Hemispheric differences for feature migrations

Acta Psychologica North-Holland

179

83 (1993) 179-201

Hemispheric migrations

differences for feature

John Polich The Scripps Research Institute, La Jolla, USA Accepted

June

1992

Hemispheric differences for feature migrations were investigated in two experiments. Stimulus displays consisting of five small squares arranged in a single row were presented tachistoscopically with the subject instructed to state in which square a horizontal ‘tick’ mark was located. Ticks could occur in any of the three middle squares, with half of the ticks presented on the inside and half presented on the outside of the square relative to the fovea. Experiment 1 presented each array of five squares to the right or left of fixation and manipulated the spacing between the squares. Experiment 2 presented arrays at two different distances from the fovea. In general, ticks migrated toward the fovea more than away from the fovea. Fewer errors were made with stimuli presented to the left visual field/right hemisphere compared to the right visual field/left hemisphere. When ticks migrated away from the fovea, stimuli presented to the right visual field/left hemisphere tended to produce fewer errors than those presented to the left visual field/right hemisphere, although this effect was not robust. Experiment 1 found that increasing the distance between the squares did not affect hemispheric differences reliably. Experiment 2 found that increasing the distance from the fovea increased hemispheric differences as well as the overall error rate. The data suggest that hemispheric differences for featural processing begin very early during sensory analysis.

Theoretical interpretations of cerebral laterality differences have been cogently characterized in a recent review by Sergent and Corballis (1991). In this formulation, hemispheric differences are categorized in terms of each hemisphere’s capacity for particular types of information (verbal/visuospatial), processing operations (analytic/holistic), and with respect to the nature of the representation of information within each hemisphere (high/low spatial frequencies). However, despite the descriptive utility the various theoretical dichotomies Correspondence to: .I. Polich, Dept. of Neuropharmacology TPC-10, The Scripps Institute, 10666 N. Torrey Pines Road, La Jolla, CA 92037, USA. Fax: 619-554-6393; [email protected]

OOOl-6918/93/$06.00

0 1993 - Elsevier

Science

Publishers

B.V. All rights reserved

Research E-mail:

180

J. Polich / Hemispheric feature migration

provide, it is apparent that any one such characterization cannot account for the variegated and ephemeral findings of many laterality studies. Indeed, these difficulties have led some researchers to examine more fundamental visual processes in an effort to understand the origins of laterality effects (Kitterle and Christman 1991; Sergent 1983; Cummings 19851, so that a more comprehensive and perhaps compelling portrait of the theoretical nature of these differences might emerge (cf. Corballis 1989; Kosslyn 1987; Tucker and Williamson 1984). In this context, one approach to elucidating the origins of laterality effects has been to determine whether visual hemispheric differences arise early or late during representational development of a stimulus. This view derives from studies that demonstrate that the two cerebral hemispheres are specialized with respect to their relative capabilities for extracting the basic attributes of visual stimuli, rather than for the type of information per se or for the operations performed after the stimulus attributes have been acquired. Thus, even though task requirements define the critical stimulus features, the left and right hemispheres appear to be differentially sensitive to the acquisition of stimulus qualities (Christman 1989; Sergent 1982, 1983). A variety of data support this assertion. For example, both letter and face stimulus materials are processed more efficiently by the right visual field/left hemisphere (RVF/LH) when a specific feature must be processed, although these same items demonstrate superior left visual field/right hemisphere (LVF/RH) performance when the global attributes of the entire stimulus must be used to perform the task (Hellige et al. 1984; Patterson and Bradshaw 1975; Polich 1978; Polich et al. 1990). A similar pattern of findings has emerged with visually degraded stimulus items: RVF/LH presentations produce superior performance when the physical structure of the stimulus coerces the use of a specific feature to perform the task, otherwise LVF/RH presentations produce superior performance (Hellige 1976; Hellige and Webster 1979; Polich et al. 1986; 1988). Taken together, these results imply that hemispheric differences for visual stimuli occur relatively early during stimulus analysis when the system is engaged in extracting fundamental stimulus attributes (cf. Christman 1989, 1990; Kitterle et al. 1990; Pike and Polich 1988). If visual laterality effects are dependent upon the specific featural aspects used to perform the required task, then hemispheric differences may be determined by those variables that control featural

J. Polich / Hemispheric feature migration

181

processing in general. For example, several non-laterality studies have found that straightlined and curved materials produce feature confusions, with stimuli sharing like-shaped attributes demonstrating poorer performance than items with dissimilar features (Chastain 1981, 1985; La Heij and Van der Heijden 1983; White 1981). Similarly, since items presented in the periphery are degraded more than centrally located items, task performance is often poorer for eccentric relative to fovea1 presentations because more feature migration or perturbation occurs (Banks and White 1984; Banks et al. 1977; Chambers and Wolford 1983; Krueger and Gott 1985; Wolford 1975; Wolford and Chambers 1983). Although the majority of these reports were not concerned with visual field/hemispheric differences, at least a few studies have noted such effects under conditions that encouraged feature perturbation (Chastain 1982, 1986a,b; Wolford and Hollingsworth 1974). When viewed in the context of the laterality data reviewed above, these results suggest that visual field/hemispheric differences are produced at least in part from the perturbation of stimulus features caused by featural similarity and non-fovea1 presentations. The present studies were designed to assess this hypothesis by employing stimuli that have been used to investigate feature migrations (Wolford and Shum 1980). Linear arrays consisting of five small squares with a single horizontal tick in one of the three center squares were projected to the left or right visual fields (see fig. 1). These stimuli were chosen not only because they produce consistent featural migrations, but also because they minimize the potential hemispheric language bias that may operate when letters are used as stimuli. In addition, given the relatively complex patterns previously observed for laterality findings stemming from specific stimulus features (cf. Bryden and Allard 1976; Hellige 1983; Polich 1982), hemispheric differences only may emerge when task processing is sensitive to changes in featural migration per se (Jonsson and Hellige 1986; Polich et al. 1988). Subjects were therefore instructed simply to indicate the location of the tick within the array in order to ascertain possible visual field/hemispheric effects with a task that requires specific featural processing. Two experiments were performed to determine the potential influence of variables that have been shown to affect feature perturbation: experiment 1 manipulated the inter-square distance within the stimulus arrays; experiment 2 manipulated the distance of the arrays from

182

.I. Polich / Hemispheric feature migration

the fovea. Thus, by employing non-letter stimuli known to produce feature migrations, a task sensitive to feature migrations, and manipulations that affect featural processing, any consistent hemispheric differences for feature perturbation should be revealed.

