Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

Transcription

1 Semantic Component of a Cross-Modal Stroop-like Task Author(s): David M. Stuart and Marisa Carrasco Source: The American Journal of Psychology, Vol. 106, No. 3 (Autumn, 1993), pp Published by: University of Illinois Press Stable URL: Accessed: 25/03/ :37 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. University of Illinois Press is collaborating with JSTOR to digitize, preserve and extend access to The American Journal of Psychology.

2 Semantic component of a cross-modal Stroop-like task DAVID M. STUART and MARISA CARRASCO Wesleyan University Three experiments showed that the pattern of interference of singlemodality Stroop tests also exists cross-modally. Distractors and targets were either pictures or auditory words. In a naming task (Experiment 1), word distractors from the same semantic category as picture targets interfered with picture naming more than did semantically unrelated distractors; the semantic category of picture distractors did not differentially affect word naming. In a categorization task (Experiment 2), this Stroop-like effect was reversed: Picture distractors from the same semantic category as word targets interfered less with word categorization than picture distractors that were semantically unrelated; the semantic category of word distractors did not differentially affect picture categorization. Experiment 3 replicated these effects when each subject performed both tasks; the task, naming or categorizing, determined the pattern of interference between pictures and auditory words. The results thus support the existence of a semantic component of a cross-modal Stroop-like effect. The Stroop effect (Stroop, 1935) shows that the visual and verbal processing systems do not act independently: Subjects' ability to name the ink color in which an incongruent "color word" is printed is highly inhibited compared with their ability to name the ink color of a nonword color patch. However, reading the color word is not inhibited by the incongruent color ink. For example, subjects' ability to name the color of green ink in which the word red is printed would be inhibited; but under the same conditions, subjects' ability to read the word red would be unaffected. This is the Stroop asymmetry phenomenon that has been used to examine the processing of lexical and semantic information. Numerous variations of the Stroop test have confirmed the Stroop asymmetry, and the strength of the effect has prompted extensive inquiry (for reviews see Dyer, 1973; MacLeod, 1991). Different experimental Stroop-like tasks, within the visual modality, have shown that word processing and pictorial processing can interfere with each other (e.g., Glaser & Diingelhoff, 1984; Smith & Magee, AMERICAN JOURNAL OF PSYCHOLOGY Fall 1993, Vol. 106, No. 3, pp ? 1993 by the Board of Trustees of the University of Illinois

3 384 STUART AND CARRASCO 1980). The present study assessed the interference effects of auditory words on subjects' ability to process pictures, and vice versa, so as to investigate whether interference is found in a cross-modal Stroop-like situation. Although a cross-modal Stroop interference has been re- ported (Cowan 1989a, 1989b; Cowan & Barron, 1987), these findings have been challenged (Miles & Jones, 1989; Miles, Madden, & Jones, 1989). Furthermore, the experiments reported here are the first to explore whether cross-modal interference has a semantic component. Its existence would suggest that a common semantic code underlies both the visual and auditory modes of representation and that some memory encoding processes are not specific to a particular modality. Auditory analogs of the Stroop test have also been performed. Because color patches cannot be represented in the auditory modality, dimensions such as pitch have been manipulated instead. Shor (1975) found that subjects could distinguish more quickly between high- and low-pitched voices saying the words high or low when instructed to attend to the word rather than to the pitch of the voice. Furthermore, a congruent match (i.e., a high-pitched voice saying high, or a lowpitched voice saying low) led to a faster response than when the distractor was just a pitch or a word pronounced in a neutral voice. A cross-modal Stroop-like effect has been reported recently by Cowan and Barron (1987). Their results indicate that incongruent auditory color words interfere with a verbal response to visual color patches, whereas auditory noncolor words or music do not. These findings agree with research indicating that distractors from the response set produce maximal interference (La Heij, 1988) and that the response modality affects the magnitude of the interference (McClain, 1983). One of the most important aspects of the Stroop test, for the purposes of this study, is its semantic component. Words bearing a semantic relation to a particular color (e.g., lemon or grass) can interfere with subjects' ability to name the ink color in which the incongruent words are printed. In fact, there is a semantic gradient with regard to the magnitude of the interference that incongruent color words exert on the naming of the ink color: Color words themselves produce the most interference, followed by color-related nouns, frequent nouns, rare nouns, and pronounceable nonwords (Klein, 1964). This semantic gradient has been replicated and extended by several authors. For instance, it has been found that the amount of interference in color naming is a function of degree of color association for the irrelevant words (Scheibe, Shaver, & Carrier, 1967). This semantic gradient of interference, however, occurs in an inkcolor naming task, but not in a Stroop task variation that pairs two words (one distractor and one target) where the task is to read aloud

4 CROSS-MODAL STROOP-LIKE TASK 385 one of the words. Glaser and Glaser (1989) claim that the semantic gradient will be observed only when a semantic component (i.e., the color patch) is accessed or a semantic decision is made. In a reading task the word does not necessarily access its semantic code, because the task requires only articulatory information. A color patch, in contrast, necessarily accesses its semantic code. Hence, the semantic gradient is observed in color-naming tasks but not in word-reading tasks. The Stroop paradigm has also been used to study the processing and storing of semantic information of pictures and words (Glaser & Dingelhoff, 1984; Glaser & Glaser, 1989; La Heij, 1988; McClain, 1983; Smith & Magee, 1980). When subjects are asked to name either part of a stimulus composed of a picture and a written word, a picture distractor does not inhibit the subjects' ability to name (read) the word, whereas a word distractor does inhibit their ability to name the picture. However, when subjects are asked to categorize a stimulus, the Stroop effect is reversed: A picture distractor interferes with word categorizing, but picture categorizing is not inhibited by a word distractor (Smith & Magee, 1980). In the picture-word Stroop task, the distractor may be incongruent for a naming task (e.g., the word apple superimposed on a picture of a pear), in which case the picture-naming response would be inhibited but the word-naming response would not be inhibited. However, this same stimulus pair would be considered congruent for a categorization task, because both stimuli belong to the same semantic category. In this case, the word-categorization response would be facilitated, whereas the picture-categorization task would be unaffected. The idea is that the categorization task requires subjects to access the semantic code of a picture or a word, whereas the naming task requires them to access only a lexical code (Glaser & Glaser, 1989). Words are faster than pictures at accessing articulatory information, but pictures are faster than words at accessing semantic information (Smith & Magee, 1980). Furthermore, the source of the interference is assumed to be the information that is available from an already completed process (i.e., the articulatory information from the word or the semantic information from the picture). An important finding regarding the Stroop picture-word analog is the element of semantic interference, which has not yet been explored in a cross-modal situation. This phenomenon is studied in the present article for the first time.' La Heij (1988) found that in a picturenaming task, distractors that were both semantically related and members of the response set produced more interference than distractors that were only semantically related, and that these, in turn, produced

