- Email: [email protected]
Right-hemisphere interactions in picture-word processing
BRAIN
AND
COGNITION
4, 273-286 (1985)
Right-Hemisphere
Interactions in Picture-Word
Processing
GEOFFREY UNDERWOOD AND ALISON WHITFIELD University
of Nottingham
The processing of pictures was investigated in three experiments which eliminated the response effects involved in naming. When a categorization task was used, clear advantages in response latency and accuracy were observed for left visualfield (LVF) presentations. This was in contrast to previous investigations which have used a naming task and which have reported right visual-field (RVF) advantages. When a distracting word was added to the display, the pattern of influence also changed from that reported previously. The use of naming tasks has indicated predominantly left-hemisphere effects, with demonstrations of interactions between pictures and words in the RVF. With a categorization task in Experiment 2, however, the only effective words were those related in meaning to the picture, and only when they were projected to the right hemisphere. The third experiment confirmed the LVF advantage for picture processing with masked displays, but found no reliable asymmetry with unmasked presentations. The pattern of semantic facilitation was also confirmed with the masked displays, but when the mask was removed an inhibition effect replaced the facilitation effect. These effects are interpreted as indicating that picture recognition is localized within the right cerebral hemisphere. It is suggested that the facilitating effect of related words is restricted to the left hemisphere because it is an effect upon recognition processes, whereas the inhibition effect reflects response competition. It is also suggested that previous reports of left-hemisphere interference effects are due to effects of response competition in naming tasks. Q 1985 Academic Press, Inc.
How is the processing of pictures laterialized, and how does the presence of a printed word affect the normal pattern of lateralization? These questions were addressed by three experiments, in an attempt to separate the influences of the verbal and nonverbal components in a pictorial version of the Stroop effect. This research was supported by Medical Research Council Grant G8127736N. We are also grateful to E. A. Maylor and two anonymous reviewers for their comments on an earlier draft of this report. The subjects used in Experiment 3 participated with the cooperation of the Headmistress and Staff of Chilwell Comprehensive School, Nottingham. We are very grateful to them and the participants for their help. Correspondence and requests for reprints should be addressed to Geoffrey Underwood, Department of Psychology, University of Nottingham, NG7 2RD Nottingham, England. 273 0278-2626185 $3 .OO Copyright 0 1985 by Academic Press, Inc. All rights of reproduction in any form reserved.
274
UNDERWOOD
AND
WHITFIELD
Lupker and Sanders (1982) have recently suggested that “the ability of the right hemisphere to process pictorial stimuli may be nearly equivalent to that of the left hemisphere” and that “these stimuli can be processed equally rapidly when presented in either visual field” (p. 383). An earlier report of an experiment using picture naming supports this conclusion with no difference in the naming latency for pictures in the left or right visual field (Underwood, 1976). However, a report by Wuillemin, Krane, and Richardson (1982) did find that pictures projected to the left hemisphere were associated with faster naming responses than those projected to the right hemisphere. Lupker and Sanders (1982) also found a small but statistically unreliable advantage in the naming of pictures projected to the left hemisphere. These authors concluded that the weak and inconsistent appearance of a left-hemisphere advantage is an indication of an equivalence between the hemispheres in the processing of pictures. If pictures can be considered as spatial representations, then the conclusion is surprising, for visuo-spatial processing has gained a righthemisphere advantage in a number of tasks. Left visual-field (LVF) advantages have been reported for dot localization (Kimura, 1969), the categorization of drawings of figures and people (Witelson, 1977), and face identification (Marcel & Rajan, 1975; Pirozzolo & Rayner, 1979). The conflict is not difficult to resolve, however, for whereas the experiments using pictures employed a vocal naming response, these other experiments employed either vocal categorization (“same/different”) or manual pointing responses. If the generation of picture names is a process restricted to the left cerebral hemisphere, then a right visual-field (RVF) advantage would be expected. In Experiment 1, therefore, the input processing of pictures is separated from the output demands of naming by employing a vocal categorization task. Subjects were instructed to say “yes” or “no” to indicate whether or not each picture was an animal. The purpose of gaining evidence of the recognition of pictorial stimuli was to help in the understanding of interactions between words and pictures in a modified Stroop task. When words are presented at the same time as pictures, the responses to the pictures vary as a function of the semantic relationship between the two stimuli. This effect can be demonstrated with picture-naming responses when the word is presented in the opposite visual field to the picture (Underwood, 1976, 1977; Underwood & Briggs, 1984; Wuillemin, Krane, & Richardson, 1982) and when it is superimposed over the picture (Lupker & Sanders, 1982; Wuillemin et al., 1982). A variation in response time is associated with a variation in the semantic relationship between a word and a picture, indicating that the meaning of the word is processed even though no output processing is required. These effects argue for the automatic processing of word meanings, and whereas output processing is clearly the prerogative of the left hemisphere, the data are more equivocal con-
RIGHT-HEMISPHERE
PROCESSING
275
cerning the recognition of word meanings. Underwood (1976, 1977) found that words were more disruptive when they were projected to the left hemisphere (with pictures to the right hemisphere), and both Lupker and Sanders (1982) and Wuillemin et al. (1982) found no semantic interference with words and pictures presented to the right hemisphere. Presentation of the word in the RVF appears to be necessary for its meaning to be effective, but this conclusion is taken from studies in which the picture had to be named. The use of a naming task might not only confound the inferred pattern of lateralized processing of the picture, but might also induce a naming “set” which could influence the processing and therefore the effectiveness of the word. Bryden (1978), Hellige (1978), and Kinsbourne (1973) have discussed the influence which spatial or verbal modes of thought can have upon the strategies of processing of incoming stimuli, and the requirement to output vocally the name of the picture may set the subject to process all stimuli with a verbal strategy. Indeed, when a nonvocal lexical decision task was used to observe the processing of lateralized distracting words upon centrally fixated target words, Underwood, Rusted, and Thwaites (1983) found equivalent interference from words printed to the left and to the right of fixation. In this task, attention was directed to the meaning of the fixated word, without vocal output, and evidence of right-hemisphere processing of meaning was established. In a slightly different task, in which a fixated word was named as it appeared at the end of a spoken sentence, we found that distracters in the RVF did result in more interference (Underwood, Whitfield, & Winfield, 1982). More was varied between these two lexical decision and naming experiments than the response requirements only, but one explanation of the apparently contradictory results is that the naming task induces a strategy which inhibits the processing of meaning in the right hemisphere. After determining the pattern of lateralization for pictures using a categorization task in Experiment 1, the second experiment uses the same task, but with distracting words now presented in the same visual field. EXPERIMENT 1
This Experiment picture recognition
was simply a determination of the lateralization processes using a categorization response.
of
Method Subjecfs. Sixteen adult subjects were each paid fO.50 for their participation in this experiment. All had normal or corrected-to-normal vision, and all were right-handed according to their responses to Annett’s (1970) questionnaire. Materials and equipment. The pictures used in this experiment were all taken from the Snodgrass and Vanderwart (1980) standardized set. Sixty-four pictures were used from this set, thirty-two animals and thirty-two man-made and other inanimate objects. Each picture was photographed twice-once to appear in each visual field of a slide. When
276
UNDERWOOD
AND WHITFIELD
projected in the experiment, the pictures subtended horizontal and vertical angles between 1.5 and 3.5”, and the distance between the center of the visual field and the center of each picture subtended 4.5”. The sixty-four pictures were selected on the basis of having a name-agreement value greater than 65%, and a complexity value greater than 2.5. Slides of pictures were displayed with a Kodak Carousel projector modified by Forth Instruments for tachistoscopic projection (the shutter rise time was 1.5 msec). A second similar projector was used to display the masking stimulus. All slides were displayed in the same area of a white back-projection screen. The mask occupied the maximum area of projection of the pictures, and so the center of the screen was blank and illuminated. The mask was composed of a random arrangement of mixed letter features, and can be described as a pattern mask. The projectors were controlled by a Forth Instruments pulse generator (Model FI 272). Operation of this generator terminated the display of the fixation/mask field and started the display of the target field. Termination of the target field was accompanied by the reappearance of the fixation/mask field. Operation of the generator also started a Forth Instruments millisecond timer (Model FI 11 l), which was stopped by the activation of an Electronic Developments voice switch. The microphone for the voice switch was located on a table in front of the subject, who sat with chin resting upon a support and directly facing the back-projection screen. Procedure. Subjects fixated on the center of the projection screen, and said “yes” or “no” according to whether or not a picture could be considered to depict an animal. The first within-subjects factor was therefore the response set, and had two levels in that the response to the picture could be positive or negative. To investigate visual-field effects each of the 64 pictures was displayed once in each visual field. Visual field was therefore the second within-subjects factor. The display of a picture was preceded by a vocal “ready” warning, and exposure was set at 150 msec throughout the experiment. Order of presentation of slides was randomized for each subject. Subjects were instructed to fixate the center of the screen, and to say “yes” or “no” according to whether the picture showed an animal or not, and to respond as quickly and accurately as possible. The vocal response time to each stimulus was recorded, and at the end of the experiment subjects were asked whether they found pictures easier to categorize when they were presented in the right or left visual field.