Experiment

1

A major factor that affects feature migration is the distance between stimulus items. These effects have been studied in the context of lateral masking paradigms and indicate that closely spaced items promote feature migration, whereas other visual processes such as gestalt grouping begin to affect feature perturbation when stimulus items are spaced widely (Banks and White 1984; Wolford and Chambers 1983). Because several studies have reported hemispheric differences for laterally masked stimuli (Hellige 1983; Polich 1978; Polich et al. 1986; 19881, the effects of stimulus spacing were investigated in experiment 1 to determine how the spacing between stimulus items contribute to the laterality differences for feature location.

Method

Subjects A total of 32 (16 male, 16 female) undergraduates (mean age = 20.3, SD = 1.8) from the University of California, San Diego served as subjects for course credit. All subjects were right handed as assessed by a short questionnaire (Bryden 1977) and observation of writing. All had normal or corrected-to-normal vision.

Stimuli and design The stimulus materials for the first experiment are illustrated in the top portion of fig. 1 and were modelled after those employed by Wolford and Shum (1980) in size and arrangement. The squares were drawn in black ink on white cards, with the square position numbers presented in fig. 1 only for illustrative purposes and not included on the actual stimulus cards. The ‘close’ and ‘wide’ arrays were positioned 0.2” respectively, to the right or left of fixation as measured from the square side nearest the fixation point. As portrayed in the figure, the arrays were arranged so that the middle squares (positions 2, 3, 4) increased in distance from the fovea systematically across the visual field. In each array a single horizontal ‘tick’ was drawn on one of the vertical sides of either the 2nd, 3rd, or 4th square. The tick was the same width as a side of the square. The tick could occur on either the inside or outside of the square on which it was drawn relative to the fovea. Hence, a total of 24 cards were constructed from the combination of visual field/hemisphere (left vs. right), square spacing (close vs. wide), tick position (2nd, 3rd, or 4th square), and tick side (inside vs. outside).

183

J. Polich / Hemispheric feature migration Procedure

Stimuli were presented in a Gerbrands three-channel tachistoscope, with a display field luminance of 12.4 cd/m*. After a brief introduction to the equipment, the subject read the task instructions silently. These were reiterated by the experimenter and any questions answered. Subjects were told that on each trial a small red dot would appear in the center of the white viewing field. They were to fixate on the dot and await the stimulus presentation that occurred randomly to the left or right of fixation. Subjects were instructed to respond verbally by stating (1) which square contained the tick and (2) on which side of the square it occurred. The subjects were not told that the ticks could occur only in the three center squares. A diagram was used to indicate that the number of the squares increased from the center outward regardless of whether the arrays were presented in the left or right visual field. The experimenter emphasized the importance of being accurate even though the stimulus arrays 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 1 set later, the trial was begun with a red centerfield fixation dot that was illuminated for 500 ms then removed and followed by the stimulus array presented in the left or right visual field for 100 ms. Subjects responded by stating the number of the square (1, 2, 3, 4, or 5) and the side of the square relative to fovea (inside or outside) in which they thought the tick occurred. No feedback was given. This procedure was repeated for all trials with short breaks provided periodically. The 24 stimulus cards were presented in a random order, reshuffled and presented again for a total of five repetitions of the entire card set. Subjects received eight practice trials before the actual experiment to illustrate the various stimulus items and conditions. The data from the practice trials were excluded from all analyses.

EXPERIMENT

1

EXPERIMENT

2

Close

Wide

4 Near Far Fig. 1. Stimulus

b

4

i -,,4n ’ I-

b

[7

0

&cl

items used in experiments

m”

q

1 and 2. See text for details.

184

_I.Polich / Hemispheric feature migration

Results The overall percentage of error for each stimulus card was computed over the five presentations of that card for each subject. A four-factor (visual field/hemisphere X square spacing X tick side X tick position) analysis of variance was applied to the error percentages obtained for each condition from each subject. Greenhouse-Geisser procedures were used to correct for violations of sphericity inherent in repeated measures designs by adjusting the df appropriately; only the probability values based on these corrections are reported. The results of this analysis are described below. In addition to the analysis of the overall error rates, the error data were assessed for specific error types that were based on previous feature perturbation and hemispheric studies. These error type classifications were adopted here to determine if hemispheric differences would be obtained for specific feature migration error types as well as to facilitate comparisons across studies (cf. Chastain 1986a; Polich et al. 1988; Wolford and Shum 1980). These analyses are important for theoretical reasons as well since it may be the case that specific error types are more prevalent in one hemisphere or the other. If this is the case, then a detailed analysis of hemispheric feature migrations patterns should help pinpoint such effects and provide additional support for the early hemispheric specialization approached proffered here. The various error types are illustrated in fig. 2 with inside and outside tick side stimuli presented separately: whole transposition errors were defined as errors on which the tick was reported to be in a square other than the one in which it was actually presented, although tick side was reported correctly. Feature perturbation errors were defined as errors on which the tick side was perceived incorrectly, i.e. inside ticks were reported as occurring on the outside of a square and vice versa. Hence, a given feature migration was classified as either a whole transposition or feature perturbation error but could not be both. Toward-fovea migration errors were counted when the subject reported that the tick occurred in a position that was nearer to the fixation point than it actually was. Away-from-fovea migration errors were counted when the subject reported that the tick occurred in a position that was farther away from the fixation point than it actually was. Thus, four different possible errors could occur for each stimulus card. Overall error rate The mean error rates for experiment 1 from each visual field/hemisphere and stimulus condition are plotted as a function of actual tick position in fig. 3. As suggested by the error rate profiles, LVF/RH stimulus presentations produced fewer errors than RVF/LH stimulus presentations (43.2% vs. 48.0%), with F(1,31) = 4.8, p < 0.05. However, as is also apparent this effect interacted with all of the independent variables to yield a significant four-way interaction, F(2,62) = 5.7, p < 0.005. These results indicate that although the visual field/hemisphere of stimulus projection does affect feature identification, the magnitude of this effect is influenced by square spacing, tick side, and tick position. To examine the source of this complex interaction, additional three-factor (visual field/hemisphere X tick side X tick position) analyses of variance were performed on the error data from the close and wide stimulus conditions separately. These analyses revealed that the main effect for visual

