Journal of Psycholinguistic Research, Vol. 21, No. 2, 1992

Limits on Literal P r o c e s s i n g D u r i n g Idiom Interpretation W i l l i a m P. N e e d h a m ~ Accepted September 5, 1990 Two experiments investigated the extent to which literal processing occurs in comprehending figurative idiomatic expressions. Subjects read stories on a cathode-ray tube (CRT). Target phrases, some of which were idioms, contained nouns which were potential anaphors of previously mentioned referents. A method developed by Dell, McKoon and Ratcliff (1983) was used to determine whether subjects carried out semantic processing resulting in activation of the referents of those anaphors. In Experiment 1 the targets consisted of either an idiom or a literal phrase, each including the same potential anaphor, or a control phrase. Results suggest that the preceding referent was activated by the anaphor in the literal phrase, but not by the potential anaphor in the idiomatic phrase. Experiment 2 showed that these results were not due to differences in the materials used. These results are interpreted as supporting the hypothesis that when an idiomatic phrase is interpreted figuratively full literal semantic processing of that phrase is not necessarily carried out.

Idioms are common phrases whose intended meanings are different from their literal interpretations. Since their meanings cannot be derived by regular syntactic and semantic analyses, the comprehension of idioms has posed a problem for models of the language comprehension process. Many idiomatic phrases are potentially subject to both literal and figurative interpretations, and a number of researchers have addressed the question of how those two interpretive processes relate to one another~ This research was conducted while the author was a graduate student at the Department of Psychology, Northeastern University. I would like to thank James J. Carroll and Joanne Miller for their c o m m e n t s on earlier versions of this work. Division of Natural Sciences, State University of New York at Purchase, Purchase, New York 10577.

0090-6905/92/0100-0001506.50/0 9 1992 Plenum Publishing Corparation

2

Needham

Swinney and Cutler (1979), summarizing previous linguistic and psycholinguistic work, proposed two general idiom processing models. The first of these was termed the Idiom List Hypothesis. According to this hypothesis, idioms are stored in a special list which is not part of the lexicon. When an idiomatic phrase is encountered, normal literal analysis takes place, and if the literal interpretation fails to be satisfactory, processing is shifted into an "idiom m o d e , " and the figurative meaning is recovered from the list. The second model suggested by Swinney and Cutler (1979) was termed the Lexical Representation Hypothesis. According to this hypothesis, idioms are stored in the form of single long words in the standard lexicon. When a phrase is being interpreted, individual words are recovered, and then standard syntactic and semantic processing is performed. Meanwhile, lexical analysis of the entire idiomatic phrase is being performed. This model thus suggests that literal and figurative interpretations of an idiomatic phrase are performed simultaneously, rather than in sequence. In order to test these two hypotheses, Swinney and Cutler (1979) asked subjects to judge whether or not strings of words constituted acceptable English phrases. Their results, that decision times were faster for idiom strings than literal ones, were taken as support for the Lexical Representation Hypothesis. A number of other investigators (Estill & Kemper, 1982; Gibbs, 1980; Glass, 1983) have also provided evidence suggesting that people do not need to process the literal meanings of idiomatic phrases before finding their figurative meanings, further supporting the hypothesis of simultaneous literal and figurative interpretation. Gibbs (1985,1986) has taken this idea one step further, suggesting that when people encounter an idiomatic expression, they will usually process the figurative meaning first, and will only construct the literal interpretation if the figurative meaning is not satisfactory. In particular, Gibbs raises the question of whether, when context induces a figurative interpretation of an idiomatic phrase, literal interpretation of the phrase is actually carried out. Gibbs (1986) presents experimental results suggesting that, in that case, literal interpretation is not completed. In that study, subjects read short passages, one sentence at a time. Each passage ended with a sentence that constituted an idiom. The body of the passage induced either the literal or the figurative interpretation of that sentence. After reading a passage, the subject was presented with a string of words, and was asked to decide whether or not the string constituted a meaningful English

