- Email: [email protected]
Movements of attention in the three spatial dimensions and the meaning of “neutral” cues
Neuropsychologia, Printed in Great
Vol. 25. No. IA. pp. 19-29, Bntain
MOVEMENTS DIMENSIONS
0028-3932187 $3.00+0.00 Pergamon Journals Ltd.
1987
OF ATTENTION IN THE THREE SPATIAL AND THE MEANING OF “NEUTRAL” CUES
Lutz DE GONZAGA GAWRYSZEWSKI,*
LUCIA
CARLO Istituto
di Fisiologia
RIGGIO,
GIACOMO
RIZZOLATTI~
and
UMILTA
Umana,
Universita
di Parma,
Italy
Abstract-Six experiments were conducted to examine the effect of various attentional manipulations on reaction time to visual stimuli. The first three experiments compared the responses to stimuli presented in the depth (Experiment 1 ), along the horizontal (Experiment 2), and vertical (Experiment 3) meridians in a valid condition (stimulus presented in the cued position), an invalid condition (stimulus presented in the alternative position to the cued position) and a neutral condition (no information on stimulus position). The most interesting result was the demonstration that attention can be moved along the sagittal plane in the absence of vergence eye movements and that when attention is focused on a certain point, unattended points between this point and the observer (i.e. near points) are responded faster than unattended points beyond it (i.e. far points). In the frontal plane no asymmetry was found between the responses to unattended points above or below the fixation, whereas a certain, albeit non-constant, advantage was present for unattended stimuli on the right of the fixation point in respect to those on the left of it. The second series of experiments was similar to the first one, except that a new situation was introduced in which the fixation point was cued and stimuli could appear either in correspondence to it or in a peripheral position (invalid condition with attention at the fixation point). The results showed that in this new situation the responses to unattended stimuli are much longer than they are under neutral conditions, and as long as they are under conventional invalid condition. It is suggested that the so called neutral condition is a condition of diffuse attention and an attempt is made to explain it in terms of a premotor theory of attention.
INTRODUCTION THERE IS clear evidence that attention can be allocated to different points in the visual field in the absence of saccadic eye movements [4, 5, 8, 15; see also 141. The present research addresses two questions related to this finding. The first question concerns a fundamental human skill. Can people selectively attend to points located at different depths? The second question concerns the effect of cueing the fixation point. Does this procedure impair reaction time to peripheral stimuli relative to a condition (neutral condition) where, supposedly, attention is on the fixation point, but stimuli in correspondence to it are never presented? The first issue is important because visual exploration normally involves continuous fixation changes along the sagittal dimension. Thus, if covert orientation of attention is a normal way of exploring space, this capacity should not be confined to the frontal plane, but should also include the depth plane. Furthermore, physiological data show that the external
* Author’s present address: Universidade Federal Fluminense, RJ, Brasil. t Correspondence to be addressed to: G. Rizzolatti, Istituto Gramsci 14, 43 100 Parma, Italy. 19
Departamento di Fisiologia
de Neurobiologia, Umana,
Universita
Niteroi,
24000
di Parma,
via
space is not a unitary structure but its perception is mediated by different neural circuits [19, 201. In particular, the space around the observer (peripersonal space) is partly controlled by circuits responsible for the organization of reaching mouth and arm movements and of vergence eye movements, and partly by circuits controlling saccadic eye movements. By contrast, space that is distant from the observer (far space) is under the control of circuits which are responsible for conjugate eye movements. It is therefore interesting to see if there is a directional bias in the depth plane congruent with this anatomofunctional organization. The second aim of this paper is to clarify the location of attention under the so-called “neutral” condition. According to POSNER [14], a “neutral” condition is one in which subjects expect stimuli with equal probability in two locations equidistant from the fixation point. Posner argued that since there is no advantage to orient attention towards one or the other location, attention should remain in register with the eyes and therefore at the fixation point. This point of view has been generally accepted in practice and the advantages (benefits) and the disadvantages (costs) of attending or not attending to a given stimulus location have usually been measured relative to “neutral” conditions of this type. The assumption however that in “neutral” condition attention is aligned with the eyes is by no means proved. For example it is possible that under neutral condition subjects switch attention randomly or systematically between the two locations where stimuli can appear. If this is true, the results obtained under “neutral” conditions are nothing else but the arithmetic average of the responses given to the attended and non-attended stimuli. A second possibility is that under “neutral” condition subjects do not focus attention on any particular location but distribute attention over a large region of the visual field using the diffuse mode of attention postulated