J. Polich / Hemispheric feature migration

185

RVF/LH

LVF/RH

WHOLE TRANSPOSITION INSIDE TOWARD

cl

0

D

iI--

0

TlCK +

0

--El

a -

AWAY

clL3DOO

+ OUTSIDE

q

0

IJmEl-I30 TICK

TOWARD

AWAY

FEATURE

PERTURBATION

INSIDE TOWARD

AWAY

OD-fr!cl -

TICK +

O--DO

0

+

q

a)-4300

cll+lL~

OUTS/DE TOWARD

q !30

0

+

TlCK 0

O--DO

0

AWAY

Fig. 2. Illustration of whole transposition and feature perturbation error types for toward-fovea and away-from-fovea error migrations. The errors are illustrated for stimulus arrays in each visual field and whether the target tick stimulus was located on the inside or outside of the square relative to fovea.

field/hemisphere was produced primarily by the close spaced stimuli, F(1,31) = 4.5, p < 0.05, as was the interaction between the three factors, F(2,62) = 3.3, p < 0.05, since no significant effects involving the visual field/hemisphere variable were obtained for the wide spaced stimuli. Thus, only when the stimulus squares were spaced relatively closely were reliable visual field/hemispheric effects observed. As the distance of the tick from the fovea increased more errors were produced overall, F(2,62) = 232.7, p < 0.001. In addition, as the distance of the tick from the fovea increased, more errors were made when the tick was on the square’s inside compared to when the tick was on the square’s outside with respect to the fovea as portrayed in fig. 2, F(2,621= 35.2, p < 0.02. This relationship also was affected by the square spacing variable that demonstrated no main effect but did interact with tick side and tick position: wide-spaced squares yielded less of a difference between inside and outside ticks as tick position increased compared to close-spaced squares, F(2,62) = 5.1, p < 0.01.

J. Polich / Hemispheric feature migration

186

WIDE

CLOSE

;‘c-.i i-/x::;

100

100

0

0 2

3

TICK

POSITION

4

2

3

TICK

4

POSITION

(left vs. Fig. 3. Overall error rates as a function of tick position for each visual field/hemisphere right), array spacing type (close vs. wide), and migration error direction (toward-fovea vs. away-from-fovea) for inside and outside ticks (experiment 1).

Whole transposition and feature perturbation errors Based on the classification scheme outlined above, the error rate for each error type was computed for each stimulus condition over the five presentations of that card for each subject. Overall, 37.4% of errors were whole transpositions (33.7% toward-fovea, 3.7% away-from-fovea) and 62.7% were feature perturbations (46.1% toward-fovea, 16.6% away-from-fovea). These overall error rates for each error type were then used to compute the error rates each subject produced for each error type using the following formula: (each subject’s number of errors for each error type/total number of errors for that error type) x 100. These error rates represent the normalized error rate for each specific error type with respect to the total number of errors so that hemispheric differences between conditions can be examined with the gross, overall effects of the different types of errors held constant. These error rates are expressed as ‘Percent of Errors’ in figs. 4 and 5. The means of each error rate type computed across subjects in experiment 1 for each visual field/hemisphere and stimulus condition are presented as a function of tick position for the whole transposition errors in fig. 4 and for the feature perturbation errors in fig. 5. Preliminary analysis indicated that the pattern of errors originating from the inside and outside ticks differed appreciably for both error types. Hence, separate analyses for the errors from inside and outside tick stimuli were employed in

J. Polich / Hemispheric feature migration

.-c-

RVFILH-TOWARD

--&-

LVFiRH-TOWARD

- -.-

-

RVFILH-AWAY

- -I-

-

LVF/RH-AWAY

WHOLE

TRANSPOSITION

CLOSE

WIDE INSIDE

2

3

TICK

187

TICK

2

4

POSITION

3

TICK

OUTSIDE

4

POSITION

TICK 80

40

20

.0 60 : 2

3

TICK

Fig. 4. Percent field/hemisphere

4

POSITION

of whole transposition errors as a function (left vs. right), array spacing type (close

-m._

---______,__1::=_

2

3

TICK

4

POSITION

of tick position for each visual vs. wide), and migration error

direction (toward-fovea vs. away-from-fovea) for ticks occurring on the inside and outside of the square (experiment 1).

order to specify the locus of hemispheric effects in an efficacious fashion. Feature migrations that originated with ticks located on the inside of the square relative to the fovea are illustrated in the top portion of the figure, and those that originated with ticks located on the outside of the square relative to the fovea are illustrated in the bottom portion of the figure. The results of these analyses will be discussed by presenting the visual field/hemispheric effects first, with summaries of the effects for other stimulus input variables presented second.

J. Polich / Hemispheric feature migration

18X

FEATURE

PERTURBATION

CLOSE

WIDE INSIDE

co

5 g

TlCK

8o oo-

w

b

40-

5 02

3

TICK

4

2

POSITION

OUTS/DE

2

POSITION

TICK

POSITION

4

TICK

4

3

TICK

3

TICK

a

2

POSITION

4

Fig. 5. Percent of feature perturbation errors as a function of tick position for each visual field/hemisphere (left vs. right), array spacing type (close vs. wide), and migration error direction

(toward-fovea

vs. away-from-fovea)

for ticks occurring

square (experiment

on the inside and outside of the

1).