5 386 S'I'UAR'I' AND CARRASCO more interference than nonsemantically related distractors. These findings support results indicating that semantically related distractor words interfere with target stimuli at a semantic level, and not only because they are relevant to the task at hand (Glaser & Diingelhoff, 1984; Klein, 1964). Recently, La Heij, Happel, and Mulder (1990) have shown that semantic interference effects do not appear when, instead of a picture-naming task, a word-reading task is used. There are three hypotheses concerning the possible locations at which the interference between the processing of pictures and words may occur: 1. The perceptual stage hypothesis attributes the interference between pictures and words or between colors and words to the attraction of attention by the distractor, thereby reducing the processing capacity available for target encoding (Hock & Egeth, 1970). 2. The semantic decision stage hypothesis or conceptual encoding hypothesis, which issued from experimental studies that tried to isolate response-related effects from semantic-encoding and decision effects, states that the locus of interference is a point at which both components of the stimulus (i.e., both picture and word, color and word, or pitch and word) are being processed semantically (e.g., Glaser & Diingelhoff, 1984; Seymour, 1977; Shimamura, 1987). 3. The response competition hypothesis holds that reading (the irrelevant word of the stimulus pair) occurs more rapidly than color nam- ing, and therefore dominates and delays the naming response (e.g., Cowan & Barron, 1987; Dyer, 1973; Keele, 1972; Smith & Magee, 1980). Because the incongruent word is accessed prior to the color name, the result is competition for the articulatory code. This competition can be suppressed by requiring subjects to provide a manual response, as opposed to the usual verbal response, thus reducing the strength of the Stroop effect (e.g., Keele, 1972). Cowan and Barron (1987) support the response competition hypothesis and postulate the existence of a prespeech buffer from which the subjects select the correct response. Any articulatory information must enter this buffer before it can be used to produce a response. This hypothesis accounts for the Stroop effect in the following way: Unwanted items (distractors) enter this buffer, and a selection mechanism traces the origin of each response before a decision is reached. Hence, if the prespeech buffer is bypassed, as in a button-press response task, the Stroop effect will not occur. In fact, Cowan and Barron attribute the failure of cross-modal interference previously reported by Thackray and Jones (1971) and Dyer (1973) to the button-press response that those investigators had utilized. Cowan and Barron found that auditory color words interfered with visual words, but

6 CROSS-MODAL STROOP-LIKE TASK 387 auditory noncolor words and music did not. They proposed that the selection mechanism in the prespeech buffer processes the items in parallel, and that parallel processing causes interference only between phonetically or semantically related stimuli. Cowan and Barron's (1987) findings are central to the present article, but leave open the questions that are explored here: 1. The fact that cross-modal interference occurred suggests that one cannot selectively attend to one modality. This needs to be confirmed to examine the cross-modal Stroop effect further. 2. Their claim that they found cross-modal Stroop interference is not complete. The Stroop effect is the finding that there is an asymmetrical pattern of interference: Words interfere with the processing of colors but colors have no effect on the processing of words. Cowan and Barron did not examine the effect in both directions for the obvious reason that colors cannot be presented auditorily. The present study examined the cross-modal interference effects of the Strooplike picture-word effect for both naming and categorization tasks, to test whether such an asymmetry exists cross-modally. 3. Their finding that color words caused more interference than noncolor words indicates that there are different degrees of crossmodal interference. They correctly conclude, however, that their experiment provides no indication that this differential interference involves a semantic component, because their noncolor word condition was a repetition of the word the, which is not semantically related to colors, not relevant to the task, and has no meaning per se. Furthermore, their experiment does not satisfactorily address their hypothesis that stimuli from both modalities enter a common prespeech processing unit in which semantically related interference cannot be easily rejected. A more accurate test is needed to assess the role of a possible semantic component; the experiments reported here include interference of relevant semantic categories so as to explore the existence of a cross-modal semantic component. The present study focuses on the pattern of interference obtained when the target and distractor in a Stroop-like task are not within the same perceptual modality. The experiments involve two tasks, naming (Experiments 1 and 3) or categorizing (Experiments 2 and 3), with pictures and auditory words. The main question is whether a dual-modality presentation of the target and distractor will yield a pattern of semantic interference that parallels that obtained by Glaser and Diingelhoff (1984) within a single modality. Research in the area of cross-modal semantic interference of visual and auditory stimuli provides a method of studying the nature of semantic representation. Moreover, by examining the pattern of interference with picture pro-

