Hemispheric Asymmetry for Components of Spatial Processing

Cerebral Asymmetries in Sensory and Perceptual Processing S. Christman (Editor) 9 1997 Elsevier Science B.V. All rights reserved.

83

Chapter 4

Hemispheric Asymmetry for Components of Spatial Processing Joseph B. Hellige University of Southern California The two cerebral hemispheres of the human brain do not have identical ability to localize stimuli in space. A predominant view has been that the right hemisphere is superior to the left for identifying spatial relations among objects. During the last decade, however, it has become apparent that such right-hemisphere superiority does not extend to the identification of all aspects of spatial relations. Instead, it has been hypothesized that each hemisphere is dominant for identifying different and, to some extent, complementary aspects of spatial relations. In the present chapter, I outline a distinction that has been made between "categorical" and "coordinate" spatial relations and review investigations of hemispheric asymmetry for identifying these different types of spatial relations, with a view toward identifying mechanisms that might underlie such asymmetry. Accordingly, the chapter begins with a brief review of the categorical/coordinate distinction and relevant studies of hemispheric asymmetry. This is followed by a review of recent experiments that have tested various hypotheses about the underlying mechanisms. I then turn to experiments that have attempted to extend the categorical/coordinate distinction beyond the domain of spatial relationships. The chapter ends with a reconsideration of underlying mechanisms and a discussion of directions for future studies of hemispheric asymmetry for spatial processing.

84 Hellige

The Categorical/Coordinate Distinction During the last 10 years or so, a considerable amount of theoretical and empirical work has beenconducted within a framework proposed by Kosslyn and colleagues in which the brain is hypothesized to compute two kinds of spatial-relation representations (e.g., Kosslyn, 1987). One type of representation ("categorical") is used to assign a spatial relation between two stimuli to a category such as "touching" versus "not touching" and "above" versus "below." The other type of representation ("coordinate") preserves location information using something like a metric coordinate system in which distances are specified effectively. Recent experiments suggest that the right hemisphere makes more effective use of this coordinate or metric distance information about spatial relationships whereas there is either no hemispheric asymmetry or there is left-hemisphere superiority for processing information about categorical spatial relationships (for reviews, see Hellige, 1993, 1995, 1996; Kosslyn, Koenig, Barrett, Cave, Tang & Gabrieli, 1989; Kosslyn & Koenig, 1992). Results consistent with the general conclusions just outlined have been reported for a variety of visual half-field experiments from a number of different laboratories. In such experiments, visual stimuli on each trial are presented briefly to either the left visual field and right hemisphere (LVF/RH trials) or to the right visual field and left hemisphere (RVF/LH trials) and performance is examined as a function of which hemisphere receives the stimulus information directly. For example, in an experiment with neurologically normal right-handed observers, Chikashi Michimata and I (Hellige & Michimata, 1989) had observers indicate whether a dot was above or below a line (a categorical Above/Below task) or, on a different block of trials, indicate whether the dot was within 2 cm of the line (a coordinate, Near/Far task). Figure 1 illustrates the possible positions of the dot relative to the line. The results are shown in Figure 2a. Consistent with the hypothesis outlined above, for the Above/Below task a RVF/LH advantage approached statistical significance and for the Near/Far task there was a highly significant LVF/RH advantage--producing a highly significant Task by Hemisphere interaction. We (Hellige, Bloch, Cowin, Eng, Eviatar & V. Sergent, 1994) have replicated this effect for a new group of right-handed observers (see Figure 2b), and also found the Task by Hemisphere interaction to be absent in left-handed observers. For additional examples of

Spatial Processing

85

0 0 0 0 0 0 0 0 0 0 0 0 Figure 1. Illustration of the line and dot positions used by Hellige and Michimata (1989). On each triM, observers saw the line and a single one of the 12 dots.

this type of Task by Hemisphere interaction for fight-handed observers, see Kosslyn (1987), Kosslyn et al. (1989), J. Sergent (1991; but only for a low level of luminance); Koenig et a1.(1992), Rybash and Hoyer (1992); Cowin and Hellige (1994, 1995), and Crebolder and Bryden (1996); see also Laeng, Peters and McCabe (1996). In addition, in some studies, this Task by Hemisphere interaction disappears with practice, possibly because observers learn to perform the distance judgment in a more categorical way as they become familiar with the stimuli (e.g., Kosslyn et al., 1989, Cowin & Hellige, 1994). Note from Figures 2a and 2b that, for the tasks developed by Hellige and Michimata (1989), the categorical task was easier than the coordinate task. If this were always the case, an alternative interpretation of the Task by Hemisphere interaction could be formulated in terms of task difficulty. However, in several experiments reported by Kosslyn and colleagues (e.g., Kosslyn, 1987; Kosslyn et al., 1989), the specific categorization tasks used (e.g., is a dot touching an irregular blob or not--an "on/off" task) were more difficult than the specific coordinate distance judgment tasks (e.g., is the dot within 2 mm of the blob).

86

Hellige

760 740

Hellige & Michimata

(1989)

720 700 i RT

680 ~ 660 640 620

LVF/RH RVF/LH

600 580 560

I

Above/Below

I

Task

Near/Far

Figure 2a. Reaction time for a categorical spatial task (Above/Below) and for a coordinate spatial task (Near/Far) for LVF/RH and RVF/LH trials. Results come from Hellige and Michimata (1989).

Nevertheless, there was a RVF/LH advantage for the on/off task and a LVF/RH advantage for the distance judgment task. Thus, the Task by Hemisphere interaction seems to be related to the categorical versus coordinate nature of the task and not simply to the difficulty of the spatial discrimination, at least as difficulty is measured by overall performance. The categorical/coordinate distinction has also been extended to tasks that require the generation and processing of visual images. For example, Kosslyn and his colleagues have suggested that the left

Spatial Processing 87 hemisphere is dominant for generating and processing visual images that require the correct categorical arrangement of parts or that are 760

740

,,

Hellige et al. ( 1 9 9 4 )

720

700 680 RT 660 640 620

--

LVF/RH F/LH

600

580 560

I

Above/Below

I

Task

Near/Far

Figure 2b. Reaction time for a categorical spatial task (Above/Below) and for a coordinate spatial task (Near/Far) for LVF/RH and RVF/LH trials. Results come from Hellige et al. (1994). generated in a categorical, part-by-part manner; whereas the right hemisphere is dominant for generating and processing visual images in terms of precise metric coordinates. Thus, Kosslyn, Holtzman, Farah and Gazzaniga (1985) report that the right hemisphere of split-brain patient J. W. could not perform tasks that required him to make categorical decisions about parts of imaged objects (e.g., Do a hog's ears protrude above the top of the skull?), though the left hemisphere's performance was nearly perfect. In addition, J. W.'s right hemisphere could perform perfectly tasks that required a decision about the overall shape or size of

88

Hellige

an imaged object (e.g., Is a book higher than it is wide?). In a series of visual half-field experiments with neurologically intact individuals, Kosslyn (1988) has also reported a RVF/LH advantage for a task that is believed to use categorical relations to arrange letter segments within a grid and a LVF/RH advantage for a task that requires the arrangement of letter segments without benefit of a grid and believed to use coordinate relations to arrange the segments (see also, Kosslyn, Maljkovic, Hamilton, Horwitz, & Thompson, 1995). Michimata (1997) required right-handed observers to make categorical and coordinate decisions about the long and short hands of a clock face. For the categorical task, observers indicated whether both of the hands were above or below the horizontal midline of the clock face. For the coordinate task, observers indicated whether the angle formed by the two hands was greater or smaller than 60 degrees. In a perceptual condition, analog clock faces were presented to either the LVF/RH or RVF/LH on each trial. In an imagery condition, time was shown on a digital clock face (e.g., 3:25) in either the LVF/RH or RVF/LH, so that observers would have to generate images of analog clock faces in order to perform the tasks. In both the perceptual and the imagery conditions, there was a significant LVF/RH advantage for the coordinate task and a nonsignificant trend in the opposite direction for the categorical task. Thus, the Task by Hemisphere interaction does not require external stimulation of visual pathways. The similarity of perceptual and imagery results is consistent with hypothesized isomorphism between the processing of visual stimuli and the processing of visual images and indicates that the categorical/coordinate distinction is relevant for both. As this brief review suggests, the Spatial Task by Hemisphere interaction is sufficiently robust to merit further empirical and theoretical investigation. It is particularly important to consider what sort of computational and neural mechanisms might underlie these aspects of hemispheric asymmetry. With this in mind, I turn to the search for such mechanisms.

The Search for Underlying Mechanisms of Hemispheric Asymmetry for Spatial Processing In this section, I review the evolution of hypotheses regarding the computational and neural mechanisms that underlie hemispheric asymmetry for processing categorical and coordinate spatial relation-

Spatial Processing 89 ships. This section begins with a discussion of the speech/attention-shift hypothesis introduced by Kosslyn (1987). One issue that arises in considering possible mechanisms is the extent to which the two types of spatial relationships are processed independently of each other or in neural subsystems that do not interact with each other. Accordingly, in the second part of this section I consider this issue of independence. Hemispheric asymmetry for many aspects of visual information processing has been shown to be influenced by the nature of visual input and by which aspects of the visual input are most relevant for efficient performance of a task. In the third part of this section, I consider theoretical and empirical work that illustrates how the nature of task-relevant visual information is also important for spatial processing. The

Speech/Attention-Shift Hypothesis

Kosslyn's original hypothesis about hemispheric asymmetry for processing spatial relations was based on the assumption that the left hemisphere is specialized for the control of speech and the right hemisphere is specialized for the control of rapid shifts of attention across space (Kosslyn, 1987). He proposed that these initial specializations provided a "seed" function for each hemisphere, which would operate in the following way. As new skills are added during the course of phylogenetic or ontogenetic development, those skills become lateralized to one hemisphere or the other to the extent that they can be performed better by the neurological substrata laid down in one hemisphere compared to the other (see Hellige, 1993, for additional discussion of this type of developmental "seeding" idea). For a variety of reasons, Kosslyn argued that the neurological substrata for speech control and for rapid shifts of attention would be well-adapted for categorical and coordinate spatial processing, respectively. While not e x p l i c i t l y abandoning this speech/attention-shift hypothesis, more recently Kosslyn and colleagues have emphasized differences in the nature of the visual information that is most useful for computing categorical versus coordinate spatial information (e.g., Kosslyn. Chabris, Marsolek & Koenig, 1992). I will discuss this newer hypothesis in some detail later. Before doing so, however, it is useful to consider certain experimental results that point to limitations of the speech/attention-shift hypothesis and that must be considered in the evaluation of alternative hypotheses.

