Characteristics of Perceptual Grouping in...

Journal of Comparative Psychology Copyright 1997 by the American Psychological Association, Inc. 1997, Vol. 111, No. 2, 126-134 0735-7036/97/$3.00

Characteristics of Perceptual Grouping in Rats

Danie l D. Kury lo , Jess ica Van Nest , and B r y a n K n e p p e r Bowdoin College

Parametric analysis was made of the characteristics by which proximity and alignment serve as cues for perceptual grouping in rats. Rats were initially conditioned to discriminate a series of horizontal lines from vertical lines. Following training, rats were presented with test stimuli that consisted of bistable arrays of disjunct dots. A grouping cue (greater proximity, greater alignment, or both) was randomly assigned to either the horizontal or vertical orientation. The effectiveness of the cues was based on behavioral responses to the cued orientation. Results indicated that proximity served as a cue for perceptual grouping. The effectiveness of the proximity cue was less for rats than found previously in humans and, unlike humans, diminished with increased stimulus scale. Rats did not respond to alignment cues when used in isolation, although alignment facilitated grouping when used in conjunction with proximity cues. Diminished effectiveness of grouping cues likely reduces object recognition abilities, particularly for complex visual stimuli.

Rats possess a highly developed visual system to which a large cortical region is devoted (Sefton & Dreher, 1985). The visual system of rats is thought to serve in the detection of predators and prey, as well as several aspects of naviga- tion (as discussed in Dean, 1990). Investigations of visual function in rats have focused on detection sensitivity and the neural mediation of form discrimination. Absolute thresholds in rats are comparable with those of humans (Birch & Jacobs, 1979; Kurylo, Hansen, & Fulton, 1991; Muntz, 1967; Rosenberger & Ernest, 1971; Tedo, Herreros de Tejada, & Green, 1994; Trejo & Cicerone, 1987), whereas acuity is much poorer (Dean, 1978, 1981), with diminished contrast sensitivity that is restricted to low spatial frequencies (Birch & Jacobs, 1979; Legg, 1984). Rats are never- theless capable of discriminating complex visual patterns (e.g., Lashley, 1930, 1938). Investigations of the effects of cortical lesions on visual pattern discrimination have indicated that multiple cortical systems within occipto-temporal areas mediate form perception in rats (Lashley, 1931; Lavond et al., 1978; McDaniel, Wildman, & Spears, 1979; Tompson & Bachman, 1979; Wuensch, Broome, Means, & Harris, 1978).

Perceptual grouping is a fundamental component of object recognition in which elements of the visual scene are segregated into unified forms. Few investigations, however, have examined perceptual grouping in rats. Lashley (1938) reported that rats trained to discriminate the orientation of lines transferred the discrimination to lines interrupted by

Daniel D. Kurylo, Jessica Van Nest, and Bryan Knepper, De- partment of Psychology, Bowdoin College.

We are grateful to Robert Stephens for the construction of the operant conditioning chamber and to James Garner for data analysis. Jessica Van Nest was supported by a Langbein Research Fellowship from Bowdoin College.

Correspondence concerning this article should be addressed to Daniel D. Kurylo, Department of Psychology, Bowdoin College, 6900 College Station, Brunswick, Maine 04011-8469. Electronic mail may be sent via Internet to [email protected].

gaps without a significant decline in performance. He further reported that rats conditioned to discriminate solid figures, such as crosses, circles, or triangles, did not differ in their rate of learning from rats conditioned to discriminate similar figures composed of disjunct dots. Although these results indicate that rats are sensitive to stimulus configurations, they do not directly address the question of whether separate elements are perceptually combined to form unified figures. Krechevsky (1938) addressed this issue more directly by first conditioning rats to select one of two patterns of disjunct elements, then testing whether discrimination transferred to an identical pattern or to a solid figure derived from the grouping of the elements. When the stimuli used for training differed distinctively, discrimination did not transfer to the solid figure. However, when the stimuli used for training differed only slightly in the spatial relationships among elements, discrimination transferred to the solid figure. Krechevsky interpreted these results as indicating that rats are able to base discrimination on the global organization of the elements, as opposed to the pattern of the individual elements. This interpretation has been supported by similar discrimination transfer studies; for example, the discrimination of line orientations composed of proximal elements appears to be based on the configural organization of the stimulus and not on the pattern of disjunct elements (Dodwell, Ferguson, & Niemi, 1976; Dodwell, Niemi, & Ferguson, 1976).