A four-factor (visual field/hemisphere x tick position X Mole transposition errors. square spacing x migration direction) analysis of variance was applied to the error percentages obtained for each condition from each subject. These are illustrated in top portion of fig. 4. Inside ticks. Stimulus presentations to the LVF/RH produced fewer errors than those directed to the RVF/LH (13.3% vs. 19.1%), indicating that the right hemisphere localized tick placements more readily than the left hemisphere when whole

.I. Polich / Hemispheric feature migration

189

transposition migrations alone were considered, F(1,31) = 9.3, p < 0.005. As illustrated in the top half of fig. 4, the toward-fovea errors demonstrated hemispheric effects more strongly than away-from-fovea errors such that visual/field hemisphere interacted significantly with migration direction, F(1,31) = 12.3, p < 0.002. In addition, the visual field/hemispheric difference increased as the tick position changed, with fewer errors produced by LVF/RH compared to RVF/LH presentations when the tick increased its distance from fixation (i.e. from the 2nd to the 3rd to the 4th square), to yield a significant interaction between visual field/hemisphere and tick position, F(2,62) = 3.7, p < 0.05. As also suggested by the error patterns in the top half of fig. 4, the other major input variables produced significant effects. More errors were obtained for toward-fovea compared to away-from-fovea errors and as tick position increased, with these two variables interacting significantly as well (p < 0.001 in all cases). However, no effects for square spacing were obtained. Outside ticks. The bottom portion of fig. 4 presents the error patterns obtained for the stimuli containing ticks on the outside of the square. No reliable effects of visual field/hemisphere were obtained for errors produced by these stimuli. More errors were obtained again for toward-fovea compared to away-from-fovea errors and as tick position increased, with these two factors producing a significant interaction as well (p < 0.001 in all cases). In addition, square space also interacted with migration direction as well as tick position (p < 0.02 in both cases). The same four-factor (visual field/hemisphere X tick perturbation errors. position x square spacing x migration direction) analysis of variance was applied to the error percentages calculated for the feature perturbation errors. These data are illustrated in fig. 5. Inside ticks. In contrast to the outside tick error patterns, when the tick was located on the inside of the square relative to the fovea, fewer errors were made with stimuli presented to the RVF/LH compared to the LVF/RH (17.1% vs. 25.0%) to yield a strong main effect for visual field/hemisphere, F(1,31) = 8.7, p < 0.01. In addition, more errors were made overall as tick position increased in distance from the fovea (p < 0.001) with tick position and migration direction yielding a significant interaction (p < 0.001). Outside ticks. The bottom portion of fig. 5 illustrates the patterns from the outside tick stimuli for feature perturbation errors. Fewer errors were made when stimuli were presented to the LVF/RH compared to the RVF/LH for this error category (25.5% vs. 30.5%) although this main effect was only marginally significant, F(1,31) = 3.3, p = 0.08. More errors were made for the toward- compared to away-from-fovea migrations, with migration direction interacting with tick position (p < 0.001 in both cases). Wide square spacing produced more errors than close square spacing (p < 0.02) and the spacing factor interacted with migration direction (p < 0.02). Feuture

Discussion Hemispheric

feature

perturbation

The present findings demonstrate that visual field/hemispheric differences feature migrations exist and are influenced by the tick side and tick position,

for with

more errors made when ticks migrated toward the fovea in agrccmcnt with previous feature perturbation studies (Banks et al. 1977; Chastain 1986a.b; Wolford 1975; Wolford and Hollingsworth 1974; Wolford and Shum 1980). Away-from-fovea errors were much lower and yielded far less robust visual field/hemispheric differences, although this factor did interact with the other stimulus presentation variables. Despite previous suggestions of visual field/hemispheric lateral masking diffcrcnccs for the stimulus spacing variable (Hellige 19X3: Polich 197X), only relatively weak effects were observed for this factor. However, given the general laterality findings obtained, it is reasonable to conclude that reliable visual ficld/hemisphcric differences were obtained for the prcscnt stimulus arrays and feature location task. As is clear from the error patterns portrayed in figs. 4 and 5. the type of feature migration produced specific visual ficld/hcmisphcric effects as did the specific qualities of the stimulus display - a common finding for fcaturc migration studies (Banks and White 1984; Chastain 1981, 1985, 1986a,b; White 19X1). In addition, specific hemispheric effects were obtained for whole transposition errors compared to feature perturbation errors, with toward- compared to away-from-fovea feature movements interacting with whether the tick was originally presented on the inside compared to outside of the target stimulus square relative to the fovea. The complex nature of these effects suggests that several factors are contributing to the observed hemispheric patterns. These factors include: (1) whether stimulus variables produce pronounced rather than weak feature migration (toward- vs. away-from-fovea, tick distance from the fovea); (21 how the pcrccption of feature migration is affected by the physical relationship between the specific stimulus and the surrounding items (whole transposition vs. feature perturbation); and (3) if task performance can be best accomplished primarily by either isolating the target stimulus feature or employing the spatial relationships bctwccn the target fcaturc and its stimulus context (left vs. right hemisphere advantage). Indeed, the interaction of these factors conspired to produce opposite and/or eliminate hemispheric effects for the whole transposition compared to feature perturbation error category depending on whether the tick was on the inside or the outside of the square. However, given that both types of errors appear to occur at the feature identification stage (Wolford and Shum 19801, it seems likely that the obscrvcd hemispheric effects originate very early during visual information processing. Scanning

bias effects

An alternative explanation for these results might be couched in terms of scanning biases. Assuming that the well-learned reading tendencies for left-to-right scanning would be employed consistently regardless of which visual field contained the stimulus arrays, it might be supposed that stimulus arrays presented to the RVF/LH would be scanned moving away from the fovea and stimulus arrays presented to the LVF/RH would be scanned moving toward the fovea. Hence, more errors might be obtained for the RVF/LH stimuli since the scan would be in the direction of peripheral vision with the opposite finding obtained for LVF/RH stimuli since the scan would be toward the fovea. This account cannot explain the present findings for several reasons: (I) If scanning biases were the major cause underlying the visual field differcnccs obtained.

.I. Polich / Hemispheric feature migration

191

consistent scanning effects should be observed regardless of variation in stimulus feature conditions. As is apparent from figs. 4 and 5 this result was not obtained, since substantial differences for visual field/hemisphere error patterns were found between the various experimental conditions as well as error migration classifications. (2) A left-to-right scan of the arrays also might produce an interaction between the visual field/hemisphere of presentation and tick side, since ticks occurring on the inside would be favored for RVF/LH presentations, while ticks occurring on the outside would be favored for LVF/RH presentations. No such consistent interaction was obtained. (3) If scanning biases were operating to a significant degree, then scanning direction might interact with feature migration direction. In this case, toward-fovea errors should be greater in the LVF/RH, whereas away-from-fovea errors should be greater in the RVF/LH, since the scanning direction might augment error production by contributing to the propensity for feature perturbation. No such error pattern was observed, with many of the findings demonstrating exactly the opposite outcome. Thus, the general lack of any consistent effects or appropriate interactions across experimental conditions in the present study, when taken in conjunction with previous feature perturbation studies that controlled for scanning direction and found only minimal influence of scanning on feature migration (Wolford and Hollingsworth 1974; Wolford and Shum 1980), suggest that a scanning bias explanation for the present findings is not tenable. Theoretical