90

Hellige

Based on the idea of individual differences in seeding of the two hemispheres, Kosslyn (1987) suggested that individuals who show a relatively large LVF/RH advantage for coordinate spatial processing should also show a relatively large RVF/LH advantage for categorical spatial processing. This assertion received some tentative support from a study in which larger laterality effects of both sorts were obtained for strongly right-handed individuals than for ambidextrous individuals, with the assumption being that different seeding of the two hemispheres is more likely in the strongly right-handed group (Kosslyn, 1987; Kosslyn et al., 1989). However, other results suggest that, in fact, the two types of asymmetry are not correlated with each other. For example, Hellige and Michimata (1989) and Hellige et al. (1994) had the same individuals perform both a categorical task (is a dot above or below a line) and a coordinate task (is a dot within 2 cm of a line), so that they could examine the relationship between the perceptual asymmetries for the two tasks. In addition, Hellige et al. (1994) included left-handed as well as fight-handed individuals. To the extent that individuals who show a relatively large LVF/RH advantage for the coordinate task also show a relatively large RVF/LH advantage for the categorical task, the L V F RVF difference scores for the two tasks should correlate negatively. This was clearly not the case, as none of the correlations even approached statistical significance. In fact, Hellige and Michimata reported a nonsignificant positive correlation of .16 (the dependent variable was reaction time of correct responses). For right- and left-handed individuals combined, Hellige et al. (1994) reported a correlation o f - . 0 1 (with the correlations for fight- and left-handed groups being .19 and -. 15, respectively). Thus, laterality for the two types of spatial relations tasks seem to be uncorrelated (see also Laeng & Peters, 1995). While this does not necessarily invalidate the concept of seeding of the two hemispheres in different ways, it does suggest that the seeds that create a bias toward efficient categorical versus coordinate processing are sown independently of each other. Multi-task investigations of individual differences in hemispheric asymmetry have also provided information about the extent to which either of these two aspects of laterality for spatial processing arise from a more primitive left-hemisphere superiority for speech. If, for example, a common seed underlies hemispheric asymmetry for speech processing and for processing categorical spatial relationships, then we might expect an appropriate correlation between hemispheric asymmetry for

Spatial Processing 91 speech perception and hemispheric asymmetry for categorical spatial processing. In a multi-task study of individual variation in hemispheric asymmetry, Hellige et al. (1994) included the categorical and coordinate line and dot tasks used by Hellige and Michimata (1989), as well as a dichotic listening task requiring the identification of stop-consonants and a visual half-field task that required the identification of nonword trigrams. For present purposes, the interesting finding was that hemispheric asymmetry for both categorical and coordinate spatial processing was unrelated to ear asymmetry for the verbal dichotic listening task or to the visual field asymmetry for identifying nonword trigrams (though the latter two types of asymmetry were significantly correlated). Studies by Boles (1991, 1992) also suggest that laterality for spatial processing is independent of laterality for verbal processing. Thus, hemispheric asymmetry for processing spatial relationships does not seem to be based on a more primitive asymmetry for speech processing.

Are Categorical and Coordinate Spatial Relationships Processed Independently? In considering mechanisms that might contribute to the processing of categorical and coordinate spatial relationships, it is useful to consider the extent to which the two types of spatial relationships are processed independently of each other. To be sure, the existence of Task by Hemisphere interactions suggests that somewhat different neural mechanisms underlie the two types of spatial processing or that different aspects of visual information are important for different tasks. However, this does not necessarily mean that these mechanisms or different aspects of visual information are completely independent of each other in the sense that they do not interact. For example, Justine Sergent (1991) reported that absolute distance between two stimuli sometimes influences performance in categorical tasks that require individuals to make decisions about their relative location. Issues of independence have also been addressed by Niebauer (1996). In one experiment, he presented line-and-dot stimuli similar to those described earlier to the center of the visual field (CVF) on each trial. With the CVF presentation, Niebauer found that the time to perform a coordinate task (near/far judgment) was influenced by a prime stimulus that informed the observer whether the upcoming dot would be above or below the line (a categorical prime). That is, a valid

92

Hellige

prime decreased reaction time relative to an invalid prime. Interestingly, a coordinate prime (indicating whether the upcoming dot was near the line or far from the line) had no effect on making a categorical decision (Is the dot above or below the line?). Based on these priming effects, Niebauer suggests that computation of categorical information precedes, and serves as input to, the computation of coordinate spatial information--at least on CVF trials on which both hemispheres receive the stimulus information. This is an interesting idea that merits additional investigation. As has been the case in many previous experiments, reaction times in Niebauer's studies were faster for the categorical task than for the coordinate task. Among other things, it is important to determine whether this same asymmetry in priming would be found when the difficulty of categorical and coordinate tasks is reversed. The priming effects found by Niebauer (1996) on CVF trials disappeared when the line-and-dot stimuli were presented to the LVF/RH or RVF/LH on each trial. This may indicate that the priming effects on CVF trials reflect interactions between the two, equivalently stimulated, hemispheres. It is instructive, however, to consider alternative possibilities. For example, suppose that priming is obtained to the extent that the prime allows the observer to focus attention on a single critical "boundary" when a prime is presented (especially when compared with the no-prime condition). For the Above/Below task, the boundary is defined by the line that is actually presented on the screen. For the Near/Far task, there are two "invisible" boundaries that separate near from far stimuli, one above the line and one below the line. If either the prime does not permit attention to be restricted to a single, critical boundary or such a restriction is already possible in the no-prime condition, then no priming is obtained. For the Above/Below task on CVF trials, attention can be directed to the boundary line regardless of whether there is a prime. Thus, a Near/Far cue produces no additional information about the boundary and no priming is predicted. For the Near/Far task, in the no-prime condition the observer must split attention between two near/far boundaries, one above the line and one below the line. In this case, an Above/Below cue allows attention to be restricted to a single one of these boundaries, and priming results. By way of contrast, when the stimuli are directed randomly to the LVF/RH or RVF/LH on each trial, attention can never be restricted to a single critical boundary because there is always uncertainty about which visual field will be stimulated. Hence, all priming disappears. Although this

Spatial Processing

93

explanation is admittedly speculative, it could be tested in an experiment that randomly intermixed CVF and lateralized presentations, with the prediction being that priming would disappear on CVF trials. Even though additional studies are needed to choose among specific interpretations of the priming effects, the existence of such effects is generally consistent with hypothesized interaction between the two types of spatial processing. However, Kosslyn et al. (1992) found that a neural-network model that codes only categorical information (Is a dot above or below a line?) is sensitive to the distance between the dot and the line. Specifically, the further the dot from the line, the better the performance of the model. This demonstrates that a network devoted to the processing of categorical spatial relationships need not be completely immune to the effects of distance (a type of coordinate spatial information). Thus, the presence of distance effects in a categorical task (e.g., J. Sergent, 1991) does not necessarily mean that a network devoted to categorical processing interacts with a network devoted to coordinate processing. In view of Niebauer's (1996) priming results, it would be interesting to use similar neural-network models to simulate the various types of priming (categorical-on-coordinate and coordinate-on-categorical). For example, suppose that a network devoted to coordinate processing performs more efficiently with a valid categorical prime than with an invalid categorical prime. This would suggest that the existence of priming in and of itself does not necessarily indicate any interaction between computational mechanisms devoted to the two types of spatial processing. The Nature of Task-Relevant Visual Information

Hemispheric asymmetry for the identification of visual stimuli is sensitive to the precise nature of the visual input and to the aspects of visual information that are relevant for efficient performance of the task that the observer is required to perform. It is useful to consider how the nature of task-relevant visual information might also influence hemispheric asymmetry for determining the location of visual stimuli in space. Detailed review of the stimulus identification studies is beyond the scope of the present chapter. However, a brief review of the relevant stimulus identification findings provides an important context for considering potential mechanisms that underlie hemispheric asymmetry for stimulus localization.

94 Hellige

J J J JJJJJJJJ J J J

J J J J J J

Figure 3. Example of an hierarchical visual pattern composed of small letters (the local level) arranged into the shape of a large letter (the global level).