Results The median response times were calculated and the means of those medians, together with the error data, are presented in Table 1. Both the medians and the error scores were submitted to analyses of variance. The informal enquiry of preferred visual field obtained too few data for analysis, but the result is of passing interest. Four subjects claimed to find RVF presentations easier and nine found LVF presentations easier. TABLE MEAN RESPONSE LATENCIES PERCENTAGES FROM EXPERIMENT TURE WAS DISPLAYED ALONE)
1 (IN
msec) AND 1 (IN WHICH
ERROR A PIC-
Picture displayed Left visual field 729 (8.26%)
Right visual field 784 (12.62%)
RIGHT-HEMISPHERE
PROCESSING
277
The remaining three subjects suggested that there was no difference in ease of response. The analysis of variance of median response times indicated that responses to the positive set (animals) were faster than those to the negative set (F(1, 15) = 22.8, p < .OOl). This set effect is of limited interest and is not discussed further. The analysis also indicated that responses to stimuli in the LVF were faster than those to stimuli in the RVF (F(1, 15) = 6.5, p < .025). The interaction between these two factors was not reliable (F(1, 15) = 2.85). The advantage gained by pictures in the LVF, suggested by subjects’ preferences and by response latencies, was confirmed by the analysis of variance performed upon the error data. More errors were made in the decisions made about pictures in the RVF than the LVF (F(1, 15) = 10.2, p < .Ol). This suggests that the difference in response latencies was not caused by a speed-error trade-off. No other effects or interactions were reliable. Discussion
The conclusions to be drawn from this experiment are straightforward. The use of a picture-categorization response demonstrates a righthemisphere advantage, both in terms of categorization latencies and accuracy of recognition. Previous reports of weak and inconsistent lefthemisphere advantages (Lupker & Sanders, 1982; Wuillemin et al., 1982) appear to have confounded the lateralization of input and output processes, and relate more to the requirement to generate the names of pictures than to recognition processes. Given that the categorization task indicates a left visual-field advantage for picture processing, we can now investigate the change in response latency as a function of the introduction of distracting words. EXPERIMENT 2
To observe the influence upon picture categorization, of the meaning of a word which was not required for output, this experiment varied the relationship between the two stimuli when they were both presented in the same visual field. With similar displays of words and pictures, Lupker and Sanders (1982) reported that semantically related words lead to slower responses than unrelated words, at least for RVF presentations with a picture-naming task. As we have seen from Experiment 1, this may not be the pattern when a categorization task is used. A pronounceable nonword was used as the control stimulus in Experiment 2. The other conditions were compared to this condition rather than a picture-alone condition to maintain the extent of nonlexical visual distraction which is present when a word is used. The only difference between a nonword and an unrelated word is thus the existence of a lexical entry for the
278
UNDERWOOD
AND WHITFIELD
word, and any effects of visual distraction are equated. Two conditions were used in which related words were presented. The word was either a member of the positive set, in that it was the name of an animal (“decision related”), or it was related to the meaning of the object in the specific picture (“picture related”). Method Subjects. Sixteen adult subjects were paid fO.50 for their participation in this experiment. They were recruited from the same population as used previously. All subjects had normal or corrected-to-normal vision, and administration of the Annett (1970) questionnaire indicated that they were all right-handed. Materials and equipment. The equipment used previously was again used here. The same pictures were also used, and the only difference between experiments was in the presentation of words which accompanied the pictures. For each picture four words were selected according to their relatedness to the picture and the task: (i) Decision-related words were all animal names. They were the names of the pictures used in the experiment (e.g., the picture of the SWAN with the word “tiger,” and LEAFpig). (ii) Picture-related words were semantic associates of the particular picture (e.g., FOXhunt, PEN-ink). (iii) Unrelated words were of similar frequency and length of other words used in the experiment, but were not obviously related to the pictures (e.g., OWL-cannon, GUITARkite). They were the names of the nonanimal pictures used elsewhere in the experiment. (iv) The final condition was a control in which pronounceable nonwords were presented (e.g, CAMEL-wristle, VASE-wibe). They were created by changing one or two letters of the picture-related words used elsewhere. All letter strings varied between three and seven letters in length. The words were printed onto white cards using Kroy Murotype lower-case lettering (Helvetica 24-pt. style), and each word was positioned either above or below its accompanying picture. Half the words in each condition were printed above the picture, and half below, but always so that they appeared in the same visual field as the picture. For each picture, eight slides were prepared, with the words from each of the four conditions appearing in each visual field. Procedure. Subjects were instructed to decide whether each picture was an animal or not, and to say “yes” or “no” accordingly. They were instructed to ignore the words completely, and to look only at the center of the screen. Each subject saw any specific picture only twice-once in each visual field, and with the word belonging to the same condition on each occasion. This requirement resulted in a counterbalancing design which used four groups of subjects, with any specific picture belonging to one of the four word conditions for each group. Within each group of subjects, and within each pair of slides corresponding to each picture, half the subjects saw the RVF picture first and half saw the LVF picture first. This resulted in two blocks of slides for each subject, with half the slides in each block containing RVF pictures, and half containing LVF pictures. Order of presentation of slides, within the constraints of this design, was randomized for each subject. The procedure was otherwise similar to that used in Experiment 1.