implications

The obtained laterality error patterns indicate a differential sensitivity to feature processing by each cerebral hemisphere and may account for the variegated laterahty effects found in previous feature perturbation studies wherein left hemisphere (Chastain 1981, 1982, 1986b; Wolford and Chambers 19831, right hemisphere (Chastain 1986a; Wolford and Shum 1980), and no hemispheric superiorities (Chastain 1985; Krueger and Gott 1985; White 1981) have been reported. In the present case, however, the obtained LVF/RH advantage for many of the conditions may have resulted from specific requirements involved for the tick localization task. Under the conditions of relative perceptual degradation (non-fovea1 rapid stimulus presentation), determining that square contained the tick could be accomphshed more readily by attempting to localize the target stimulus in terms of its spatial relationship with the surrounding items, thereby favoring a right hemisphere advantage (Hellige et al. 1984; Polich et al. 1990). For conditions in which this type of processing was inappropriate, a weak left hemisphere effect was obtained concomitantly with many more awayfrom-fovea feature perturbation errors. Hence, the stimulus conditions appeared to contribute to the overall hemispheric differences observed by affecting the manner in which the features migrated when tick localization was required. This interpretation implies that hemispheric effects originate at the level of featural extraction during visual information processing and support the theoretical view that how the stimulus representation is formed is a critical component in defining hemispheric specialization (Sergent and Corballis 1991). Moreover, such effects are not specific to the present task situation, since similar Iaterality differences have been reported in a variety of stimulus/task situations such that RVF/LH superiorities are obtained when a specific feature can be used for task performance,

J. Polich / Hemispheric feature migration

192

and LVF/RH advantages employed for processing 1976, 1983; Hellige et al. and previous hemispheric differences can emerge at 1989, 1990; Kitterle et al.

Experiment

are obtained when global attributes of the stimulus are degraded visual stimuli (Bryden and Allard 1976; Hellige 1984; Polich et al. 1986, 1988). Given the present pattern results, it seems reasonable to conclude that hemispheric a very early stage during perceptual processing (Christman 1990; Pike and Polich 1988).

2

If it is the case that hemispheric differences exist for featural processing, then variables that influence the degree of featural perturbation also should affect the degree of visual field/hemispheric difference obtained. A major factor that contributes to feature migration is distance from the fovea, with stimuli presented at greater distances demonstrating more featural migrations than those presented closer to the fovea (Banks et al. 1977; Chastain et al. 1986; Wolford 1975). Hence, the second experiment manipulated this variable to determine whether fovea1 distance would influence hemispheric differences for feature migration. If migration errors do increase as retinal distance increases, it might be expected that the visual field/hemispheric differences also would become more pronounced. Stimuli similar to those used previously were employed, with a condition in which the arrays of squares were presented further from the fovea than was done in the previous study. Method A new sample of 32 subjects (mean age = 20.4, SD = 2.3) was recruited from the same source with the same characteristics as the first study. The stimulus items employed are illustrated in the bottom portion of fig. 1. The linear array of squares were similar to those used in the first experiment for the ‘near’ condition, with the fovea1 distance for the ‘far’ condition manipulated as illustrated in fig. 1. The ‘near’ stimulus. arrays were presented with the square nearest the fixation point 0.2” away from fixation, as was done in the ‘close’ condition of the first experiment. The ‘far’ stimulus arrays were presented with the square nearest the fixation point 1.25” away from fixation. A total of 24 stimulus cards were constructed from the combination of visual field (left vs. right), array distance (near vs. far), tick position (2nd, 3rd, or 4th square), and tick side relative to the fovea (inside vs. outside). All other aspects of the second experiment were identical to those employed in experiment 1. Results and discussion Overall error rate The percentage of error for each stimulus card was computed over the five presentations of that card for each subject. The mean error rates for experiment 2 from each visual field/hemisphere and stimulus condition are plotted as a function of actual tick position in fig. 6. A four-factor (visual field/hemisphere x array distance

J. Polich / Hemispheric feature migration

193

FAR

NEAR

o’

0

3

2

TICK

POSITION

4

’ 3

2

TICK

4

POSITION

Fig. 6. Overall error rates as a function of tick position for each visual field/hemisphere (left vs. right), array distance (near vs. far), and migration error direction (toward-fovea vs. away-fromfovea) for inside and outside ticks (experiment 2).

x tick side x tick position) analysis of variance was applied to the error percentages obtained for each condition from each subject. Greenhouse-Geisser corrections procedures again were applied and only the probability values based on these corrections are &ported. The general pattern of overall error rates was similar to that obtained in experiment 1. Left visual field/right hemisphere stimulus presentations again produced overall fewer errors than right visual field/left hemisphere stimulus presentations (58.6% vs. 60.7%). As tick position increased in distance from the fovea, fewer errors were made by LVF/RH compared to RVF/LH presentations to yield a significant interaction between visual field/hemisphere and tick position, F(2,62) = 7.3, p < 0.005. In addition, visual field/hemisphere of presentation also interacted significantly with tick side and array distance, F(1,31) = 7.9, p < 0.01, as well as producing a four-factor interaction between these three factors and tick position, F(2,62) = 4.5, p < 0.02. These interactions indicate that the error rates for each visual field/hemisphere were influenced by the major independent variables such that array distance appreciably modified the effects of tick side and tick position. To examine the source of these interactions, additional three-factor (visual field/hemisphere x tick side x tick position) analyses of variance were performed on the error data from the near and far

194

J.

