more quickly when preceded by literally biasing contexts than by nonliterally biasing contexts, F1(1, 39) = 3.96, F2(1, 22) = 5.97. Last word of proverb. The main effect of context was significant by subjects and marginal by items, F1(1, 39) = 4.23, F2(1, 22) = 3.53, p < .08. Familiar proverbs were still read more quickly than unfamiliar proverbs, F1(1, 39) = 25.28, F2(1, 22) = 18.88. Both of these effects were modified by the two-way interaction between context and familiarity, F1(1, 39) = 5.35, F2(1, 22) = 6.52. Planned comparisons revealed that familiar proverbs were read at the same rate in the literal and nonliteral context (both Fs < 1), whereas unfamiliar proverbs were read more quickly when they were preceded by literally biasing contexts than by nonliterally biasing contexts, F1(1, 39) = 8.43, F2(1, 22) = 9.79. Word after target. The two-way interaction between context and familiarity was not significant, both Fs < 1. Familiar proverbs were read slightly faster in literal biasing contexts than in nonliteral biasing contexts, but this effect was only significant in the subject analysis, F1(1, 39) = 5.33, F2 < 1. Unfamiliar proverbs were not read significantly differently in the two types of contexts, F1(1, 39) = 2.56, p > .11, F2 < 1. Overall, familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 39) = 15.43, F2(1, 22) = 22.81, and people read proverbs presented in literal contexts slightly faster than in nonliteral contexts, but this effect only reached significance in the subject analysis, F1(1, 39) = 5.60, F2(1, 22) = 1.87, p > .18. In summary, the results of Experiment 2 indicate that familiar proverbs placed in nonliteral contexts are (a) initially read more slowly than when placed in literal contexts, and (b) read at the same rate regardless of contextual bias by the end of the proverbs. Moreover, readers are somewhat slower to read the first word of the subsequent sentence, suggesting that readers had some difficulty integrating the familiar proverbial statements in nonliteral contexts. The results also indicate that unlike the proverbial marker used in Experiment 1, the nonliteral marker proverbially speaking combined with the context to eliminate the difference in reading times for unfamiliar proverbs used literally versus nonliterally by the beginning of the first word of the sentence following the proverbs. Thus, forewarning about the nature of the items by a proverbial marker eliminated the spill over effect found in Experiment 1 (when an underspecified marker was employed) and also found in Katz and Ferretti (2001), when no markers were employed. Finally, the beginning of the proverbs (familiar and unfamiliar) were read much more slowly when preceded by the more specific nonliteral marker versus than when preceded by the
LINGUISTIC MARKERS IN PROVERBS
35
same literal marker used in Experiment 1. Recall that in Experiment 1, proverbs preceded by the nonliteral general marker in a manner of speaking were initially read more quickly than the same proverbs preceded by the literal marker. The initial elevation in reading times for the specific marker in Experiment 2 indicates that readers are having more difficulty initially integrating them into the contexts than the more general marker, but in the case of unfamiliar proverbs, eventually facilitated the complete integration of the unfamiliar proverbs into the discourse contexts.
EXPERIMENT 3 The results of Experiment 2 indicate that a more specific marker had a stronger influence in the integration of proverbs in discourse than did the more general marker employed in Experiment 1. However, in both experiments we always indexed the strength of the nonliteral marker relative to a condition in which a specific literal marker was used, making it difficult to investigate how the nonliteral marker strengthens nonliteral contexts in relation to the case in which no corresponding literal marker was used. For the same reason, the first two experiments do not provide any direct evidence about the influence of the literal marker literally speaking on the comprehension of the proverbs in literally biased contexts. Experiment 3 addressed these problems by employing the same nonliteral marker used in Experiment 2, but in Experiment 3 the literal marker was omitted. Thus, Experiment 3 examined directly how the nonliteral marker facilitates proverb comprehension relative to an established literal baseline (i.e., Katz & Ferretti, 2001, Experiment 1) in which no warning is given about the intended meaning of the upcoming statements, and at the same time provided a way to examine the influence of the literal marker more by comparing the current results with Experiment 2, in which the literal marker was used and compared against the same nonliteral marker condition. The predictions for Experiment 3 were similar to Experiment 2, with the exception that we predicted that if the literal marker actually facilitated the reading of proverbs in the nonliteral conditions of Experiments 1 and 2, then the advantage gained by reading these markers should be reduced relative to the same proverbs read in nonliteral biasing contexts. Method Participants. Thirty-eight individuals participated for course credit, 19 per list. All participants were native English-speaking psychology undergraduates from the University of Western Ontario who had normal or corrected-to-normal visual acuity.