Literal Processing of Idioms

sentence. The meaningful strings were either paraphrases of the figurative or the literal meaning of the idiom, or unrelated sentences. Control studies showed that, in isolation, strings of those three types did not differ in how quickly subjects judged them to be acceptable. In the condition in which the story induced the figurative interpretation of the idiom, the figurative paraphrase string was judged as acceptable more quickly than the unrelated string, but the literal paraphrase string was not. Gibbs (1986) interpreted these results to mean that in that case the subjects had not actually processed the complete literal interpretation of the idiomatic expressions. The interpretation of these results is limited by the fact that the task Gibbs (1986) used was not an on-line measure of processing. The subjects could read each sentence of the passage for as long as they wanted before going on to the next sentence or to the test string. What the results do appear to demonstrate is that, in the case where the passage leads to the figurative interpretation of the idiom, the literal meaning of the phrase is not available to aid in the processing of the literal paraphrase test string. But the results say nothing precise about the extent, or lack of extent, to which literal processing of the figuratively biased phrases was actually carried out. Another way to approach this question would be to identify particular stages of processing that would be presumed to occur during literal interpretation, and use on-line tests to see whether they in fact occur when the context induces a figurative interpretation. One early stage of literal processing is the lexical retrieval of the literal meanings of the individual words in a phrase. Research dealing with ambiguous words has found that directing context cannot prevent initial access to all meanings of such words (Seidenberg, Tanenhaus, Leiman, & Bienkowski, 1982; Swinney, 1979). Those results would suggest that, in the present case, context which induces a figurative interpretation of an idiom would not terminate literal processing prior to lexical access of the literal meanings of the words. There appears to be little published work directly addressing this question. In one study, Cacciari and Tabossi (1988) used a cross-modal lexical decision task to test for priming of words associated with the literal senses of words in idiomatic phrases. In their Experiment 1, they did not find significant priming. However, in this case the idiomatic phrases used did not admit of normal literal meaning (e.g. in seventh heaven); and they provided no direct evidence that the associations they used (e.g. heaven and saint) produced priming in other contexts. In their Experiment 2 they- used phrases that could have either literal or figurative

4

Needham

interpretations, and found, at a zero delay interval, significant priming for literal associates of the final word in the idiomatic phrase. The results of Experiment 2 are, however, of limited applicability to the present question, as no biasing context was used prior to the idiom. Thus there appears to be no strong evidence that the principle of mandatory lexical access to all senses of words does not also apply to the case of words in idiomatic phrases. If literal processing of idiomatic phrases does, in fact, include lexical access of the literal meanings of the words, it would seem reasonable to ask whether further stages of literal processing are then carried out. Following lexical access, the next stages of semantic processing of discourse include the construction of propositions and the specification of referents of pronouns and other anaphors (van Dijk & Kintsch, 1983). In a passage in text or discourse, an item such as John's new sedan may be introduced; later, that item may be referred to anaphorically with a definite noun phrase such as the car, the Chevy, or the vehicle. When a reader encounters such an anaphor, one aspect of the semantic processing of that item consists of searching to find the referent of the anaphor. Dell, McKoon, and Ratcliff (1983) have provided evidence that such a search for the referent of an anaphor begins almost immediately, and that the referent may be identified as early as 250 to 500 msec after the anaphor is encountered. The question addressed in the current experiments is whether this stage of literal processing, the search for and identification of the referent of an anaphor, will take place when the anaphor in question is part of an idiomatic phrase whose figurative interpretation is induced by context. For example, when encountering the literal phrase she dropped the beans, one will search the representation of the previous material to identify the referent of the beans. The question here is whether, when one encounters a phrase such as she spilled the beans, and interprets it figuratively, the process of literal interpretation will continue as far as the stage of searching for the referent of the beans, or whether literal processing will be terminated prior to that stage. The experiments reported here which attempt to answer that question utilize the techniques developed by Dell et al. (1983). An example of the passages used in their study is given in Table I. In the experimental condition (anaphor), the final sentence contains a term that is an anaphor to a referent in the first sentence. In the Control condition, no such anaphor is present. The sentences in the passages were presented one word at a time, at a fixed rate, with each new word being added to the right of the

Literal Processing of Idioms

Table I. Example of the Stories Used by Dell, McKoon, and Ratcliff (1983) Sentence No.: 1:

4a: 4b:

2: 3: Anaphor: Control:

A burglar surveyed the garage set back from the street. Several milk bottles were piled at the curb. The banker and her husband were on vacation The criminal slipped away from the streetlamp. A cat slipped away from the streetlamp.