by JONIDES [l I]. Finally, a third possibility is that under “neutral” conditions attention is “disengaged” from its alignment with gaze and ready to move to a new position. According to this point of view deployment of attention involves a multistage process [16]. One stage could be involved in releasing attention from the stimulus (disengagement), a second stage could control the actual movement of attention, and a third stage could be responsible for capturing the stimulus (engagement). In order to test whether attention is focused on the fixation point under neutral conditions we forced the subjects to keep their attention on the fixation point by presenting stimuli with high probability of occurrence at that point, and we compared the responses to peripheral stimuli in this condition with those obtained under “neutral”conditions. The results showed that responses to peripheral stimuli were much longer in the new condition, than under “neutral” conditions. Before reporting the experiments, an additional theoretical point requires clarification. Orienting of attention is usually described as the movement of a sort of an internal eye, which can be directed to different parts of the visual space, thus enhancing information processing efficiency at the selected position [IO]. As explained in the companion paper [21] we believe however that the notion that attention moves is a crude metaphor to describe the fact that attention can be allocated to different parts of the space and that it should be avoided. Our proposal is that redeployment of attention depends on motor programs which compute the next target position for the eyes or for other effecters in the external space. Orienting of attention is a consequence of motor programming. For convenience however we will observe convention and we will use the term “movements of attention”, but in an exclusively descriptive way and without any implication of an internal eye or any type of mechanisms of this kind.
ATTENTION
MOVEMENTS
IN THE THREE
EXPERIMENTS
SPATIAL
DIMENSIONS
21
1-3
Methods In these experiments we studied the capacity of normal subjects to shift attention in the three spatial dimensions. Since a different set-up was used in each experiment, they are described separately. Male students of the University of Parma served as paid subjects in each study. They were right-handed according to Edinburgh Inventory [13], had normal or corrected vision and were nai’ve as to the purpose of the experiment. Also eye dominance for sighting was assessed. Experiment 1 (Depth). Eight subjects were tested. They sat in a dimly illuminated cubicle in front of a horizontal platform having three vertical bars inserted in it. These bars (15 cm high) were aligned along a sagittal axis and were located at 19, 38 and 57 cm from the subject’s eyes (Fig. I). An adjustable head- and chin-rest assured that this distance remained constant. On the upper extremity of each bar there was a green LED. The LED on the far bar (57 cm) and on the near bar (19 cm) were used for generating the stimuli, the LED on the intermediate bar (38 cm) for producing the cues and the fixation point. The LEDs used for generating the stimuli were screened except for a central hole that differed in diameter for the two LEDs. The resulting stimuli were circles with 0.1 mm dia. (near stimulus) and 0.3 dia. mm (far stimulus). Because of their different linear dimensions the two stimuli covered retinal regions of the same dimension (circles of about 0.3” dia. The vertical location of the stimuli and of the fixation point were adjusted for each subject and were positioned at the eye level. Stimulus duration was 10 msec. On the intermediate bar, beside the upper LED, that served as the fixation point, there were other two LEDs. They were partially screened and, according to the way they were turned on, the following cues appeared: a cross (arms 0.5” long, 0.1” wide), a cross plus an arrowhead pointing up or a cross plus an identical arrowhead pointing down. The arrowhead was 0.4” high and 0.2” wide. The subjects were instructed to press a microswitch as quickly as possible at the appearance of the stimuli, using the right thumb. Each trial began with the presentation of one of the three cues. The cross indicated that the stimulus could occur with equal probability at the near or far position. The cross with the arrowhead pointing up indicated a high probability (80%) of occurrence of the far stimulus, the cross with the arrowhead pointing down indicated a high probability (80%) for the near stimulus. The three cues occurred at random and had the same probability of occurrence. The subject was invited to be ready to respond to either stimulus in the case of the cross and to shift his attention (but not his eyes) to the stimulus position indicated by the arrow in the case of the other two cues. The interval between the onset of the cue and the stimuli varied randomly between 1500 and 2000 msec. The cue remained on until the subject responded. Stimulus presentation was controlled by a microcomputer which also FAR
67cm)
FP
(38cm)
NEAR
(19cm)
-
+
Y-l& -%_LJ_Jd 0
0
_d
FIG. 1. Diagram showing the position of the stimuli with respect to the eyes. The two circles indicate the location of the two LEDs used for generating the stimuli, the square the location of the fixation point (FP). Below the square the cues used during the experiment are represented. The inset shows, enlarged, the retinal points stimulated by the two LEDs: near LED 4.5” temporal retina (T); far LED 1.5” nasal retina (N).