Previous investigations of perceptual grouping in rats possess two major shortcomings. The first is a lack of stimulus specificity. Previous studies do not provide a detailed description of stimuli, including the precise spatial relationships among elements. Furthermore, the position of the stimulus in relation to the cornea was not controlled, thereby precluding accurate measurements of stimulus pa- rameters. Control of the relative position of the stimulus is necessary to specify stimulus metrics in terms of visual angle. The second major shortcoming of previous studies is that they do not provide an index that directly reflects perceptual grouping. A direct measurement of grouping

126

PERCEPTUAL GROUPING IN RATS 127

capacities, however, is difficult to achieve. Seminal work on perceptual grouping in humans was phenomenological and limited mainly to demonstrations of the conditions under which grouping occurs. Experimental investigations have generally described grouping in terms of form identifica- tion, to which perceptual grouping is subordinate. Measure- ments of the strengths and limitations of perceived grouping as a function of stimulus characteristics are necessary to identify the mechanisms that mediate perceptual organization.

The present study overcomes these shortcomings and expands the investigation of grouping capacities in rats. To gain stimulus specificity, we carefully controlled the position of the rat's head during stimulus presentation. Further- more, parametric analysis of perceptual grouping was achieved by determining the limit to which a specific stimulus feature could serve as a cue for grouping. Stimuli consisted of bistable arrays of disjunct elements. Perceptual grouping was achieved by applying a grouping cue, either greater proximity or greater alignment of stimulus elements, to one of two possible grouped arrangements. Rats responded to each stimulus by means of a visual discrimination task that reflected which of the two possible arrangements was perceived as grouped. The grouping cue was progressively weakened until it no longer biased responses toward the cued arrangement, thereby indicating the limit to which the cue could serve to produce grouping. In this manner, direct measurements of grouping capacities were acquired.

Two spatial relationships that serve as cues for perceptual grouping were explored: element proximity and element alignment. Grouping by proximity (Rush, 1937) is based on the tendency for more proximal images to produce unified configurations and elements with greater separation to lose flgural integration (Gillam & Grant, 1984). Grouping by element alignment is an example of the Gestalt principle of good continuation (Rush, 1937). Grouping by alignment reflects the tendency to group elements that form regularities in the pattern. Grouping by proximity and grouping by alignment differ qualitatively in the nature of the spatial cue from which grouping is derived and are therefore likely to be mediated by different processes. In addition to examining each grouping cue in isolation, we examined the interactive effects of concurrent proximity and alignment cues.

The grouping cues examined here are based on the spatial relationships among elements, which has been termed Type P patterns (Pomerantz, 1981), referring to the position of the elements. Local elements possess no inherent spatial quality in themselves, and therefore the nature of the local elements does not provide grouping cues. Instead, perceptual grouping is based on emergent properties of the stimuli that are derived from spatial relationships among component elements. Grouping arrangements are distinguished by the degree of difference in proximity or alignments, which serves as the carrier dimension for discriminability (Honig, 1992). The degree to which the emergent feature controls behavior is affected by the amount of difference in proximity or alignment between grouping arrangements. Emergent properties of complex stimuli have been explored in this

manner for properties such as relative numerosity, form mixture, and stimulus categories. For pigeons trained to discriminate color, size, or form (O vs. X), graded responses were obtained as a function of the relative proportions of positive and negative stimulus elements, regardless of the absolute number of elements (Honig & Stewart, 1989). Similar functions were obtained for pigeons trained to discriminate arrays of uniform elements from arrays of mixed elements. Furthermore, graded response functions were obtained for discriminations that were based on conceptual categories, such as pictures of birds versus pictures of flowers. By using this general experimental procedure, these experiments described perceived emergent properties in terms of physical dimensions of the stimuli.

A complex visual scene may be described as a hierarchy of subscenes that are associated by spatial relationships, in which properties at the top of the hierarchy are more global than those at the bottom (Palmer, 1977). Discriminating complex arrays on the basis of proximity and alignment entails processing the global organization of the stimulus, thereby placing this process high on a hierarchical scheme. Much work has investigated the conditions under which processing overall structure precedes processing the component parts, an effect known as global precedence (Navon, 1977). Global precedence describes the temporal develop- ment of the percept, in which processing global properties are completed before processing local elements. Much of the work on global precedence has been done with hierarchical patterns, in which larger figures are constructed by the spatial pattern of smaller figures. Kimchi and Palmer (1985) demonstrated that the relationship between global and local levels of hierarchical patterns is related to the number and relative size of local elements. With regard to element proximity, a global advantage has been found to occur with more proximally arranged stimuli (Martin, 1979), likely caused by degradation of the global structure in stimuli with a sparse number of elements. Alignment has also been found to influence global precedence, because it affects the quality of information at either the local or global level. In this regard, distorting local elements produced faster reaction time to the global structure of stimuli, whereas distorting the global arrangement produced faster reaction time to perceiving local elements (Hoffman, 1980). Hughes, Layton, Baird, and Lester (1984) reported that when the relative visibility is comparable for local and global cues, global precedence is obtained. Furthermore, the magnitude of global precedence is inversely related to pattern luminance, suggesting that global precedence is at least partially attributable to early visual processing. Although the present study does not directly address the issue of global precedence, measurements obtained here reflect the perceived global organization of complex arrays.