36
KATZ AND FERRETTI
Materials. The items used were identical to those used in Experiment 1 with two exceptions: first, the nonliteral contexts now employed the marker proverbially speaking… instead of in a manner of speaking, and, second, there was no introductory marker employed in the in the literally biasing condition. Procedure and design. The procedure employed, the experimental design, and the analyses employed were identical to that described for Experiment 1. Results and Discussion The mean reading times for familiar and unfamiliar proverbs placed in literal and nonliteral biasing contexts are illustrated in Figure 3. The results of Experiment 3 were remarkably similar to Experiment 2, despite the fact that the literal marker literally speaking was not used. Specifically, familiar proverbs used nonliterally (and with an explicit marker) were again read considerably more slowly in the earlier regions, and there was also no longer a large difference in reading times for unfamiliar proverbs used literally versus nonliterally at the beginning of the next sentence. Moreover, the nonliteral contextual conditions
FIGURE 3 Experiment 3 reading times.
LINGUISTIC MARKERS IN PROVERBS
37
again led to an overall slowing of reading at the beginning of the proverbs in relation to the literal conditions. The only deviations from Experiment 2 were that the difference between the two contextual conditions for the unfamiliar proverbs was now only marginal in the subject analysis, and the statistical evidence indicating that participants took longer to read the first word of the second sentences following familiar proverbs in literal versus nonliteral contexts is now nonsignificant. In summary, the proverbial marker appears again to diminish the large differences in sentence wrap-up found for unfamiliar proverbs in Experiment 1 and, further, the results indicate that the absence of a literal marker had little impact on the overall pattern of reading times. Word before proverb. Recall that for targets used nonliterally, the word before was always the word “speaking” whereas for targets used literally, the word changed for each discourse context. There was a two-way interaction between target type and context, F1(1, 37) = 11.19, F2(1, 22) = 7.97. Planned comparisons revealed that for the context that introduced familiar proverbs the word before the proverb was read more slowly in the strengthened nonliteral biasing contexts (i.e., when the word was speaking) than was the word in the literally biased context, the F1(1, 37) = 14.23, F2(1, 22) = 9.52. That is, for these contexts at least, it appears that the marker was being noted. However, the word before the unfamiliar proverbs was read similarly regardless of the bias of the contexts, both Fs < 1. There was also a main effect of context in which the word before the proverb was read more quickly when preceded by literal than by nonliteral contexts, F1(1, 37) = 4.25, F2(1, 22) = 4.57. There was no effect of familiarity, both Fs < 1. First word of target. The pattern of results for the first word of the proverbs was very similar to that observed in the previous word region. Specifically, there was a two-way interaction between target type and context, although it was marginally significant in the item analysis, F1(1, 37) = 5.38, F2(1, 22) = 3.88, p < .07. Planned comparisons indicated that the first word of the familiar proverbs was read more quickly when it was preceded by literally biasing contexts than by nonliterally biasing contexts, F1(1, 37) = 18.86, F2(1, 22) = 10.00, but unfamiliar proverbs were read similarly regardless of contextual bias, both Fs < 1. Overall, the first word of the proverbs was read more quickly when it was preceded by literal than by nonliteral biasing contexts, F1(1, 37) = 13.01, F2(1, 22) = 6.46. However, there was again no effect of the familiarity on the reading times, both Fs < 1. Second word of target. The results for the second word of the proverbs were very similar to previous word regions. The two-way interaction between target type and context was again significant, F1(1, 37) = 5.38, F2(1, 22) = 6.84. Planned comparisons revealed that familiar proverbs were read more quickly when they were preceded by literally biasing contexts than by nonliterally biasing
38
KATZ AND FERRETTI