LUIZ DE GONZAGA GAWRYSZEWSKI et al.
22
recorded RTs to the nearest msec. Visual feedback about speed and accuracy was shown on a visual display located below the fixation point after each response. RTs shorter than 140 msec or longer than 1000 msec were discarded and additional trials were conducted at the end of each trial block to complete the sequence. Each subject attended two experimental sessions on consecutive days. Each session consisted of 3 blocks of trials separated by a 3 min rest period and was preceded by a series of practice trials. During each block (consisting of 86 trials) the stimuli appeared according to a quasi-random sequence with a maximum of three consecutive presentations at each position. The sessions and the blocks were identical except for the randomized sequence of stimuli. Experiment 2. (Horizontal meridicln). Eight subjects (seven new ones and one from the previous experiment) were tested. Each subject sat in front of a cathode-ray tube controlled by an Apple II computer with his head positioned in an adjustable head-and-chin rest. Each trial began with the appearance of three squares (boxes) and a fixation point. The three boxes (1.5” x 1.5”) were horizontally aligned with their centers at a distance of lo” from each other. The fixation point was a dot positioned at the center of the central box. 500 msec after the appearance of the boxes a cue was shown above the central square. The cue could be a cross, an arrow pointing to the left or an arrow pointing to the right. The cross indicated that the probability of the occurrence of a stimulus in the peripheral boxes was equal, the arrows indicated a high probability of occurrence (75%) in the designated position. No stimuli were presented in the central box. The cues were presented in random order and occurred with the same probability. The interval between the onset of the cue and the stimuli varied between 800 and 1200 msec. The instructions to the subjects were the same as in Experiment 1. The stimulus was a “rectilinear spiral” of 1” x I”. Stimulus duration was 100 msec. The responses were executed by pressing a key of the computer keyboard (letter B) using the right index finger. Visual feedback about speed and accuracy was presented on the screen after each response. Reaction time was measured to the nearest msec by the computer. For other details see Experiment I. Experiment 3. (Vertical meridian). Eight new subjects participated in the experiment. The apparatus and the experimental procedure were identical to those of the previous experiment except that the screen was rotated by 90’ Thus the stimulus configuration that was horizontal in Experiment 2 was vertical in Experiment 3. The cues were presented to the left of central square in the first session and to the right of it in the second session for four subjects. This was reversed for the remaining subjects. Data analysis. Six median RTs were obtained for each subject in each experiment. The variables which defined the treatments were stimulus position (near far or right left or upperlower) and condition: valid (stimulus in the position indicated by the cue), invalid (stimulus in the position opposite to the one indicated by the cue), neutral (no indication about the stimulus location). The six medians were used for statistical analysis (see Results).
Results Experiment 1. (Depth). Figure 2 (left side) shows the mean RT (average of the medians across the subjects) to near and far stimuli in the three basic experimental conditions. An analysis of variance with stimulus position (near and far) and condition (valid, neutral and
Horizontal
Depth stimulus 260 _
1
near 3 position
far
Valid
Neutral
meridian
stimulus
position
right
.
.