Results and Discussion Median response times and error percentages were first calculated, and means of these figures are presented in Table 2. An analysis of variance was performed upon the median response times
RIGHT-HEMISPHERE
TABLE 2 RESFQNSE LATENCIES (IN msec) AND ERROR PERCENTGES FROM EXPERIMENT (IN WHICH THE PICTURE WAS ACCOMPANIED BY A WORD)
MEAN
Picture Relationship to picture Decision Picture
279
PROCESSING
and word
2
displayed
of word Left related
related
Unrelated Pronounceable
word nonword
visual
763 (7.8%) 733 (9.0%) 767 (9.76%) 767 (9.76%)
field
Right
visual
field
804 (17.16%) 830 (17.56%) 807 (19.92%) 811 (16.64%)
to observe the influence of three within-subjects factors-visual field, response set, and distracting word conditions. As in Experiment 1, LFV responses were faster than those to pictures in the RVF (F(1, 15) = 13.7, p < .005), and responses to pictures in the positive set were faster than those to the negative set (F(1, 15) = 30.3, p < .OOl). No effects of distraction conditions was observed (F < l), and no other effects were reliable. However, a marginal two-factor interaction between visual field and conditions (F(3, 45) = 2.4, p = .08) suggested that the effects of distracting words should be observed for each of the visual fields. To examine the effectiveness of the three experimental conditions against the control nonword condition Dunnett’s test was used. For the LVF presentations only the picture-related words (733 msec) differed from the nonword controls (767 msec; p < .05), and no differences emerged between the RVF conditions. The error data were also submitted to a three-factor analysis of variance, and this confirmed that LVF pictures were recognized more easily than RVF pictures (F(1, 15) = 33.6, p < .OOl). More errors were produced with pictures from the positive set than the negative set (F(1, 15) = 9.5, p < .Ol), and this factor interacted with visual field (F(1, 15) = 16.6, p < .OOl). The response latency data confirm the LVF advantage for the recognition of pictures when a categorization response is used, and the effects of distracting words were more apparant for LVF words than for RVF words. The effect was that of a word related to a specific picture acting to aid its response, and this effect was restricted to pictures and words presented in the LVF. However, the overall interaction involving visual fields and distraction conditions was only marginally reliable, and requires confirmation. One purpose of Experiment 3, therefore, was to repeat the
280
UNDERWOOD
AND WHITFIELD
visual fields and word conditions in an attempt to replicate the pattern of effects. The results from Experiment 2 differ considerably from those reported by Lupker and Sanders (1982), with the most notable differences being the appearance of a right-hemisphere advantage in picture recognition, and the appearance of semantic facilitation for picture-related words projected to the right hemisphere. One difference in methodology which could be responsible for these differences was that the naming task used by Lupker and Sanders was replaced by a vocal categorization task. This alone can account for the appearance of a right-hemisphere advantage if we can assume that the output lexicon (see Morton, 1980) is localized within the left hemisphere. However, an alternate explanation of the appearance of the LVF advantage involves our use of a pattern masking stimulus before and after the target display. Whereas Lupker and Sanders (1982) presented their pictures with the pre- and postexposure field blank, we had these fields occupied by a masking stimulus composed of fragments of letters. The mask has the effect of rendering stimulus recognition difficult, and we included it to reduce the probability of the subjects recognizing the distracting words. It may have done more than this, because poststimulus masks are known to be able to reverse the pattern of visual-field advantages. Whereas McKeever and Suberi (1974) reported a RVF advantage for letter identification with a delayed mask, Hellige and Webster (1979) and McKeever and Suberi (1974) found that a superimposed mask produced a LVF advantage. The explanation of this reversal is not clear, but it remains possible that the introduction of the mask in Experiment 2 is the factor which changes the RVF advantage in picture processing into a LVF advantage. Experiment 3 investigates this possiblity by including masking as a factor in what is otherwise a replication of Experiment 2. EXPERIMENT 3
The influence of the distracting word in Experiment 2 was an effect which did not emerge strongly. This experiment therefore repeats the design to determine whether the effect is reliable, and at the same time explores the possibility that the LVF advantage in picture processing in some way resulted from the use of a pattern masking stimulus. This was done by presenting half the stimuli with the same mask as used previously, and half the stimuli without the mask. Method Subjecfs. Twenty-four subjects were paid fO.50 for their participation in this experiment, but they were not recruited from the same population as sampled in the earlier experiments. They were all school children, aged 16 to 18, and in daily attendance at a local Comprehensive School. They were all undertaking pre-university courses. All had normal or corrected-
RIGHT-HEMISPHERE
281
PROCESSING
to-normal vision, and by the administration of the Annett (1970) questionnaire it was determined that they were all right-handed. Materials, equipment, and procedure. The same equipment was used here as was used in Experiment 2, and the same slides were also used. The experimental design differed in only one respect, and that was that half the slides seen by each subject were masked, as before, and half were unmasked. Each picture was again seen twice by each subject, once in each visual field, and on one occasion it was masked and on the other it was unmasked. Whereas in Experiment 2 the two blocks of slides were both presented with the masking stimulus, in the present experiment it was presented with only one of the blocks. Half the subjects saw the pictures masked first, and half had the mask introduced half way through the experiment. When the mask was not used the pre- and postexposure fields were blank and illuminated. Subjectively, both the picture and the word were then clearly visible. The experiment used a completely within-subjects design, therefore, with four factors: visual field (two levels), response set (two levels), word conditions (four levels), and masking (two levels).
Results and Discussion
Median response latencies and error percentages were collected, and the means of these figures are presented in Table 3. The within-subjects analysis of variance on the latency data indicated that LVF pictures were recognized more easily than RVF pictures (F( 1, 23) = 21.4, p < JOI), that unmasked presentations lead to faster responses than masked presentations (F( 1, 23) = 125.1, p < .OOl), and that pictures from the positive set gained faster responses than those from the negative set (F(1, 23) = 17.0, p < .OOl>. As in Experiment 2, there was no main effect of distraction conditions (F < 1). Apart from the response set effect which did not interact with any other factor, the main effects were all involved in interactions. Masking interacted with visual field of presentation (F(1, 23) = 6.7, p < .025),
TABLE MEAN
RESPONSE
LATENCIES (IN msec) AND (IN WHICH THE PICTURE WAS
3 ERROR PERCENTAGES FROM EXPERIMENT ACCOMPANIED BY A WORD)
3
Picture and word displayed Relationship to picture
of word
Unmasked
Decision related Picture related Unrelated word Pronounceable
Left visual field
nonword
724 (4.9%) 752 (6.8%) 739 (3.1%) 714 (3.1%)
Masked 822 (8.3%) 770 (12.9%) 826 (11.9%) 812 (15.3%)
Right visual field Unmasked 764 (6.2%) 743 (7.3%) 737 (4.1%) 744 (8.3%)
Masked 858 (14.6%) 890 (10.9%) 848 (12.9%) 861 (8.3%)
282
UNDERWOOD
AND
WHITFIELD
and analyses of the simple main effects indicated that whereas masking was effective in both visual fields (both p < .OOl), the LVF advantage was reliable only with masked displays (p < .OOl). Although there was a 1Cmsec LVF advantage with unmasked displays this was not reliable. There was also a three-way interaction involving masking, visual field, and distraction conditions (F(3, 69) = 5.2, p < .005), and this was interpreted by investigating the distraction effects in the four conditions of masked and unmasked LVF and RVF displays. In each of these four conditions Dunnett’s test was used to compare each of the experiment conditions against the nonword control condition. With masked displays the same pattern was observed as had been seen in Experiment 2: picturerelated words (770 msec) facilitated the categorization response to LVF pictures in comparison with nonwords (812 msec; p < .05); and for masked RVF words there were again no differences between the distracting word conditions. A different pattern of results emerges from the unmasked conditions. With unmasked pictures in the LVF there is a difference between picture-related words (752 msec) and the nonword control (714 msec; p < .05), but this is a 38-msec inhibition effect, in contrast with a 42-msec facilitation effect for masked LVF pictures. With masked RVF pictures a homogenous set of latencies emerged, and no contrasts were reliable. The error data were also examined with an analysis of variance, and this indicated only one main effect, that of masking increasing the number of errors made (F(l) 23) = 19.1, p < .OOl). No other effects were reliable. This experiment has confirmed the results from Experiment 2 in that a LVF advantage was found for masked pictures, and in that the effectiveness of a distracting word is restricted to a picture-related word with LVF presentations. The LVF picture processing advantage was only reliable with masked presentations however, and so may be assumed to emerge only when feature extraction is made difficult. It should be noted that there was no indication of a RVF advantage with unmasked displays, however, contrary to the hypothesis that the effects in Experiment 2 differed from those of Lupker and Sanders (1982) because of our use of a masking stimulus. The pattern of distraction effects changed when the mask was removed, with the picture-related facilitation effect becoming an inhibition effect. GENERAL
DISCUSSION
The speed and accuracy of responses to isolated pictures in Experiment 1 indicates a considerable right-hemisphere advantage which might be taken to suggest that spatial stimuli such as pictures are most easily recognized when presented in the left visual field. The LVF advantage for picture categorization was confirmed in Experiments 2 and 3. The inthtence of words upon pictures was also restricted to LVF presentations.