7 388 STIUART AND CARRASCO cessing rather than color processing, a broader range of semantic categories can be tested. The discovery of a cross-modal semantic interference similar to that found by Glaser and Diingelhoff could indicate that regardless of differences in perceptual processes activated by different modalities, a common semantic code underlies both the visual and auditory modes of representation. EXPERIMENT 1 This experiment explored two issues: (a) Would the same pattern of interference found within a single modality in the Stroop pictureword task (Smith & Magee, 1980) be observed cross-modally? If that were the case, one would expect an asymmetrical effect: The auditory words should interfere with visual picture naming, but pictures should have no interference effect on naming auditory words; (b) Is there a semantic component to this cross-modal interference, as there is within the visual modality? This experiment measured each subject's reaction times (RTs) to naming pictures while hearing words and to naming auditory words while seeing pictures. To examine the semantic component of this task, we used three types of interfering words or pictures: categorycongruent (distractor and target belonged to same semantic category), miscellaneous (distractors chosen from three semantic categories unrelated to the targets), and a control condition (distractor was noise: either a block of X's or white noise). If the Stroop-like effect occurs cross-modally for pictures and words, and if there is a semantic component to this asymmetrical interference, the following results would be expected: (a) Significant differences would be found among the interfering effects of category-congruent, miscellaneous, and control word distractors on the naming of pictures; the category-congruent words should have the greatest interference effect, whereas the control words should have a minimal effect; and (b) no difference would be found between the category-congruent and miscellaneous picture distractors on the naming of auditory words, but the control condition would produce less interference. METHOD Subjects Sixteen students from the introductory psychology class at Wesleyan University participated to fulfill a course requirement. All had normal or corrected-to-normal vision, and were naive as to the purpose of the experiment.

8 CROSS-MODAL STROOP-LIKE TASK Apparatus and stimuli A Macintosh IIX was used to present the stimuli and record RTs. To use the exact pictures from previous studies, pictures were scanned with a Thunderscan. Spoken words (a male voice) were digitized for auditory presentation using MacRecorder, Soundeditor, and Hypersound. Stimuli were selected from Snodgrass and Vanderwart (1980). Two categories were used-fruits and tools. From each category, we chose three targets that were neither the three most exemplary nor the three least exemplary items. The pictures of these exemplars were chosen for having high name agreement, image agreement, and familiarity ratings, but low complexity ratings (see Table 1). Each target had at least a 98% name agreement and, on a scale from 1 to 5, image agreement ratings of 3.98 or higher, familiarity ratings of over 3.25, and complexity ratings of below These values were controlled for to assure that the stimuli would be readily processed. The examplars from the fruit category were banana, lemon, and pear; the exemplars from the tool category were ruler, screwdriver, and nail. All the distractors were selected following the aforementioned criteria as closely as possible. The category-congruent distractors were selected from the remaining exemplars in each category. For the fruit category, the distractors were apple, orange, and peach. For the tool category, the distractors were hammer, chisel, and saw. The miscellaneous distractors were selected from six remaining categories. The miscellaneous distractors paired with the fruit targets were drum, sock, andfork; those paired with the tool targets were cat, arm, and bus. The pictures were approximately 10 x 10 cm subtending a visual angle of 10? x 10?. The control condition consisted of a block of X's (10 x 10 cm) when auditory words were targets, and white noise when pictures were targets. Procedure 389 Subjects were tested individually. The experiment took place in a dark room. Subjects were seated 57 cm away from the computer monitor. The subjects dark adapted while they listened to the instructions. They were told that they would listen to words or noise while they watched pictures on the computer monitor and that they would name either the word or the picture as fast and as accurately as they could, striking the enter key immediately after articulating their answer. This advanced the program to the next stimulus pair. The computer presented the word and the picture simultaneously. Subjects were told to keep their eyes on the fixation point (x), which appeared in the center of the screen between pictures for 500 ms. They were also told to allow the entire word to be said by the computer before naming the target, the word, or the picture. Otherwise, one of two situations could have precluded a possible interference effect: (a) When the word was the target, subjects could have begun to name the word as soon as they heard the initial phoneme, thus preventing processing of the visual distractor; (b) when the picture was the target, they could have masked the auditory

9 390 STUART AND CARRASCO Table 1. Mean name agreement (NA), image agreement (IA), familiarity (F), and complexity (C) ratings for each of the target stimuli, categorycongruent distractors, and miscellaneous distractors Stimuli NA(%) IA F C Target stimuli Lemon Pear Banana Nail Ruler Screwdriver Category-congruent distractors Apple Orange Peach Saw Chisel Hammer Miscellaneous distractors Fork Drum Sock Bus Arm Cat Note. After Snodgrass and Vanderwart (1980). Copyright 1980 by the American Psychological Association. Adapted by permission. distractor word with their response, precluding the processing of the auditory distractor. An unconnected apparatus was set on the computer monitor to appear as if the subjects' responses were being acoustically recorded; moreover, an experimenter was present and monitored every session. Six types of experimental blocks were presented: the fruit or tool category paired with category-congruent distractors, miscellaneous distractors, or the control stimuli. Each block consisted of 27 presentations of the target pictures or words from either the fruit or tool category. The targets were presented in random order nine times each. The two experimental tasks were picture naming and word naming. There were 4 practice blocks and 48 experimental blocks. All subjects performed 2 practice blocks before each type of task. The order of presentation of the 6 types of experimental blocks was randomized. In each half of the experiment, 4 sets of 6 blocks were presented. The task order was counterbalanced so that half of the subjects named words in the first half of the experiment