Invalid
Valid
Vertical stimulus
II
Neutral
meridian position
upper
Invalid
Valid
Neutral
Invalid
FIG. 2. The results of Experiment 1 (Depth), 2 (Horizontal meridian) and 3 (Vertical meridian) are shown respectively in the three panels. In each panel mean RTs to peripheral stimuli in the three basic conditions (valid, neutral and invalid) are presented. Empty bars indicate the responses to near, right and upper stimuli. Filled bars indicate the responses to far, left and lower stimuli.
ATTENTION
MOVEMENTS
IN THE THREE SPATIAL
23
DIMENSIONS
invalid)
as within-subjects factors showed a significant effect of stimulus position, P~0.03 and condition, F (2, 14)= 15.20, P
Table 1.
Post-hoccomparisons
involving
type of condition
and stimulus
Comparison Valid Invalid
position
for Experiments
Neutral
Invalid
Analysis Condition
Analysis Stimulus position
Valid
Neutral
Overall Near Far
205 205 204
223* 223t 224t
205 205 204
239t 225t 252t
223 223 224
239* 225 252t
Overall Valid Invalid Neutral
2.-Horizontal meridian
Overall Left Right
283 284 282
294* 290 299
283 284 282
326t 335t 317t
294 290 299
326t 335t 317*
Overall Valid Invalid Neutral
3.-Vertical meridian
Overall Upper Lower
281 282 279
290 291 288
281 282 279
290 291 288
304* 308; 300
Overall Valid Invalid neutral
Experiment
1 .-Depth
* Significant
at alpha =0.05;
t significant
304t 308t
3w
at alpha =O.Ol. Other comparisons
l-3
Comparison Near Far 218 205 225 223 Lef 303 284 335 290 Upper 294 282 308 291
221* 204 252t 224 Right 299 282 317* 299 Lower 289 279 300 288
not significant.
Experiment 2. (Horizontal meridian). The results of the experiment are presented in the middle part of Fig. 2. An analysis of variance with stimulus position (right and left) and condition (valid, neutral and invalid) as within-subjects factors showed that only condition was significant, F (2, 14)=46.37, Pt0.001. The difference between responses to right (299 msec) and left (303 msec) stimuli were not significant, whereas the interaction between stimulus position and condition reached significance, F (2, 14) = 4.3 1, P < 0.05. The post-hoc comparisons are presented in Table 1 (2-Horizontal meridian). The most interesting result is the asymmetry of the responses to left and right under invalid condition. No asymmetry is present under valid and neutral treatments. Experiment 3. (Vertical meridian). Figure 2 (right side) shows the results of this experiment. An analysis of variance with stimulus position (upper and lower) and condition (valid, neutral and invalid) as main factors showed that only condition was significant, F (2, 14)= 7.15, P~0.01. The post-hoc comparisons are shown in Table 1 (3-Vertical meridian). Significant differences were found between the invalid and the two other conditions but not between the valid and the neutral condition. The advantage of valid over neutral conditions was of 9 msec. Since it was present in 7 out of 8 subjects, it is possible that the absence of significance between the valid and neutral conditions represents a type 2 error. The difference between the responses to upper and lower stimuli (294 vs 289 msec) was not significant.
24
LUIZ IX
GONZAGA
GAWRYSZEWSKI
EXPERIMENTS
et 01.