RIGHT-HEMISPHERE
PROCESSING
283
Using masked displays Experiment 2 found that picture categorization was aided by related words in the same visual field, an effect replicated in Experiment 3. This effect contrasts with that reported by Lupker and Sanders (1982), who used a naming response with unmasked displays and who found an inhibition effect with RVF pictures. Experiment 3 investigated this inconsistency by presenting half the pictures with no masking stimulus. No RVF effects were found, and the semantic facilitation associated with masked LVF pictures became a semantic inhibition effect for unmasked LVF pictures. Whereas the unmasked naming task used by Lupker and Sanders (1982) produced semantic inhibition for RVF displays, our equivalent unmasked categorization task produced inhibition for LVF displays and no differences between the RVF conditions. These experiments point to the importance of the response made in the investigation of hemispheric asymmetries, and in particular suggest that the results of Lupker and Sanders are associated with their use of a naming task. The use of categorization responses, with similar display conditions, eliminates the effectiveness of related words in the RVF, and gives rise to an effect with LVF words. This change in the asymmetry of inhibition effects, associated with the change from a naming task to a categorization task, and the appearance of a LVF advantage with difficult viewing conditions raises questions about the lateralization of the processes necessary for the categorization of pictures. It is to these questions that the discussion now turns. The speed and accuracy data suggest that the Picture-Categorization System postulated by Warren and Morton (1982) is a property of the right hemisphere. The proposed isolation, or cognitive impenetrability, of the Picture-Categorization System is not supported by data showing an interaction between pictures and words. Accordingly, the Warren and Morton model must suggest that our interactions are localized in postrecognition stages. Our interpretation of the effects in Experiment 2 contradicts this suggestion and views the influence of the picture-related word upon the picture as being an influence during picture recognition. Given the right-hemisphere advantage in picture categorization, it is perhaps not surprising that the strongest interaction between pictures and words was also recorded with LVF presentations. Day (1977) and Underwood et al. (1983) have presented data to suggest that the meanings of words can be recognized when they are projected to either hemisphere, and so the absence of a semantic interaction with RVF presentations must be attributed to a combination of two factors-the absence of a picture-naming response, and the reduced ability of the left hemisphere to recognize the meanings of pictures. The third possibility-that the RVF advantage would return when the masking stimulus was removed-was discounted in Experiment 3. The LVF advantage was considerably reduced in the unmasked conditions,
284
UNDERWOOD
AND
WHITFIELD
and this appears to be a result similar to that reported by Sergent and Lorber (1982). They found no asymmetry in a picture-categorization task with clearly visible displays, but a LVF advantage when both the exposure duration and luminance were reduced. Our pattern mask appears to have had the same effect as reducing the stimulus energy in their experiment, confirming Sergent’s (1983) conclusions that the quality of the input is critical in determining asymmetries in recognition. No reversal of asymmetry was suggested by the data, however, and the LVF advantage was reduced to unreliability rather than giving way to a RVF advantage for picture processing. Lupker and Sanders (1982) reported a strong RVF advantage in their unmasked picture-naming experiment, and we can conclude that employing a naming task produces a RVF advantage, but that stimulus imput factors are of consequence. When the picture is seen with difficulty (i.e., slower responses accompanied by more errors), then a LVF advantage becomes apparant. The changed pattern of semantic influence is more difficult to explain. Like Lupker and Sanders we found a semantic inhibition effect with words upon unmasked pictures, but unlike Lupker and Sanders our inhibition was upon LVF pictures. This is presumably due to the change from a naming to a categorization task. An influence in the left hemisphere might be expected if picture naming is localized here, and the picturenaming inhibition is understandable, but why does picture categorization suffer a penalty from related words when they are in the LVF? One possiblity is that categorization processes are localized in the right hemisphere, and this supposition gains support from the evidence that our three experiments reported faster and more accurate responses to LVF pictures. The pattern of intluencefacilitation or inhibition-provides an important clue for understanding the interactions between the two stimuli. A tentative proposal from the effects observed here is that semantic facilitation effects occur when the word aids either the identification of the picture or the picture category. Furthermore, we suggest that inhibition effects emerge when the word is processed to a stage where the processing of the picture is impeded by the necessity to keep separate the two stimuli. The facilitation effect for LVF presentations could then emerge as follows. The word provides additional information about the picture, in that features are shared by the picture and the word in the picture-related condition, and this information is assumed to aid the collection of featural data about the picture. When sufficient data have been collected, the response can be initiated. The word does not cause interference because, although it might be processed further than the input lexicon, this processing is executed via the left-hemisphere output lexicon, whereas the response to the picture does not require this processor except for the control of the “yes” or “no” response. These responses do not conflict with activation
RIGHT-HEMISPHERE
285
PROCESSING
of the word name and no inhibition is observed. In the Lupker and Sanders experiment (1982), no effects of LVF stimuli were observed, but this follows from their use of a picture-naming task. The naming task changes the processing of the picture, and eliminates the righthemisphere advantage. It also allows the appearance of inhibition effects in the output lexicon. Whereas Experiment 2 reports a small and unreliable inhibition effect with our categorization task for RVF presentations, Lupker and Sanders (1982) found a robust inhibition effect with their naming task. The appearance of an inhibition effect in the unmasked conditions of Experiment 3 may also be associated with response competition effects. The unmasked words may have been available for report and may have acted to divert resources from the categorization decision. When the mask was presented the same words facilitated the categorization response. A related phenomenon was reported in Underwood’s (1977) picture-naming experiment in which subjects were asked to report the distracting word whenever possible. Picture-related words were associated with facilitation when the words were unreported and inhibition when they were reported. Experiment 1 demonstrated that early picture processing is executed best with stimuli projected to the right cerebral hemisphere. In order to observe this advantage, it is necessary to separate recognition processes from the confounding influences of naming processes which can lead to the appearance of a left-hemisphere advantage. Using the categorization task to observe interactions between words and pictures, Experiments 2 and 3 demonstrated a facilitated response for LVF pictures accompanied by a word which was a strong associate. This effect was attributed to the priming effect of the word upon the recognition of the masked features of the picture, and is not observable if unmasked displays are used. This may be due to changed processing strategies when picture recognition is difficult, and when the information provided by the word can aid feature identification. The experiment also provides another demonstration of the right-hemisphere processing of the meaning of words, in that for facilitation to occur, the distractor must be related in meaning to the picture. REFERENCES Annett, M. 1970. A classification of hand preference by association analysis. British Journal of Psychology, 61, 303-321. Bryden, M. P. 1978. Strategy effects in the assessment of hemispheric asymmetry. In G. Underwood (Ed.), Strategies of information processing. New York/London: Academic Press. Day, J. 1977. Right-hemisphere language processing in normal right-handers. Journal of Experimental
Psychology:
Human
Perception
and Performance,
3, 518-528.
Hellige, J. B. 1978. Visual laterality patterns for pure- versus mixed-list presentations. Journal
131.
of Experimental
Psychology:
Human
Perception
and Performance,
4, 121-
286
UNDERWOOD
AND WHITFIELD
Hellige, J. B., & Webster, R. 1979. Right-hemisphere superiority for initial stages of letter processing. Neuropsychologia, 17, 653-660. Kimura, D. 1969. Spatial localization in left and right visual fields. Canadian Journal of Psychology,
23, 445-448.
Kinsboume, M. 1973. The control of attention by interaction between the cerebral hemispheres. In S. Komblum (Ed.), Attention and performance, Vol. 4. New York/London: Academic Press. Lupker, S. J., & Sanders, M. 1982. Visual-field differences in picture-word interference. Brain
and Cognition,
1, 381-398.
McKeever, W., & Suberi, M. 1974. Parallel but temporally displaced visual half-field metacontrast function. Quarterly Journal of Experimental Psychology, 26, 258-265. Marcel, T., & Rajan, P. 1975. Lateral specialization for recognition of words and faces in good and poor readers. Neuropsychologia, 13, 489-497. Morton, J. 1980. The logogen model and orthographic structure. In U. Frith (Ed.), Cognitive processes in spelling. New York/London: Academic Press. Pirozzolo, F. J., & Rayner, K. 1979. Cerebral organization and reading disability. Neuropsychologia,
17, 485-491.
Sergent, J. 1983. Role of input in visual hemispheric asymmetries. Psychological
Bulletin,
93, 481-512.
Sergent, J., & Lorber, E. (1983). Perceptual categorization Brain
and Cognition,
in the cerebral hemispheres.
2, 39-54.
Snodgrass, J. G., Jr Vanderwart, M. 1980. A standardised set of 260 pictures: Norms for name agreement, image agreement, familiarity and complexity. Journal of Experimental Psychology: Human Learning and Memory, 6, 174-215. Underwood, G. 1976. Semantic interference from unattended printed words. British Journa/ of Psychology,
Underwood,
67, 327-338.
G. 1977. Attention, awareness and hemispheric differences in word recognition.
Neuropsychologia,
Underwood, Journal
15, 717-727.
G., & Briggs, P. 1984. The development of word recognition processes. British
of Psychology,
75, 243-255.
Underwood, G., Rusted, J., & Thwaites, S. 1983. Parafoveal words are effective in both hemi-fields: Preattentive processing of semantic and phonological codes. Perception, 12, 213-221. Underwood, G., Whitfield, A., & Winfield, J. 1982. Effects of contextual constraints and non-fixated words in a simple reading task. Journal of Research in Reading, 5, 8999.
Warren, C., & Morton, Journal
Witelson, Science
of Psychology,
J. 1982. The effects of priming on picture recognition.
S. F. 1977. Developmental (Washington,
British
73, 117-129. D. C.),
dyslexia: Two right hemispheres and none left.
195, 309-311.
Wuillemin, D. B., Krane, R. V., & Richardson, B. L. 1982. Hemispheric differences in picture-word interference. Brain and Language, 16, 121-132.
