Hemispheric differences for feature detection

CQ28-3932,86 $3.00+0.00 Pergamon Journals Ltd.

Neuropsychologia, Vol. 24, No 4, pp. 591-595, 1986. Printed in Great Bntam

NOTE HEMISPHERIC

DIFFERENCES

FOR FEATURE

DETECTION*

JOHN POLICH,~ WILLIAM H. LENTZ~ and DEBORAH L. CK~SSMAN~ tDivision

of Preclinical SDepartment

Neuroscience

and Endocrinology, Scripps Clinic and Research Foundation, CA 92037, U.S.A. of Psychology, San Diego State University, San Diego, CA 92182. U.S.A.

La Jolla,

(Accepted 4 January 1986) Abstract-The role of hemispheric processing for visual features was explored by tachistoscopically presenting subjects with stimulus displays composed of the letter I, displays in which a single T was embedded in an array of I’s, or displays in which the letter 0 was embedded in an array of I’s, The number of array elements was also manipulated (4, 16. or 36) to assess the effects of display size on featural detection for each visual field. Subjects verbally indicated whether array elements were all the same, or whether a T or 0 was present. Hemispheric error rates varied as a function of the type of letter to be detected, with left-hemisphere presentations producing superior performance for the detection of T’s, right-hemisphere presentations yielding superior performance for the detection of O’s, and Same displays demonstrating hemispheric patterns consistent with these results. The findings suggest that the analysis of perceptual features during visual stimulus processing may determine hemispheric outcomes in a variety of task situations.

INTRODUCTION EARLY observations

of laterality phenomena led to a verbal/visuospatial hemispheric dichotomy since most right visual field-left hemisphere (RVF-LH) performance superiorities were obtained with letters or words, while most left visual field-right hemisphere (LVF-RH) performance superiorities were obtained with geometric and random forms or faces (e.g. [9, 11, 12, 223). Later studies, however, sometimes obtained just the opposite result such that some ‘verbal’ stimuli demonstrated more rapid and accurate processing with LVF-RH presentations and some ‘visuospatial’ stimuli demonstrated more rapid and accurate processing with RVF-LH presentations, (e.g. [3, 10, 17, 21, 301). These contradictory findings may originate from the different capabilities of each visual field/hemisphere to extract featural information. For example, letter and face processing are more efficient for RVF-LH presentations when the task requires the detection of a specific feature, but yield LVF-RH superiorities when the global attributes of the entire stimulus must be used (e.g. [ 16,20,2426]). Such results could occur because stimulus materials with well-defined angular features generally demonstrate superior performance for RVF LH presentations, while curved lined and/or relatively ill-defined features obtain superior performance for LVF-RH presentations [S, 10, 19, 27, 33, 341. Moreover, degraded stimulus items often produce better performance for RVF-LH presentations 114, 231 if the physical structure of the stimuli is such that the degradation process coerces the use of a specific feature to perform the task, otherwise better performance for LVF-RN presentations typically results [13, 15, 261. The origins of such hemispheric featural preferences may thus stem from very early stages of information processing as suggested by studies which have found visual laterality effects to be sensitive to input variables [29, 317. In particular, the spatial frequency of the target stimulus may help determine hemispheric differences such that RVF--LH presentations yield superior performance for stimuli with relatively high spatial frequencies, while LVF-RF presentations yield better performance for low spatial frequencies [30]. Hence, the specific featural attributes of a given stimulus display could determine laterality outcomes because of hemispheric preferences for small or large features. Some evidence for this assertion has been obtained from tachistoscopic conditions which encouraged featural migration or perturbation across the stimulus display to produce superior RVF-LH performances [28, 36, 381. While letters or other straight-lined stimuli are often employed in these studies, curved

*Address Foundation,

correspondence to: John Polich, Preclinical Neuroscience-BCRl, 10666 N. Torrey Pines, La Jolla, CA 92037, U.S.A.