4-6
These experiments were aimed at clarifying the concept of neutral condition. The main difference with the previous three experiments was that stimuli could appear at the fixation point. Right-handed 1131 male students of the University of Parma served as paid subjects. They had a normal or corrected vision and were naive as to the purpose of the experiment. Euperimmt 4. (Dephl. Eleven subjects participated in the experiment. The same experimental apparatus as m Experiment I was used. The only change consisted of a modification of the LEDs system on the central bar which allow,ed presentation of a stimulus (a vertical bar 0. I wide. 0.5 high) at the fixation point. The near and far stimuli were the same as in Experiment I. The duration of all stimuli was 10 msec. Each trial began with the appearance of one of the following cues: an arrowhead pointing up (see for its dimension Experiment l), an arrowhead pointing down or two arrowheads pointing in opposite directions. The arrowhead pointing up indicated a high probability of stimulus occurrence (SO”%) in the far position and a low probability of a stimulus occurrence (20%) in the near position. The arrowhead pointing down indicated the opposite (80”/0 for the near position .20% for the far positionl. Presentation of two arrowheads indicated that the probability of occurrence of a stimulus at the lixation point was 60”/0 and that the probability of occurrence of a stimulus at the far and near positions was 20% for each position. In seven subjects eye position was controlled using an infrared eye movement monitor (Mode1 200 Eye Movement Monitor, G. W.. Applied Science Laboratories). The operating principle consists in detecting the change in reflected infrared light between the sclera and the iris. The resolution of the system was better than 0.5 The output of the monitor as well as a mark corresponding to the onset ofthe stimulus was recorded on a polygraph. At theend ofeach experimental session, the records were analysed and the trials in which eye movements occurred identified and discarded. The number of trials in which the subjects did not maintain fixation was negligible (less than 3%). E.qwrimrnt 5. (Horizontalmrridian). Eight subjects (six new ones and two from Experiment 4) participated in the experiment. The stimulus arrangement followed Experiment 2. The only difference consisted in the cue instructions. Presentation ofa cross indicated that the probability ofoccurrence ofa stimulus at the fixation point was 500/u, while the probability of occurrence of a stimulus to the right or to the left of fixation point was 25% for each position. Presentation of an arrow pointing to the right or to the left indicated a high probability (75%) of occurrence of a stimulus in the designated position and a low probability (25%) in the opposite one. Other instructions to the subjects and data analysis were the same as in Experiment 2. Exuperimenr 6. (C’wtical mrridim). Eight new subjects participated in Experiment 6. The stimulus arrangement followed Experiment 3. The cue instructions were identical to those of Experiment 5. Presentation of a cross indicated that the probability of occurrence of a stimulus at the fixation point was 5O”/u, while the probability of occurrence of a stimulus above or below the fixation point was 25”/; for each position. Presentation of an arrow pointing up or down indicated a high probability (75”/“) ofoccurrence of a stimulus in the designated position and a low probability (25%) in the opposite one. Other instructions and data analysis were the same as in Experiment 3.
Results Experiment 4. (Depth). Figure 3 (left side) shows the results of the experiment. The RTs were slowest when the subject paid attention to the central location and the stimulus appeared in a peripheral location (invalid condition with attention at the fixation point, IC). They were slow, but to a lesser degree, when the subject paid attention to a peripheral location (near or far) and the stimulus appeared in the opposite. peripheral location (invalid condition with attention at a peripheral point, IP) and fastest when the cue and the stimulus corresponded (valid condition). The responses to near stimuli were faster than those to far stimuli. An analysis of variance of RTs with stimulus position (near and far) and condition (valid, IP and IC) as within-subjects factors showed significant main effects of stimulus position, F (1, 10) = 5.06, P < 0.05 and condition, F (2, 20) = 17.60. P < 0.001. The post-hoc comparisons are shown in Table 2 (4Depth). In this and in the following experiments there were two types of valid trials, those in response to an attended peripheral stimulus (traditional valid trials) and those in response to an attended central stimulus. For compatibility with the previous experiments we decided to use the traditional valid trials in the analysis. In the present experiment RT for the two types of valid trials was identical
ATTENTION
MOVEMENTS
Horizontal
Depth stimulus 290 ii $ 6
IN THE THREE SPATIAL
stimulus
position
near
r:
far
a
right
3407
Vertical
meridian position a
25
DIMENSIONS
meridian
stimulus 350-I
position
upper
r
lower
a
270
; ._ 5 ._ ;; : L
250
230
Valid
IP
IC
Valid
IP
IC
Valid
IP
IC
FIG. 3. The results of Experiments 4,5 and 6 are shown respectively in the three panels. In each panel mean reaction times to peripheral stimuli in the three basic conditions are presented. The basic conditions were: valid, invalid with attention at a peripheral point, IP; invalid with attention at the fixation point, IC. All conventions as in Fig. 2.