AND
COGNITION
4, 273-286 (1985)
Right-Hemisphere
Interactions in Picture-Word
Processing
GEOFFREY UNDERWOOD AND ALISON WHITFIELD University
of Nottingham
The processing of pictures was investigated in three experiments which eliminated the response effects involved in naming. When a categorization task was used, clear advantages in response latency and accuracy were observed for left visualfield (LVF) presentations. This was in contrast to previous investigations which have used a naming task and which have reported right visual-field (RVF) advantages. When a distracting word was added to the display, the pattern of influence also changed from that reported previously. The use of naming tasks has indicated predominantly left-hemisphere effects, with demonstrations of interactions between pictures and words in the RVF. With a categorization task in Experiment 2, however, the only effective words were those related in meaning to the picture, and only when they were projected to the right hemisphere. The third experiment confirmed the LVF advantage for picture processing with masked displays, but found no reliable asymmetry with unmasked presentations. The pattern of semantic facilitation was also confirmed with the masked displays, but when the mask was removed an inhibition effect replaced the facilitation effect. These effects are interpreted as indicating that picture recognition is localized within the right cerebral hemisphere. It is suggested that the facilitating effect of related words is restricted to the left hemisphere because it is an effect upon recognition processes, whereas the inhibition effect reflects response competition. It is also suggested that previous reports of left-hemisphere interference effects are due to effects of response competition in naming tasks. Q 1985 Academic Press, Inc.
How is the processing of pictures laterialized, and how does the presence of a printed word affect the normal pattern of lateralization? These questions were addressed by three experiments, in an attempt to separate the influences of the verbal and nonverbal components in a pictorial version of the Stroop effect. This research was supported by Medical Research Council Grant G8127736N. We are also grateful to E. A. Maylor and two anonymous reviewers for their comments on an earlier draft of this report. The subjects used in Experiment 3 participated with the cooperation of the Headmistress and Staff of Chilwell Comprehensive School, Nottingham. We are very grateful to them and the participants for their help. Correspondence and requests for reprints should be addressed to Geoffrey Underwood, Department of Psychology, University of Nottingham, NG7 2RD Nottingham, England. 273 0278-2626185 $3 .OO Copyright 0 1985 by Academic Press, Inc. All rights of reproduction in any form reserved.
274
UNDERWOOD
AND
WHITFIELD
Lupker and Sanders (1982) have recently suggested that “the ability of the right hemisphere to process pictorial stimuli may be nearly equivalent to that of the left hemisphere” and that “these stimuli can be processed equally rapidly when presented in either visual field” (p. 383). An earlier report of an experiment using picture naming supports this conclusion with no difference in the naming latency for pictures in the left or right visual field (Underwood, 1976). However, a report by Wuillemin, Krane, and Richardson (1982) did find that pictures projected to the left hemisphere were associated with faster naming responses than those projected to the right hemisphere. Lupker and Sanders (1982) also found a small but statistically unreliable advantage in the naming of pictures projected to the left hemisphere. These authors concluded that the weak and inconsistent appearance of a left-hemisphere advantage is an indication of an equivalence between the hemispheres in the processing of pictures. If pictures can be considered as spatial representations, then the conclusion is surprising, for visuo-spatial processing has gained a righthemisphere advantage in a number of tasks. Left visual-field (LVF) advantages have been reported for dot localization (Kimura, 1969), the categorization of drawings of figures and people (Witelson, 1977), and face identification (Marcel & Rajan, 1975; Pirozzolo & Rayner, 1979). The conflict is not difficult to resolve, however, for whereas the experiments using pictures employed a vocal naming response, these other experiments employed either vocal categorization (“same/different”) or manual pointing responses. If the generation of picture names is a process restricted to the left cerebral hemisphere, then a right visual-field (RVF) advantage would be expected. In Experiment 1, therefore, the input processing of pictures is separated from the output demands of naming by employing a vocal categorization task. Subjects were instructed to say “yes” or “no” to indicate whether or not each picture was an animal. The purpose of gaining evidence of the recognition of pictorial stimuli was to help in the understanding of interactions between words and pictures in a modified Stroop task. When words are presented at the same time as pictures, the responses to the pictures vary as a function of the semantic relationship between the two stimuli. This effect can be demonstrated with picture-naming responses when the word is presented in the opposite visual field to the picture (Underwood, 1976, 1977; Underwood & Briggs, 1984; Wuillemin, Krane, & Richardson, 1982) and when it is superimposed over the picture (Lupker & Sanders, 1982; Wuillemin et al., 1982). A variation in response time is associated with a variation in the semantic relationship between a word and a picture, indicating that the meaning of the word is processed even though no output processing is required. These effects argue for the automatic processing of word meanings, and whereas output processing is clearly the prerogative of the left hemisphere, the data are more equivocal con-
RIGHT-HEMISPHERE
PROCESSING
275
cerning the recognition of word meanings. Underwood (1976, 1977) found that words were more disruptive when they were projected to the left hemisphere (with pictures to the right hemisphere), and both Lupker and Sanders (1982) and Wuillemin et al. (1982) found no semantic interference with words and pictures presented to the right hemisphere. Presentation of the word in the RVF appears to be necessary for its meaning to be effective, but this conclusion is taken from studies in which the picture had to be named. The use of a naming task might not only confound the inferred pattern of lateralized processing of the picture, but might also induce a naming “set” which could influence the processing and therefore the effectiveness of the word. Bryden (1978), Hellige (1978), and Kinsbourne (1973) have discussed the influence which spatial or verbal modes of thought can have upon the strategies of processing of incoming stimuli, and the requirement to output vocally the name of the picture may set the subject to process all stimuli with a verbal strategy. Indeed, when a nonvocal lexical decision task was used to observe the processing of lateralized distracting words upon centrally fixated target words, Underwood, Rusted, and Thwaites (1983) found equivalent interference from words printed to the left and to the right of fixation. In this task, attention was directed to the meaning of the fixated word, without vocal output, and evidence of right-hemisphere processing of meaning was established. In a slightly different task, in which a fixated word was named as it appeared at the end of a spoken sentence, we found that distracters in the RVF did result in more interference (Underwood, Whitfield, & Winfield, 1982). More was varied between these two lexical decision and naming experiments than the response requirements only, but one explanation of the apparently contradictory results is that the naming task induces a strategy which inhibits the processing of meaning in the right hemisphere. After determining the pattern of lateralization for pictures using a categorization task in Experiment 1, the second experiment uses the same task, but with distracting words now presented in the same visual field. EXPERIMENT 1
This Experiment picture recognition
was simply a determination of the lateralization processes using a categorization response.
of
Method Subjecfs. Sixteen adult subjects were each paid fO.50 for their participation in this experiment. All had normal or corrected-to-normal vision, and all were right-handed according to their responses to Annett’s (1970) questionnaire. Materials and equipment. The pictures used in this experiment were all taken from the Snodgrass and Vanderwart (1980) standardized set. Sixty-four pictures were used from this set, thirty-two animals and thirty-two man-made and other inanimate objects. Each picture was photographed twice-once to appear in each visual field of a slide. When
276
UNDERWOOD
AND WHITFIELD
projected in the experiment, the pictures subtended horizontal and vertical angles between 1.5 and 3.5”, and the distance between the center of the visual field and the center of each picture subtended 4.5”. The sixty-four pictures were selected on the basis of having a name-agreement value greater than 65%, and a complexity value greater than 2.5. Slides of pictures were displayed with a Kodak Carousel projector modified by Forth Instruments for tachistoscopic projection (the shutter rise time was 1.5 msec). A second similar projector was used to display the masking stimulus. All slides were displayed in the same area of a white back-projection screen. The mask occupied the maximum area of projection of the pictures, and so the center of the screen was blank and illuminated. The mask was composed of a random arrangement of mixed letter features, and can be described as a pattern mask. The projectors were controlled by a Forth Instruments pulse generator (Model FI 272). Operation of this generator terminated the display of the fixation/mask field and started the display of the target field. Termination of the target field was accompanied by the reappearance of the fixation/mask field. Operation of the generator also started a Forth Instruments millisecond timer (Model FI 11 l), which was stopped by the activation of an Electronic Developments voice switch. The microphone for the voice switch was located on a table in front of the subject, who sat with chin resting upon a support and directly facing the back-projection screen. Procedure. Subjects fixated on the center of the projection screen, and said “yes” or “no” according to whether or not a picture could be considered to depict an animal. The first within-subjects factor was therefore the response set, and had two levels in that the response to the picture could be positive or negative. To investigate visual-field effects each of the 64 pictures was displayed once in each visual field. Visual field was therefore the second within-subjects factor. The display of a picture was preceded by a vocal “ready” warning, and exposure was set at 150 msec throughout the experiment. Order of presentation of slides was randomized for each subject. Subjects were instructed to fixate the center of the screen, and to say “yes” or “no” according to whether the picture showed an animal or not, and to respond as quickly and accurately as possible. The vocal response time to each stimulus was recorded, and at the end of the experiment subjects were asked whether they found pictures easier to categorize when they were presented in the right or left visual field.