In addition to examining the effectiveness of proximity and alignment as cues for grouping, we also explored the effect of the separation among elements. The overall density of elements, referred to as the stimulus scale, may influence the effectiveness of grouping cues. Dense arrays may facil- itate computations for grouping assignment, benefiting from the close spatial proximity of elements. Alternatively, stim-

128 KURYLO, VAN NEST, AND KNEPPER

uli in which elements are highly separated require the integration of broader spatial regions. For low-density arrays in which elements are highly separated, the association among elements is weaker and estimations of relative spatial relationships are more difficult. To test these predictions, we measured the effectiveness of proximity and alignment on grouping across a range of stimulus scales.

M e t h o d

Subjects

Six male Long-Evans hooded rats served as subjects. Training began at approximately 40 days of age. The animal housing facility maintained a 12-hr light-dark cycle.

Apparatus

Behavioral measurements were made in a customized operant conditioning chamber (see Figure 1) designed to maximize stimulus specificity in behaving animals. The front panel of the chamber was 5 cm wide and 20 cm high and supported a glass funnel that extended 3 cm outside of the chamber. Rats had access to the funnel though a 5-cm diameter hole, centered 2.5 cm from the chamber floor. Placement of the rat's head completely within the funnel disrupted an infrared (IR) light beam connecting an emitter-detector pair that was mounted outside the front of the

/ Computer Monitor

Rewa~ Well ~ eward Well

lO cm

Figure 1. Schematic diagram of operant conditioning chamber; top view, represented to scale. Rats initiated a trial by placing their head in the glass funnel. Stimuli appeared on the computer monitor, which was viewed by the left eye. Rats responded in the visual discrimination task by placing their head into either the left or the right reward well.

chamber. The positions of IR components were adjusted to ensure that interruption of the beam occurred when a rat's head was in a highly specified position. The position of the funnel and peripheral equipment allowed a wide angle of unoccluded viewing.

The front panel was bordered on each side by panels, 10 cm wide and 20 cm high, each of which angled toward the back of the chamber by 45 °. Each side panel possessed a 2.5-cm opening, centered 2.5 cm from the chamber floor, through which rats had access to an enclosed drinking well. Drinking wells extended 2.5 cm beyond the chamber walls. A small hole through which water was delivered was positioned on the back wall of each drinking well. IR emitter-detector pairs, which monitored placement of the rat's head near the drinking spout, were mounted outside each well. Measured amounts of water were delivered to each well by means of solenoid-driven valves.

A small lamp was positioned on the side of the chamber near the back wall, which served as the house light. Luminance of house light, as measured on the monitor, was less than 0.015 cd/m 2 and appeared relatively uniform. Except for the house light, the experimental room was not illuminated. Circuitry for each of the three IR pairs, each of the two solenoids, and the house light were interfaced via solid-state relay switches (CyberResearch CYR- DIO) to a computer (DECpc LPv 433dx). Data collection, trial events, stimulus presentation, and contingency algorithms were controlled in real time by computer.

Stimuli

Stimuli were presented on a computer monitor (DEC model PC7XV-DE) that was centered 24 cm from the rat's left eye and angled 65 ° from the longitudinal axis of the rat, thereby position- ing the monitor's surface tangent to the surface of the left eye (see Figure 1). Stimuli fell within a square field that was 32 ° on a side.

Training phase. Stimuli used during the conditioning phase consisted of three parallel lines, 1.2 ° × 32.9 °, and separated by 14.9 ° , which were oriented either horizontally or vertically.

Experimental phase. Stimuli for the experimental conditions consisted of an array of illuminated white squares, 1.2 ° on a side, on a darkened background. For each trial, a stronger grouping cue (either greater proximity or greater alignment of the elements) was randomly assigned to either the horizontal or the vertical orientation. Testing occurred under three conditions: proximity cue, alignment cue, and concurrent proximity and alignment cues. For each condition, measurements were made at three scales of the stimulus arrays, in which the mean separation of dots along the noncued orientation was 5.7 °, 12.1 °, and 16.8 °. Higher density stimuli contained more elements and were of greater mean luminance. The overall size of the stimulus array remained constant across all three stimulus scales.

Proximity cue. Elements were spaced at regular increments and were perfectly aligned along the horizontal and vertical orientations. The horizontal and vertical orientations differed in amount of separation between elements (see Figure 2a). The magnitude of difference in element separation (between the two orientations) ranged from 11% to 86%, in increments of 11%. For each trial, the more proximal elements were randomly assigned to either the horizontal or the vertical orientation.