Visual stimuli can contain many levels of embedded structure, with smaller local patterns or parts contained within larger global patterns (see Figure 3). With respect to stimulus identification, there is a great deal of evidence that the fight hemisphere is superior for the processing of global levels of visual information (e.g., the large H in Figure 3) whereas the left hemisphere is superior for the processing of local levels of visual information (e.g., the small J's in Figure 3) [e.g., Delis, Robertson & Efron, 1986; Ivry & Robertson, 1997; Lamb, Robertson & Knight, 1989, 1990; Martin, 1979; Robertson, 1995; Robertson, Lamb & Knight, 1988, 1991; J. Sergent, 1982; van Kleeck, 1989]. There is a clear relationship between global and local aspects of a visual stimulus and what is referred to as low versus high spatial frequency. A single spatial frequency consists of a regular sinusoidal variation of luminance across space and looks somewhat like alternating dark and light bars with fuzzy borders. Spatial frequency refers to the number of dark-light cycles per unit of space--the more cycles per unit of space, the higher the spatial frequency. The concept of spatial frequency has generated considerable interest because it is possible, in principle, to represent any complex image as a set of spatial frequencies in specific orientations, phase relationships and so forth. Under normal viewing conditions, information about the larger, global aspects of a visual stimulus is carried by a lower range of spatial frequencies than is

Spatial Processing 95 information about the smaller, local aspects of the same stimulus. Thus, it may not come as a surprise that, at some level of processing beyond the sensory cortex, the fight and left hemispheres are biased toward efficient use of lower and higher spatial frequencies, respectively. In fact, a variety of evidence suggests that at least three aspects of spatial frequency influence hemispheric asymmetry for stimulus identification: (1) the absolute range of spatial frequencies contained in the stimulus; (2) the range of spatial frequencies that is relevant for the task being performed or that is attended to by the observer; and (3) whether the attended frequencies are high or low relative to other frequencies contained in the stimulus (for empirical examples and reviews, see Christman, 1989, 1990, Christman, Kitterle & Hellige, 1991, Hellige, 1993, 1995, 1996; Ivry & Robertson, 1997; Kitterle & Christman, 1991; Kitterle, Christman & Conesa, 1993; Kitterle, Christman & Hellige, 1990; Kitterle & Selig, 1991; J. Sergent, 1983, 1987; J. Sergent & Hellige, 1986). In view of the fact that the nature of presented and attended visual information exerts powerful influences on hemispheric asymmetry for stimulus identification, it is important to consider how the nature of visual information might contribute to hemispheric asymmetry for localizing stimuli in space. As noted earlier, Kosslyn et al. (1992; see also Koenig & Kosslyn, 1992) have provided a conceptualization of the mechanisms that underlie hemispheric asymmetry for spatial processing that emphasizes the nature of the visual information that is most useful for computing categorical versus coordinate information. In a set of neural-network simulations, Kosslyn et al. (1992) found that networks receiving input that had been filtered through units with relatively large, overlapping "receptive fields" computed coordinate spatial information better than networks that received input that had been filtered through units with relatively small, nonoverlapping "receptive fields." Exactly the reverse was found for the computation of categorical spatial information (for critique and additional discussion of these neural network simulations, see Cook, Fruh & Landis, 1995 and Kosslyn, Chabris, Marsolek, Jacobs & Koenig, 1995). To account for hemispheric differences in categorical versus coordinate processing, Kosslyn and colleagues hypothesize that the left hemisphere is predisposed toward efficient use of information from visual channels with small, nonoverlapping receptive fields whereas the right hemisphere is predisposed toward efficient use of information from visual channels

96 Hellige with large, overlapping receptive fields. In support, Kosslyn et al. (1992) suggest that magnocellular ganglia (which tend to have relatively large receptive fields) may project preferentially to the right hemisphere. In addition, they note the evidence mentioned earlier that, at some level of processing beyond the sensory cortex, the left and fight hemispheres are dominant for processing visual information carried by channels tuned to relatively high and low spatial frequencies, respectively. Although these neural network simulations have been criticized (Cook et al., 1995), and alternative interpretations of the simulations are possible, the hypotheses that were generated by the simulations can be subjected to empirical test. There are two types of predictions about the effects of experimental manipulations that serve to accentuate or attenuate processing in visual streams that have characteristics similar to those attributed to the input filtering units of the neural network models. One type of prediction has to do with the effect of experimental manipulations on categorical versus coordinate processing tasks--and this type of prediction does not directly involve hemispheric asymmetry. For example, an experimental manipulation that attenuated processing along the magnocellular visual pathway would be expected to disrupt coordinate spatial processing more than categorical spatial processing. The other type of prediction has to do with the Task by Hemisphere interaction. For example, to the extent that the two hemispheres utilize different aspects of the visual information to perform efficiently, the attenuation of one type of information should have a greater detrimental effect on the hemisphere that is more dependent on that type of information (for examples of this logic in the domain of visual identification, see Hellige, 1993; Jonsson & Hellige, 1986). With these types of predictions in mind, it is instructive to consider recent examinations of the effects of perceptual manipulations on categorical and coordinate spatial processing. It has been hypothesized that the processing of visual information in primates is accomplished by two parallel visual pathways with different spatial and temporal characteristics (for discussion, see Breitmeyer & Williams, 1990; Breitmeyer, May & Heller, 1991; Livingstone & Hubel, 1984, 1987, 1988; Schiller & Malpeli, 1978; Shapley, 1994; Van Essen, 1985). In general, the magnocellular system is most sensitive to low spatial frequencies, has high temporal resolution and responds quickly and transiently to moving targets. This system is thought to be involved in such things as brightness discrimination, the perception of motion

Spatial Processing 97 and depth, the localization of visual stimuli in coordinate space and in the global analysis of visual scenes. By way of contrast, the parvocellular system is most sensitive to high spatial frequencies, has a long response persistence and responds in a sustained fashion to stationary targets. This system is thought to be involved in such things as the identification of visual patterns, especially small local details, and in color perception. Given the characteristics attributed to these two visual pathways, one possible interpretation of the neural network results described earlier is that categorical and coordinate spatial processing depend relatively more on information carried by the parvocellular and magnocellular pathways, respectively. Elizabeth Cowin Roth and I (Cowin and Hellige, 1994) examined the effects of dioptric blurting on categorical (above/below) and coordinate (near/far) spatial processing tasks using line and dot stimuli similar to those described earlier. Dioptric blurting selectively impairs processing of relatively high visual spatial frequencies and, according to the hypotheses outlined, such blurting should be particularly disruptive of categorical spatial processing. In fact, we found that dioptric blurring consistently increased reaction time and error rate for a categorical task that required observers to indicate whether a dot was above or below a line. However, the amount of dioptric blurring that we used had no consistent effect on either reaction time or error rate for a coordinate task that required observers to indicate whether the dot was within 3 mm of the line. On an initial block of trials, we found significantly fewer errors on LVF/RH than on RVF/LH trials for the coordinate processing task and this LVF/RH advantage was independent of whether the stimuli were clear or blurred. While we did not design this experiment with parvocellular and magnocellular pathways in mind, a dioptric blurring manipulation might be expected to differentially attenuate processing along the parvocellular pathway. Thus, our results suggest that processing along this pathway is more critical for categorical than for coordinate spatial processing. More recently, Roth and I (Cowin & Hellige, 1995; Roth & Hellige, 1997) have attempted to examine the effects of attenuating processing along the magnocellular visual pathway. We did so in the following way. Breitmeyer and his colleagues (e.g., Breitmeyer & Williams, 1990; Breitmeyer et al., 1991; Williams, Breitmeyer, Lovegrove & Guitierrez, 1991) have reported that both metacontrast masking and the perception of stroboscopic motion are considerably weaker when stimuli are

98

Hellige

presented on a red background than on an isoluminant green background. Neurophysiologically based models of metacontrast and stroboscopic motion indicate that both phenomena result from the interaction of two visual subsystems: the sustained subsystem and the transient subsystem. As Breitmeyer and colleagues note, the parvocellular and magnocellular pathways in monkey may be neural analogs of the more functionally defined sustained and transient channels, respectively. Within the context of a sustained-transient channel approach, metacontrast masking is produced when the faster responding transient channels activated by the subsequent mask inhibit the slower responding sustained channels activated by the target. The perception of stroboscopic motion is thought be the result of response integration within the transient visual system. Thus, the fact that a red background reduces both metacontrast masking and the perception of stroboscopic motion indicates that the activity of transient visual channels is attenuated by red relative to green backgrounds. From this perspective, it is interesting that a subpopulation of magnocellular neurons have receptive fields that are characterized by a red-dominant surround mechanism (e.g., Wiesel & Hubel, 1966; DeMonasterio, 1978; DeMonasterio & Schein, 1980; Livingstone & Hubel, 1984; Marrocco, McClurkin & Young, 1988). As noted by Breitmeyer and Williams (1990), this may be the reason why diffuse red light has been found to provide tonic suppression of activity in certain neurons contained in the magnocellular pathway (Dreher, Fukuda & Rodieck, 1976; Livingstone & Hubel, 1984; Van Essen, 1985). In view of the results reported by Breitmeyer and colleagues, we reasoned that the use of green stimuli on an isoluminant red background would attenuate processing along the magnocellular pathway relative to the parvocellular pathway. To the extent that processing along the magnocellular pathway is more important for coordinate spatial processing than for categorical spatial processing, coordinate processing should be more disrupted by the use of green-on-red compared to red-on-green stimulus conditions. In our experiments with colored stimuli, the stimuli on each trial consisted of a horizontal line and two dots, with the dots being on the same horizontal level as each other (see Figure 4). The line varied in length from trial to trial as did the horizontal distance between the two dots. The categorical task required observers to indicate whether the dots were above or below the line whereas the coordinate task required observers to indicate whether or not the line on that trial could fit