Scripps

Clinic

and

Research

592

NOTE

stimulus materials have demonstrated similar results. Stimuli sharing like-shape featural attributes (i.e. straight or curved) have yielded more errors than items with dissimilar featural construction [6,7, 18,353, altough no laterality effects have yet been noted. Because stimulus retinal locus is also a critical variable in these studies, in could readily contribute to lateralityeffects by degrading the target stimulus [I, 371 to affect the items’spatial frequency and visual field effects [7,26, 38, 391. Thus, straight and curved-feature stimuli may produce differential hemispheric outcomes because of lateralized sensitivities to the spatial attributes of the stimulus features and their retinal locus. If hemispheric differences for featural detection are a contributing factor to visual information processing in the manner suggested by these laterality, masking, and feature perturbation studies, then detection of straight-lined features should be better performed by RVF-LH presentations, while the detection of curved-lined and whole patterns should be better performed by LVF-RH presentations. The present study was designed to examine systematically hemispheric differences for featural analysis by presenting both straight-lined and curved stimulus features which would encourage stimulus migrations in a tractable fashion. Stimulus display size was manipulated to increase the amount of stimulation to the peripheral portions of the retina and promote opportunities for featural interaction. Use of various display sizes in this context will also permit comparison of the findings with previous hemispheric studies which suggest that variegated featural perception increases performance variability when larger numbers of array elements ate presented [26-281.

METHODS Subjects Thirty-two undergraduate students (16 male, 16 female) served as subjects for course credit. handed as assessed by a short questionnaire [4] and had normal or corrected-to-normal vision. Stimuli

All were right-

and apparatus

The stimulus materials consisted of rectangular arrays which were composed entirely of the upper-case letter I, arrays which contained a single upper-case T embedded in the array of I’s or arrays which contained a single upper case 0 embedded the array of 1’s. All letters were typed with an IBM Orator typeball (lo-pitch) with a single space between rows, and 1.5 spaces between columns. Stimulus displays contained either 4, 16, or 36 letters in equal numbers ofrows and columns and subtended visual angles of0.4” x 0.6”, 1.1” x 1.5” and 1.7” x 2.3’, respectively. All arrays were presented with the fovea1 edge starting at 1.0” to the right or left of fixation. The aberrant item (T or 0) was randomly located within the array such that each quadrant of the display matrix contained equal numbers ofthe target letter across stimulus repetitions and sizes. All stimuli were presented in a Gerbrands three-channel tachistoscope. Design

Stimulus arrays were presented to the left or right of fixation for 100 msec duration. One third of the arrays contained all I’s (Same), one third of the arrays contained a single T (Detect-T), and one third of the arrays contained a single 0 (Detect-O), Eight replications of each stimulus type for each of the three size conditions were presented to each of the two visual fields for a total of 144 trials within an experimental session. All subjects participated in a single 45-min session which was preceded by practice trials that were illustrative of all the experimental conditions. Procrdurr

Subjects were run individually. After a brief introduction to the equipment, the subject read the task instructions silently. These were reiterated by the experimenter and any question answered. The subjects were told that on each trial a small red dot would appear in the center of the tachistoscope’s white viewing field. They were to fixate on the dot and await the stimulus presentation which occurred randomly to the left or right of fixation. The subjects were instructed to respond ‘Same’ if all the elements in the array were identical, or to respond ‘T’ or ‘0’ if either letter was present in the array. The experimenter emphasized the importance of responding accurately even though the stimulus materials would be presented quickly and hard to see. The sequence of events for each trial was as follows: the experimenter placed a stimulus card in the stimulus channel of the tachistoscope and said ‘ready’. Approximately I set later, the trial was begun with the fixation dot illuminated with a randomly chosen stimulus array. The subject responded by indicating verbally whether all the elements of the array were the same or whether a T or 0 had been presented. The experimenter recorded the response and prepared the equipment for the next trial. This procedure was repeated for all trials with short breaks provided periodically. After the completion of the session, subjects were debriefed about the nature of the experiment.