contexts, F1(1, 37) = 17.07, F2(1, 22) = 10.74, and unfamiliar proverbs were read similarly regardless of the bias of the contexts, both Fs < 1. Overall, the second word of the proverbs was read more quickly when it was preceded by literal than by nonliteral biasing contexts, although this effect only reached marginal significance in the items analysis, F1(1, 37) = 9.46, F2(1, 22) = 3.59, p < .08. Once again there was no effect of the familiarity of the proverb, both Fs < 1. Middle region of target. Unlike the previous three word positions, the twoway interaction between target type and context was not significant, both Fs < 1. Planned comparisons revealed that familiar proverbs were still read more quickly when they were preceded by literally biasing contexts than by nonliterally biasing contexts, F1(1, 37) = 6.87, F2(1, 22) = 4.42. However, unlike the previous word regions, now the unfamiliar proverbs, were read 12 ms more quickly when they were preceded by literally biasing contexts than by nonliteral biasing contexts, but this effect was only marginal in the subject analysis, F1(1, 37) = 3.29, p < .08, F2 < 1. Overall, the first words of the proverbs were read more quickly when they were preceded by literal than by nonliteral biasing contexts, F1(1, 37) = 8.22, F2(1, 22) = 4.88. However, there was now a strong effect of familiarity, as familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 37) = 15.44, F2(1, 22) = 8.46. Second last word of proverb. The two-way interaction between context and familiarity was not significant, both Fs < 1. Unlike the preceding word regions read to this point, familiar proverbs were read at a similar rate regardless of the contextual bias, both Fs < 1. Unfamiliar proverbs were now read more quickly when preceded by literal biasing contexts, an effect that was only marginal in the items analysis, F1(1, 37) = 1.95, p > .17, F2(1, 22) = 3.00, p < .10. Overall, familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 37) = 28.73, F2(1, 22) = 8.65, and there was still a trend for people to read proverbs presented in literal contexts more quickly than in nonliteral contexts, but this effect was only marginal, F1(1, 37) = 2.54, p = .11, F2(1, 22) = 4.13, p < .06. Last word of proverb. The results for the last word of the proverbs were similar to that observed in the previous word regions. The two-way interaction between context and familiarity was nonsignificant, both Fs < 1. Familiar proverbs were read at a similar rate regardless of the contextual bias, F1(1, 37) = 1.10, p > .3, F2 < 1. Unfamiliar proverbs were read marginally more quickly when preceded by a literal biasing context than a nonliteral biasing context, F1(1, 37) = 3.74, p = .06, F2(1, 22) = 2.12, p > .15. Overall, familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 37) = 20.88, F2(1, 22) = 10.51, and people read proverbs presented in literal contexts more quickly than in nonliteral contexts, but this effect was only marginally significant in the item analysis, F1(1, 37) = 5.95, F2(1, 22) = 2.56, p < 13.
LINGUISTIC MARKERS IN PROVERBS
39
Word after target. The two-way interaction between context and familiarity was not significant, both Fs < 1. Familiar proverbs were read slightly faster in literal biasing contexts than in nonliteral biasing contexts, although this effect did not reach significance, F1(1, 37) = 2.12, p > .15, F2(1, 22) = 1.94, p > .17. Importantly, unfamiliar proverbs were again read similarly regardless of contextual bias, F1 < 1, F2(1, 22) = 1.51, p > .23. Overall, once again, familiar proverbs were read more quickly than unfamiliar proverbs, F1(1, 37) = 22.56, F2(1, 22) = 12.46, and people read proverbs presented in literal contexts marginally faster than those presented in nonliteral contexts, F1(1, 37) = 2.95, p < .10, F2(1, 22) = 3.12, p < .10. In summary, this experiment was an exact replication of Experiment 2, with the exception that the literal marker literally speaking was removed. The results are quite clear. The nonliteral marker again slowed the reading of the familiar proverbs (relative to the same sentences used literally) until about the middle of the item, at which time the reading of familiar proverbs was about the same regardless of contextual bias. Importantly, we found no statistical evidence showing reading time advantages for the literal biasing contexts for the first word following the familiar proverbs. For unfamiliar targets, marking an upcoming sentence as a proverb reduced the last word wrap-up effect found in Katz and Ferretti (2001) and in Experiment 1, and, like Experiment 2, eliminated the large spill over effect that is also observed in those two experiments. However, unlike Experiment 2, we now see some evidence that this reduction in reading times is occurring earlier (i.e., at the last word of the proverb). These findings suggest that the comprehension of unfamiliar proverbs was completed by the end of the sentences, at least when participants know that the item is used proverbially. Thus, using a more specific marker reduces difficulty associated with integrating proverbial statements on nonliteral contexts, and, indirectly, the pattern of results suggests that there was some, albeit small, advantage gained for interpreting the proverbs in literal biasing contexts when they included the marker literally speaking in Experiments 1 and 2.