10 CROSS-MODAL STROOP-LIKE TASK 391 and pictures in the second half; this order was reversed for the other subjects. There was a 5-min break between tasks. The RTs for each block were recorded by the computer, and errors were recorded manually by the experimenter. Even though a voice key has been shown to be a precise device for recording RT to individual stimulus pairs, this experiment was more akin to a conventional Stroop test in which a global RT is recorded for a block of stimulus pairs. Global RT has also been measured for picture-word Stroop-like tasks (e.g., Smith & Magee, 1980). Furthermore, Smith and Magee replicated these Stroop-like results when they presented slides of stimuli with a tachistoscope and recorded the RT to individual stimulus pairs with a voice-key mechanism. RESULTS AND DISCUSSION Mean RTs were obtained for each of the 12 conditions for each subject. Approximate mean time for naming a single item may be calculated by dividing the total RT by 27 (number of trials per block). Table 2 shows the mean RTs across both categories of targets because there were no significant differences between categories. A 2 (Category: Fruit vs. Tools) x 2 (Target: Picture vs. Word) x 3 (Interference: Congruent vs. Miscellaneous vs. Control) within-subjects design anal- ysis of variance (ANOVA) was performed on the mean RT for each condition. There was a significant three-way interaction among category, target, and interference, F(2, 30) = 9.40, p <.001. The interaction between target and category was significant, F(1, 15) = 28.7, p <.001; this was expected because of the difference in lengths of the words, ease of articulation, or both. Relevant for the present hypotheses is that there was a significant two-way interaction between the target and interference variables, F(2, 30) = 5.0, p <.02, and that there was no interaction between category and interference, F(2, 30) =.74. A simple effects analysis showed that the RTs to word and picture targets were significantly different at all types of interference (p <.02), and that interference overall had a significantly different effect on words and pictures (p <.001).2 A Tukey HSD post hoc test showed that there were significant differences between all naming conditions (p <.05) except between the word-naming conditions with category-congruent and miscellaneous interference (p >.10). This test also showed that RTs to each target were significantly different from one another at each level of interference (p <.01). An ANOVA of the number of errors on each block showed that neither the interactions nor the main effects were significant (p >.2); the range of mean error rate on each block was only 0.2% to 1.63%.

11 392 STUART AND CARRASCO Table 2. Mean response time (ms per item) for naming and categorizing target words and pictures as a function of type of distractor: Congruent (category congruent), Miscellaneous (category incongruent), and Noise (control) Distractor Task Congruent Miscellaneous Noise Experiment 1 Word naming Picture naming Experiment 2 Word categorizing Picture categorizing Experiment 3 Word naming Picture naming Word categorizing Picture categorizing In sum, Experiment 1 found an asymmetric cross-modal Strooplike effect: Subjects' RTs to name pictures were significantly slower when word distractors were category-congruent compared with when word distractors were category-incongruent, but there was no difference between RTs for naming words with category-congruent and category-incongruent picture interference. These results indicate that there is a cross-modal Stroop effect with picture-word processing analogous to that observed with color-word processing (Cowan & Barron, 1987). In a naming task, moreover, auditory word interference does have a semantic component, whereas visual pictorial interference does not. These results support previous findings of the semantic gradient effects of words on picture processing (Glaser & Diingelhoff, 1984; La Heij, 1988); distractor words of the same semantic category as the target picture cause more interference than miscellaneous words in the picture-naming task. However, when Glaser and Glaser (1989) tested the interference effect of distractor words on target words, they found no semantic gradient, indicating that a word-naming task does not require access to semantic information. The present results confirm that conclusion in that there was not a significant difference between the interference effects of category-congruent and miscellaneous pictures on the word-naming task. The conclusion is that

12 CROSS-MODAL STROOP-LIKE TASK 393 pictorial accessing of different semantic information had no bearing on the processing of auditory words in a naming task. A complete immunity of words to pictorial interference has been reported. Smith and Magee (1980) found no significant difference between subjects' RTs for reading a word alone and reading a word superimposed on an incongruent picture. The control in these tasks was naming the picture or word alone with no distracting word or picture. In the present experiment, nonetheless, both categorycongruent and miscellaneous distractors interfered with naming both auditory and visual targets more than the control condition did. The control conditions, where the distractors were white noise or a matrix of X's, were expected to elicit faster RTs than those of the miscellaneous conditions. In the former, visual and auditory distractors were repetitive and meaningless nonverbal material; in the latter, one of three meaningful pictures or auditory words was presented during each trial, which are exemplars of verbal material (i.e., words). EXPERIMENT 2 This experiment investigated whether a reverse Stroop effect in a cross-modal test is present when the task is to categorize words or pictures, thus furthering Smith and Magee's (1980) findings of a reverse Stroop effect within a single modality. Given that Experiment 1 suggested the existence of a cross-modal semantic component and that the Stroop semantic asymmetry was obtained cross-modally for the naming task, it was expected that the categorization task would yield the reverse asymmetry. Because pictures are known to have more rapid access to the semantic code than the words naming the pictures (Smith & Magee, 1980), the miscellaneous picture should interfere with the word's accessing its semantic code. The miscellaneous au- ditory word distractor, however, should have no effect on categorizing the visual picture. Furthermore, given that a categorization task, unlike a naming task, requires the subject to access semantic information that is common to both the target and distractor, subjects' RTs should be faster when the picture is semantically related to the word than when it is unrelated. In fact, when both the picture and the word access the same category set, they could be considered as a congruent pair. This experiment, like Experiment 1, required subjects to make a decision based on words or pictures while perceiving three types of distracting pictures or auditory words: category-congruent, miscel-