Polich / Hemispheric feature migration

stimulus conditions separately. For the near stimulus arrays, inside ticks produced about the same proportion of errors for each visual field/hemisphere, whereas outside ticks produced significantly fewer errors for RVF/LH compared to LVF/RH presentations, F(1,31) = 5.8, p < 0.05. Also, RVF/LH presentations demonstrated fewer errors than LVF/RH presentations as tick position increased in distance from the fovea to produce a significant interaction between visual field/hemisphere and tick position, F(2,62) = 3.9, p < 0.05. However, for far stimulus arrays, just the opposite interaction between visual field/hemisphere and tic position was obtained, F(2,62) = 4.7, p < 0.05. Thus, even though the visual field/hemisphere of stimulus projection does affect feature identification, the magnitude of this influence is determined strongly by array distance, tick side, and tick position. Substantially more errors were made with arrays presented far compared to near the fovea, F(1,31) = 118.5, p < 0.001. As the distance of the tick from the fovea increased, more errors were produced overall, F(2,62) = 111.8, p < 0.001. Fewer errors were made when the tick was located on the inside of the square relative to the outside position, F(1,31) = 5.3, p < 0.05. Whether the tick was on the inside or outside of the square with respect to the fovea interacted with tick position, F(2,62) = 31.1, p < 0.001, and position interacted with array distance, F(2,62) = 16.4, p < 0.001. Whole transposition and feuture perturbation errors The tick location identification errors were computed in the same fashion as in the previous study. The overall rates for whole transposition compared to feature perturbation errors were 31.8% and 46.7% for toward-fovea errors, with 5.4% and 16.1% for away-from-fovea errors, respectively. The mean error rates for experiment 2 from each visual field/hemisphere and stimulus condition are plotted as a function of actual tick position for the whole transposition errors in fig. 7 and feature perturbation errors in fig. 8. Feature migrations that originated with ticks located on the inside of the square relative to the fovea are illustrated in the top portion of the figures, and those that originated with ticks located on the outside of the square relative to the fovea are illustrated in the bottom portion of the figures. mole trilnsposition errors. A four-factor (visual field/hemisphere x tick position x array distance x migration direction) analysis of variance was applied to the whole transposition error percentages obtained for each condition. These error patterns are portrayed in top portion of fig. 7. Inside ticks. The overall hemispheric differences were similar to those observed in the first study. In the present case, opposite hemispheric effects were obtained between the toward-and away-from-fovea errors: LVF/RH superiorities were obtained for the toward-fovea effects as in experiment 1, with a tendency for RVF/LH superiority obtained for the away-from-fovea effects. These results were sufficiently robust to produce a significant interaction between visual field/hemisphere and error migration direction, F(1,31) = 17.2, p < 0.001. Thus, LVF/RH presentations yielded fewer toward-fovea errors. Main effects for tick position and migration direction were obtained (p < 0.0011, with the additional finding of more overall errors obtained for the ‘far’ compared to ‘near’ stimulus arrays (p < 0.01). Tick position and migration direction also interacted

195

J. Polich / Hemispheric feature migration

-@-

RVF/LH-TOWARD

--If

LVFIRH-TOWhRD

- -.-

-

RVF/LH-AWAY

--¤-

-

LVF/RH-AWAY

WHOLE

TRANSPOSITION FAR

NEAR

60-

40-

20-

o3

2

TICK

POSITION

OUTS/DE

3

2

TICK

POSITION

4

TICK

POSITION

TICK

POSITION

4

TICK

2

3

4

Fig. 7. Percent of whole transposition errors as a function of tick position for each visual field/hemisphere (left vs. right), array distance (near vs. far), and migration error direction (toward-fovea vs. away-from-fovea) for ticks occurring on the inside and outside of the square (experiment 2).

significantly as found previously (p < 0.001). These results were similar to those of the first experiment, except that array distance appeared to affect both hemispheric and the other input variables more strongly than did the inter-square spacing variable of experiment 1. Outside ticks. The bottom portion of fig. 7 presents the error patterns obtained for the stimuli containing ticks on the outside of the square. The only effect of visual field/hemisphere for these stimuli was a marginal interaction with the array distance variable, F(1,31) = 3.6, p = 0.07. Again, more errors were found for toward-fovea

J. Polich / Hemispheric feature migration

196

compared to away-from-fovea errors and as_ tick position increased, with these two factors producing a significant interaction (p < 0.001 in all cases>. No other effects for array distance were obtained. This pattern of effects was virtually identical to those obtained in experiment 1 for this conditfon. The same four-factor (visual field/hemisphere X tick Feature perturbation errors. position x array distance x migration direction) analysis of variance was applied to the error percentages calculated for the feature perturbation errors. These data are illustrated in fig. 8.

PERTURBATION

FEATURE NEAR

FAR

INSIDE

TICK

01 3

2

TICK

POSITION

OUTSIDE

3

2

TICK

POSITION

4

3

2

4

TICK

POSITION

TICK

POSITION

4

TICK

3

2

4

Fig. 8. Percent of feature perturbation errors as a function of tick position for each visual field/hemisphere (left vs. right), array distance (near vs. far), and migration error direction (toward-fovea vs. away-from-fovea) for ticks occurring on the inside and outside of the square (experiment 2).

.I. Polich / Hemispheric feature migration

197

Inside ticks. No significant visual field/hemispheric effects were obtained for this condition, although the overall pattern was somewhat similar to that observed in experiment 1 with respect to the increased proportion of away-from-fovea migration errors. Main effects for array distance, with overall more errors obtained for the ‘far’ compared to ‘near’ arrays (p < 0.05) and tick position were found (p < 0.01). The interactions between migration direction and tick position as well as that between array distance and tick position were significant (p < 0.001). Thus, no hemispheric effects were observed, but the other stimulus input variables produced effects similar to those obtained previously. Outside ticks. The bottom portion of fig. 8 illustrates the patterns from the outside tick stimuli for feature perturbation errors. The major hemispheric effect observed for these data was the general RVF/LH advantage obtained for the ‘near’ arrays, with a LFV/RH superiority for the toward migration errors of the ‘far’ arrays and the opposite tendency for the away-from-fovea errors. Tick position also influenced these results in the usual manner for the toward- and away-from-fovea feature migrations to contribute to a three-way interaction between visual field/hemisphere, error migration direction, and array distance, F(1,31) = 3.9, p < 0.03. That array distance affected the hemispheric outcomes appreciably also was demonstrated by a marginally significant interaction between the visual field/hemisphere, array distance, and migration direction factors, F&31) = 3.6, p = 0.07. To examine the source of these interactions, additional three-factor (visual field/hemisphere x tick position X migration direction) analyses of variance were performed on the error data from the near and far stimulus conditions separately. For the near stimulus arrays, RVF/LH presentations demonstrated significantly fewer errors compared to LVF/RH presentations as tick position increased away from the fovea to yield a significant interaction between these two factors, F(2,62) = 4.4, p < 0.05. As indicated by the lower left portion of fig. 8, only when the ticks were in the position nearest to the fovea were strong visual field/hemispheric differences obtained. For the far stimulus arrays, visual field/hemisphere and migration direction interacted significantly, F(1,31) = 5.6, p < 0.05, as is apparent in the lower right portion of fig. 8. This pattern is somewhat different from that obtained in experiment 1, since the ‘near’ arrays produced RVF/LH advantages with superiorities for this visual field/hemisphere of presentation condition also obtained for the away-from-fovea migration errors - results in essentially the opposite direction as found for the ‘close’ condition of the first experiment (cf. fig. 5, lower left). The reasons for this discrepancy between similar conditions of the two studies are not clear. However, given the overall similarity and relative consistency between analogous conditions for the remaining error types (cf. figs. 4 and 5 vs. 7 and 8), it seems likely that the obtained feature perturbation errors for outside ticks in this context may be an aberration. Current studies with similar stimuli support this view, since results similar to the ‘close’ condition of experiment 1 have been obtained. The same overall effects of error migration direction, array distance, and tick position were obtained (p < 0.001 in all cases). The array distance factor interacted with tick position as well as error migration direction (p < 0.005 in both cases). Hence, array distance appears to have influenced the hemispheric and other visual input variables to a greater degree than did square spacing of experiment 1.