GENERAL DISCUSSION The main goal of this research was to investigate the role played by explicit markers that highlight whether an upcoming sentence should be interpreted literally or nonliterally. In Experiment 1, this was done by use of a phrase that suggested the nonliteralness of the upcoming sentence without informing the reader about the proverbial nature of that sentence: Proverbial usage was explicitly marked in Experiments 2 and 3. Literal usage was also marked in two of the three studies. The findings are clear. Using an underspecified marker (in a manner of speaking) produced a strikingly similar pattern of reading times to that observed in the
40
KATZ AND FERRETTI
reading of unfamiliar proverbs (in literal and proverbial biasing contexts) when no marker was employed (cf. Katz & Ferretti, 2001). This strongly indicates that just knowing that an upcoming sentence is being used nonliterally will not eliminate the reading advantage of unfamiliar proverbs used literally over the same proverbs used proverbially. In contrast, when the same items are preceded by a marker that specifies not only that the upcoming sentence is nonliteral but also that it is a proverb, there are marked reductions in the reading time between unfamiliar proverbs used literally and the same items used proverbially, both at the last word wrap-up position (Experiment 2) and no indication of spill over into the next sentence (Experiments 2 and 3). In fact, the difference in reading time between these conditions for reading the first word of the sentence following the trope was almost twice as large (relative to Experiment 2) and over three times as large (Experiment 3) when the underspecified marker (in a manner of speaking) was employed compared to when the explicit marker was used (proverbially speaking). The fact that the participants came from samples tested in different years, a large and reliable overall difference in reading times across the three studies and the limited number of targets employed make a direct statistical comparison inappropriate. The interaction of experiment by marker type, nonetheless, approached marginal significance for the contrast between Experiments 1 and 3, F1(1, 86) = 1.95, p < .17, F2 < 1. Naturally, these data are consistent with the notion that providing interpretively useful information in the preceding discourse context allows for the integration of nonliteral meaning as the person is comprehending the proverb online. It should be noted that the effect of the explicit proverb marker occurred whether the literal usage was explicitly marked (Experiment 2) or not (Experiment 3), although the advantages gained by the nonliteral marker were more evident in Experiment 3. The current experiments are the first to document the role played by explicit markers during the online interpretation of literal and nonliteral statements. Although the main goal of this research was an empirical investigation of the role that explicit markers play during online language comprehension, we also take these findings to have implications for the controversy between direct-access and obligatory-first models of nonliteral language processing. Recall that proponents of obligatory-first models argue that one must initially access the literal (e.g., Grice, 1975; Honeck, 1997) or is obligated to access the salient (Giora, 1999) meaning of a statement. Access is modular in the sense that the arousal of sentence meaning proceeds independent of context. We take these models to predict in general that the reading of literal meaning should be faster than the reading of nonliteral meaning, and that context effects should not be evident at the earliest moments of processing, such as during the comprehension of the proverb itself. Specifically, as noted above, we get clear evidence for contextual effects in all three studies. Moreover, in opposition to predictions of the literal first model, we
LINGUISTIC MARKERS IN PROVERBS
41
do not find faster reading of either unfamiliar or familiar proverbs used literally in Experiment 1. Clearly, this class of models was not supported by our data. The contrasting theoretical position is that context can set up an interpretive structure that permits direct access of the nonliteral meaning (see Gibbs, 1994). That is, this position posits that there is no principled reason to expect faster reading of sentences used literally rather than nonliterally. In contrast, with equivalent contextual support, one should find equally fast reading for items used literally or nonliterally. In this research, as in Katz and Ferretti (2001, see Table 1), we made an effort to make the contexts equally supportive of the literal and nonliteral interpretations and, at least as indicated by our offline ratings, were quite