Spatial Processing 99 In these stimuli, the dots are above and the line fits between the gap: m

m

m

mm

~

m

m

m

In these stimuli, the dots are below and the line fits between the gap: m

m

In these stimuli, the dots are above and the line does not fit between the gap: mm

mm

m

m

m

m

m

m

In these stimuli, the dots are below and the line does not fit between the gap: m

m

mm

~

mm

~

mm

mm

m

Figure 4. Illustration of line and dot patterns used by Cowin and Hellige (1995). between the two dots. These stimuli and tasks were patterned after those used by Rybash and Hoyer (1992) and shown in previous experiments to produce a robust Task X Hemisphere interaction. For some observers, green lines and dots were presented on an isoluminant red background. For other observers, red lines and dots were presented on an isoluminant green background. The stimuli on each trial were presented briefly to either the LVF/RH or RVF/LH. Figure 5 shows reaction time for categorical and coordinate tasks as a function of background color. As shown in the figure, there was a robust Task by Color interaction. For the coordinate task, reaction time was significantly longer in the green-on-red condition than in the redon-green condition. For the categorical task, there was a nonsignificant trend in the opposite direction. Note that this interaction is consistent with the hypothesis that the coordinate task is more dependent on magnocellular processing than is the categorical task. Although Figure 5 presents the results in terms of background color, it is not clear from this experiment how much the color-condition effect is actually attributable to the background color and how much might be attributable to the color of the stimuli. In view of the fact that metacontrast masking and perception of stroboscopic motion are both reduced when black stimuli are presented against a red background, we suspected that much of the effect is likely produced by background color. In

100

Hellige 900

800 I I

700 RT

600

--

500

m

CAT

COOR

400

Green

Red Background

Color

Figure 5. Reaction time for categorical (CAT) and coordinate (COOR) spatial

processing tasks as a function of background color in the red-on-green versus greenon-red experiment. order to separate effects of background color from effects of stimulus color, we have recently completed an additional experiment in which color is presented only in the background or only in the stimulus. That is, there are four color conditions: black lines and dots presented on a red background; black lines and dots presented on a green background; red lines and dots presented on a black background; and green lines and dots presented on a black background. Although Figure 5 presents the results in terms of background color, it is not clear how much the color-condition effect is actually attributable to the background color and how much might be attributable to the color of the stimuli. In view of the fact that metacontrast masking and perception of stroboscopic motion are reduced when black stimuli

Spatial Processing

101

900

800

700 RT

B'-"""

"

600 .

CAT

. I

500

400

,_,'-I

,

I

COOR

!

Green

Red Background

Color

Figure 6. Reaction time for categorical (CAT) and coordinate (COO) spatial

processing tasks as a function of background color in the black-on-green versus black-on-red experiment.

are presented against red backgrounds, we suspected that much of the effect is likely produced by background color. In order to separate the effects of background versus stimulus color, we have recently completed an additional study in which color is presented only in the background or only in the stimulus (i.e., there are four color conditions: black lines & dots presented on a red background; black lines & dots presented on a green background; red lines & dots presented on a black background; and green lines & dots presented on a black background). Figure 6 shows reaction time for categorical and coordinate tasks as a function of background color in the case where the line and dot stimuli were black. Note that the Task by Color interaction is very similar to (and not significantly different from) that shown in Figure 5.

102

Hellige 900

800

700

RT 600 1-----'-'1 .... El m

500

400

I

CAT COOR

I

Green

Red

Background Color Figure 7. Reaction time for categorical (CAT) and coordinate (COOR) spatial processing tasks as a function of stimulus color in the green-on-black versus redon-black experiment.

There were also differential effects of stimulus color on the two spatial processing tasks, though the direction of the effect is opposite what would be expected if stimulus color contributed to the Task by Color interaction shown in Figure 5. Figure 7 shows reaction time for both tasks as a function of stimulus color for black backgrounds. Note that, for the coordinate task, reaction time was longer to red stimuli than to green stimuli, whereas stimulus color had no effect on performance of the categorical task. Thus, performance of the coordinate (but not the categorical) task is disrupted when the only color in the display is red, regardless of whether the color is restricted to the background or restricted to the stimulus. When different colors are present in the

Spatial Processing

103

Figure 8a. Illustration of a low-pass version of a stimulus from Figure 4. Although this figure provides a reasonable illustration of a band-pass filtered stimulus, the illustration is not completely identical to the stimuli as they appem~ on the viewing screen. stimulus and in the background, the background color seems to be far more important--possibly because the continuously present background color is far more important for determining the relative efficiency of processing along magnocellular and parvocellular visual pathways. Elizabeth Roth has also presented observers with spatially filtered versions of the stimuli shown in Figure 4. Of particular importance is the distinction between low-pass stimuli, which only contain spatial frequencies lower than or equal to 2 cycles per degree (cpd) of visual angle, and high-pass stimuli, which only contain spatial frequencies equal to or higher than 8 cpd (see Figures 8a [low-pass] and 8b [highpass]). The high-pass condition (low spatial frequencies are eliminated) is similar to the use of a red background, in the sense that it is likely to attenuate processing along the magnocellular pathway--at least relative

104

Hellige

Figure 8b. Illustration of a high-pass version of a stimulus from Figure 4. Although this figure provides a reasonable illustration of a band-pass filtered stimulus, the illustration is not completely identical to the stimuli as they appemed on the viewing screen. to the low-pass condition. From this perspective, it is interesting that reaction time is significantly longer in the high-pass condition than in the low-pass condition for the coordinate task, but not for the categorical task (see Figure 9). However, the different effects for categorical and coordinate tasks must be treated with caution, as the Task by Filtering Condition interaction was not statistically significant. The results from the foregoing set of experiments can be summarized in the following way. For the categorical processing tasks, only dioptric blurring (which is likely to be more disruptive of processing along the parvocellular pathway than of processing along the magnocellular pathway) had a significant detrimental effect. There was no such effect of dioptric blurring for a coordinate processing task. However, a red background (and red stimuli, when the background was

Spatial Processing 105 800

700

RT

600it ~

--

,

I"!

~

C O O R .

500

CAT

m

I

.

.

.

.

,,,

,,,

I

High-Pass

Low-Pass Filtering

Condition

Figure 9. Reaction time for categorical (CAT) and coordinate (COOR) spatial processing tasks using low-pass versus high-pass versions of the stimuli. black) was significantly disruptive relative to a green background (and green stimuli, when the background was black). In addition, the use of high-pass stimuli was significantly disruptive relative to the use of lowpass stimuli. Although other interpretations of these effects may be possible, it is interesting that the stimulus conditions that produced significant disruption for the coordinate task were chosen to attenuate processing along the magnocellular visual pathway. With respect to hemispheric asymmetry for processing different types of spatial information, it is important to note that all of these experiments produced a Task by Hemisphere interaction similar to that reported in previous experiments. That is, for the coordinate task,

106 Hellige reaction time was faster on LVF/RH trials than on RVF/LH trials whereas the reverse pattern was found for the categorical task. However, none of these perceptual manipulations influenced the magnitude of the Task by Hemisphere interaction or the magnitude of the hemispheric differences for either the categorical or coordinate tasks considered individually. That is, none of the Task by Hemisphere by Color condition interactions and none of the Task by Hemisphere by Filtering Condition interactions even approached statistical significance 1. With respect to the hypotheses derived from Kosslyn's computational model (Kosslyn et al., 1992), the results are somewhat mixed. The pattern of perceptual effects on the two tasks is consistent with the hypothesis that the magnocellular pathway is more important for coordinate than for categorical spatial processing. In view of what is known about receptive field sizes, this is consistent with the simulation models indicating that coordinate spatial information is computed more efficiently when input is filtered through units with relatively large receptive fields than through units with relatively small receptive fields, whereas the opposite may be the case for computing categorical spatial information. At the same time, the perceptual manipulations do not change the hemispheric asymmetries. Or, to put it in a slightly different way, the perceptual manipulations seem to have the same effect on both hemispheres. This suggests that, for both categorical and coordinate tasks, the two hemispheres rely on the same visual information and on the same computational mechanisms as each other--though they do not always use that information with equal efficiency. For example, a red background interferes with coordinate spatial processing equally in both hemispheres. This suggests that coordinate processing depends on the magnocellular visual pathway to the same extent in both hemispheres. Nevertheless, there is a clear LVF/RH advantage for the coordinate processing tasks. The fact that this fight-hemisphere advantage does not change with the perceptual manipulations that we have used suggests that it arises at a level of processing beyond the early sensory level, a general conclusion that is consistent with what is known about a variety of other types of hemispheric asymmetry (for discussion, see Bradshaw, 1989; Hellige, 1993; Moscovitch, 1986). To be sure, we must be cautious in speculating about the anatomical locus of these hemispheric asymmetries for components of spatial processing. It is worth noting, however, that diffuse red light has been found to suppress activity of certain neurons (so-called Type IV

Spatial Processing

107

neurons) not only in the retinal ganglia and in cells within the lateral geniculate nucleus but also in the magnocellular layers of Area 17 in the primate visual cortex (Livingstone & Hubel, 1984). In view of the fact that the color manipulations had similar effects on LVF/RH and RVF/LH trials, this suggests that the hemispheric asymmetries of the sort we have studied arise at some level subsequent to Area 17. The pattern of results just described is also generally consistent with other recent findings that question low-level, wired-in interpretations of visual laterality. For example, based on visual half-field experiments in which observers indicated whether two successive line segments had the same orientation, Kosslyn, Anderson, Hillger and Hamilton (1994) reject the idea of hemispheric differences in strength of the projection of the magnocellular ganglion cells or that there are wired-in hemispheric differences in the sizes of receptive fields. Instead, they argue that, while there appears to be a bias for the right hemisphere to attend to larger patterns and the left hemisphere to attend to smaller patterns, this is an attentional bias that can be overcome by specific task demands. With all of the foregoing results in mind, it is useful to consider an interesting alternative account of hemispheric asymmetry for processing visual information, including information about categorical and coordinate spatial relationships. Ivry and Robertson (1997) have proposed a general model of perceptual asymmetry referred to as Double Frequency by Filtering or DFF. While a detailed account of the DFF model is beyond the scope of the present chapter, I will summarize briefly its critical elements as applied to vision and its application to hemispheric asymmetry for spatial processing. An important feature of the DFF model is the proposal that hemispheric asymmetries are sensitive to relative rather than absolute spatial frequencies. In addition, the emphasis on attentional selection is especially interesting in view of the conclusions reached by Kosslyn et al. (1994). According to the DFF model, initial sensory processing is identical in the two cerebral hemispheres, so that asymmetries arise at a postsensory level. An attentional mechanism is proposed to select taskrelevant information by choosing the region of the sensory spectrum to be enhanced for further processing. In vision, this selection is proposed to be in terms of spatial frequency and in audition this selection is proposed to be in terms of temporal frequency. This first frequencyfiltering stage is hypothesized to be identical for the two hemispheres. Hemispheric asymmetries are hypothesized to arise in a subsequent,