198

J. Polich / Hemispheric

featuremigration

Taken together, these findings again indicate that visual field/hemispheric differences for feature migration exist. In contrast to experiment 1, however, LVF/RH presentations did not demonstrate generally consistent advantages over RVF/LH presentations for the tick localization task. Rather, the stimulus arrays of the present study effected much stronger away-from-fovea error patterns than those observed previously. In addition, array distance from the fovea influenced performance in a manner consistent with previous feature migration studies (Banks and White 1984; Wolford and Chambers 1983) and contributed to the visual/field hemispheric patterns obtained as also has been observed (Polich et al. 1986, 1988).

General discussion The present experiments demonstrate that feature migrations can occur with visual field/hemisphere specificity. LVF/RH presentations were generally superior relative to RVF/LH presentations for feature migrations toward the fovea. Because the visual field/hemispheric differences occurred in conjunction with inter-item spacing and retinal locus effects (Banks and White 1984; Banks et al. 1977; Chambers and Wolford 1983; Wolford 1975; Wolford and Chambers 1983), it is reasonable to assume that laterality differences for feature processing are a fundamental attribute of the visual system in the same fashion as has been observed for spatial frequency (Kitterle et al. 1990; Sergent 1982, 19831, stereoscopic depth (Grabowska 19831, and edge discrimination (Pike and Polich 1988; Polich and Aguilar 1990). Although the mechanisms underlying these laterality effects are not yet well understood and are actively being pursued (e.g., Hellige et al. 1989; Kitterle and Kaye 1985; Peterzell et al. 1989), it is clear that feature perturbation appears to contribute to hemispheric differences in a significant fashion when task performance requires specific featural processing. Thus, at least some portion of hemispheric differences for visual information processing must originate in the early stages of stimulus representation formation (Sergent and Corballis 1991; Sergent 1983) rather than later during cognitive-perceptual processing (cf. Kosslyn 1987; Sergent 1991). The LVF/RH effects found for the tick location task were consistent across a variety of perceptual variables, but exactly how the two hemispheres differ in their feature extraction capabilities is still an open question. The current task situation required the identification of a specific stimulus location under presentation conditions that

J. Polich / Hemispheric feature migration

199

made localization of the migrating tick feature visually very difficult perceptual and task demands that are critical for the emergence of hemispheric differences (Kitterle et al. 1990; Hellige and Sergent 1986). In general, when a specific feature can be perceived correctly and employed to perform the task, RVF/LH advantages usually are observed; when the whole stimulus must be used - often because it is degraded through lateral or stimulus masking - LVF/RH advantages typically are obtained (Hellige 1983; Hellige et al. 1984; Moscovitch and Radzins 1987; Polich 1978, 1982; Polich et al. 1986, 1990). A few feature migration studies also have reported similar hemispheric effects such that RVF/LH superiorities are obtained for specific feature utilization (Chastain 1982, 1986b; Wolford and Hollingsworth 1974) and LVF/RH superiorities are obtained for gestalt stimulus processing (Chastain 1986a; Wolford and Shum 1980). However, despite this similarity of the present hemispheric effects to previous findings with degraded stimuli (Hellige 1983; Hellige and Webster 1979; Polich 1978; Polich et al. 19881, precisely what stimulus conditions will produce which hemisphere advantage is still uncertain even though these laterality effects for feature migration appear quite strong. Conclusions In sum, the hemispheric differences for feature perturbation observed in the present studies support the hypothesis that cerebral laterality effects begin early during visual information processing. For migrations toward the fovea, this process was best effected by the LVF/RH under most stimulus arrangements and may have obtained because the response demands required highly spatial processes (i.e. localization of the tick in the array) - conditions usually conducive to right hemisphere effects, rather than requiring a description or stimulus identification as has often been used in feature perturbation studies. The present findings therefore must be viewed in the context of the feature localization task when comparing them to previous feature migration studies that have reported visual field/hemispheric effects. Thus, given the consistent patterns observed across these and previous experiments, it can be concluded that visual field/hemispheric differences do exist for feature migrations and that such hemispheric effects originate very early during stimulus processing.