Results The median response times were calculated and the means of those medians, together with the error data, are presented in Table 1. Both the medians and the error scores were submitted to analyses of variance. The informal enquiry of preferred visual field obtained too few data for analysis, but the result is of passing interest. Four subjects claimed to find RVF presentations easier and nine found LVF presentations easier. TABLE MEAN RESPONSE LATENCIES PERCENTAGES FROM EXPERIMENT TURE WAS DISPLAYED ALONE)
1 (IN
msec) AND 1 (IN WHICH
ERROR A PIC-
Picture displayed Left visual field 729 (8.26%)
Right visual field 784 (12.62%)
RIGHT-HEMISPHERE
PROCESSING
277
The remaining three subjects suggested that there was no difference in ease of response. The analysis of variance of median response times indicated that responses to the positive set (animals) were faster than those to the negative set (F(1, 15) = 22.8, p < .OOl). This set effect is of limited interest and is not discussed further. The analysis also indicated that responses to stimuli in the LVF were faster than those to stimuli in the RVF (F(1, 15) = 6.5, p < .025). The interaction between these two factors was not reliable (F(1, 15) = 2.85). The advantage gained by pictures in the LVF, suggested by subjects’ preferences and by response latencies, was confirmed by the analysis of variance performed upon the error data. More errors were made in the decisions made about pictures in the RVF than the LVF (F(1, 15) = 10.2, p < .Ol). This suggests that the difference in response latencies was not caused by a speed-error trade-off. No other effects or interactions were reliable. Discussion
The conclusions to be drawn from this experiment are straightforward. The use of a picture-categorization response demonstrates a righthemisphere advantage, both in terms of categorization latencies and accuracy of recognition. Previous reports of weak and inconsistent lefthemisphere advantages (Lupker & Sanders, 1982; Wuillemin et al., 1982) appear to have confounded the lateralization of input and output processes, and relate more to the requirement to generate the names of pictures than to recognition processes. Given that the categorization task indicates a left visual-field advantage for picture processing, we can now investigate the change in response latency as a function of the introduction of distracting words. EXPERIMENT 2
To observe the influence upon picture categorization, of the meaning of a word which was not required for output, this experiment varied the relationship between the two stimuli when they were both presented in the same visual field. With similar displays of words and pictures, Lupker and Sanders (1982) reported that semantically related words lead to slower responses than unrelated words, at least for RVF presentations with a picture-naming task. As we have seen from Experiment 1, this may not be the pattern when a categorization task is used. A pronounceable nonword was used as the control stimulus in Experiment 2. The other conditions were compared to this condition rather than a picture-alone condition to maintain the extent of nonlexical visual distraction which is present when a word is used. The only difference between a nonword and an unrelated word is thus the existence of a lexical entry for the
278
UNDERWOOD
AND WHITFIELD
word, and any effects of visual distraction are equated. Two conditions were used in which related words were presented. The word was either a member of the positive set, in that it was the name of an animal (“decision related”), or it was related to the meaning of the object in the specific picture (“picture related”). Method Subjects. Sixteen adult subjects were paid fO.50 for their participation in this experiment. They were recruited from the same population as used previously. All subjects had normal or corrected-to-normal vision, and administration of the Annett (1970) questionnaire indicated that they were all right-handed. Materials and equipment. The equipment used previously was again used here. The same pictures were also used, and the only difference between experiments was in the presentation of words which accompanied the pictures. For each picture four words were selected according to their relatedness to the picture and the task: (i) Decision-related words were all animal names. They were the names of the pictures used in the experiment (e.g., the picture of the SWAN with the word “tiger,” and LEAFpig). (ii) Picture-related words were semantic associates of the particular picture (e.g., FOXhunt, PEN-ink). (iii) Unrelated words were of similar frequency and length of other words used in the experiment, but were not obviously related to the pictures (e.g., OWL-cannon, GUITARkite). They were the names of the nonanimal pictures used elsewhere in the experiment. (iv) The final condition was a control in which pronounceable nonwords were presented (e.g, CAMEL-wristle, VASE-wibe). They were created by changing one or two letters of the picture-related words used elsewhere. All letter strings varied between three and seven letters in length. The words were printed onto white cards using Kroy Murotype lower-case lettering (Helvetica 24-pt. style), and each word was positioned either above or below its accompanying picture. Half the words in each condition were printed above the picture, and half below, but always so that they appeared in the same visual field as the picture. For each picture, eight slides were prepared, with the words from each of the four conditions appearing in each visual field. Procedure. Subjects were instructed to decide whether each picture was an animal or not, and to say “yes” or “no” accordingly. They were instructed to ignore the words completely, and to look only at the center of the screen. Each subject saw any specific picture only twice-once in each visual field, and with the word belonging to the same condition on each occasion. This requirement resulted in a counterbalancing design which used four groups of subjects, with any specific picture belonging to one of the four word conditions for each group. Within each group of subjects, and within each pair of slides corresponding to each picture, half the subjects saw the RVF picture first and half saw the LVF picture first. This resulted in two blocks of slides for each subject, with half the slides in each block containing RVF pictures, and half containing LVF pictures. Order of presentation of slides, within the constraints of this design, was randomized for each subject. The procedure was otherwise similar to that used in Experiment 1.