Experiment 5. (Horizontal meridian). The results of this experiment are illustrated in the central part of Fig. 3. The analysis of variance with stimulus position (right and left) and condition (valid, IP and IC) as main factors, showed that only condition was significant, F (2, 14) = 59.67, P
DISCUSSION The purpose of the present experiments was two-fold: first, to study how attention is oriented in the three dimensions ofspace (Experiments l-3), and second, to examine how the alignment of attention with fixation affects the speed of reaction time to stimuli in nonattended positions (Experiments 4-6). The results of the first experiments showed that attention can be moved in depth as well as in the other spatial dimensions. When a subject expects a stimulus at a certain location in the sagittal plane his responses to it are faster than to stimuli presented in other locations in the
LUIZ DE GONZAGA GAWRYSZEWSKIet ul.
26
Table 2. Posr-hoc Experiment
comparisons
S.-Horizontal meridian
6.-Vertical meridian
type of condition
Analysis Stimulus
4.-Depth
involving
position
and stimulus
position
Comparison
for Experiments Analysis Condition
Near
Far
245 222 242 271
255* 228 256* 282
Valid
IC
Valid
IP
IC
IP
Overall Near Far
225 222 228
277t 271t 282T
225 222 228
249t 242+ 256i.
217 271 2x2
249t 242t 256t
Overall Valid IP IC
Overall Left Right
280 281 219
3237 320t 326T
280
327t 327t 3277
323 320 326
321 321 327
Overall Valid IP IC
Overall
289 292 286
324t 316T 331t
289 292 286
317t
324 316 331
317 321 314t
Overall Valid IP IC
Upper Lower
* Significant at alpha=0.05; tsignificant attention at a peripheral point; IC=invalid
281 279
at alpha=O.Ol. with attention
3217 314t
446
Other comparisons not significant. at the fixation point.
Comparison
LfJff 309 281 327 320 upper 310 292 321 316 IP=
Right 311 219 321 326 LOWW 310 286 314 331t
Invalid with
same plane. Similarly, if the subject is not informed where in depth the stimulus will appear his responses are slower than when he knows the exact stimulus location. That changes in eye vergence are not responsible for the effect has been already shown by DOWNING and PINKER [2] and Experiment 4 clearly indicates that attentional effects with stimuli presented at different distances from the observer can be obtained in the absence of eye movements. Thus there is no doubt that attention can be dissociated from vergence eye movements. An interesting aspect of the depth experiments (Experiments 1 and 4) is the asymmetry between the capacity to detect near stimuli when attention is allocated to a far point and that of detecting far stimuli when attention is allocated to a near point. When a subject expects a near stimulus, and a far stimulus is presented, the subject needs 27 msec more to respond to it than when he expects a far stimulus, and a near stimulus is presented. If this phenomenon is measured with reference to the neutral condition, the cost for moving attention towards the near stimulus is negligible (2 msec), but the cost for moving it to the far stimulus amounts to 28 msec. DOWNING and PINKER [2] reported a qualitatively similar effect but of smaller magnitude than that reported here. This could be due to methodological differences between the experiments. For example binocular depth cues were present in our experiments but not in those of Downing and Pinker. An attempt to provide an explanation of the observed asymmetry is presented at the end of the discussion. There is another point concerning depth experiments that deserves some comments. The above discussion rests on the assumption that the subjects did use depth for discriminating stimulus position. One might argue, however, that, due to suppression of the input from the non-dominant eye, the stimuli were in fact aligned along the horizontal meridian. For example, if the input from the left eye were to have been suppressed, the near stimulus would have been to the left of the right eye fovea and the far stimulus to the right of it. This interpretation seems to be unlikely because (a) the ambient light was strong enough to provide the subjects with other depth cues, besides retinal disparity and (b) the suppression of one eye is not a common phenomenon among normal subjects. Moreover, one feature of our data argues against the eye-suppression explanation. One prediction from this explanation is
ATTENTION
MOVEMENTS
IN THE THREE
SPATIAL
DIMENSIONS
27