successful in doing so. Nonetheless we still found reading differences in the Katz and Ferretti (2000) experiments, and our attempts to create even stronger contextual constraints in the three studies reported here did not produce equivalencies in reading. Instead what we found was that explicitly marking the literal or nonliteral nature of the upcoming target actually led to large reading differences (such as with the familiar proverbs in Experiments 2 and 3); differences not found when markers were absent (see Katz & Ferretti, 2001, Experiment 2). Indeed, we had expected that presenting an explicit proverbial marker might eliminate the reading differences found earlier with unfamiliar proverbs in which these items are read more slowly when used nonliterally than when used literally. That is, if people are slower in reading unfamiliar proverbs in a context that forces them to understand the material nonliterally because they do not have an expectation that a nonliteral interpretation is required, then one should predict that the reading of the unfamiliar proverbs would mimic the reading of the familiar proverb in which this expectation is the default: This we clearly did not find. In all our studies, regardless of context or markers, familiarity effects emerge quite early in processing, by the second or so word of the target. However, explicit markers did have an effect in eliminating spill over effects suggesting that constraining interpretation to a proverbial form can be completed by the last word of the sentence. An effect that emerges across studies is the changes in reading time of the exact same target sentences in the same discourse contexts when specific markers are added. For instance, the reading of familiar proverbs used proverbially is slower initially when the marker proverbially speaking is added (relative to the same proverbs used literally, Experiments 2 and 3), but is faster when we add the marker in a manner of speaking (Experiment 1). This pattern of results is not clearly explained by either the extant direct-access or most obligatory-first type models. These data, and other data that have emerged in the language comprehension literature (e.g., Spivey-Knowlton & Sedivy, 1995), suggest that the current mode of contrasting obligatory-first and direct-access models might prove to be the less profitable route to take in understanding the comprehension of nonliteral language. Instead, we suggest that it might be better to consider in depth a single
42
KATZ AND FERRETTI
mechanism in which comprehension is based on utilizing all the sources of information that a person has at his command at any one instant. Various sources of information have already been identified in the literature, and we would argue that informational constraints regarding the nonliteralness of text requires the same sort of in-depth identification. In recent years, the general approach that we suggest here has been instantiated in constraint satisfaction models (e.g., McRae et al., 1998). These models work by assuming that different sources of information (e.g., syntactic, lexical, conceptual) compete for activation in parallel over time. Constraints interact to provide probabilistic evidence in support of the various interpretative alternatives: Competition duration is a function of the strength of the various alternatives such that resolution of meaning is delayed to the extent that the different alternatives provide support for different interpretations. As applied to these data presented here, the slowing of familiar proverbs used nonliterally suggests that the marker proverbially speaking was a constraint that invited an interpretation different than that induced by other elements of the discourse or by the fixed expression aspects of familiar items. Recent work indicates that the proverbial marker provides multiple (and possibly conflicting) types of information. On the one hand, as the studies collated by Mieder (1982, 1990) indicated, the term is used to signal the proverbial nature of the upcoming text. However, recent corpus-based work suggests that the term proverbial is often used to indicate other idioms or metaphorical use of clichés or fixed expressions (Moon, 1998). Arguably, the slowing in reading of initial words of familiar proverbs reflects these competing alternatives. Also arguably, the resolution of the nonliteral interpretation of the unfamiliar proverbs used nonliterally when the proverbial marker is employed, at least compared to the case when the marker is not present (i.e., Katz & Ferretti, 2001), suggests the proverb signaling aspects of the marker do constrain comprehension. A constraint-based model, in principle, also could explain the differences in reading time we observe here when we change the