Results and Discussion Median response times and error percentages were first calculated, and means of these figures are presented in Table 2. An analysis of variance was performed upon the median response times
RIGHT-HEMISPHERE
TABLE 2 RESFQNSE LATENCIES (IN msec) AND ERROR PERCENTGES FROM EXPERIMENT (IN WHICH THE PICTURE WAS ACCOMPANIED BY A WORD)
MEAN
Picture Relationship to picture Decision Picture
279
PROCESSING
and word
2
displayed
of word Left related
related
Unrelated Pronounceable
word nonword
visual
763 (7.8%) 733 (9.0%) 767 (9.76%) 767 (9.76%)
field
Right
visual
field
804 (17.16%) 830 (17.56%) 807 (19.92%) 811 (16.64%)
to observe the influence of three within-subjects factors-visual field, response set, and distracting word conditions. As in Experiment 1, LFV responses were faster than those to pictures in the RVF (F(1, 15) = 13.7, p < .005), and responses to pictures in the positive set were faster than those to the negative set (F(1, 15) = 30.3, p < .OOl). No effects of distraction conditions was observed (F < l), and no other effects were reliable. However, a marginal two-factor interaction between visual field and conditions (F(3, 45) = 2.4, p = .08) suggested that the effects of distracting words should be observed for each of the visual fields. To examine the effectiveness of the three experimental conditions against the control nonword condition Dunnett’s test was used. For the LVF presentations only the picture-related words (733 msec) differed from the nonword controls (767 msec; p < .05), and no differences emerged between the RVF conditions. The error data were also submitted to a three-factor analysis of variance, and this confirmed that LVF pictures were recognized more easily than RVF pictures (F(1, 15) = 33.6, p < .OOl). More errors were produced with pictures from the positive set than the negative set (F(1, 15) = 9.5, p < .Ol), and this factor interacted with visual field (F(1, 15) = 16.6, p < .OOl). The response latency data confirm the LVF advantage for the recognition of pictures when a categorization response is used, and the effects of distracting words were more apparant for LVF words than for RVF words. The effect was that of a word related to a specific picture acting to aid its response, and this effect was restricted to pictures and words presented in the LVF. However, the overall interaction involving visual fields and distraction conditions was only marginally reliable, and requires confirmation. One purpose of Experiment 3, therefore, was to repeat the
280
UNDERWOOD
AND WHITFIELD
visual fields and word conditions in an attempt to replicate the pattern of effects. The results from Experiment 2 differ considerably from those reported by Lupker and Sanders (1982), with the most notable differences being the appearance of a right-hemisphere advantage in picture recognition, and the appearance of semantic facilitation for picture-related words projected to the right hemisphere. One difference in methodology which could be responsible for these differences was that the naming task used by Lupker and Sanders was replaced by a vocal categorization task. This alone can account for the appearance of a right-hemisphere advantage if we can assume that the output lexicon (see Morton, 1980) is localized within the left hemisphere. However, an alternate explanation of the appearance of the LVF advantage involves our use of a pattern masking stimulus before and after the target display. Whereas Lupker and Sanders (1982) presented their pictures with the pre- and postexposure field blank, we had these fields occupied by a masking stimulus composed of fragments of letters. The mask has the effect of rendering stimulus recognition difficult, and we included it to reduce the probability of the subjects recognizing the distracting words. It may have done more than this, because poststimulus masks are known to be able to reverse the pattern of visual-field advantages. Whereas McKeever and Suberi (1974) reported a RVF advantage for letter identification with a delayed mask, Hellige and Webster (1979) and McKeever and Suberi (1974) found that a superimposed mask produced a LVF advantage. The explanation of this reversal is not clear, but it remains possible that the introduction of the mask in Experiment 2 is the factor which changes the RVF advantage in picture processing into a LVF advantage. Experiment 3 investigates this possiblity by including masking as a factor in what is otherwise a replication of Experiment 2. EXPERIMENT 3
The influence of the distracting word in Experiment 2 was an effect which did not emerge strongly. This experiment therefore repeats the design to determine whether the effect is reliable, and at the same time explores the possibility that the LVF advantage in picture processing in some way resulted from the use of a pattern masking stimulus. This was done by presenting half the stimuli with the same mask as used previously, and half the stimuli without the mask. Method Subjecfs. Twenty-four subjects were paid fO.50 for their participation in this experiment, but they were not recruited from the same population as sampled in the earlier experiments. They were all school children, aged 16 to 18, and in daily attendance at a local Comprehensive School. They were all undertaking pre-university courses. All had normal or corrected-
RIGHT-HEMISPHERE
281
PROCESSING
to-normal vision, and by the administration of the Annett (1970) questionnaire it was determined that they were all right-handed. Materials, equipment, and procedure. The same equipment was used here as was used in Experiment 2, and the same slides were also used. The experimental design differed in only one respect, and that was that half the slides seen by each subject were masked, as before, and half were unmasked. Each picture was again seen twice by each subject, once in each visual field, and on one occasion it was masked and on the other it was unmasked. Whereas in Experiment 2 the two blocks of slides were both presented with the masking stimulus, in the present experiment it was presented with only one of the blocks. Half the subjects saw the pictures masked first, and half had the mask introduced half way through the experiment. When the mask was not used the pre- and postexposure fields were blank and illuminated. Subjectively, both the picture and the word were then clearly visible. The experiment used a completely within-subjects design, therefore, with four factors: visual field (two levels), response set (two levels), word conditions (four levels), and masking (two levels).
Results and Discussion
Median response latencies and error percentages were collected, and the means of these figures are presented in Table 3. The within-subjects analysis of variance on the latency data indicated that LVF pictures were recognized more easily than RVF pictures (F( 1, 23) = 21.4, p < JOI), that unmasked presentations lead to faster responses than masked presentations (F( 1, 23) = 125.1, p < .OOl), and that pictures from the positive set gained faster responses than those from the negative set (F(1, 23) = 17.0, p < .OOl>. As in Experiment 2, there was no main effect of distraction conditions (F < 1). Apart from the response set effect which did not interact with any other factor, the main effects were all involved in interactions. Masking interacted with visual field of presentation (F(1, 23) = 6.7, p < .025),
TABLE MEAN
RESPONSE
LATENCIES (IN msec) AND (IN WHICH THE PICTURE WAS
3 ERROR PERCENTAGES FROM EXPERIMENT ACCOMPANIED BY A WORD)
3
Picture and word displayed Relationship to picture
of word
Unmasked
Decision related Picture related Unrelated word Pronounceable
Left visual field
nonword
724 (4.9%) 752 (6.8%) 739 (3.1%) 714 (3.1%)
Masked 822 (8.3%) 770 (12.9%) 826 (11.9%) 812 (15.3%)
Right visual field Unmasked 764 (6.2%) 743 (7.3%) 737 (4.1%) 744 (8.3%)
Masked 858 (14.6%) 890 (10.9%) 848 (12.9%) 861 (8.3%)
282
UNDERWOOD
AND
WHITFIELD
and analyses of the simple main effects indicated that whereas masking was effective in both visual fields (both p < .OOl), the LVF advantage was reliable only with masked displays (p < .OOl). Although there was a 1Cmsec LVF advantage with unmasked displays this was not reliable. There was also a three-way interaction involving masking, visual field, and distraction conditions (F(3, 69) = 5.2, p < .005), and this was interpreted by investigating the distraction effects in the four conditions of masked and unmasked LVF and RVF displays. In each of these four conditions Dunnett’s test was used to compare each of the experiment conditions against the nonword control condition. With masked displays the same pattern was observed as had been seen in Experiment 2: picturerelated words (770 msec) facilitated the categorization response to LVF pictures in comparison with nonwords (812 msec; p < .05); and for masked RVF words there were again no differences between the distracting word conditions. A different pattern of results emerges from the unmasked conditions. With unmasked pictures in the LVF there is a difference between picture-related words (752 msec) and the nonword control (714 msec; p < .05), but this is a 38-msec inhibition effect, in contrast with a 42-msec facilitation effect for masked LVF pictures. With masked RVF pictures a homogenous set of latencies emerged, and no contrasts were reliable. The error data were also examined with an analysis of variance, and this indicated only one main effect, that of masking increasing the number of errors made (F(l) 23) = 19.1, p < .OOl). No other effects were reliable. This experiment has confirmed the results from Experiment 2 in that a LVF advantage was found for masked pictures, and in that the effectiveness of a distracting word is restricted to a picture-related word with LVF presentations. The LVF picture processing advantage was only reliable with masked presentations however, and so may be assumed to emerge only when feature extraction is made difficult. It should be noted that there was no indication of a RVF advantage with unmasked displays, however, contrary to the hypothesis that the effects in Experiment 2 differed from those of Lupker and Sanders (1982) because of our use of a masking stimulus. The pattern of distraction effects changed when the mask was removed, with the picture-related facilitation effect becoming an inhibition effect. GENERAL
DISCUSSION
The speed and accuracy of responses to isolated pictures in Experiment 1 indicates a considerable right-hemisphere advantage which might be taken to suggest that spatial stimuli such as pictures are most easily recognized when presented in the left visual field. The LVF advantage for picture categorization was confirmed in Experiments 2 and 3. The inthtence of words upon pictures was also restricted to LVF presentations.