13 394 STUART AND CARRASCO laneous, or control. The expected outcome was that category-congruent and miscellaneous auditory word distractors would have no differential effect on the categorization of pictures; on the other hand, miscellaneous pictures were expected to interfere more than categorycongruent pictures on the categorizing of words. METHOD Subjects Sixteen students from the introductory psychology class at Wesleyan University participated to fulfill a course requirement. All had normal or corrected-to-normal vision. They had not participated in Experiment 1 and were naive as to the purpose of the experiment. Apparatus and stimuli The same apparatus and stimuli were used as in Experiment 1. The target stimuli set, however, consisted of five presentations of each of the three tools and four presentations of each of the three fruits when the target was a picture or word from the tool category. Conversely, five presentations of each of the three fruits and four presentations of each of the three tools made up the target stimuli when the picture or word target was a fruit. This was done so that each block still included 27 stimuli, but the response elicited by the stimuli was not the same for every trial in the same block. There were three types of distractors. The control for each target and the miscellaneous distractors were the same as those of Experiment 1. In this experiment, it was not desirable that all trials in the same block be category-congruent, as they were in the naming task. Had this been the case, every stimulus would have elicited a yes response because the subjects' task was to say if each target belonged to a given category. Therefore, the distractors of the category-congruent condition were congruent with the targets only 56% of the time (e.g., 15 of the 27 trials in a fruit block), and 44% of the time the distractors were from the other category (e.g., tools). Procedure The procedure was like that of Experiment 1, except that the subjects' task was to make a categorization decision. Subjects were told that they would be listening to words or noise while they watched pictures on the monitor, and that they would categorize either the word or the picture. They were instructed to say yes if the picture or word belonged to the category that was announced at the beginning of each block (e.g., yes if the picture is a fruit) and no if it did not, as fast and as accurately as they could. As in Experiment 1, subjects had to wait for the entire word to be said by the computer before categorizing the words or pictures. All subjects were given two practice blocks before each type of task, picture categorizing and word categorizing.

14 CROSS-MODAL STROOP-LIKE RESULTS AND DISCUSSION IASK 395 Mean RTs were obtained for each of the 12 conditions for each subject. Approximate mean time for naming a single item was calculated by dividing the total RT by 27 (number of trials per block). Because there were no significant differences between categories, Table 2 shows the mean RTs across both categories. As in Experiment 1, a 2 (Category: Fruit vs. Tools) x 2 (Target: Word vs. Pictures) x 3 (Interference: Congruent vs. Miscellaneous vs. Control) withindesign ANOVA was performed on the mean RTs for each type of block. There was no three-way interaction among the category, target, and interference, F(2, 30) = 1.88, p >.1. There was neither a twoway interaction between category and target, F(1, 15) =.21, nor between category and interference, F(2, 30) = 2.09, p >.1. But the relevant interaction for the present experiment, between the target and interference, was significant, F(2, 30) = 8.63, p <.001. A simple effects analysis revealed that the RTs to words and pictures were significantly different at each type of interference (p <.001), and that there was a significantly different effect of the three different types of interference at both pictures and words (p <.001). A Tukey HSD post hoc test revealed that all categorizing conditions had significantly different RTs (p <.05) except for category-congruent and miscellaneous interference for pictures (p >.05). The range of error rate on each block was 0.2% to 2.93%. As in Experiment 1, an ANOVA of the number of errors showed that the differences between conditions were not significant (p >.2). In sum, Experiment 2 found that there was no difference between the subjects' RTs for categorizing pictures with category-congruent and miscellaneous auditory word interference. On the other hand, subjects' RTs for categorizing auditory words were significantly faster when distractors were category-congruent compared with when distractors were miscellaneous pictures. Unlike what occurred in the naming task, in the categorization task, category-congruent interfer- ence may have acted in a "facilitatory" fashion, because both the distractor and the target would elicit the same category response. Actually, given that for both targets, the control condition (white noise or a matrix of X's) elicited significantly faster RTs than the miscellaneous or category-congruent interference conditions, it may be more reasonable to refer to facilitation in the category-congruent condition as reduced interference. Because previous experiments (e.g., Dalrymple-Alford, 1972; Sichel & Chandler, 1969) have shown that it may not be possible to speed up the already rapid response elicited

15 396 STUART AND CARRASCO by the control condition, the RTs in the category-congruent condition were expected to be faster than those in the miscellaneous condition but not faster than in the control condition. The extent of apparent facilitation may be a function of the choice of control condition (MacLeod, 1991). There was no significant difference between the picture categorization RTs with congruent and miscellaneous interference. This indicates that although auditory words can interfere with pictures in a naming task, they seem to have no effect on the categorizing of pictures. This result supports the idea that words, regardless of their modality of presentation, lack rapid access to their semantic code, whereas pictures have rapid access to their semantic code (e.g., Glaser & Glaser, 1989; Smith & Magee, 1980). Therefore, pictures that quickly access incongruent semantic information will interfere with the word's ability to access its relevant semantic information. This finding of an interference effect for word targets but not for picture targets in a categorizing task is a reverse of the asymmetry that occurred in a naming task (Experiment 1)--a difference in the interference effect for picture targets but not for word targets. These results indicate that the pattern of interference found by Smith and Magee (1980) within the visual modality also occurs cross-modally. A between-design three-way ANOVA was performed as a post hoc analysis of Experiments 1 and 2: [2 (Task: Naming vs. Categorizing) x 2 (Target: Picture vs. Word) x 3 (Interference: Congruent vs. Miscellaneous vs. Control)]. The three-way interaction was significant, F(2, 60) = 7.06, p <.005, as were all two-way interactions: task and target, F(1, 30) = 19.88, p <.001; task and interference, F(2, 60) = 3.55, p <.05; and target and interference, F(2, 60) = 7.92, p <.001. A simple effects analysis of the task and target interaction revealed that RTs for categorizing pictures were significantly faster than for naming pictures (p <.001), but there was no difference in RTs between categorizing and naming words (p >.1). Overall, RTs were significantly faster for pictures than for words (naming, p <.005; categorizing, p <.001). The other two-way interactions also showed that the type of interference was significantly different for both targets (words, p <.001; pictures, p <.001), as well as for both tasks (naming, p <.001; categorizing, p <.001). EXPERIMENT 3 Experiment 3 was carried out using a within-subject experimental design to confirm the findings of Experiments 1 and 2 as well as those