108 Hellige second frequency-filtering stage. During this second filtering stage, low spatial and temporal frequencies are enhanced in the right hemisphere and high spatial and temporal frequencies are enhanced in the left hemisphere. Because this second filtering stage operates only on the information selected by the first filtering stage, the DFF model predicts hemispheric asymmetry in terms of relative rather than absolute spatial and temporal frequencies; specifically, left hemisphere superiority for processing relatively high spatial and temporal frequencies and right hemisphere superiority for processing relatively low spatial and temporal frequencies. Ivry and Robertson review several examples of hemispheric asymmetry that are consistent with the DFF model. For present purposes, it is particularly important to consider a DFF account of hemispheric asymmetry for categorical and coordinate spatial processing. As Ivry and Robertson (1997) note, the DFF model emphasizes the spatial frequency information required to perform the different spatial relationship tasks without any particular regard for whether the task would be described as categorical or coordinate. To the extent that one task requires higher spatial frequencies than does the other task, a RVF/LH advantage is predicted for the first task and a LVF/RH advantage is predicted for the second task. As an example of how this would operate, they consider two tasks employed by Kosslyn et al. (1989) presenting a dot and an amorphous blob on each trial. As discussed earlier, the categorical task required observers to indicate whether the dot was touching the blob and the coordinate task required observers to indicate whether the dot was within 2 mm of the blob. As Ivry and Robertson note, in the Kosslyn et al. experiment, the spacing between the dot and the blob was 0, 1 or 10 mm. Thus, the 0 mm and 10 mm stimuli were assigned to different response categories for both tasks, but the assignment of the 1 mm stimulus changed across tasks. For the categorical task, the 1 mm stimulus was assigned to the same category as the 10 mm stimulus, so that the most difficult discrimination was between the 0 mm and 1 mm stimuli. For the coordinate task, the 1 mm stimulus was assigned to the same category as the 0 mm stimulus, so that the most difficult discrimination was between the 1 mm and 10 mm stimuli. Note that, in this case, a much finer perceptual discrimination was required for the categorical task than for the coordinate task. Perhaps for this reason, overall performance was worse for the categorical task than for the coordinate task. If we assume that a finer

Spatial Processing

109

discrimination depends on higher spatial frequencies than does a less fine discrimination, then the DFF model provides an alternative account of the Task by Hemisphere interaction without regard to the categorical versus coordinate nature of the tasks. In support of the DFF account, Ivry and Robertson (1997) describe an experiment by Ivry, Prinzmetal and Hazeltine that is similar to the one just described, but that included two different spacing conditions. The first condition was similar to that used by Kosslyn et al. (1989), with the spacing between the dot and the blob being 0, 1 or l0 mm. As was the case in the Kosslyn et al. experiment, the categorical task required observers to make one response to the 0 mm stimuli and a different response to the 1 and l0 mm stimuli and the coordinate task required observers to make one response to the 0 and 1 mm stimuli and a different response to the l0 mm stimulus. In the second spacing condition, the spacing between the dot and the blob was 0, 8 and l0 mm. A categorical task required observers to make one response to the 0 mm stimuli and a different response to the 8 and l0 mm stimuli. A coordinate task required observers to make one response to the 0 and 8 mm stimuli and a different response to the 10 mm stimuli. Note that, in this second spacing condition, the coordinate task required a finer discrimination than the categorical task, so the DFF model predicts a reversal of the hemispheric asymmetries. In fact, Ivry and colleagues found exactly the sort of Spacing Condition by Task by Hemisphere interaction that would be expected--though this experiment alone does not indicate whether the relevant variable is absolute or relative spatial frequency and all laterality effects were restricted to the first trial block (cf., Kosslyn et al., 1989). If overall performance is taken as a measure of the difficulty of the perceptual discrimination, then the DFF model would seem to predict a RVF/LH advantage for whichever task leads to worse performance and a LVF/RH advantage for whichever task leads to better performance. As noted earlier, this is clearly not the case. In fact, for most of the studies cited earlier, coordinate tasks have led to considerably worse overall performance than categorical tasks (e.g., Cowin & Hellige, 1994; Hellige & Michimata, 1989; Kosslyn et al., 1989, line-and-dot tasks; Michimata, 1997). Despite this, the coordinate tasks have produced a significant LVF/RH advantage and the categorical tasks have produced a trend toward a RVF/LH advantage. However, while overall performance may be a reasonable measure of task difficulty, it is not necessarily the best

1 10 Hellige measure of the extent to which a task depends on lower versus higher spatial frequencies--and it is the frequencies that are more relevant for generating predictions from the DFF model. For example, although Cowin and Hellige found that their Near/Far task led to more errors and longer reaction times than their Above/Below task, dioptric blurring (which impairs processing of relatively high spatial frequencies) interfered with performance only for the Above/Below (categorical) task. Thus, there is a dissociation of overall performance and whether one task is more or less dependent than another task on high spatial frequencies. Consequently, differences in the overall level of perfor, mance of two tasks is of limited value in generating predictions from the DFF model. Ivry and Robertson (1997) offer additional support for the plausibility of a DFF account of hemispheric differences in spatial processing in the form of a set of neural network simulations. The simulations are based on the sort of line-and-dot stimuli described earlier and, in fact, use dot spacings patterned after those used by Hellige and Michimata (1989) and illustrated in Figure 1. Among other things, these simulations demonstrate that hemispheric differences in spatial processing tasks can emerge as a byproduct of hemispheric differences in processing relatively high versus relatively low spatial frequencies. In addition, these simulations demonstrate that such things as the actual spacing of stimuli in a set can be expected to influence the hemispheric asymmetry that is observed. Presentation of the details of the neural network simulations is beyond the scope of the present chapter. However, a critical portion of the neural network architecture can be summarized in the following way. Input is filtered through spatial frequency channels by using input layers whose units differ in the size of their receptive fields--with smaller receptive field sizes being associated with higher spatial frequencies than larger receptive field sizes. Although this is accomplished in different ways by the DFF simulation and by the neural network simulations presented by Kosslyn et al. (1992), both types of simulation rely on input units with different size receptive fields. A critical difference is the extent to which these two models emphasize hemispheric differences in processing absolute versus relative ranges of spatial frequency. By suggesting that the hemispheres may differ in receptive field size, the model presented by Kosslyn et al. emphasizes absolute spatial frequency differences in the processing properties of the two hemispheres. In contrast, in the DFF model, the hemispheres are

Spatial Processing 111 identical in their processing of different ranges of absolute spatial frequency, but selective attention operates in a way that makes the hemispheres differ in the processing of relative spatial frequencies. Although more theoretical and empirical work is needed to separate the DFF model from other computational possibilities, it is instructive to reconsider the effects of perceptual manipulations (e.g., blurring, background color, spatial frequency filtering) from the perspective provided by the DFF model. From this perspective, it is interesting that manipulations which should be more disruptive of low spatial frequency information (e.g., red background, high-pass stimuli) selectively disrupted a coordinate spatial processing task whereas a manipulation (dioptric blurring) that should be more disruptive of high spatial frequency information selectively disrupted a categorical spatial processing task. It is clear that these perceptual manipulations would be expected to influence the absolute spatial frequencies that are available for processing. The influence on relative frequency is more complicated. To be sure, such things as filtering operations affect relative frequencies within a stimulus in the following way. Consider a complex stimulus containing spatial frequencies between 0.1 cpd and 32 cpd. If such a stimulus is filtered to remove all frequencies below 8 cpd, the 8 cpd component becomes a relatively low frequency but if the stimulus is filtered to remove all frequencies above 8 cpd, the 8 cpd component becomes a relatively high frequency. However, in view of the fact that all of these stimuli contain a range of spatial frequencies, the filtering operation does not necessarily change whether the spatial frequencies required by two tasks are high or low relative to each other. With this in mind, it is interesting that the Task by Hemisphere interaction was not influenced by the perceptual manipulations outlined earlier. While these results can be accommodated by the DFF account, it must be noted that these experiments were not designed to be a test of the DFF model or to discriminate between the DFF model and other possibilities. The new results and simulations summarized in this section suggest that hemispheric asymmetry for categorical versus coordinate spatial processing is related to more general differences between the two hemispheres in aspects of visual information processing. Further consideration of these mechanisms and others that might underlie hemispheric asymmetry for processing spatial relationships can benefit from recent extensions of the categorical/coordinate distinction beyond the domain

112

Hellige

of explicit spatial localization. Accordingly, I turn now to a review of several relevant studies.