The test word was either burglar or garage, and was presented at various times during sentence 4. m

previous material. At some point during a sentence the screen went blank, and a test word appeared. The subject's task was to decide as quickly as possible whether or not the test word had occurred in the story. In the experimental stories the test occurred at various intervals after the display of the anapho;r (criminal) or the control item (cat). The test word was either the referent of the anaphor (burglar) or another word in the same sentence (garage). Subjects were faster in verifying the test word when it followed the anaphor than when it followed the control item. When the referent was used as the test word, this advantage relative to the control was found when the tes~ word was displayed as early as 250 msec after the anaphor. When the alternative item from the same sentence was used as the test word, an advantage was found when it was displayed 500 msec after the anaphor. The authors interpreted these results to mean that the search for the referent of the anaphor caused the referent, and the entire sentence in which it occurred, to be activated in memory. In the current experiments the same basic procedure was used. An example of the passages employed is given in Table II. Each passage terminates, and a test word is presented, shortly following an idiomatic phrase that is also subject to a literal interpretation. If the phrase is interpreted literally, the final word of the phrase constitutes an anaphor referring to a previously mentioned referent. As the question of interest here is not whether or not subjects wilt interpret the idiomatic phrase figuratively, but rather to what extent the literal processing will proceed when the phrase is interpreted figuratively, elements of each passage were designed to bias the subjects toward the figurative interpretation of the idiomatic phrase. Three forms of the final phrase of each passage were used, constituting three conditions. In the idiom condition, the final noun is a po-

Needham

Table II. Example of the Stories Used in Experiment 1 i

Sentence No: Title: 1: 2: 3:

6a: 6b: 6e:

4: 5: Idiom: Anaphor: Control:

CAROL LETS OUT A SECRET Carol was cooking dinner for Bob. After dinner, there was going to be a surprise birthday party for him. She was putting some vegetables in a pan. She had poured some drinks for the two of them. She got nervous talking to Bob. She spilled the beans when * She spilled the carrots when * She spilled the beer when *

The test word for all three cases waspan, presented at the *

tential subordinate anaphor for an item previously mentioned (beans/ vegetables), and the phrase itself constitutes an idiom (spill the beans). In the anaphor condition, that noun is a different potential anaphor for the same previous item (carrots~vegetables), but the phrase is not an idiom (spill the carrots). In the control condition, the noun is not an anaphor to the s a m e previous item, and the phrase is not an idiom. The test w o r d used in all three conditions is a w o r d f r o m the same sentence as the referent of the anaphor. It w a s expected that, as in the Dell et al. (1983) experiments, verification of the test word would be facilitated in the anaphor condition as c o m p a r e d to the control condition. The main question to be tested was whether, in the idiom condition, literal processing would continue to the stage of activating the referent of the anaphor and the other items in the same sentence, including the test word. If such processing did occur, verification of the test w o r d should be facilitated in the idiom condition to the same extent as in the anaphor condition. If literal processing were terminated prior to activation of the anaphor, verification of the test word should not be facilitated in the idiom condition as c o m p a r e d to the control condition. Three aspects of the design and choice of materials warrant further discussion. First, O ' B r i e n , Duffy, and M y e r s (1986) have pointed out a potential p r o b l e m with the control condition used b y Dell et al. (1983). In place of the anaphoric noun, a previously unmentioned item (such as cat) was used. The longer response times in this condition m a y be due to the fact that the term used introduced a new topic, and thus m a y have m a d e the previous material less available. Following a suggestion made b y O ' B r i e n et al., the nouns in the control passages in the present ex-

Literal Processing of Idioms

periments were potential anaphors for items previously mentioned other than the original referents. For the example in Table II, the term in the control phrase is beer, and the control referent is drin.~. Second, O'Brien et al. (1986) also point out that the fast recognition times found by Dell et al. (1983) for the test word when it was the referent (burglar) of the anaphoric term just displayed (criminal) could be due in part to lexical priming due to semantic relations between those two concepts. For this reason, rather than using the referent of the anaphor as the test item, another word from the same sentence as the referent was used. Since Dell et al. found advantages for such test words from the same sentence as the referent when they were presented 500 msec after the anaphoric term, that interval was used in the present experiments. Finally, the referent/anaphor pairs which were used by Dell et al. (1983) consisted of category terms as the anaphors (criminal), and subordinate members of those categories as the preceding referents (burglar). Unfortunately, there are few American English idioms that have a category term as their final item. There are, however, many idioms that end with basic noun terms, such as spill the beans and kick the bucket, which are members of categories. Thus in this experiment such category members were used as the potentially anaphoric nouns in the target phrases, and a category term or phrase (such as vegetables or cleaning gear) was used as the preceding referent. In order to determine that activation takes place with this arrangement of terms, the results for the anaphor and control conditions will be compared. Superior verification of the test word in the anaphor condition as compared to the control condition would demonstrate that a subordinate category member used as an anaphor can serve to activate a preceding category referent.