16 CROSS-MODAL STROOP-LIKE TASK 397 of the post hoc analysis. That is, the same subjects both named and categorized pictures and words. It was expected that the patterns of interference observed in the previous experiments would be replicated: The pattern of interference would depend on the task, and the semantic component would play a role accordingly. METHOD Subjects Thirty-one students from the introductory psychology class at Wesleyan University participated to fulfill a course requirement. All had normal or corrected-to-normal vision. They had not participated in the previous ex- periments and were naive as to the purpose of this study. Apparatus and stimuli The same apparatus and stimuli were used as in Experiments 1 and 2. Procedure The procedures of Experiments 1 and 2 were combined in this experiment. Subjects performed only 2 blocks of trials in each condition instead of 4 because they were tested in both naming and categorizing tasks in one experimental session (1 hr). The order for both experimental tasks, naming and categorizing, as well as the order for targets, words and pictures, was counterbalanced. Each subject performed 48 blocks (as in the previous experiments). RESULTS AND DISCUSSION Mean RTs were obtained for each of the 24 conditions for each subject. Approximate mean time for naming or categorizing a single item was calculated by dividing the total RT by 27 (number of trials per block). Because there were no significant differences between categories, Table 2 shows the mean RTs across both categories. A 2 (Task: Naming vs. Categorizing) x 2 (Target: Word vs. Picture) x 2 (Category: Fruit vs. Tools) x 3 (Interference: Congruent vs. Miscellaneous vs. Control) within-design ANOVA was performed on the mean RT for each type of block. The four-way interaction among task, target, category, and interference was not significant, F(2, 60) = 1.18, p >.3, nor were the following three-way interactions: target, category, and interference, F(2, 60) = 1.26, p >.2; task, category, and interference, F(2, 60) = 2.99, p >.05; and task, target, and category, F(1, 30) =.22. However, the predicted three-way interaction among task, target, and interfer-

17 398 STUART AND CARRASCO ence was significant, F(2, 60) = 6.80, p <.005, and all its two-way interactions were significant as well: task and target, F(1, 30) = 94.16, p <.001; task and interference, F(2, 60) = 4.57, p <.02; and target and interference, F(2, 60) = 7.03, p <.005. A simple effects analysis of the task and target interaction revealed that subjects had significantly faster RTs for naming words than for categorizing words (p <.001). But here was only a marginally significant difference between naming pictures and categorizing pictures (p <.10). On the other hand, there was a significant difference between categorizing words and categorizing pictures (p <.001), but not between naming words and naming pictures (p >.2). The other two-way interactions also showed that type of interference was significantly different for both targets (words, p <.001; pictures, p <.001), as well as for both tasks (naming, p <.001, categorizing, p <.001). A Newman-Keuls post hoc test of the three-way interaction between task, target, and interference revealed that there was a significantly different effect between naming pictures with congruent auditory interference and with miscellaneous auditory interference (p <.01); however, there was no significant difference between naming words with congruent pictorial interference and with miscellaneous pictorial interference (p >.05). This post hoc test also showed that there was a significant difference between categorizing words with congruent interference and with miscellaneous interference (p <.05); however, there was no difference between categorizing pictures with congruent interference and with miscellaneous interference (p >.05). There were significant differences for all the tasks (picture naming, word naming, picture categorizing, and word categorizing) between noise interference and the other two types of interference (p <.05). As in the previous experiments, an ANOVA of the number of errors showed that the differences between conditions were not significant (p >.2); the range of mean error rate was from 0.12% to 1.78%. In sum, the cross-modal Stroop effect and the asymmetrical semantic effects found in Experiments 1 and 2 were replicated in Experiment 3: Subjects' RTs for pictures (but not words) were slower in a naming task with category-congruent auditory distractors than with miscellaneous auditory distractors, whereas RTs for words (but not pictures) were slower in a categorization task with miscellaneous visual distractors than with category-congruent visual distractors. This confirms that there is a semantic component to the interference that occurs cross-modally when one is processing pictures and auditory words.

18 CROSS-MODAL STROOP-LIKE TASK GENERAL DISCUSSION 399 The most important finding of these three experiments is that a cross-modal Stroop-like task yields results that parallel the pattern of semantic interference that Glaser and Dingelhoff(1984) observed in a single-modality test. Experiment 1 revealed that the semantic category of picture distractors did not differentially affect the interference on auditory word naming, whereas the semantic category of auditory word distractors did differentially affect picture naming; category-congruent stimuli interfered more than miscellaneous stimuli did. Experiment 2 reversed this effect: In a categorization task, the semantic category of word distractors did not differentially affect picture categorizing, whereas word categorizing was affected by the semantic category of the pictorial interference; miscellaneous stimuli interfered more than category-congruent stimuli did. Experiment 3 confirmed these findings with a within-subject experimental design. The magnitude of the effects is comparable to those reported by Cowan and Barron (1987) and by Cowan (1989a). A number of issues not addressed by Cowan and Barron (1987) were explored here. First, there was the question of whether selective attention to one modality was possible in a cross-modal Stroop task. The interference effects that both pictures and auditory words produced in this study suggest that selective attention to one modality was not possible. That the control-interference conditions (i.e., a repeated block of X's or white noise) yielded significantly faster RTs than the congruent and miscellaneous interference conditions indicates that the latter types of verbal distractors interfere cross-modally, thus preventing selective attention. The second issue was whether the full extent of the Stroop effect could be obtained cross-modally. Recently, controversy has arisen regarding the existence and nature of the cross-modal Stroop effect. Miles et al. (1989) failed to replicate Cowan and Barron's (1987) findings, and they dispute the existence of cross-modal interference by auditory words on color naming. Cowan (1989b) replied, however, that this failure to replicate his original results can be attributed to some methodological flaws: The rate of word-interference presentation was slower than that of the original study, and given that the cross-modal effect is much smaller than the conventional Stroop effect, a valid failure to replicate the effect would require a larger sample of subjects (they had tested 12 subjects in one experiment and 8 in another). In a second exchange between these authors, Miles and Jones (1989) acknowledged that there were some methodological dif-