Alignment cue. The mean amount of separation between elements was equivalent for the horizontal and vertical orientations. Elements along one of the orientations were perfectly aligned, whereas elements along the other orientation were offset laterally and were randomly distributed within a dispersion field whose border extended to 100% of the mean separation between elements (see Figure 2b). For each trial, the perfectly aligned elements were


Figure 2. Example of stimuli and their associated metrics for each of the three conditions. For each trial, grouping cues were randomly assigned to either the horizontal or the vertical orientation. In the examples shown here, the grouping cue is applied to the vertical orientation. (A) Proximity cue: Grouping is established by less separation between elements. Metrics (D) are described in terms of the more proximal separation (in this example, the vertical separation) divided by the less proximal separation (in this example, horizontal separation). (B) Alignment cue: Grouping is established by the alignment of elements to a straight line. The proximity of elements is equivalent for the two orientations (vertical and horizontal separation). Metrics (E) are described in terms of the misalignment divided by the separation. (C) Proximity and alignment cues: Grouping is established by greater proximity as well as alignment that are concurrently applied to the same orientation (in this example, the vertical). Alignment is fixed, and metrics (F) are described in the same manner as the proximity cue condition: the more proximal separation divided by the less proximal separation.

randomly assigned to either the horizontal or the vertical orientation.

Concurrent proximity and alignment cue. As with the proximity cue alone, the horizontal and vertical orientations differed in the amount of separation between elements. Furthermore, the orientation with less proximal elements was misaligned, such that elements were laterally displaced and distributed within a dispersion field, as was done for the alignment cue condition (see Figure 2c). For each trial, the perfectly aligned more proximal elements were randomly assigned to either the horizontal or the vertical orientation.

Stimulus metrics. Grouping cue strength is described in terms of the metrics of the stimulus array and is independent of obtained data. Strong grouping cues correspond to stimuli in which elements along one of the orientations was highly proximal, highly

aligned, or both. Weak grouping cues correspond to stimuli in which the metrics along each orientation were similar. For the proximity cue alone and the concurrent cue conditions, stimulus metrics are described in terms of element separation along the more proximal orientation in relation to the less proximal orientation and is referred to as the relative separation (see Figures 2D and 2F). For example, a relative separation of 5% indicates that elements along the more proximal orientation are separated by 5% of the less proximal orientation, thereby providing a strong grouping cue. A relative separation of 100% indicates that element separation is equivalent along each orientations, thereby providing no grouping cue. The relative separation is therefore inversely related to the strength of separation cue.

For the alignment cue alone, stimulus metrics are described in terms of the degree of lateral dispersion in relation to the amount


of separation between elements (see Figure 2E). For example, for an orientation in which elements were separated by 17 °, a lateral dispersion of 50% indicates that elements were randomly distributed within a field whose borders extended by 8.5 °.

Procedure

Rats were deprived of water for 22 hr before each session and were allowed water ad lib for 1 hr following sessions. Training or data collection was performed 5 or 6 days each week, and rats were allowed water ad lib on 1 day each week. Weights were monitored daily in young rats to ensure normal growth.

Training phase. Rats were initially trained on a visual discrimination task, in which solid horizontal fines became associated with approaching one of the drinking wells and solid vertical lines became associated with approaching the alternative well. Rats were initially trained to place their heads completely into the central funnel, which initiated a trial, and to remain in place until a stimulus appeared on the monitor. Stimuli appeared after a randomly selected delay, which ranged from 500 to 2,500 ms after trial initiation. Line orientation (horizontal or vertical) was paired with the delivery of reward to either the left or the right drinking well, and pairing remained fixed throughout the experiment. The pairing of line orientation to response side was counterbalanced across subjects.

Initially, water was delivered automatically to the appropriate drinking well 500 ms after stimulus onset, at which time the stimulus was removed. Across sessions, the delay before automatic reward delivery was progressively increased. In doing so, the percentage of trials in which rats responded (by entering a drinking well) before automatic reward delivery also increased. For trials in which rats responded before automatic reward delivery, correct responses resulted in immediate reward delivery, whereas incorrect responses resulted in the withholding of reward and a 7-s time-out (the termination of the house light during which time trials could not be initiated). Initially, rats discriminated stimuli at approximately 50% correct (for those trials in which responses were made before automatic reward delivery). To maintain an adequate response rate (approximately 80 trials per 1-hr session), we controlled the percentage of trials in which automatic reward delivery occurred. For each session, the delay before automatic reward delivery was estimated, based on the previous session, to produce rewarded trials on approximately 85% of the trials (based on correct responses as well as automatic reward). Conditioning was considered complete when all trials in a session were response driven and rats performed correctly on 85% of the trials.

Experimental phase. Rats initiated trials by placing their heads within the central funnel. Stimuli appeared 1 s after trial initiation. Rats were allowed 2 s to respond to the stimulus by placing their heads into the correct drinking well, in which case reward was delivered. Incorrect responses, or the absence of a response within the allowed period, was scored as an error and resulted in a 7-s time-ont. In this manner, responses were made by means of a two-alternative forced-choice procedure, thereby controlling for the effects of response criteria. Absence of response occurred rarely, accounting for less than 1% of trials, as calculated across sessions. The intertrial interval for correct respo0ses was set to 2 s, although rats spent several seconds drinking from the wells when rewarded. Experimental sessions lasted for a maximum of 1 hr or a maximum of 150 trials, whichever occurred first.