explicit marker. Thus, the marker in a manner of speaking plays very little role in providing evidence for or against a proverbial interpretation, unlike the marker proverbially speaking. It must be noted that much of these data reported here are also compatible with the most recent version of Giora’s graded saliency model, a model that is basically an obligatory-first type of theory. In recent years, Giora (2003; see Peleg, Giora, & Fein, 2001) has modified her original theory to incorporate, in addition to her original modular component, a separate non-modular context-sensitive mechanism; however, she maintains her claim that context-driven facilitation of less salient meaning cannot be done at a cost to the access of salient meaning. As such, away from the obligatory-first models, Giora’s model, on the surface, can accommodate the findings reported here of early acting context effects. Nonetheless, it is important that the theoretical differences of a constraint satisfaction approach
LINGUISTIC MARKERS IN PROVERBS
43
with the approach taken by Giora not to be underestimated. Constraint satisfaction models do not give processing priority to any fixed element, such as the saliency of language. Rather, these models view language processing as a function of the ongoing interaction of multiple sources of information, each of which constrains the interpretation that will emerge to various degrees. Constraints that are very biasing, such as the saliency of familiar proverbial meanings, may appear to be obligatory because other sources of constraint are not strong enough to overcome the dominance of saliency. Moreover, unlike Giora’s saliency model, the constraint-based model would hold that increasing the strength of opposing constraints should lead to a processing cost in the activation of salient meanings. Importantly, the constraint satisfaction approach would take the position that it is possible in principle to find that the less salient meanings might be activated to a greater degree than salient meanings and, if the constraints indicating the less salient interpretation are so persuasive, then a highly salient alternative may not be activated at all (see Vu, Kellas, & Paul, 1998 for such a demonstration for word access). Thus, in principle, the position favored by Giora and the one taken here are empirically distinguishable, though, given the dominance of saliency as a constraint, some measure more sensitive than reading time might be necessary to disentangle the two theoretical positions. One possibility is the use of event-related brain potentials (ERPs), a measure shown, because of its multidimensional nature, to be sensitive to subtle differences in the processing of language and other cognitive processes (see Osman, 1998 for a review). Studies employing ERP methodology are currently planned in our labs for the study of proverbs and of irony, another instance of nonliteral language. A final finding in the studies reported here deserves brief comment, namely the indication of “list-wise” effects. For instance, the effect of the marker proverbially speaking differs when it is present in a list that also contains the marker literally speaking (Experiment 2 vs. Experiment 3), and the effect of the marker “literally speaking” differs when it is incorporated in a list that contains the marker proverbially speaking (Experiment 2) compared to a list that contains the marker in a manner of speaking (Experiment 1). Taken together, these data indicate that our comprehension and integration of text (in this case, target proverbs) is highly sensitive to differences in contextual constraints defined by the list-wise conditions, that is, by factors external to the specific text in which a given target is embedded. In principle, extra-textoid effects, such as indicated by list effects, also would include pragmatic knowledge. Research in identifying pragmatic constraints includes identifying when a person might employ a given expression (such as a proverb), socioeconomic knowledge about language usage (e.g., Pexman et al., 2000), and gender-related effects (e.g., Katz, Piasecka, & Toplak, 2001). In essence, a constraint-based approach requires an in-depth examination of the reasons why a given expression might be used (e.g., Toplak & Katz, 2000, for sarcastic irony) and when
44
KATZ AND FERRETTI
it might be employed (e.g., Drew & Holt, 1995, for idioms). Research on these questions is only beginning to emerge.
ACKNOWLEDGMENTS This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) operating grant to the first author (Grant 06P007040) and an Ontario Graduate Scholarship (OGS) to the second author. We would like to thank Richard Honeck and Wolfgang Mieder for helpful advice regarding the markers employed in these studies.