Extensions of the Categorical/Coordinate Distinction In the studies reviewed so far, observers have been required to make explicit judgments of categorical or coordinate spatial relationships. In this section, I review evidence for Categorical/Coordinate by Hemisphere interactions that have arisen in studies in which the spatial judgments are either implicit or completely absent. Following this brief review, I will consider the extent to which these various results can be treated as different manifestations of the same underlying hemispheric asymmetry. Laeng (1994) examined categorical and coordinate processing in patients with unilateral stroke and in a control group with no known neurological damage. In one experiment observers were first shown, on each trial, a drawing of one or more objects (e.g., a large cat to the left of a small cat). Following a delay of approximately 5 s, the observers were shown the original drawing and a drawing in which the objects were transformed in either a categorical way (e.g., a large cat to the fight of a small cat) or a coordinate way (e.g., a large cat to the left of a small cat, with the distance between them being larger than in the original drawing). The observer was instructed to choose the drawing that was identical to the original. In a second experiment, the same observers were shown an original drawing followed by two choices, neither of which was identical to the original. One choice contained a categorical change (e.g., left-right reversal) and the other contained a coordinate change (e.g., distance between the objects). The task of the observer was to indicate which version looked more similar to the original drawing. Note that neither of these tasks requires an explicit identification of spatial relations. In the first experiment, patients with left-hemisphere stroke tended to mistake the categorical transformation for the original whereas patients with right-hemisphere stroke tended to mistake the coordinate transformation for the original. In the second experiment, patients with left-hemisphere stroke judged the categorical transformation to be more similar than the coordinate transformation to the original stimulus whereas patients with right-hemisphere stroke showed the opposite bias.

Spatial Processing

1 13

Similar stimuli have been used in visual half-field studies with neurologically intact right-handed and left-handed observers (e.g., Laeng & Peters, 1995). Specifically, observers indicated whether a brief, lateralized drawing was identical to a drawing seen immediately before in free vision. When the two drawings were not identical, the new drawing contained the same objects as the original with either a categorical or a coordinate aspect of the drawing changed. In general, right-handed observers were faster to identify categorical changes on RVF/LH trials and coordinate changes on LVF/RH trials. There were no visual field differences for left-handed observers. Christman (1997) has provided preliminary evidence that the categorical/coordinate distinction extends to the processing of dynamic visual input. On each trial, observers were presented with a dot in one visual field that appeared to either "shrink" or "grow" and to change size either quickly or slowly. That is, over a 167 ms time period, the diameter of the dot changed by either 1.0 degrees of visual angle ("quickly") or 0.5 degrees of visual angle ("slowly"). Thus, the perceived rate of change was perfectly correlated with the spatial magnitude of the change. Christman hypothesized that indicating whether the dot shrunk or grew involved categorical processing, whereas indicating whether the rate of size change was fast or slow involved coordinate processing. The percentage of errors was significantly smaller for the categorical task than for the coordinate task and, more importantly, there was a significant Task by Hemisphere interaction. For the coordinate task there was a significant LVF/RH advantage and for the categorical task there was a nonsignificant trend in the opposite direction. The Task by Hemisphere interaction was not significant for reaction time of correct responses, perhaps because the error rates in some conditions were quite high (e.g., 40% or so). Niebauer and Christman (1996) have attempted to extended the categorical/coordinate distinction to a task that requires judgments of spatial frequency relations in sinusoidal gratings. Observers indicated whether the second of two successively presented gratings was higher or lower in spatial frequency than the first (after Kitterle & Selig, 1991). In one condition, the two gratings differed by 1.0 octave. Because a difference of 1.0 octave could be easily characterized (e.g., each grating could be easily categorized as "thick" o r "thin"), Niebauer and Christman consider this to be a categorical condition. In a second condition, the two gratings differed by only 0.125 octave. Because this

114 Hellige involves a much finer discrimination and would require processing of precise spatial frequencies, Niebauer and Christman consider this to be a coordinate condition. From this perspective, it is interesting that there was a significant LVF/RH advantage for the coordinate task and no significant hemispheric asymmetry for the categorical task. Weiner and Christman (1994) have also examined hemispheric asymmetry for the processing of what they refer to as categorical and coordinate auditory pitch relations. The results of this study are especially interesting because the tasks do not involve visual stimuli and do not demand the processing of spatial relationships (either explicitly or implicitly). Listeners were presented with a baseline tone (800 Hz) binaurally followed by a comparison tone to one ear or the other (the ear not receiving the comparison tone received white noise). The comparison tone was either 450, 600, 1067 or 1356 Hz. Note that two of the comparison tones (1067 and 1356 Hz) were higher in pitch than the baseline tone and two of the comparison tones (450 and 600 Hz) were lower in pitch than the baseline tone. In addition, two of the comparison tones (600 and 1067 Hz) were "near" in pitch to the comparison tone and two of the comparison tones (450 and 1356 Hz) were "far" in pitch from the comparison tone. For the categorical task, listeners indicated whether the pitch of the comparison tone was higher or lower than the pitch of the baseline tone. For the coordinate task, listeners indicated whether the pitch of the comparison tone was near to or far from the pitch of the baseline tone. Weiner and Christman reported a highly significant Task by Hemisphere (Ear) interaction, with there being a significant right-ear/left-hemisphere advantage for the Above/ Below task and a significant left-ear/fight-hemisphere advantage for the Near/Far task. Based on these pitch processing results, Weiner and Christman suggest that modes of categorical versus coordinate processing reflect general aspects of hemispheric asymmetry that are not specifically tied to visual stimuli. To be sure, it might be possible for listeners to use visual imagery to represent the pitch of the tones to themselves as points on a unidimensional vertical continuum, with higher pitches represented by points higher on the continuum. But the use of an imagery analog to the line-and-dot stimuli is not required. The DFF model could account for the Weiner and Christman results to the extent that the categorical task required processing of relatively higher temporal frequencies and the coordinate task required processing of

Spatial Processing

1 15

relatively lower temporal frequencies. However, it is not obvious (at least to me) whether this is the case. It is also useful to consider hemispheric asymmetry for visual shape processing tasks that differ in ways that seem analogous to the categorical/coordinate distinction with respect to spatial relationships. For example, Marsolek (1995) had observers learn to categorize novel visual stimuli by presenting them with specific exemplars that were variations on a set of unseen prototypes. After an initial learning phase, a single item was presented to the LVF/RH or RVF/LH on each trial and the observer was required to place it into the correct category. Marsolek found a LVF/RH advantage for the classification of the "old" exemplars (those used during training) and a RVF/LH advantage for the classification of the previously unseen prototypes that had been used to define the categories. In addition, on LVF/RH trials observers responded faster to old exemplars than to new, previously unseen, exemplars, with the results for prototypes being intermediate. On RVF/LH trials, observers responded faster to the previously unseen prototypes than to either type of exemplar, with there being no difference between old and new exemplars. Based on this pattern of results and on a variety of other findings, Marsolek suggests that the right hemisphere is superior for processing specific shape information of the sort that would be needed to distinguish among the exemplars within a single category whereas the left hemisphere is superior for extracting and identifying the categorydefining prototype. Jacobs and Kosslyn (1994) have extended the neural network models of Kosslyn et al. (1992) to examine the possible importance of receptive field sizes in the encoding of shape information. They found that networks that received input from units with large, overlapping receptive fields coded the identity of specific shapes better than networks that received input from units with small, nonoverlapping receptive fields. Exactly the opposite was found for the assignment of shapes to categories (see also Brown & Kosslyn, 1995). Thus, there would seem to be a connection between categorical spatial processing and the assignment of shapes to categories without regard for distinctions among the exemplars within a category and a connection between coordinate spatial processing and the ability to distinguish among the specific exemplars. In addition, Marsolek, Kosslyn and Squire (1992) have reported that, at least for words, a highly formspecific type of visual priming is restricted to the fight hemisphere (see

1 16 Hellige also Marsolek, Schacter & Nicholas, 1996). Specifically, in word-stem completion priming tasks there is more priming if the case of the letters remains the same across all phases of the experiment than if the letter case changes--but only on LVF/RH and not RVF/LH trials. Jacobs and Kosslyn relate this effect to their neural network models by suggesting that form-specific priming depends on information filtered through units with large, overlapping receptive fields. What common thread, if any, connects the experiments reviewed in this section to each other and to studies that require explicit judgments about spatial relationships? It is not difficult to see a connection between studies in which the spatial judgments are explicit and studies in which the spatial judgments are implicit (e.g., Laeng, 1994; Laeng & Peters, 1995; Christman, 1997). It is not so clear, however, what thread might connect these studies to studies of pitch relations in audition (e.g., Wiener & Christman, 1994) and to certain types of shape processing (e.g., Marsolek, 1995). One possibility is the distinction between representations that treat a wide range of stimuli as an equivalence class (a "categorical" representation) and representations that preserve discriminable differences among stimuli within an equivalence class ("coordinate" representations). While this distinction applies to the two types of spatial relationships that we have discussed, it also applies to the other dimensions illustrated in the present section. Indeed, it is this more general distinction, or something close to it, that motivated much of the research reviewed in this section and that leads the various authors to connect their work to the categorical/coordinate distinction in spatial processing. In fact, this general conceptual distinction contributed to Kosslyn's (1987) original formulation of the categorical/coordinate distinction in spatial processing. In view of this thread that runs through the various studies, it is useful to conclude the present chapter by considering what type of general mechanism could underlie all of these results and to consider promising directions for future research. Concluding Comments: More on Mechanisms and Future Directions

In this final section, I reconsider the mechanisms that may underlie hemispheric asymmetry for spatial processing in view of the additional findings just reviewed. In doing so, I also point out some of the important questions that need to be resolved and what would seem to be fruitful directions for future research.