EXPERIMENT 1

MeNod Subjects. Sixty-three Northeastern University undergraduates participated in the experiment for course credit. All subjects were native speakers of English and none reported any visual deficiencies. Three subjects pressd illegal keys for at least one of the experimental items, and were replaced. Materials. Thirty-six idiomatic phrases ending in nouns were presented to 20 subjects, who were asked to indicate whether or not they knew the phrase as an idiom, and to give its literal meaning. Eighteen

8

Needham

idioms which were accurately identified by at least 80% of the subjects were chosen for use. Eighteen experimental passages were constructed having the same form as the passage shown in Table II. Each passage consisted of five full sentences and a final partial sentence followed by the test item. Each passage began with a title that included a paraphrase of the figurative meaning of the idiomatic phrase. The first sentence in each passage introduced a scene and mentioned the primary character(s). The second sentence mentioned some aspect of the scene that set the occasion for the use of the idiomatic phrase. The third and fourth sentences mentioned specific activities performed by the primary character(s); the third sentence included a category term that was designated as the Experimental Referent, and the fourth sentence included a different category term that was designated as the Control Referent. The fifth sentence referred to some thought or decision of the primary character(s) that could lead to the activity denoted by the figurative meaning of the idiom. For each story, three versions of the final, partial sentence containing the target phrase were constructed. The three versions were identical except for the final noun of the target phrase. In the idiom condition the noun was a member of the category named by the Experimental Referent and the phrase as a whole constituted an idiom. In the anaphor condition the noun was another member of the same category, but the phrase was not an idiom. In the control condition the noun was a member of the category named by the Control Referent and the phrase was not an idiom. In all versions of the final sentence, one further word followed the final noun in the target phrase, so that the test word would be displayed at the proper interval. The test word was identical for all three versions of each passage, and was an item that occurred in the third sentence, along with the Experimental Referent. The Experimental Referents and the final partial sentences used in the three conditions in are listed in the Appendix. Three sets of materials were constructed. Each set contained eighteen experimental passages; six of those passages were in each of the conditions: idiom, anaphor, and control. Across the three sets, each passage appeared once in each experimental condition. Thirty-two filler passages and 10 practice passages wer~ also constructed. The filler pasages served two purposes. Since the correct response for the presence of the test item in each of the experimental passages was positive, seventeen of the filler stories were constructed so that the test word did not occur in the passage, and the correct answer was negative. Since the test item was always presented during the sixth sentence in the experimental pas-

Literal Processing of Idioms

ages, the filler passages were constructed so that the test item occurred anywhere from the third to the eighth sentence. Procedure. Subjects were tested individually. Materials were presented on a DEC VT101 CRT controlled by a D E C 11/23 minicomputer that also recorded the response times. Each subject read 10 practice passages before reading the 18 experimental and 32 filler passages. The order o f the 50 passages was determined randomly for each subject. The subject initiated the presentation o f each passage in response to a ready signal on the CRT. The first word of the title line then appeared in the upper left corner o f the screen. After 250 msec the next word appeared to the right o f the first word on the same line, separated b y a space. The first word remained on the screen. The other words in the title appeared in the same manner. No title or sentence was longer than one screen line. After the last word of the title was displayed, the entire title remained on the screen for 500 msec. Then the title disappeared and, 500 msec later, the first sentence of the passage was displayed in the same manner. Subjects were told that at some point while they were reading each passage the current unfinished sentence would disappear and a test word would appear in the position where the next word would have been displayed. The test word was displayed in capital letters and underlined by asterisks. T h e y were told to decide as quickly as possible whether or not the test word had occurred in the passage, and to press one of two labeled keys on the terminal keyboard. On the practice trials the subjects received feedback as to whether their answers were correct.