REFERENCES Cohen, J. D., MacWinney, B., Flatt, M., & Provost, J. (1993). PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using MacIntosh computers. Behavior Research Methods, Instruments, and Computers, 25, 257–271. Drew, P., & Holt, E. (1995). Idiomatic expressions and their role in the organization of topic transition in conversation. In M. Everaert, E.J. van der Linden, A. Schenk, & R. Schreuder (Eds.), Idioms; Structural and psychological perspectives (pp. 117–132). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Gibbs, R. W. (1994). The poetics of mind. Cambridge, England: Cambridge University Press. Gibbs, R. W. (1995). What proverb understanding reveals about how people think. Psychological Bulletin, 118, 133–154. Giora, R. (2003). On our mind: Salience, context and figurative language. Oxford, England: Oxford University Press. Giora, R. (1999). On the priority of salient meanings: Studies of literal and figurative language. Journal of Pragmatics, 31, 919–929. Glucksberg, S. (2001). Understanding figurative language: From metaphors to idioms. Oxford, England: Oxford University Press. Grice, H. P. (1975). Logic and conversation. In P. Cole and J. Morgan (eds.) Syntax and semantics. Vol. 3: Speech Acts (pp. 41–58). New York: Academic Press Honeck, R. (1997). A proverb in mind. Mahwah: NJ: Lawrence Erlbaum Associates, Inc. Katz, A. N., & Ferretti, T. R. (2001). Moment-by-moment comprehension of proverbs in discourse. Metaphor and Symbol, 16(3 & 4), 193–221. Katz, A.N., Paisecka, I., & Toplak, M. (2001). Comprehending the sarcastic comments of males and females. Poster presented at the 42nd Annual Meeting of the Psychonomic Society, Orlando, Florida. McRae, K., Spivey-Knowlton, M. J., & Tanenhaus, M. K. (1998). Modeling the influence of thematic fit (and other constraints) in on-line sentence comprehension. Journal of Memory and Language, 38, 283–312. Mieder, W. (1982). International proverb scholarship: An annotated bibliography, Vol 1. New York: Garland. Mieder, W. (1990). International proverb scholarship: An annotated bibliography, Vol 2. New York: Garland. Moon, R. (1998). Fixed expressions and idioms in English: A corpus-based approach: Oxford: Clarendon Press.
LINGUISTIC MARKERS IN PROVERBS
45
Osman, A. (1998). Brainwaves and mental processes: Electrical evidence of attention, perception and intention, In D. Scarborough and S. Sternberg (Eds.), An invitation to cognitive science, Volume 4: Methods, models, and conceptual issues (pp.865–915). Cambridge, MA: MIT Press. Peleg, O., Giora, R., & Fein, O. (2001). Salience and context effects: Two are better than one. Metaphor and Symbol, 16(3 & 4), 173–192. Pexman, P., Ferretti, T. R., & Katz, A. (2000). Discourse factors that influence online reading of metaphor and irony. Discourse Processes, 29, 201–222. Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition. (2nd ed.). Oxford: Blackwell. Spivey-Knowlton, M. J., & Sedivy, J. (1995). Resolving attachment ambiguities with multiple constraints. Cognition, 55, 227–267. Temple, J., & Honeck, R. (1999). Proverb comprehension: The primacy of literal meaning. Journal of Psycholinguistic Research, 29, 41–70. Toplak, M., & Katz, A. (2000). On the uses of sarcastic irony. Journal of Pragmatics, 32, 1476–1488. Turner, N. (1989). The role of novelty in the comprehension of figurative and literal meaning. Unpublished master’s dissertation, The University of Western Ontario, London, Canada. Turner, N., & Katz, A. (1997). The availability of conventional and of literal meaning during the comprehension of proverbs. Pragmatics and Cognition, 5, 199–233. Vu, H., Kellas, G., & Paul, S. (1998). Sources of constraint on lexical ambiguity resolution. Memory and Cognition, 26, 979–1001.
APPENDIX Familiar Proverbs 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
As you make your bed, so you must lie in it. We’re not out of the woods yet. You can lead a horse to water but you can’t make it drink. There are plenty of fish in the sea. Man cannot live on bread alone. Lightning never strikes the same place twice. I bang my head against a brick wall. The grass is always greener on the other side of the fence. Don’t count your chickens before they hatch. It never rains, but it pours. Don’t put all your eggs in one basket. You can’t get blood from a stone.
Unfamiliar Proverbs 1. 2. 3. 4. 5. 6.
There are no birds in last year’s nest. Raw leather will stretch. But, while the grass grows, the horse lies dying. A creaking door hangs longest. Hard rocks are hollowed out by soft water. Straight trees have crooked roots.
46
KATZ AND FERRETTI
7. 8. 9. 10. 11. 12.
Blue are the faraway hills. White silver draws black lines. A river needs a spring. An empty sack cannot stand upright. A straight stick looks crooked in the water. Empty bottles make the most sound.