Spatial Processing

1 17

As noted at the outset, a considerable amount of theoretical and empirical work has dealt with the distinction between categorical and coordinate spatial relations. I believe that much of this work has been fruitful, even though specific interpretations continue to evolve and significant theoretical issues remain. Several years ago, Justine Sergent and I (J. Sergent & Hellige, 1986) argued that, in building theories of hemispheric asymmetry, it is important to specify the characteristics of the input on which the brain operates and to move toward descriptions of hemispheric asymmetry in terms of the operations on that input that are actually performed by the brain. From this perspective, I believe that we have made progress in understanding hemispheric differences in spatial processing by acknowledging that different spatial tasks require different aspects of perceptual input to be performed with maximum efficiency. Thus, hemispheric asymmetry for the different types of spatial processing can be viewed as manifestations of underlying mechanisms that have the potential to account for other aspects of hemispheric asymmetry as well. It has been suggested that the type of asymmetries reviewed in the immediately preceding section are manifestations of the same underlying mechanisms that produce Task by Hemisphere interactions when explicit judgments about spatial location are required. To the extent that this is the case, the various perceptual manipulations (e.g., blurring, background colors, frequency-filtering) would be expected to have similar effects across these various experimental paradigms--including those in which judgments about spatial location are implicit rather than explicit (e.g., Laeng & Peters, 1995; Niebauer & Christman, 1996) and those in which there may be no spatial judgments at all (e.g., Marsolek, 1995). In designing such studies, it will be important to include conditions that have the potential to discriminate between the effects of absolute and relative spatial frequency (cf., Kosslyn et al., 1992; Ivry & Robertson, 1997). It is also important to know whether the distinction between categorical and coordinate spatial relationships is important for determining hemispheric asymmetry outside the visual modality. To be sure, Weiner and Christman (1994) report an analog of sorts for auditory processing, but their task involved pitch relations rather than spatial relations. It would be useful to have studies that make explicit, spatial processing demands. For example, within both auditory and tactile domains it is possible to present lateralized stimuli that come from

1 18 Hellige different locations in space and manipulate whether the required spatial judgment is categorical (e.g., in front of versus in back of) or coordinate (within a specified distance). Finding similar Task by Hemisphere interactions across stimulus modalities would motivate the search for mechanisms that transcend any single modality. It is difficult, of course, to speculate about the sort of mechanisms that might transcend specific modalities or that might connect the various findings reviewed in the present chapter. Nevertheless, I think there are plausible candidates that deserve careful consideration. For example, Kosslyn et al. (1992) were motivated to vary receptive field size in the input units of their neural network simulations (and the extent to which the receptive fields of different units overlap) by considering whether there might be hemispheric differences in what has been termed coarse coding. In general, coarse coding refers to a situation in which coarse processing at one level results in very precise processing at another level. A prototypical example is color vision, in which precise coding of color emerges from only three types of cones in the retina with overlapping distributions of sensitivity centered at different wavelengths of light. There are many examples of this sort of coarse coding within the sensory domains (e.g., Hinton, McClelland & Rumelhart, 1986). In view of the fact that coarse coding of various sorts is used widely as a computational strategy by the brain, any systematic hemispheric differences in the use of coarse coding could account for the fact that subtle, complementary hemispheric asymmetries occur for a broad range of domains including motor performance, language, spatial processing and emotion (see Hellige, 1993). And if it is the case that coarse coding provides an effective computational mechanism for preserving discriminable differences among stimuli within an equivalence class, as some neural network simulations have suggested (e.g., Jacobs & Kosslyn, 1994; Kosslyn et al., 1992), then hemispheric differences in the use of coarse coding could underlie the sort of results reviewed in the immediately preceding section of the present chapter. In fact, it has even been hypothesized that hemispheric differences in semantic processing come about because of relatively coarse coding of semantic information in the right hemisphere and relatively fine coding of semantic information in the left hemisphere (Beeman, Friedman, Grafman, Perez, Diamond & Lindsay, 1994). I began the present chapter by observing that the two hemispheres do not have identical ability to localize stimuli in space. Research in the

Spatial Processing

119

last 10 years has done a great deal to refine our conceptualization of hemispheric asymmetry for spatial processing and to focus attention on m e c h a n i s m s that might underlie the asymmetries that have been documented. There are, at the present time, several promising leads that need to be explored. We are fortunate to have at our disposal an increasing range of converging operations to assist in that exploration; techniques for the study of behavioral asymmetries in neurologically intact individuals and in patients, neural-network modeling techniques, a variety of structural and functional brain-imaging techniques, and so forth. Each of these techniques is useful for differentiating among the hypotheses considered here and together they provide a powerful means of exploring the leads that have been identified. Although some of the leads are likely to take us to dead ends and others will lead us to more complexity than we imagine, we will undoubtedly learn a great deal about spatial processing and hemispheric asymmetry along the way. Notes 1. Cowin and Hellige (1995) reported that, for the green-on-red versus red-on-green experiment, there was a Background Color by Hemisphere interaction on the very first 24-trial block with reaction time as the dependent variable. Specifically, averaged across the two spatial tasks, there tended to be a LVF/RH advantage when the background was red and a RVF/LH advantage when the background was green. We are not inclined to treat this effect as reliable for the following reasons. In an overall analysis from the same experiment that included trial block as an independent variable, there was no hint whatsoever of either a Background Color by Hemisphere interaction or a Trial Block by Background Color by Hemisphere interaction. That is, the omnibus analysis provided no support for the claim that there is a Background Color by Hemisphere interaction at all on any block of trials. Furthermore, when the background was red, an analysis restricted to the first trial block indicated that there were significantly fewer errors on RVF/LH than on LVF/RH trials--a visual field difference exactly opposite that found with reaction time as the dependent variable. In addition, even when analysis was restricted to the first trial block, there was no hint of any Color Condition by Hemisphere interaction (or Filtering Condition by Hemisphere interaction) for any of our subsequent experiments reported in the present section--and this was the case for both reaction time and percentage of errors. That is, we have not replicated this particular effect. Finally, in none of the experiments was there ever any indication that the theoretically relevant Task by Hemisphere interaction changed with perceptual manipulations.

120 Hellige References

Beeman, M., Friedman, R. B., Grafman, J., Perez, E., Diamond, S., & Lindsay, M. B. (1994) Summation priming and coarse coding in the fight hemisphere. Journal of Cognitive Neuroscience, 6, 26-45. Boles, D. B. (1991) Factor analysis and the cerebral hemispheres: Pilot studyand parietal functions. Neuropsychologia, 29, 59-92. Boles, D. B. (1992) Factor analysis and the cerebral hemispheres: Temporal, occipital and frontal functions. Neuropsychologia, 30, 963 -988. Bradshaw, J. L. (1989) Hemispheric specialization and psychological function. Chichester, England: Wiley. Breitmeyer, B. G., & Williams, M. C. (1990) Effects of isoluminantbackground color on metacontrast and stroboscopic motion: Interactions between sustained (P) and transient (M) channels. Vision Research, 30, 1069-1075. Breitmeyer, B. G., May, J. G., & Heller, S. S. (1991) Metacontrast reveals asymmetries at red-green isoluminance. Journal of the Optical Society of America, 8, 1324-1329. Brown, H. D., & Kosslyn, S. M. (1995) Hemispheric differences in visual object processing: Structural versus allocation theories. In R. J. Davidson & K. Hugdahl (Eds.), Brain Asymmetry (pp. 77-122). Cambridge MA: MITPress. 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 a~yammetries in temporal integration. Neuropsychologia, 28, 3619, J

I

-'1"o

Christman, S. (1997) Hemispheric asymmetry in categorical versus coordinate processing of dynamic visual input. Paper presented at the Annual Meeting of the International Neuropsychological Society, Orlando, FL. Christman, S., Kitterle, F. L., & Hellige, J. B. (1991) Hemispheric asymmetry in the processing of absolute versus relative spatial frequency. Brain and Cognition, 16, 62-73. Cook,N. D., Fruh H., & Landis T. (1995) The cerebral hemispheres and neural network simulations: Design considerations. Journal of

Experimental Psychology: Human Perception and Performance, 21,

410-422. Cowin E. L., Hellige J. B. (1995) Spatial processing and hemispheric asymmetry: Contribution of the magnocellular system. Paper presented at the Annual Meeting of the Psychonomic Society, Los Angeles, CA. Cowin, E. L., & Hellige, J. B. (1994) Categorical versus coordinate spatial processing: Effects of blurring and hemispheric asymmetry. Journal of Cognitive Neuroscience, 6, 156-164. Crebolder, J. M., & Bryden, M. P. (1996) Assessing verbal and spatial hemispheric performance with identical material. Brain and Cognition, 30, 383-385.

Spatial Processing

121

Delis D. C., Robertson L. C., & Efron R. (1986) Hemispheric ecialization of memory for visual hierarchical stimuli. uropsychologia, 24, 205-214. DeMonasterio, F. M., & Schein, S. J. (1980) Protan-like spectral sensitivity of foveal Y ganglion cells of the retina of macaque monkeys. Journal of Physiology, London, 299, 385-396. DeMonasterio, F. M. (1978) Center and surround mechanisms of opponent-color X and Y ganglion cells of retina of macaques. Journal of Neurophysiology, 41, 1418-1434. Dreher, B., Fukuda, Y., & Rodieck, R. W. (1976) Identification, classification and anatomical segregation of cells with X-like and Ylike properties in the lateral genlculate nucleus of old-world primates. Journal-of Physiology, London, 258, 433-452. Hellige, J. B. (1993) Hemispheric Asymmetry: What's Right and What's Left.. Cambridge: Harvard University Press. Hellige, J. B. (1995) Hemispheric asymmetry for components of visual information processing. In R. J. Davidson & K. Hugdahl (Eds.), Brain Asymmetry (pp. 99-121). Cambridge, MA: MIT Press. Hellige, J. B. (1996) Hemispheric asymmetry for visual information processing. Acta Neurobiologiae Experimentalis, 56, 485-497. Hellige, J. B., & Michimata, C. (1989)Categorization versus distance: Hemispheric differences for processing spatial information. Memory & Cog~nition, 17, 770-776. Hellige, o. B., Bloch, M. I., Cowin, E. L., Eng, T. L., Eviatar, Z., & Sergent, V. (1994) Individual variation in hemispheric asymmetry: Multitask study of effects related to handedness and sex. Journal of Experimental Psychology: General, 123, 235-256. Hellige J. B., & Sergent Ji (1986) Role of task factors in visual field asymmetries. Brain and Cognition, 5, 200-222. Hinton, G. E., McClelland, J. L., & Rumelhart, D. E. (1986) Distributed representations. In D. E. Rumelhart & J. L. McClelland (Eds.),

Parallel distributed processing: Explorations in the microstructure of cognition, Vol. 1 (Pp. 77-109). Cambridge, MA: MIT Press. Ivry, R. & Robertson, L. C. (1997) The two sides of perception.