RESULTS The subject’s responses were scored as an error if they claimed to detect a T or an 0 when all the elements of the array were identical. if they failed to detect a T or an 0 when one was present, or if they reported an incorrect

593

NOTE

detection by stating that they had seen a T when an 0 had been presented and vice versa. All analyses were performed on the number of errors made by each subject for each condition. A preliminary analysis of variance was conducted on the overall number of errors produced by each response (i.e. the number of times ‘Same’, ‘T’, or ‘0’ was given across stimulus conditions) to assess for possible response bias. While subjects said ‘Same’ or ‘T’ significantly more often than ‘0’ (15.3%, 17.8% and 7.3%, respectively), no differences were obtained for the number of ‘same’ compared to ‘T’ responses. Thus, the response rates for all three error types were well below chance (33.3%), and the two most frequently occurring errors demonstrated the same rate. Given this lack of overall response bias in conjunction with the tripartite report procedure and the uncertainty of the resulting error distribution shapes (making the application of standard signal detection procedures problematic), it was assumed that the systematic hemispheric effects obtained stemmed from perceptual errors during stimulus detection rather than response selection pressures. A second preliminary analysis of variance comparing male and female subjects revealed that female subjects made significantly more ‘T’ errors for the Same array condition across all array sizes compared to male subjects (40.5% vs. 27.7%). However, because subject sex was unrelated to the other experimental variables of this condition, or any other stimulus error pattern, it was not considered further. The mean percent error for each experimental condition and error type over all subjects is presented in Fig. I. A four-factor (Hemisphere x Display Condition x Array Size x Error Type) analysis of variance was performed on the number of errors made by each subject in each experimental condition. This analysis substantiated the trends illustrated in Fig. I. Hemisphere of projection interacted significantly with Display Condition [F(2, 62)= 5.4; P=O.O07] and yielded a significant three-way interaction with Display Condition and Array Size F(4, 124)=2.7; P=O.O4]. Display Condition (Same, Detect-T, Detect-O) also produced significant differences in error rate [F(2, 62)=65.4; P
SAME

DETECT

T

DETECT0

r

1 Right Hem --m--

1

“T-“SAME”

NUMBER OF ARRAY ELEMENTS FIG. 1 .Percentage of error trials taken over all subjects (n = 32) for the Same (arrays composed only of the letter I), Detect-T (arrays containing a single T), and Detect-O (arrays containing a single 0) conditions plotted as a function of array size for each hemispheric presentation condition.

Same displays. Despite the patterns of hemispheric differences for arrays in which only the letter I was presented, the three-way interaction between Hemisphere, Array Size and Error Type apparent in the Same portion of Fig. I was only marginally significant (P= 0.08). This may reflect the operation of processing differences between the male and female subjects which could have decreased the overall hemispheric effects. However, analysis of individual error rates suggests that these effects were caused by a few female subjects who produced a disproportionate number of errors in this condition. Array size did yield a significant effect with the mean percentage of error decreasing (23.1%, 22.2% 18.9%) as the size of the array increased [F(2, 62)= 3.8; P=O.O3]. Error Type also demonstrated a significant difference in error rate (34.1%, vs 8.6%). with F(I, 31)= 54.4; P
594

Nom

Detect-Tdisplayr When arrays with a single T were presented, RVF-LH presentations yielded significantly fewer errors than did LVF-RH presentations (15.8% vs 19.5%). with F(1, 31)=7.3; P=O.Oll. Hemisoheric differences also tended to increase with increases in array size, but this effect was only marginally significant (P=O,O8). Array Size produced an increase in error rate as the number of display elements increased (3.5%, 20.6%, 28.8%), with F(2, 62)=116.9; P
DISCUSSION These results suggest that hemispheric differences exist for the detection offeatural properties in stimulus patterns, The left hemisphere appeared more sensitive to angular features as reflected by superior performance under the Detect-T display condition, while the right hemisphere appeared more sensitive to curved or circular features as indicated by superior performance under the Detect-O display condition. Hence, fewer errors of each type (i.e. failure to detect the aberrant element or claiming to see the incorrect letter) were made by the left and right hemispheres for the Detect-T and Detect-display conditions, respectively. In addition, array size consistently influenced the magnitude of the laterality effect such that a greater number of display eiements yielded larger hemispheric differences and a higher error rate relative to small array sizes [2&28]. Although the array size effect may have resulted from the increased numbers ofelements in a more peripheral retinal position, given the generally consistent hemispheric effects obtained, it seems more likely that increases in array size promoted feature perturbations and therefore increased laterality differences. The same display condition demonstrated relatively weak hemispheric effects (perhaps due to the peculiar differences in error patterns observed as a function of subject sex). However, the resulting error patterns were consonant with the overall hemispheric effect observed in the other display conditions: more left-hemisphere errors for the incorrect detection ofan angular T when none was present and more right-hemisphere errors for the incorrect detection ofan 0 when none was present. The hemisphere most sensitive to a particular type offeature was also most likely to error in the direction of its sensitivity when no aberrant stimulus was present-an effect observed in several previous laterality reports 127,281. These hemispheric differences appeared to originate from perturbations of the stimulus features during the early stages of sensory data processing and were augmented with increased opportunities for misperception as evidenced by the effects ofarray size [3 1,321. The findings are consistent with tachistoscopic studies which have found superior performance by RVF LH presentations for well-defined angular features and superior performance by LVF-RH presentations for curved-lined or ill-defined stimulus features [S, 10, 19, 27, 33, 341. Because of the similarity between the hemispheric data and the feature perturbation results, it is tempting to suppose that the left hemisphere is sensitive to straight or well-defined stimulus materials of a relatively high spatial frequency, whereas the right hemisphere is sensitive to curved or ill-defined stimuli with relatively low spatial frequencies [30, 311. Although the specificity of each hemisphere for different types of stimulus features reported previously [I. 7, 18, 351 was again observed in the current study, future research should expressly manipulate the shape of the target stimulus features relative to the background items to determine the genesis of hemispheric differences for feature detection.