Results The two dependent measures used were response time and error rate for verification of the presence of the test word in the passage. Only the response times for correct responses were analyzed. Response times greater than 4 sec were not included in the analysis; these items represented 1.7% of all responses. In two cases, a subject did not give any correct responses in one of the three conditions. The missing values for these response time means were replaced according to the method described by Winer (1971). Repeated-measures analyses of variance were performed on both response time and error rate. For each variable both an F1 test, based on the mean for each condition for each subject, and an F2 test, based on the mean for each condition for each passage, were performed. The mean response times and error rates, based on subject means, are presented in Table IIl.

10

Needham

Table III. Mean Response Times (msec) and Percentage Error for Test

Verification in Experiment 1" Condition

Response time

Percentage error

Idiom Anaphor Control

1644 (473) 1576 (420) 1627 (531)

33.1 (24.4) 24.4 (19.0) 33.3 (20.8)

aNote: Standard deviations in parentheses.

Although the response times in the anaphor condition were slightly faster than those in the idiom and control conditions, the analysis by subjects was not significant, F1 < t..0, and the analysis b y materials was only marginally significant, F 2 ( 2 , 34) = 2.98, M S e = 54,504, p < .07. For the error analysis, however, the main effect of condition was significant across both subjects and materials, F l ( 2 , 118) = 4.94, M S e = 309.9, p < .01; F 2 ( 2 , 34) = 3.33, M S e = 135.7, p < .05. The error rates for each condition were compared using the Bonferroni t test. The error rate for the anaphor condition was significantly lower than the error rates for either the idiom condition, t(59) = 2.87, p < .01, or the control condition, t(59) = 2.77, p < .01. The rates for the idiom and control conditions did not differ significantly from one other, t < 1. Discussion

Subjects made fewer errors on the anaphor items than on the control items, and this difference was not balanced by a tradeoff with higher reaction times. These results suggest that the mention of the subordinate anaphor did cause its referent, and the accompanying material in the same sentence, to be activated. There was no difference, however, in either error rate or response times between the idiom items and the control items. These data suggest that the referents of the potential anaphors in the idiom phrases were not activated. Thus it appears that, when the subjects read the idiom phrases, any literal processing which took place terminated prior to this stage in the semantic interpretation o f the phrases. The response times and error rates were somewhat larger than those found in the Dell et al. (1983) study. The mean response times of their subjects were in the range of 700 to 750 msec and the error rates ranged from 14% to 20%. This difference m a y have been due to the fact that the passages used in the present experiment were both longer and more complicated than the ones used by Dell et al., containing material that

Literal Processing o f I d i o m s

11

made both the figurative and the literal interpretations of the idiomatic phrases possible. The difference may also have been due to the reversal in the relation of the terms used for the anaphor and the referent. In the Dell et al. study, facilitation in the anaphor condition appeared in the form of faster response times. Here, facilitation appeared in the form of lower error rates. This may also have been due to the greater complexity of the passages, which made the subjects" task more difficult. Before the central interpretation of the results obtained in the present study can be accepted, one potentially confounding factor must be considered. It is possible that the category members used as potential anaphors in the idiom condition (such as beans) were not as closely associated with the category terms used as referents (vegetables) as were the category members used in the anaphor condition (such as carrots). If the terms used as subordinate anaphors differed in this way, then the search for the referent would be more difficult in the idiom case than in the anaphor case, possible leading to the lack of facilitation observed in the idiom condition. A control experiment was therefore conducted to test this possibility. In this control study all materials in the stories remained unchanged except for the verbs in the target phrases. That verb in each passage was changed so that the target phrase did not constitute an idiom in any of the three conditions. An example of the altered phrases is given in Table IV. Since the passages in what was originally termed the idiom condition no longer contained an idiom, the name of that condition was changed to the " i d i o m * " condition. If the category members used as potential anaphors in the idiom condition in the first experiment and the idiom* condition in this exper-

Table IV. Example of the Stories Used in Experiment 2 Sentence No.: Title: 1: 2:

6a: 6b: 6c:

3." 4: 5: Idiom*: Anaphor: Control:

CAROL LETS OUT A SECRET Carol w a s c o o k i n g dinner for Bob. After dinner, there w a s g o i n g to be a surprise birthday party for h i m . She w a s putting s o m e v e g e t a b l e s in a pan. She h a d poured s o m e drinks for the two of t h e m . She got n e r v o u s talking to Bob. She dropped the b e a n s w h e n * She dropped the carrots w h e n * She dropped the beer w h e n *

T h e test w o r d for all three c a s e s w a s p a n , presented at the *

12

Needham

iment are less closely related to the referent category terms than the category members used in the anaphor conditions are, then the anaphor items should still show greater facilitation than the idiom* items. If, however, the category members are equally closely related to the category terms, then equal facilitation should be found in the idiom* and the anaphor conditions.