During the experimental phase, rats' performance was monitored by means of control trials that occurred randomly throughout each session on 17% of the trials. Stimuli for control trials consisted of solid lines (as were presented during the training phase).

Sessions in which the mean performance on control trials was below 85% correct (which was the criterion for baseline performance) were excluded from analysis and resulted in returning rats to training procedures until performance reached criterion.

Threshold measurements. Grouping thresholds were measured by the method of constant stimuli. For each measurement, five to seven stimuli that differed in their level of grouping cue strength were used. Levels of cue strength were produced by varying the relative separation (for the proximity cue alone and the concurrent cue conditions) or the amount of lateral dispersion (for the alignment cue alone condition) of the elements. Across levels of grouping cue strength, performance ranged from nearly 100% correct (for strong grouping cues) to near chance level (for weak grouping cues). The grouping threshold is defined as the grouping cue strength (described in terms of stimulus metrics) that produced 75% correct responses. In each session, levels of grouping cue strength were presented in random order. The percentage of correct responses was based on 70 to 90 trials, accumulated across multiple sessions. More than 16,000 experimental trials were therefore required to obtain measurements from all conditions from all animals. The percentage of trials in which the cued orientation was selected was plotted as a function of cue strength, thereby deriving a psychometric function of each stimulus condition. Probit analysis (McKee, Klein, & Teller, 1985) was used to determine the 75% point of the psychometric functions.

Resu l t s

For the proximity cue alone and the concurrent cues conditions, rats were biased to select the orientation that contained the grouping cue. When the g roup ing cue was strong, rats selected the orientation that contained' the cue at greater than 85% correct (performance comparable with the control condition). Progressive reduction in the strength of the grouping cue resulted in a progressive/reduct ion in selection of the orientation that contained the/cue, until rats selected either orientation with equal probal~ility. The sys- tematic relationship between stimulus metrics and orientation selection allowed psychometric functions to be derived for the proximity cue and the concurrent~ue conditions. For the alignment cue alone condition, biafed selection was not obtained with even maximal cue strea/gth, indicating that the alignment cue alone did not serve/as an adequate cue for / perceptual grouping. / /

Proximity Cue Alone and Concurrent Proximity and Alignment Cues jJ

/

The mean percentages of trials in which the orientation with more proximal elements was selected as a function of relative separation (strength of the cue) are shown in Figure 3 (proximity cue alone) and Figure 4 (concurrent proximity and alignment cues). In all cases, the orientation with more proximal elements was chosen on greater than 85% of the trials when the grouping cue was strong (i.e., the relative separation was approximately 10%) and fell progressively to chance level (selection of either orientation with equal probabili ty) as the cue strength was diminished (relative separation increased).

Grouping thresholds were determined individually for


a8 Q

80-

7O

5 0 -

5O 0

100-

90 -

8 0 .

70-

6 0 .

50 0

12.1

1o0.

90.

80.

70,

60,

50 0

5.7 °

• i . . . . | . . . . , . . . . i

25 50 75 IO0

o i

ca

Q u i

0 Q

0

4) W. O .~= 0

0

- | . . . . , . . . . . . . . . .

16.8 o

• . , . . . . o

25 .50 75 IOO

Relative Separation (%)

Figure 3. Mean psychometric functions derived from the proximity cue condition. The mean percentage of trials in which the cued orientation was selected is plotted as a function of the ratio of element separation between the cued and noncued orientation. Smaller relative separations correspond to stronger grouping cues (i.e., the cued orientation was highly proximal compared with the noncued orientation). Psychometric functions are depicted for three stimulus size scales, in which the separation of elements along the cued orientation was 5.7 °, 12.1 °, and 16.8 ° visual angle.

each subject. The mean grouping thresholds for the proximity cue alone and for the concurrent proximity and alignment cues across stimulus scales are shown in Figure 5. For the proximity cue alone, grouping thresholds occurred at 51%, 40%, and 25% for stimulus scales of 5.7 °, 12.1 °, and 16.8 °, respectively. For the concurrent cue condition, grouping thresholds occurred at 54%, 50%, and 40% at each stimulus scale, respectively. Lower grouping thresholds indicate that the cue was less effective at eliciting grouping. On the basis of a repeated measures analysis of variance, the main effect of stimulus scale was significant, F(2, 10) = 16.19, p < .05. This result indicates that increasing the stimulus scale (increasing the overall separation among elements within the array) produced progressively lower grouping thresholds, indicating that the effectiveness of the grouping cue diminished for stimuli in which elements were separated by broader regions. The main effect of stimulus condition (proximity alone vs. concurrent cues) was also found to be significant, F(1, 5) = 13.47, p < .05, indicating that more proximal elements were more likely to be grouped when the alternative (less proximal) orientation was misaligned. An interactive effect of stimulus scale by cue condition was not significant, F(2, 10) = 2.87, p > .1.