19 400 STUAR'T AND CARRASCO ferences between Miles et al. (1989) and Cowan and Barron (1987). In another experiment, however, in which they minimized differences with Cowan and Barron's procedure, Miles and Jones obtained results that were consistent with Miles et al. (1989) but not with Cowan and Barron (1987). Cowan (1989a) subsequently questioned how similar the procedures really were and briefly described another experiment in which a cross-modal Stroop effect was found. In any event, in their original experiment, Cowan and Barron (1987) tested only the effect of auditory distractor words on naming of visual stimuli (colors). The present study allowed the testing of the crossmodal effect in both directions: auditory stimuli on visual naming and categorizing, as well as visual stimuli on auditory word naming and categorizing. The findings of the three experiments reported here indicate that the full conventional Stroop effect can be replicated cross-modally. That is, the pattern of interference was determined by the task: Auditory distractors interfered with subjects' RTs at naming pictures, but did not interfere with subjects' RTs at categorizing pictures, whereas picture distractors did not interfere with subjects' RTs at naming words, but did interfere with subjects' RTs at categorizing words. The final question not addressed by the Cowan and Barron (1987) study, or to our knowledge by any other study, was whether the crossmodal Stroop effect involved a semantic component. As discussed above, the pattern of semantic interference found in the cross-modal Stroop task is the same as that found in the color-word Stroop test (Klein, 1964) and in the picture-word Stroop test (Glaser & Diingelhoff, 1984). Also, the existence of a semantic component demonstrates that the cross-modal interference cannot be fully attributed to perceptual distraction. Furthermore, the semantic component suggests that regardless of perceptual modality (visual vs. auditory) and regardless of representational modality (pictorial vs. verbal), the semantic information connected to a concept is stored in a common semantic code. This semantic component also suggests that some coding effects in memory are not specific to a particular modality. Hypotheses regarding the locus of the Stroop interference were outlined in the introduction to this article. The perceptual stage hypothesis can be ruled out because it fails to account for the semantic component of the interference effect reported previously (e.g., Klein, 1964; Glaser & Diingelhoff, 1984) and confirmed by the present study or for the cross-modal effect previously reported by Cowan and Barron (1987) and extended here. The two loci that have received the most attention are the response selection (output) stage and the semantic decision (input) stage. The response selection explanation relies on

20 CROSS-MODAL STROOP-LIKE TASK 401 the notion that a conflicting stimulus dimension, which is processed more quickly, will interfere with the stimulus dimension which is processed more slowly, and therefore there is competition for response activation at the output stage. The source of interference in the response competition hypothesis is the different time courses associated with word and pictorial processing depending on the experimental task (i.e., word naming occurs more quickly than picture naming and therefore interferes with the output of the picture name). The semantic decision stage hypothesis, also known as the conceptual encoding hypothesis, states that interference occurs prior to output, at an encoding stage during which both components of the stimulus (i.e., both picture and word, color and word, or pitch and word) are being processed semantically. Recent research and the results of this study support this hypothesis (Glaser & Diingelhoff, 1984; Mayor, Sainz, & Gonzalez-Marques, 1988; Seymour, 1977; Shimamura, 1987). The failure of a stimulus-onset-asynchrony (SOA) to eliminate the Stroop effect has been considered as evidence against the response selection hypothesis and as support for the conceptual encoding hypothesis. According to the former hypothesis, if a picture were presented at various SOAs prior to the word, the response conflict would be eliminated because the picture would have prior access to its articulatory code. Nonetheless, even when a picture was presented at various SOAs prior to the word, the same pattern of interference occurred (Glaser & Diingelhoff, 1984). That is, the words were still able to interfere with picture naming; the conflict must therefore occur at some point prior to the response selection. Mayor et al. (1988) examined the time course of interference between pictures and words. Although, as SOAs increased, the magnitude of interference between pictures and words in naming and categorization tasks decreased, the semantic relationship between the target and distractor elicited the same pattern of interference that has been found between pictures and words (i.e., Smith & Magee, 1980). The results of the experiments presented here are similar to the findings of Mayor et al. That is, a variation in the perceptual modality of the target and distractor altered the ability of words and pictures to interfere with one another in the naming and categorizing tasks: When the modalities of the target and distractor were reversed (auditory word vs. picture), semantic interference was still observed, albeit in the alternative type of task (categorizing or naming) and with an inverse ordering of RTs in congruent and miscellaneous conditions. The results of this study support the hypothesis that the locus of Stroop interference is the encoding stage during which the target and distractor are semantically processed. Furthermore, the results also

21 402 STUAR'I AND CARRASCO lend support to the proposition by Cowan and Barron (1987), which they had not explicitly tested, that the selection mechanism processes the items in parallel and that parallel processing causes interference between semantically related stimuli. It should be pointed out, however, that both semantic decision and response competition may coexist. La Heij (1988; La Heij et al., 1990) has recently proposed that semantic effects may be located at the response-selection stage. The implications of the present data are not limited to the issue of the locus of interference; they also support a naming versus categorizing model recently proposed by Glaser and Glaser (1989). They tested modally pure stimuli (i.e., color-color, word-word, and picturepicture) that were necessarily presented with an SOA, instead of the usual modally mixed color-word or picture-word stimuli. The sequential discrimination task showed that subjects' reading process can be disrupted by a word distractor and that a color distractor can disrupt the color-naming response. Glaser and Diingelhoff (1984) had previously tested picture-word interference and found that picture naming was disrupted by a distracting word but that the reading response was not inhibited by a picture. The processing of a word was disrupted by the picture if the semantically related categorization task was given, even though picture processing was not inhibited by the word. Glaser and Glaser (1989) incorporated these findings, as well as other previous results (Glaser & Diingelhoff, 1984; Smith & Magee, 1980), into a network model of human memory. They modified the model originally proposed by Collins and Loftus (1975) to explain the pattern of semantic interference observed in the Stroop tests. According to their model, Stroop interference occurs if the distractor activates nodes or links that are closely connected with the target pathway. This activation can either help or hinder subjects' performances. It is the type of task that determines whether there is inhibition or facilitation. Words will not stimulate a concept node very rapidly, or at least not as rapidly as will a picture or a physical stimulus, because the conceptual activation plays no role in reading the word aloud. When a word categorization task is required, the semantic memory node would need to be activated, and an incongruent picture would disrupt this task. The findings of this study confirm that pictures and auditory words, regardless of being processed via different modalities, can interfere with each other in a differential way determined by the task, naming or categorizing. Furthermore, a semantic component was found in these cross-modal experiments. This finding indicates that interference cannot be fully attributed only to perceptual factors and that a