EXPERIMENT 2

Me~od Subjects. A new group of 64 Northeastern University undergraduates participated for course credit. All subjects were native speakers of English and none reported any visual deficiencies. Four subjects pressed illegal keys for at least one of the experimental items, and were replaced. Materials and Procedure. The same passages were used as in the first experiment, except that the verbs in the target phrases were replaced by other verbs, so that none of the phrases constituted an idiom. The target phrases used in Experiment 2 are presented in the Appendix. The study was conducted in the same manner as the first experiment. Results As in the first experiment, response times for incorrect responses were excluded from the analysis. Response times over 4 sec were also excluded, these items represented 2.2% of all responses. Again, in two cases a subject did not give any correct responses in one of the three conditions, and those missing values were replaced as previously specified. The mean response times and error rates, based on subject means, are presented in Table V. The subjects' response times in the three conditions did not differ

Table V. Mean Response Times (msec) and Percentage Error for Test Verification in Experiment 2~ Condition

Response time

Percentage error

Idiom* Anaphor Control

1622 (434) 1638 (478) 1657 (523)

25.8 (20.9) 27.2 (21.9) 38.9 (25.2)

aNote: Standard deviations in parentheses.

Literal Processing of I d i o m s

13

significantly, F 1 < 1.0; F 2 ( 2 , 34) --- 1.02, M S e = 34,133, n.s. There was, however, a significant effect of condition on the error rates, F l ( 2 , 118) = 9.70, M S ~ = 318.1, p < 0.01; F 2 ( 2 , 34) = 8.00, M S ~ = 115.7, p < 0.01. The Bonferroni t test showed that the error rates for the idiom* and the anaphor conditions did not differ significantly from each other, t < 1; and that each was significantly lower than the error rate for the control condition [idiom* vs. control, t(59) = 3.99, p < .001; anaphor vs. control, t(59) = 3.59, p < .001]. Discussion

Facilitation was again found in the anaphor condition, and an equal amount of facilitation was found in the idiom* condition. Thus it can be concluded that the category members used as potential anaphors in the idiom* condition were as effective in driving a search for the category referent term as the category members used in the anaphor condition were. Thus the lack of facilitation found in the idiom condition in the first experiment did not result from the nature of the particular category members used.

G E N E R A L DISCUSSION The results of the experiments presented here support the hypothesis advanced by Gibbs (1986) that people do not necessarily complete literal processing of idiomatic expressions. When subjects encountered idiomatic expressions containing potential anaphors of previously mentioned items, they did not appear to complete searches for the referents of those anaphors, a process that they did complete for matched nonidiomatic expressions. These experiments extend the previous finding by Gibbs (1986) in three ways. First, Gibbs" experiments involved measuring the comprehension time for a paraphrase sentence presented after the sentence in which the idiomatic phrase occurred. Here, an on-line task was used to ensure that the results did not reflect aspects of processing occurring after the completion of the sentence. Second, a specific stage of the comprehension process was tested, rather than the general results of comprehension. Third, Gibbs' results may have reflected a bias on the part of his subjects. In his main experiment, over two-thirds of the stories ended with sentences that potentially constituted idioms. It is possible that his subjects became aware that the experiment was concerned with idioms, and thus deliberately tried to interpret all sentences figuratively, leading

14

Needham

to the observed lack of priming effects that would stem from literal interpretations. In the main experiment in the current study, only 6 out of the 50 stories encountered by any particular subject contained potentially idiomatic expressions as the target phrase. During debriefing, none of the subjects indicated that they had become aware that the experiment concerned idioms. Thus it is unlikely that the results reported here could have been influenced by subject bias. The results of the present experiments indicate that people do not necessarily complete all stages of literal processing of idiomatic expressions. These results, however, give no indication as to which stages of literal processing people may actually carry out. As Cacciari and Tabossi (1988) point out, the extent to which literal processing of an idiom is carried out on any specific occasion may depend on actual point at which the figurative interpretation of the idiom is determined. If the figurative meaning is recognized early in the phrase, the literal processing of that phrase may be terminated at a very early stage. Further research will be necessary to determine just how quickly such literal processing of idioms can be terminated, and which aspects of that processing may or may not be terminated.