.-l[,~rl(j~./~~d!/r,~t~~,~~f Sincere thank5 arc due to Dr Alan Litrownik University for his administrative support.

of the Psychology

Department

at San Diego State

REFERENCES I. fhNI(S.

W. I’.. BA(‘IIKA(.H. K. M. ANI) I,,\KsIx. D. W. The asymmetry of Interval interference in visual letter idcntilication. P~r~c~pr. P.s~c/I~~/I~,~. 22, 232 2 19. 1977. 2. BKAIXX~W. J. L. and NI.T-II I ‘roe. N. C. The nature of hcmisphcric specialization in man. Brlrtrr. Bruin Sci. 4, 51 63. 19x2. .1 BKAUSHAW. J. L., GATI:S. A. and PATTEKSON,K. Hemispheric differences in processing visual patterns. Q. JI. rup. P.syh/. 22, 667 681, 1976. 4. BKYIX 4. M. P. Measuring handedness with questionnaires. .~r,~tr~,p.\?,c,/l~~/~~~~~~/ IS, 617 624. 1977. 5 BKYIXX. M. I’. and AI LAKI), I-. Visual hcmilicld differences dcpcnd on typcfacc. Hrc~irl Ltrml. 3, 191- 200. 1976. i: CIIASIAI~. G. A.\ymmc[ric ldcntification ofparafovcal stimulus pairs: fccaturc perturbations or failure in feature cxtraclion? (‘011. ./. /‘.\IY/Ic~/. 35. 13 23. 1981.