Alignment Cues

Element alignment used alone was not an effective grouping cue. For stimulus scales of 5.7 ° , 12.1 ° , and 16.8 ° and with a maximum cue strength, rats selected the cued orientation on 54%, 52%, and 53% of the trials, respectively. For the alignment cue condition, the mean percentages correct for control trials across stimulus scale were 89%, 87%, and 92%, respectively. As with the proximity cue alone and the concurrent cues conditions, control trials were randomly intermixed with grouping cue trials across each session.

Discussion

Element proximity was found to serve as a cue for perceptual grouping in rats. The effectiveness of proximity cues, however, was several times less than that found in humans (Kurylo, 1997a). Under similar conditions, human observers perceived grouping of more proximal elements when competing grouping arrangements differed by as little as 8%, whereas rats required a difference of 51% to 25%, depending on stimulus scale. Furthermore, the effectiveness of proximity cues for rats varied as a function of stimulus scale, diminishing with broader separation of the elements throughout the array. In humans, grouping thresholds for proximity cues remained constant across stimulus scales that ranged from 1.01 ° to 5.05 ° of element separation. For rats, differences in thresholds across size scale is not likely attributed to differences in mean luminance of stimuli, because this condition occurred with the human measurements as well. The decreased effectiveness of proximity cues across broader spatial regions may reflect a diminished capacity to compare relative separations between pairs of


I00,

o ¢=

4) i l l I m

0 " 0

¢.)

C ¢= Q =!

90.

80.

70,

60.

SO 0

I00-

90-

80-

70-

60-

50

5.7 °

i . . . . i . . . . i . . . . |

25 5O 75 I O0

12.1 o

" ~ . . . . s~ . . . . 7~ . . . . Tdo

I00

90.

80.

70.

60.

50 0

16.8

25 50 7~ I 00

0

Relative Separation (%)

Figure 4. Mean psychometric functions derived from the concurrent cues condition. The mean percentage of trials in the cued orientation (cued by proximity as well as alignment) was selected as a function of relative separation at each of three stimulus scales.

elements, thereby impeding the selection of grouping from among several possible arrangements.

Rats did not respond to alignment cues for grouping. This result is in contrast to humans, for whom an aligned orientation appears grouped when the alternative orientation is

misaligned by as little as 16% (Kurylo, 1997a). The difference in effectiveness between proximity and alignment cues in rats may reflect an increased complexity in processing alignment as a grouping cue. Grouping by proximity is likely based on the relative separation of pairs of elements, whereas grouping by alignment requires identifying pattern regularities within the global organization of the stimulus. In this regard, application of alignment cues requires more elaborate processing of the spatial relationships among elements. The greater relative complexity of grouping by alignment is reflected by the temporal characteristics of grouping found in humans. In humans, grouping by proximity required a mean of 87.6 ms to complete, whereas grouping based on alignment required a mean of 118.8 ms (Kurylo, 1997b). For rats, the increased complexity of the grouping by alignment may exceed the processing capacity of the visual system.

Alignment cues, however, facilitated grouping in rats when they occurred in conjunction with proximity cues. Under these conditions, misalignment appeared to disrupt grouping along the orientation with less proximal elements. This result indicates that although alignment alone does not serve as a cue for grouping, pattern regularity influences the perceptual organization of the stimulus. The reason that the alignment cue facilitated grouping in the concurrent cue condition but not in the alignment alone condition may stem from the minimal strength of the alignment cue. Alignment may have an impact on visual processing but is inadequate to establish grouping when presented in isolation. Compar- ison of the concurrent cue condition with the proximity alone condition (see Figure 5) supports this notion. Align- ment had the greatest impact at the stimulus scale of 16.8 ° ,

A

o'~ 100" v

, 5

0 , .

• '~ o n" 0

z Pm=lmRy cue = c o n ~ cure

, . . . . , . . . . i . . . . i

5 10 15 20

S t i m u l u s S c a l e

Figure 5. Mean grouping thresholds as a function of stimulus size scale for the proximity cue alone and the concurrent cues conditions. Thresholds are defined as the relative separation at which the cued orentafion was selected on 75% of the trials. Smaller relative separations correspond to stronger grouping cues. As the stimulus scale increased (greater separation of elements within the stimulus array), stronger grouping cues were required to reach grouping threshold.


in which the grouping threshold changed from 25% for proximity cue alone to 40% for the concurrent cue condition. This indicates that the additional cue of alignment affected thresholds by 15%, which may be inadequate to produce grouping when presented without the concurrent effect of proximity.