APPENDIX: REFERENTS AND TARGET PHRASES FROM T H E P A S S A G E S IN E X P E R I M E N T S 1 A N D 2

The term in capitals is the category-level referent. In single parentheses are the wordings used in Experiments 1 and 2, in that order. In double parentheses are the terms used in the idiom (or idiom*), anaphor, and control conditions, in that order. 1. BAR SUPPLIES: She (broke/dropped) the ((ice/bottle/chips)) 2. CLEANING GEAR: After she (kicked/grabbed) the ((bucket/mop/ chair)) 3. CLEAN LAUNDRY: He was about ready to (throw in/reach for) the ((towel/shirt/keys)) 4. GEOMETRIC FIGURES: He (drew/traced) the ((line/triangle/title)) 5. VEGETABLES: She (spilled/dropped) the ((beans/carrots/beer)) 6. TOOLS: They (buried/hid) the ((hatchet/crowbar/coins)) 7. BARRIERS: He was about ready to (climb/scale) the ((walls/fence/ ladder)) 8. MOUNTAIN CLIMBING GEAR: He asked the boss to (show him/ help him find) the ((ropes/boots/forms)) 9. FURNITURE: They would (turn/watch) the ((tables/chairs/rings))

Literal Processing of Idioms

15

10. SCRAPS FROM THE ROAST: They sat and (chewed/nibbled at) the ((fat/bones/apples)) 11. TRAFFIC SIGNALS: She had (seen/noticed) the ((light/sign/dress)) 12. OUTSIDE PARTS OF THE STORE: He had almost (hit/touched) the ((roof/wall/hedge)) 13. HORSE-DRAWN VEHICLES: He had (been on/rearranged) the ((wagon/cart/seat)) 14. TIP OF A N ISLAND: Dave said that he (saw/noticed they were even with) the ((point/cape/cabin)) 15. SLOPE: He was already (over/past) the ((hill/rise/stream)) 16. LOGGING TOOLS: The boss had (given/loaned) him the ((axe/saw/ money)) 17. LINGERIE: She (gave/handed) her the ((slip/bra/lipstick)) 18. STATIONERY: He had (stacked/placed) the ((cards/envelopes/ brushes))

REFERENCES Cacciari, C., & Tabossi, P. (1988). The comprehension of idioms. Journal of Memory and Language, 27, 668-683. Dell, G. S., McKoon, G., & Ratcliff, R. (1983). The activation of antecedent information during the processing of anaphoric reference in reading. Journal of Verbal Learning and Verbal Behavior, 22, 121-132. van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. Orlando, FL: Academic Press. Estill, R. B., & Kemper, S. (1982). Interpreting idioms. Journal of Psycholinguistic Research, 11, 559-568. Gibbs, R. W. (1980). Spilling the beans on understanding and memory for idioms in conversation. Memory and Cognition, 8, 149-156. Gibbs, R. W. (1985). On the process of understanding idioms. Journal of Psycholinguistic Research, 14, 465-472. Gibbs, R. W. (1986). Skating on thin ice: Literal meaning and understanding idioms in conversation. Discourse Processes, 9, 17-30. Glass, A. L. (1983). The comprehension of idioms. JouJ~at of Psycholinguistic Research, 12, 429-442. O'Brien, E. J., Duffy, S. A., & Myers, J. L. (1986). Anaphoric inference during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 346352. Seidenberg, M. S., Tanenhaus, M. K., Leiman, J. M., & Bienkowski, M. (1982). Automatic access of meanings of ambiguous words in context: Some limitations of knowledge based processing. Cognitive Psychology, 14, 489-537.

16

Needham

Swinney, D. A. (1979). Lexical access during sentence comprehension: (Re)consideration of context effects, Journal of Verbal Learning and Verbal Behavior, 18, 523-534. Swinney, D. A., & Cutler, A. (1979) The access and processing of idiomatic expressions. Journal of Verbal Learning and Verbal Behavior, 18, 523-554. Winer, B. J. (1971). Statistical principles in experimental design (2nd ed.). New York: McGraw-Hill.