Cambridge, MA: MIT Press. Jacobs, R. A., & Kosslyn, S. M. (1994) Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations. Cognitive Science, 18, 361-386. Jonsson, J. E., & Hellige, J. B. (1986)Lateralized effects of blurring: A test of the visual spatial frequency model of cerebral hemisphere asymmetry. Neuropsychologia, 24, 351-362. Kitterle F . L . , & Sellg L. M. (1991) Visual field effects in the discrimination of sine wave gratings. Perception & Psychophysics, 50, 15-18. Kitterle F. L., Christman S., & Conesa J. (1993) Hemispheric differences in the interference among components of compound gratings. Perception & Psychophysics, 54, 785-793. Kitterle, F. L., Christman, S., & Hellige, J. B . (1990) Hemispheric differences are found in the identification but not the detection of low vs. high spatial frequencies. Perception & Psychophysics, 48, 297-306.

122 Hellige Kitterle, F. L., Hellige, J. B., & Christman, S. (1992) Visual hemispheric asymmetries depend on which spatial frequencies are task relevant. Brain and Cognition, 20, 246-258. Koenig, O., Reiss, L. P., & Kosslyn, S. M. (1990) The development of spatial relation representations: Evidence from studies o f cerebral lateralization. Journal of Experimental Child Psychology, 50, 119130. Kosslyn S. M., Chabris C. F., Marsolek C. J., Jacobs R. A., & Koenig O. (1995) On computational evidence for different types of spatial relations encoding: Reply to Cook et al. (1995). Journal of

Experimental Psychology: Human Perception and Performance, 21,

423-431. Kosslyn, S. M. (1987)Seeing and imagining in the cerebral hemispheres: A computational approach. Psychological Review, 94, 148-175. Kosslyn, S. M. (1988) Aspects of a cognitive neuroscience of mental imagery. Science, 240, 1621-1626. Kosslyn, S. M., & Koenig, O. (1992) Wet mind: The new cognitive neuroscience. New York: Free Press. Kosslyn, S. M., Anderson, A. K., Hillger, L. A., & Hamilton, S. E. (1994) Hemispheric differences in sizes of receptive fields or attentional biases? Neurops~chology, 8, 139-147. Kosslyn, S. M., Chabris, C. r., Marsolek, C. J., & Koenig, O. (1992) Categorical versus coordinate spatial relations: Computational analyses and computer simulation. Journal of Experimental Psychology: Human Perception and Performance, 18, 562-577. Kosslyn, S. M., Holtzman, J., Farah, M. J., & Gazzaniga, M. S. (1985) A computational analysis of mental image generation: Evidence from functional dissociations in split-brain patients. Journal of Experimental Psychology_: General, 114, 311-341. Kosslyn, S. M., Koenig, O., Barrett, A., Cave, C. B., Tang, J., & Gabrieli, J. D. E. (1989) Evidence for two types of spatial representations: Hemispheric specialization for categorical and coordinate relations.

Journal of Experimental Psychology: Human Perception and Performance, 15, 723-735.

Kosslyn, S. M., Maljkovic, V., Hamilton, S. E., Horwitz, G., & Thompson, W. L. (1995) Two types of image generation: Evidence for left and fight hemisphere processes. Neuropsychologia, 33, 14851510. Laeng, B. (1994) Lateralization of categorical and coordinate spatial functions: A study of unilateral stroke patients. Journal of Cognitive Neuroscience, 6, 189-203. Laeng, B., & Peters, M. (1995) Cerebral lateralization for the processing o f spatial coordinates and categories in left- and right-handers. Neuropsychologia, 33, 421-439. Laeng, B., Peters, M., & McCabe, B. (1996) A left hemisphere s bias for encoding visuo-spatial categories? Submitted for publication. Lamb M. R., Robertson L. C., & Knight R. T. (1989) Attention and interference in the processing of hierarchical patterns: Inferences from patients with right and left temporal-parietal lesions. Neuropsychologia, 27, 471-483.

Spatial Processing 123 Lamb, M. R., Robertson, L. C., & Kn!ght, R. T. (1990) Component mechanisms underlying the processing of hierarchically organized patterns: Inferences from patients with unilateral cortical lesions.

Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 471-483.

Livingstone M. S., & Hubel D. H. (1988) Segregation of form, color, movement, and depth: Anatomy, physiology, and perception. Science, 240, 740-749. Livingstone M. S., & Hubel, D. H. (1987) Psychophysical evidence for separate channels for the perception of form, color, motion, and depth. Journal of Neuroscience, 7, 3416-3468. Livingstone, M. S., & Hubel, D. H. (1984) Anatomy and physiology of a color system in the primate visual cortex. Journal of Neurosci-ence, 4, 309-356. Marrocco, R. T., McClurkin, J. W., & Young, R. A. (1988) Spatial summation and conduction latency classification of cells in the lateral geniculate nucleus of macaques. Journal of Neuroscience, 2, 12751291. Marsolek, C. J. (1995) Abstract-visual-form representations in the left cerebral hemisphere. Journal of Experimental Psychology: Human

Perception and Performance, 21,375-386.

Marsolel~, C. J,, Kosslyn, S. M., & Squire, L. R. (1992) Form-specific visual priming in the right cerebral hemisphere. Journal of

Experimental Psychology: Learning, Memory, and Cognition, 18,

492-508. Marsolek, C. J., Schacter, D. L., & Nicholas, C. D. (1996) Form-specific yisual priming for new associations in the fight cerebral hemisphere. Memory & Cbgnition, 24, 539-556. Martin M. (1979) Hemispheric specialization for local and global processing. Neuropsychologia, 17, 33-40. Michimata, C. (1997)Hemispheric processing of categorical and coordinate spatial relations in vision and visual imagery. Brain and

Cognition.

Moscovitch, M. (1986) Afferent and efferent models of visual asymmetries: Theoretical and empirical implications. uropsychologia, 24, 91-114. Niebauer, C. L. (1996) On the nature of categorical and coordinate spatial relation representations. Unpublished Ph.D. Dissertation, The University of Toledo, Toledo, OH. Niebauer, C., & Christman, S. (1996) Categorical versus coordinate processing of spatial frequency relations. Submitted for publication. Robertson L. C. (1995) Hemispheric specialization and cooperation in processing complex visual patterns. In F. L. Kitterle (Ed.), Hemispheric communication: Mechanisms and models, (Pp. 301318). Hillsdale, N.J.: Erlbaum Press. Robertson L. C., Lamb M. R., Knight R. T. (1988) Effects of lesions of temporal-parietal junction on perceptual and attentional processing in humans. Journal of Neurosdence, 8, 3757-3769. Robertson L. C., Lamb M. R., Knight R. T. (1991) Normal global-local analysis in patients with dorsolateral frontal lobe lesions. Neuropsychologia, 29, 959-968.

124 Hellige Roth, E. & Hellige, J. B. (1997) Unpublished manuscript. Rybash, J. M., & Hoyer, W. J. (1992) Hemispheric specialization for categorical and coordinate spatial representations: A reappraisal. Memory & Cognition, 20, 271-276. Schiller, P. H., & Malpeli, J. G. (1978) Functional specificity of lateral geniculate nucleus laminae of the rhesus monkey. Journal of Neurophysiology, 41,788-797. Sergent J. (1982) The cerebral balance of power: Confrontation or cooperation? Journal of Experimental Psychology: Human Perception and Performance, 8, 253-272. Sergent J. (1983) The role of the input in visual hemispheric asymmetries. Psychological Bulletin, 93, 481-514. Sergent J. (1987) Failures to confirm the spatial-frequency hypothesis: Fatal blow or healthy complication? Canadian Journal of Psychology, 41,412-428. Sergent, J. (1991) Judgments of relative position and distance o n representations of spatial relations. Journal of Experimental Psychology: Human Perception and Performance, 17, 762-780. Sergent, J., & Hellige, J. B. (1986) Role ot" input factors in visual-field asymmetries. Brain and Cognition, 5, 174-199. Shapley R. (1994) Parallel neural pathways and visual function. In M. S. Gazzaniga (Ed.), The cognitive neurosciences, (Pp. 315-324). Cambridge, MA: MIT Press. Van Essen, D. C. (1985) Functional organization of primate visual cortex. In E. G. Jones & A. A. Peters (Eds.), Cerebral cortex, (Vol. 3, Pp. 259-329). New York: Plenum Press. van Kleeck M. H. (1989) Hemispheric differences in global versus local processing of hierarchical visual stimuli by normal subjects: New data and a meta-analysis of previous data. Neuropsychologia, 27, 11651178. Weiner, R., & Christman, S. (1994) Hemispheric asymmetry in the processing of categorical versus coordinate auditory pitch relations. Paper presented at the Annual Meeting of the Psychonomic Society, St. Louis, MO. Wiesel, T. N., & Hubel, D. H. (1966) Laminar and columnar distribution of geniculocortical fibers in the macaque monkey. Journal of Neurophysiology, 29, 1115-1156. Williams, M. C., Breitmeyer, B. G., Lovegrove, W. J., & Guitierrez, C. (1991) Metacontrast with masks varying in spatial frequency and wavelength. Vision Research, 31 , 2017-2023.