NOTE

595

and interference between members of parafoveal letter pairs. Percept. Psychophys. 1982. 8. COHEN, G. Theoretical interpretations of lateral asymmetries. In Divided Visual Field Studirs of Cerebral Organization, J. GRAHAM BEAUMONT(Editor). Academic Press, New York, 1982. 9. FONTENOT, D. J. Visual field differences in the recognition of verbal and nonverbal stimuli in man. J. romp. Physiol. Psychol. 85, 564-659, 1973. 10. FONTENOT,D. J. and BENTON, A. L. Perception of direction in the right and left visual fields. Nruropsycholoyia 10, 447452, 1972. 11. GEFFEN, G., BRADSHAW, L. L. and NETTLETON, N. C. Hemispheric asymmetry: verbal and spatial encoding of visual stimuli. J. exp. Psychol. 95, 25-31, 1972. 12. GKOSS, M. M. Hemispheric specialization for processing visually presented verbal and spatial stimuli. Percept. Psychophys. 12, 357-363, 1972. 13. HELLIGE, J. B. Changes in same-different laterality patterns as a function of practice and stimulus quality. Percrpr. Psychophys. 20, 267-273, 1976. 14. HELLIGE, J. B. Feature similarity and laterality effects in visual masking. Neuropsycho/ogia 21,633%639, 1983. 15. HELLIGE, J. B. and WEBSTER, R. Right hemisphere superiority for initial stages of letter processing. Neuropsycholoyia 17, 653-660, 1979. 16. HELLIGE,J. B., COKWIN, W. H. and JONSSON,J. E. Effects of perceptual quality on the processing of human faces presented to the left and right cerebral hemispheres. J. exp. Psych& Hum. Percept. Pe@rm. 10,9@107, 1984. 17. KLATSKY, R. L. and ATKINSON, R. C. Specialization of the cerebral hemispheres in scanning for information in short-term memory. Percept. Psychophys. 10, 335-338, 1971. 18. LA HEIJ, W. and VAN DER HEIJDEN, A. H. C. Feature-specific interference in letter identification. Acra p.sychol. 53, 37 -60, 1983. 19. LONGUEN, K., ELLIS, C. and IVERSON, S. Hemispheric differences in the discrimination of curvature. Neuropsychologia 14, 195-202, 1976. 20. MARTIN, M. Hemispheric specialization for local and global processing. NeuropsychoIogiu 17, 3340, 1979. 21. MARZI, C. A. and BERLUCCHI, G. Right visual field superiority for accuracy of recognition of famous faces in normals. Neuropsychologia 15, 751-756, 1977. 22. MCKEEVER, W. F. and HULING, M. D. Lateral dominance in tachistoscopic word recognition performances obtained with simultaneous bilateral input. Neuropsycho[ogia 9, 15-20, 1971. 23. MOXOVITCH, M. Laterality and visual masking: interhemispheric communication and the locus of perceptual asymmetries for words. Can. J. Psycho/. 37, 85-106, 1983. 24. MOSCOVIT~H, M., SCULLION, D. and CHRISTIE, D. Early vs late stages of processing and their relation to functional hemispheric asymmetries in face recognition. J. exp. Psycho!., Hum. Percept. Pet@m. 2, 401416. 1976. 25. PATTERSON, K. and BRADSHAW, J. L. Differential hemispheric mediation of nonverbal visual stimuli. J. exp. Psychol., Hum. Percept. Peform. 3, 246252, 1975. 26. POLICH, J, Hemispheric differences in stimulus identification. Percept. Psychophys. 24. 49-57, 1978. 27. POLICH, J. Hemispheric differences for visual search: serial vs parallel processing revisited. Neuropsycholoyia 20, 297 307, 1982. 28. POLICH, J. Hemispheric patterns in visual search. Brain Cognit. 3, 128-l 39, 1984. 29. SERGENT,J. Theoretical and methodological consequences of variations in exposure duration in visual laterality studies. Percept. Psychophys. 31, 451461, 1982. 30. SERGENT, J. The cerebral balance of power: confrontation or cooperation? J. rxp. Psychok, Hum. Percept. Perform. 8, 253-272, 1982. 31. SERGENT, J. Role of the visual input in visual hemispheric asymmetries. Psycho/. Bull. 93, 481-512, 1983. 33. UMILTA, C., BAGNARA, S. and SIMION, F. Laterality effects for simple and complex geometrical figures and nonsense patterns. Neuropsychologia 16, 4349, 1978. 34. UMILTA, C., SALMASO,D., BAGNARA, S. and SIMION, F. Evidence for right hemisphere superiority and for serial search strategy in a dot detection task. Cortex 15, 609 618, 1979. 35. WHITE, M. J. Feature-specific border effects in the discrimination of letter-like forms. Percept. Psychophys. 29, 156-162, 1981. 36. WOLFORD, G. Perturbation model for letter identification. Psychol. Rea. 82, 184199, 1975. 37. WOLFORD, G. and CHAMBERS,L. Lateral masking as a function of spacing. Percept. Psychophys. 33, 129-138, 1983. 38. WOLFORD, G. and HOLLINGWORTH, S. Retinal locus and string position as important variables in visual information processing. Percept. Psychophys. 16, 437442, 1974. 39. WOLFORD, G. and SHUM, K. Evidence for feature perturbations. Percept. Psychophys. 27,409420, 1980. 7. CHASTAIN, G. Confusability

32, 576-580,