22 CROSS-MODAL STROOP-LIKE TASK 403 common semantic code underlies both the visual and auditory modes of representation. Notes We thank John Seamon, James D. Fernandez, Nelson Cowan, and an anonymous reviewer for their helpful comments on an earlier version of this manuscript, and Svetlana Katz and Thomas Savel for their help at different stages of this study. This project was supported by a Faculty Project Grant of Wesleyan University. Correspondence concerning this article should be addressed to Marisa Carrasco, Department of Psychology, Wesleyan University, Middletown, CT ( MCARRASCO@WESLEYAN.BITNET). Received for publication August 5, 1991; revision received February 17, Schriefers, Meyer, and Levelt (1990), in their study of time course on lexical access in language production, found an interference effect of semantically related words on picture-naming latencies. That study, however, differs from the present study in that it included neither the effect of pictures on words nor a categorization task; hence, the asymmetrical pattern of interference and the reverse Stroop effect were not explored. 2. Pilot results showed that for a naming-picture task with continuous auditory word interference, there was a greater difference between the category-congruent and miscellaneous conditions than in this experiment. The purpose of using the paired stimuli presentation was to measure not only the effect of auditory words on the processing of pictures but also the effect of pictures on the processing of auditory words. This asymmetrical effect has not been studied prior to this experiment, and the method of presentation may account for the smaller effect. References Cowan, N. (1989a). The reality of cross-modal Stroop effects. Perception & Psychophysics, 45, Cowan, N. (1989b). A reply to Miles, Madden, and Jones (1989): Mistakes and other flaws in the challenge to the cross-modal Stroop effect. Perception & Psychophysics, 45, Cowan, N., & Barron, A. (1987). Cross-modal, auditory-visual Stroop inter- ference and possible implications for speech memory. Perception & Psychophysics, 41, Dalrymple-Alford, E. C. (1972). Associative facilitation and interference in the Stroop color-word task. Perception & Psychophysics, 11, Dyer, F. N. (1973). The Stroop phenomenon and its use in the study of perceptual, cognitive, and response processes. Memory & Cognition, 1, Glaser, W. R., & Dungelhoff, F. J. (1984). The time course of picture-word interference. Journal of Experimental Psychology: Human Perception and Performance, 10,

23 404 STUART AND CARRASCO Glaser, W. R., & Glaser, M. 0. (1989). Context effects in Stroop-like word and picture processing. Journal of Experimental Psychology: General, 118, Hock, H. S., & Egeth, H. (1970). Verbal interference with encoding in a perceptual classification task. Journal of Experimental Psychology, 83, Keele, S. (1972). Attention demands of memory retrieval. Journal of Experimental Psychology, 93, Klein, G. S. (1964). Semantic power measured through the interference of words with color naming. American Journal of Psychology, 77, La Heij, W. (1988). Components of Stroop-like interference in picturenaming. Memory & Cognition, 16, La Heij, W., Happel, B., & Mulder, M. (1990). Components of Stroop-like interference in word reading. Acta Psychologica, 73, MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109, Mayor, J., Sainz, J., & Gonzalez-Marques, J. (1988). Stroop and priming effects in naming and categorizing tasks using pictures and words. In M. Denis, J. Engelkamp, & J. T. E. Richarson (Eds.), Cognitive and neuropsychological approaches to mental imagery (pp ). Dordrecht, The Netherlands: Martinus Nijhoff. McClain, L. (1983). Stimulus-response compatibility affects auditory Stroop interference. Perception & Psychophysics, 33, Miles, C., & Jones, D. M. (1989). The fallacy of the cross-modal Stroop effect: A rejoinder to Cowan (1989). Perceptioin & Psychophysics, 45, Miles, C., Madden, C., & Jones, D. M. (1989). Cross-modal, auditory-visual Stroop interference: A reply to Cowan and Barron (1987). Perceptioln & Psychophysics, 45, Scheibe, K. E., Shaver, P. R., & Carrier, S. C. (1967). Color association values and response interference on variants of the Stroop test. Acta Psychologica, 26, Schriefers, H., Meyer, A. S., & Levelt, W. J. M. (1990). Exploring the time course of lexical access in language production: Picture-word interference studies. Journal of Memory alnd Language, 29, Seymour, P. H. K. (1977). Conceptual encoding and the locus of the Stroop effect. Quarterly Journal of Experimental Psychology, 29, Shimamura, A. P. (1987). Word comprehension and naming: An analysis of English and Japanese orthographies. American Journal of Psychology, 100, Shor, R. E. (1975). An auditory analog of the Stroop test. Journal of General Psychology, 93, Sichel, J. L., & Chandler, K. A. (1969). The color-word interference test: The effects of varied color-word combinations upon verbal response latency. Journal of Psychology, 72, Smith, M. C., & Magee, L. E. (1980). Tracing the time course of picture-