These results indicate that, compared with humans, rats have a diminished capacity to use proximity and alignment as cues for perceptual grouping. A limited capacity at the level of perceptual organization will likely reduce the effectiveness of object recognition, particularly for complex visual stimuli. Under natural conditions, however, multiple cues serve to establish grouping within a complex visual scene, including feature similarity, coherence of motion, and texture. It may be that purely spatial cues are of limited utility for perceptual organization in rats and grouping is established by means of other stimulus features. Having established a metric that directly reflects grouping capacity, a more comprehensive examination may be made of grouping capacity based on other visual cues. Grouping thresholds of competitive grouping arrays also provide a standard- ized index that can be compared across species.

Recent anatomical and physiological studies have examined cross-cortical connectivity (Hirsch & Gilbert, 1991; Kisvarday & Eysel, 1993; McGuire, Gilbert, Rivlin, & Wiesel, 1991; Schwarz & Bolz, 1991). These studies address the issue of how large numbers of neurons associate information across functionally integrated regions. Paramet- ric analysis of perceptual grouping that was developed here may be used to examine cortical integrative function at a perceptual level. Perceptual grouping represents a condition in which the visual system constructs perceived properties that are not directly contained within the stimulus. A set of disjunct visual elements that are highly proximal or highly aligned are perceived as a cohesive pattern, although the stimulus elements are not physically connected. Grouping therefore necessitates the integration of information across spatial areas.

Within retinotopically organized visual areas, adjacent spatial areas of the visual scene are represented by anatomically adjacent cortical locations. Individual visual elements that are disjunct in space are therefore represented by neural activity across anatomically disjunct sites. The process of grouping reflects perceptual integration that is established by the pattern of connections among the neural sites, and grouping thresholds reflect the limit at which these patterns are effective. Measurements of grouping thresholds provide an opportunity to monitor perceptual effects of physiological interventions, such as localized lesions or selective neural blockers. The behavioral task described here may also be applied while monitoring neural responses to a stimulus element that appears in isolation, compared with the response found when the element is contained within a perceptually grouped array. In this manner, the perceptual effects of spatial relationships among stimulus elements may be used to explore interactive function of neural populations.

References

Birch, D., & Jacobs, G. H. (1979). Spatial contrast sensitivity in albino and pigmented rats. Vision Research, 19, 933-937.

Dean, P. (1978). Visual acuity in hooded rats: Effects of superior collicular or posterior neocortical lesions. Brain Research, 156, 17-31.

Dean, P. (1981). Visual pathways and acuity in hooded rats. Behavioural Brain Research, 3, 239-271.

Dean, P. (1990). Sensory cortex: Visual perceptual functions. In B. Kolb & R. C. Tees (Eds.), The cerebral cortex of the rat (pp. 275-307). Cambridge, MA: MIT Press.

Dodwell, P. C., Ferguson, H. B., & Niemi, R. R. (1976). Anom- alous transfer in rats: Pattern element separation and discriminability. Bulletin of the Psychonomic Society, 7, 154-156.

Dodwell, P. C., Niemi, R. R., & Ferguson, H. B. (1976). Anom- alous transfer in rats: A "macropattern" phenomenon. Bulletin of the Psychonomic Society, 7, 157-159.

Gillam, B. & Grant, T., Jr. (1984). Aggregation and unit formation in the perception of moving collinear lines. Perception, 13, 659-664.

Hirsch, J. A., & Gilbert, C. D. (1991). Synaptic physiology of horizontal connections in the cat's visual cortex. Journal of Neuroscience, 11, 1800-1809.

Hoffman, J. E. (1980). Interaction between global and local levels of a form. Journal of Experimental Psychology: Human Percep- tion and Performance, 6, 222-234.

Honig, W. K. (1992). Emergent properties of complex arrays. In W. K. Honig & J. G. Fetterman (Eds.), Cognitive aspects of stimulus control (pp. 301-321). Hillsdale, NJ: Erlbaum.

Honig, W. K., & Stewart, K. E. (1989). Discrimination of relative numerosity by pigeons. Animal Learning and Behavior, 17, 134-146.

Hughes, H. C., Layton, W. L., Baird, J. C., & Lester, L. S. (1984). Global precedence in visual pattern recognition. Perception and Psychophysics, 35, 361-371.

Kimchi, R., & Palmer, S. E. (1985). Separability and integrality of global and local levels of hierarchical patterns. Journal of Ex- perimental Psychology: Human Perception and Performance, 8, 521-535.

Kisvarday, Z. F., & Eysel, U. T. (1993). Functional and structural topography of horizontal inhibitory connections in cat visual cortex. European Journal of Neuroscience, 5, 1558-1572.

Krechevsky, I. (1938). An experimental investigation of the principle of proximity in the visual perception of the rat. Journal of Experimental Psychology, 22, 497-523.