200

J. Polich / Hemispheric feature migration

References Banks, W.P., K.M. Bachrach and D.W. Larson, 1977. The asymmetry of lateral interference in visual letter identification. Perception and Psychophysics 22, 232-240. Banks, W.P. and H. White, 1984. Lateral interference and perceptual grouping in visual detection. Perception and Psychophysics 36, 285-295. Bryden, M.P., 1977. Measuring handedness with questionnaires. Neuropsychologia 15, 617-624. Bryden, M.P. and F. Allard, 1976. Visual hemifield differences depend on typeface. Brain and Language 3, 191-200. Chambers, L. and G. Wolford, 1983. Lateral masking vertically and horizontally. Bulletin of the Psychonomic Society, 21, 459-461. Chastain, G., 1981. Asymmetric identification of parafoveal stimulus pairs: Feature perturbations or failure in feature extraction? Canadian Journal of Psychology 35, 13-23. Chastain, G., 1982. Confusability and interference between members of parafoveal letter pairs. Perception and Psychophysics 32, 576-580. Chastain, G., 1985. Positional differences in performance on members of confusable and nonconfusable letter pairs. Journal of Experimental Psychology: Human Perception and Performance 11, 752-764. Chastain, G., 1986a. Figure mislocalizations as predicted from feature perturbation theory. Canadian Journal of Psychology 40, 83-96. Chastain, G., 1986b. Evidence for feature perturbations from character misidentifications. Perception and Psychophysics 39, 301-306. Christman, S., 1989. Perceptual characteristics in visual laterality research. Brain and Cognition 11, 238-257. Christman, S., 1990. Effects of luminance and blur on hemispheric asymmetries in temporal integration. Neuropsychologia 28, 361-374. Corballis, M.C., 1989. Laterality and human evolution. Psychological Review 96, 492-505. Cummings, J.L., 1985. ‘Hemispheric specialization in visual perceptual and visualspatial function’. In: D.F. Bensen and E. Zaidel (eds.), The dual brain: Hemispheric specialization in humans (pp. 233-246). New York: The Guilford Press. Grabowska, A., 1983. Lateral differences in the detection of stereoscopic depth. Neuropsychologia 21, 249-257. Hellige, J.B., 1976. Changes in same-different laterality patterns as a function of practice and stimulus quality. Perception and Psychophysics 20, 267-273. Hellige, J.B., 1983. Feature similarity and laterality effects in visual masking. Neuropsychologia 21, 633-639. 1984. Effects of perceptual quality on the Hellige, J.B., W.H. Corwin and J.E. Jonsson, processing of human faces presented to the left and right cerebral hemispheres. Journal of Experimental Psychology: Human Perception and Performance 10, 90-107. Hellige, J.B. and J. Sergent, 1986. Role of task factors in visual field asymmetries. Brain and Cognition 5, 200-222. Hellige, J.B., A. Taylor and T. Eng, 1989. Interhemispheric interaction when both hemispheres have access to the same stimulus information. Journal of Experimental Psychology: Human Perception and Performance 15, 71 l-722. Hellige, J.B. and R. Webster, 1979. Right hemisphere superiority for initial stages of letter processing. Neuropsychologia 17, 653-660. Jonsson, J.E. and J.B. Hellige, 1986. Lateralized effects of blurring: A test of the spatial frequency model of cerebral asymmetry. Neuropsychologia 24, 351-362. Kitterle, F.L. and S. Christman, 1991. ‘Symmetries and asymmetries in the processing of sinusoidal gratings’. In: F.L. Kitterle ted.), Cerebral laterality: Theory and research (pp. 224-231). Hillsdale, NJ: Erlbaum.

.I. Polich / Hemispheric feature migration

201

Kitterle, F.L., S. Christman and J.B. Hellige, 1990. Hemispheric differences are found in the identification, but not detection, of low versus high spatial frequencies. Perception and Psychophysics 48, 297-306. Kitterle, F.L. and R. Kaye, 1985. Hemispheric symmetry in contrast and orientation sensitivity. Perception and Psychophysics 37, 391-396. Kosslyn, S.M., 1987. Seeing and imaging in the cerebral hemispheres: A computational approach. Psychological Review 94, 148175. Krueger, L.E. and R. Gott, 1985. Effects of lateral masking and letter reversal on same-different judgments. Bulletin on the Psychonomic Society 23, 185-188. La Heij, W. and A.H.C. van der Heijden, 1983. Feature-specific interference in letter identification. Acta Psychologia 53, 37-60. Moscovitch, M. and M. Radzins, 1987. Backward masking of lateralized faces by noise, pattern, and spatial frequency. Brain and Cognition 6, 72-90. Patterson, K. and J.L. Bradshaw, 1975. Differential hemispheric mediation of nonverbal visual stimuli. Journal of Experimental Psychology: Human Perception and Performance 3,246-252. Peterzell, D.H., L.O. Harvey and C.D. Hardyck, 1989. Spatial frequencies and the cerebral hemispheres: Contrast sensitivity, visible persistence, and letter classification. Perception and Psychophysics 46, 443-455. Pike, J. and J. Polich, 1988. Hemispheric differences for visual evoked potentials from checkerboard stimuli. Neuropsychologia 3, 947-952. Polich, J., 1978. Hemispheric differences in stimulus identification. Perception and Psychophysics 24, 49-57. Polich, J., 1982. Hemispheric differences for visual search: Serial vs. parallel processing revisited. Neuropsychologia 20, 297-307. Polich, J. and V. Aguilar, 1990. Hemispheric local/global processing revisited. Acta Psychologica 74, 47-60. Polich, J., D.L. Crossman and D.P. DeFrancesco, 1988. Hemispheric differences for feature perception. Psychological Research 50, 12-18. Polich, J., D., DeFrancesco, J. Garon and W. Cohen, 1990. Hemispheric differences in visual search of simply line arrays. Psychological Research 52, 54-61. Polich, J., W.H. Lentz and D.L. Crossman, 1986. Hemispheric differences for feature detection. Neuropsychologia 24, 591-595. Sergent, J., 1982. Theoretical and methodological consequences of variation in exposure duration in visual laterality studies. Perception and Psychophysics 31, 4511-461. Sergent, J., 1983. Role of the input in visual hemispheric asymmetries. Psychological Bulletin 93, 481-512. Sergent, J., 1991. Judgments of relative position and distance on representations of spatial relations. Journal of Experimental Psychology: Human Perception and Performance 91, 762-780. Sergent, J. and M.C. Corballis, 1991. ‘Ups and downs in cerebral lateralization’. In: F.L. Kitterle (ed.), Cerebral laterality: Theory and research (pp. 175-200). Hillsdale, NJ: Erlbaum. Tucker, D.M. and P.A. Williamson, 1984. Asymmetric neural control systems in human self-regulation. Psychological Review 91, 185-215. White, M.J., 1981. Feature-specific border effects in the discrimination of letter-like forms. Perception and Psychophysics 29, 156-162. Wolford, G., 1975. Perturbation model for letter identification. Psychological Review 82, 184-199. Wolford, G. and L. Chambers, 1983. Lateral masking as a function of spacing. Perception and Psychophysics 33, 129-138. Wolford, G. and S. Hollingsworth, 1974. Retinal locus and string position as important variables in visual information processing. Perception and Psychophysics 16, 437-442. Wolford, G. and K.H.D. Shum, 1980. Evidence for feature perturbations. Perception and Psychophysics 27, 409-420.