RIGHT-HEMISPHERE
PROCESSING
283
Using masked displays Experiment 2 found that picture categorization was aided by related words in the same visual field, an effect replicated in Experiment 3. This effect contrasts with that reported by Lupker and Sanders (1982), who used a naming response with unmasked displays and who found an inhibition effect with RVF pictures. Experiment 3 investigated this inconsistency by presenting half the pictures with no masking stimulus. No RVF effects were found, and the semantic facilitation associated with masked LVF pictures became a semantic inhibition effect for unmasked LVF pictures. Whereas the unmasked naming task used by Lupker and Sanders (1982) produced semantic inhibition for RVF displays, our equivalent unmasked categorization task produced inhibition for LVF displays and no differences between the RVF conditions. These experiments point to the importance of the response made in the investigation of hemispheric asymmetries, and in particular suggest that the results of Lupker and Sanders are associated with their use of a naming task. The use of categorization responses, with similar display conditions, eliminates the effectiveness of related words in the RVF, and gives rise to an effect with LVF words. This change in the asymmetry of inhibition effects, associated with the change from a naming task to a categorization task, and the appearance of a LVF advantage with difficult viewing conditions raises questions about the lateralization of the processes necessary for the categorization of pictures. It is to these questions that the discussion now turns. The speed and accuracy data suggest that the Picture-Categorization System postulated by Warren and Morton (1982) is a property of the right hemisphere. The proposed isolation, or cognitive impenetrability, of the Picture-Categorization System is not supported by data showing an interaction between pictures and words. Accordingly, the Warren and Morton model must suggest that our interactions are localized in postrecognition stages. Our interpretation of the effects in Experiment 2 contradicts this suggestion and views the influence of the picture-related word upon the picture as being an influence during picture recognition. Given the right-hemisphere advantage in picture categorization, it is perhaps not surprising that the strongest interaction between pictures and words was also recorded with LVF presentations. Day (1977) and Underwood et al. (1983) have presented data to suggest that the meanings of words can be recognized when they are projected to either hemisphere, and so the absence of a semantic interaction with RVF presentations must be attributed to a combination of two factors-the absence of a picture-naming response, and the reduced ability of the left hemisphere to recognize the meanings of pictures. The third possibility-that the RVF advantage would return when the masking stimulus was removed-was discounted in Experiment 3. The LVF advantage was considerably reduced in the unmasked conditions,
284
UNDERWOOD
AND
WHITFIELD
and this appears to be a result similar to that reported by Sergent and Lorber (1982). They found no asymmetry in a picture-categorization task with clearly visible displays, but a LVF advantage when both the exposure duration and luminance were reduced. Our pattern mask appears to have had the same effect as reducing the stimulus energy in their experiment, confirming Sergent’s (1983) conclusions that the quality of the input is critical in determining asymmetries in recognition. No reversal of asymmetry was suggested by the data, however, and the LVF advantage was reduced to unreliability rather than giving way to a RVF advantage for picture processing. Lupker and Sanders (1982) reported a strong RVF advantage in their unmasked picture-naming experiment, and we can conclude that employing a naming task produces a RVF advantage, but that stimulus imput factors are of consequence. When the picture is seen with difficulty (i.e., slower responses accompanied by more errors), then a LVF advantage becomes apparant. The changed pattern of semantic influence is more difficult to explain. Like Lupker and Sanders we found a semantic inhibition effect with words upon unmasked pictures, but unlike Lupker and Sanders our inhibition was upon LVF pictures. This is presumably due to the change from a naming to a categorization task. An influence in the left hemisphere might be expected if picture naming is localized here, and the picturenaming inhibition is understandable, but why does picture categorization suffer a penalty from related words when they are in the LVF? One possiblity is that categorization processes are localized in the right hemisphere, and this supposition gains support from the evidence that our three experiments reported faster and more accurate responses to LVF pictures. The pattern of intluencefacilitation or inhibition-provides an important clue for understanding the interactions between the two stimuli. A tentative proposal from the effects observed here is that semantic facilitation effects occur when the word aids either the identification of the picture or the picture category. Furthermore, we suggest that inhibition effects emerge when the word is processed to a stage where the processing of the picture is impeded by the necessity to keep separate the two stimuli. The facilitation effect for LVF presentations could then emerge as follows. The word provides additional information about the picture, in that features are shared by the picture and the word in the picture-related condition, and this information is assumed to aid the collection of featural data about the picture. When sufficient data have been collected, the response can be initiated. The word does not cause interference because, although it might be processed further than the input lexicon, this processing is executed via the left-hemisphere output lexicon, whereas the response to the picture does not require this processor except for the control of the “yes” or “no” response. These responses do not conflict with activation
RIGHT-HEMISPHERE
285
PROCESSING
of the word name and no inhibition is observed. In the Lupker and Sanders experiment (1982), no effects of LVF stimuli were observed, but this follows from their use of a picture-naming task. The naming task changes the processing of the picture, and eliminates the righthemisphere advantage. It also allows the appearance of inhibition effects in the output lexicon. Whereas Experiment 2 reports a small and unreliable inhibition effect with our categorization task for RVF presentations, Lupker and Sanders (1982) found a robust inhibition effect with their naming task. The appearance of an inhibition effect in the unmasked conditions of Experiment 3 may also be associated with response competition effects. The unmasked words may have been available for report and may have acted to divert resources from the categorization decision. When the mask was presented the same words facilitated the categorization response. A related phenomenon was reported in Underwood’s (1977) picture-naming experiment in which subjects were asked to report the distracting word whenever possible. Picture-related words were associated with facilitation when the words were unreported and inhibition when they were reported. Experiment 1 demonstrated that early picture processing is executed best with stimuli projected to the right cerebral hemisphere. In order to observe this advantage, it is necessary to separate recognition processes from the confounding influences of naming processes which can lead to the appearance of a left-hemisphere advantage. Using the categorization task to observe interactions between words and pictures, Experiments 2 and 3 demonstrated a facilitated response for LVF pictures accompanied by a word which was a strong associate. This effect was attributed to the priming effect of the word upon the recognition of the masked features of the picture, and is not observable if unmasked displays are used. This may be due to changed processing strategies when picture recognition is difficult, and when the information provided by the word can aid feature identification. The experiment also provides another demonstration of the right-hemisphere processing of the meaning of words, in that for facilitation to occur, the distractor must be related in meaning to the picture. REFERENCES Annett, M. 1970. A classification of hand preference by association analysis. British Journal of Psychology, 61, 303-321. Bryden, M. P. 1978. Strategy effects in the assessment of hemispheric asymmetry. In G. Underwood (Ed.), Strategies of information processing. New York/London: Academic Press. Day, J. 1977. Right-hemisphere language processing in normal right-handers. Journal of Experimental
Psychology:
Human
Perception
and Performance,
3, 518-528.
Hellige, J. B. 1978. Visual laterality patterns for pure- versus mixed-list presentations. Journal
131.
of Experimental
Psychology:
Human
Perception
and Performance,
4, 121-
286
UNDERWOOD
AND WHITFIELD
Hellige, J. B., & Webster, R. 1979. Right-hemisphere superiority for initial stages of letter processing. Neuropsychologia, 17, 653-660. Kimura, D. 1969. Spatial localization in left and right visual fields. Canadian Journal of Psychology,
23, 445-448.
Kinsboume, M. 1973. The control of attention by interaction between the cerebral hemispheres. In S. Komblum (Ed.), Attention and performance, Vol. 4. New York/London: Academic Press. Lupker, S. J., & Sanders, M. 1982. Visual-field differences in picture-word interference. Brain
and Cognition,
1, 381-398.
McKeever, W., & Suberi, M. 1974. Parallel but temporally displaced visual half-field metacontrast function. Quarterly Journal of Experimental Psychology, 26, 258-265. Marcel, T., & Rajan, P. 1975. Lateral specialization for recognition of words and faces in good and poor readers. Neuropsychologia, 13, 489-497. Morton, J. 1980. The logogen model and orthographic structure. In U. Frith (Ed.), Cognitive processes in spelling. New York/London: Academic Press. Pirozzolo, F. J., & Rayner, K. 1979. Cerebral organization and reading disability. Neuropsychologia,
17, 485-491.
Sergent, J. 1983. Role of input in visual hemispheric asymmetries. Psychological
Bulletin,
93, 481-512.
Sergent, J., & Lorber, E. (1983). Perceptual categorization Brain
and Cognition,
in the cerebral hemispheres.
2, 39-54.
Snodgrass, J. G., Jr Vanderwart, M. 1980. A standardised set of 260 pictures: Norms for name agreement, image agreement, familiarity and complexity. Journal of Experimental Psychology: Human Learning and Memory, 6, 174-215. Underwood, G. 1976. Semantic interference from unattended printed words. British Journa/ of Psychology,
Underwood,
67, 327-338.
G. 1977. Attention, awareness and hemispheric differences in word recognition.
Neuropsychologia,
Underwood, Journal
15, 717-727.
G., & Briggs, P. 1984. The development of word recognition processes. British
of Psychology,
75, 243-255.
Underwood, G., Rusted, J., & Thwaites, S. 1983. Parafoveal words are effective in both hemi-fields: Preattentive processing of semantic and phonological codes. Perception, 12, 213-221. Underwood, G., Whitfield, A., & Winfield, J. 1982. Effects of contextual constraints and non-fixated words in a simple reading task. Journal of Research in Reading, 5, 8999.
Warren, C., & Morton, Journal
Witelson, Science
of Psychology,
J. 1982. The effects of priming on picture recognition.
S. F. 1977. Developmental (Washington,
British
73, 117-129. D. C.),
dyslexia: Two right hemispheres and none left.
195, 309-311.
Wuillemin, D. B., Krane, R. V., & Richardson, B. L. 1982. Hemispheric differences in picture-word interference. Brain and Language, 16, 121-132.