Kurylo, D. D. (1997a). Effects of element proximity and alignment on perceptual grouping. Manuscript submitted for publication.

Kurylo, D. D. (1997b). Time course of perceptual grouping. Per- ception and Psychophysics, 59, 142-147.

Kurylo, D. D., Hansen, R. M., & Fulton, A. B. (1991). Dark adapted threshold in young RCS rats. Vision Research, 31, 637-642.

Lashley, K. S. (1930). The mechanism of vision: I. A method for rapid analysis of pattern-vision in the rat. Journal of Genetic Psychology, 37, 453-460.

Lashley, K. S. (1931). Cerebral function in the pattern vision of the rat. Psychological Bulletin, 28, 221-222.

Lashley, K. S. (1938). The mechanism of vision: XV. Preliminary studies of the rat's capacity for detailed vision. Journal of General Psychology, 18, 123-193.

Lavond, D., Hata, M. G., Gray, T. S., Geckler, C. I., Meyer, P. M., & Meyer, D. R. (1978). Visual form perception is a function of the visual cortex. Physiological Psychology, 6, 471-477.


Legg, C. R. (1984). Contrast sensitivity at low spatial frequencies in the hooded rat. Vision Research, 24, 159-161.

Martin, M. (1979). Local and global processing: The role of sparsity. Memory and Cognition, 7, 476-484.

McDaniel, W. F., Wildman, L. D., & Spears, R. H. (1979). Posterior association cortex and visual pattern discrimination in the rat. Physiological Psychology, 7, 241-244.

McGuire, B. A., Gilbert, C. D., Rivlin, P. K., & Wiesel, T. N. (1991). Targets of horizontal connections in macaque primary visual cortex. Journal of Comparative Neurology, 305, 370 -392.

McKee, S. P., Klein, S. A., & Teller, D. Y. (1985). Statistical properties of forced-choice psychometric functions: Implica- tions of probit analysis. Perception and Psychophysics, 37, 286 -298.

Muntz, W. R. A. (1967). A behavioral study on the photopic and scotopic vision in the hooded rat. Vision Research, 7, 371-376.

Navon, D. (1977). Forest before trees: The precedence of global features in vision perception. Cognitive Psychology, 9, 353-383.

Palmer, S. E. (1977). Hierarchical structure in perceptual repre- sentation. Cognitive Psychology, 9, 441-474.

Pomerantz, J. R. (1981). Perceptual organization in information processing. In M. Kubovy & J. R. Pomerantz (Eds.), Perceptual organization (pp. 141-179). Hillsdale, NJ: Erlbaum.

Rosenberger, P. B., & Ernest, J. T. (1971). Behavioral assessment of absolute visual thresholds in the albino rat. Vision Research, 11, 199-207.

Rush, G. P. (1937). Visual grouping in relation to age. Archives of Psychology, 31(Whole No. 217).

Schwarz, C., & Bolz, J. (1991). Functional specificity of a long- range horizontal connection in cat visual cortex: A cross- correlation study. Journal of Neuroscience, 11, 2995-3007.

Sefton, A. J., & Dreher, B. (1985). Visual system. In G. Paxinos (Ed.), The rat nervous system: Vol. 1. Forebrain and midbrain (pp. 169-221). Sydney, Australia: Academic Press.

Tedo, C. M., Herreros de Tejada, P., & Green, D. G. (1994). Behavioral estimates of absolute threshold in rat. Visual Neuro- science, 11, 1077-1082.

Tompson, R., & Bachman, M. K. (1979). The occipito-striate pathway and visual discrimination performance in the white rat. Bulletin of the Psychonomic Society, 14, 433-434.

Trejo, J. T., & Cicerone, C. M. (1987). Changes in visual sensitivity with age in rats with hereditary retinal degeneration. Vision Research, 27, 915-918.

Wuensch, K. L., Broome, B., Means, L. W., & Harris. E. C., Jr. (1978). Dorsalmedial thalamic lesions and amphetamine: Acqui- sition and retention of a visual pattern discrimination escape task. Physiological Psychology, 6, 288-293.

Received February 21, 1996 Revision received July 25, 1996

Accepted July 26, 1996

New Editor Appointed for Contemporary Psychology: 1999-2004

The Publications and Communications Board of the American Psychological Association announces the appointment of Robert J. Steinberg (Yale University) as editor of Contempo- rary Psychology, for a 6-year term beginning in 1999. The current editor, John H. Harvey (University of Iowa), will continue as editor through 1998.

All reviews are written by invitation only, and neither the current editor nor the incoming editor receives books directly from publishers for consideration. Publishers should continue to send two copies of books for consideration, along with any notices of publication, to PsyclNFO Services Department, APA, Attn: Contemporary Psychology Processing, 750 First Street, NE, Washington, DC 20002-4242.

Characteristics of Perceptual Grouping in...

Documents

Transcript of Characteristics of Perceptual Grouping in...