that any right-left bias seen along the horizontal meridian should manifest itself as a near-far bias in the depth situation. Since most of our subjects in Experiments 1 and 4 were right-eye dominant (10 out of 19 subjects showed right-eye dominance; 3 were left-eye dominant and for the other 6, eye dominance was uncertain), one would have predicted an advantage for the far stimulus, given the right-side advantage observed in horizontal dimension (see below). On the contrary we found a clear advantage for the near stimulus. Movements of attention along the vertical meridian did not show any bias in favor of upper or lower stimuli. In contrast, an advantage was found for right stimuli in the case of attentional shifts along the horizontal meridian. The cost, measured with reference to the neutral condition, was 18 msec for shifting attention to the right, and 45 msec for shifting attention to the left. The advantage for right stimuli, although reported by others [7, see other ref. in 221, is not consistently observed. For example it was absent in Experiments 5 of the present paper and in the experiments reported in the companion paper [21]. If this right-left asymmetry is a reliable effect, one may speculate that it is a reflection of right hemisphere dominance in spatial attention. According to this interpretation, the right hemispace is controlled by both hemispheres, while the left hemispace is controlled exclusively by the right hemisphere. The consequence of this functional arrangement should be a greater salience of right stimuli than left stimuli. The second series of experiments (Experiments 46) dealt with the problem of the neutral condition. The question we asked was the following: if the probability of stimulus occurrence renders the fixation point attentionally important, will the RTs to peripheral stimuli differ from those obtained under neutral condition? In other words is the conventional neutral treatment genuinely neutral in both spatial and functional terms with attention focused at the fixation point? Our results showed that when there is an advantage to focus attention on the fixation, RTs to the non-attended stimuli are as long as those obtained when attention is oriented to a peripheral stimulus and much longer that thby are under neutral condition. These findings support the following conclusion. First the difference between the conventional invalid condition (with attention on a peripheral point) and neutral condition is not due to the distance that attention has to travel in order to reach the non-attended location. Secondly, and more importantly, the results indicate that attention is not oriented to the fixation point under neutral condition or, if it is, only part of its capacity is allocated to that point. If attention is not focused on the fixation point under the neutral treatment, what has happened to it? One reasonable possibility is that subjects allocate attention randomly to one or the other of the two equiprobable positions where the stimulus can occur [see for a similar hypothesis 93. If this were the case however the S.D. should be higher under the neutral condition than under the other conditions since it involves an average of valid and invalid trials. Our data show that this is not true. The S.D. in the neutral condition was slightly higher (55 msec) than in the valid condition (47 msec) and slightly lower than in the invalid condition (64 msec). The same pattern of results was obtained in the horizontal, vertical and sagittal dimensions. Thus a “gambling” interpretation of neutral condition can be ruled out. Another possibility is that subjects allocate the attention simultaneously to both of the possible stimulus locations under neutral treatment, but, given the limited capacity of the attentional resources [12], the responses are slower than with a single attended stimulus location, Although there is some evidence concerning the division and distribution of attention to various locations in the visual field [3, 17, 181, it is unlikely that this is the case
2x
LUIZ m G~NZAC;.~ GAWRY~KE~~KI
et al
under the experimental conditions used here. Under conditions similar to ours POWER et al. [15] found that subjects were unable to allocate attention to two points lying on each side of the fixation point. More recently JONIDES [ 1 l] and ERIKSEN and YEH [6] reached the same conclusion. The idea that subjects pay simultaneous and selective attention to two stimulus locations under neutral treatment appears to be rather unlikely. A plausible and more attractive interpretation of the neutral condition is that attention is not focused at all but is diffused. As suggested by JONIDES [ 1 l] there are two modes in which subjects can attend to a visual display: (a) by allocating attentional resources evenly across the display and, (b) by concentrating them in response to a cue on one display location. Our argument is that the way in which subjects are instructed to deal with the stimuli during neutral trials induces them to switch from the focused mode of processing information adopted during cued trials to a diffuse mode. In the companion paper [Zl] we proposed that in visual detection tasks the time necessary to “move” attention corresponds to time required to program a motor act, or, if this has been programmed, to modify it. How can this interpretation account for the diffuse mode of processing information? Let us consider the motor programming point of view. If one compares neutral conditions to conditions in which the stimulus is expected at the fixation point, it is clear that they are not identical. In the first condition a precise outer alignment is not necessary since no stimuli are going to effect at the fixation point. In contrast, in the second condition a continuous precise adjustment of ocular axes is essential if the subject is to optimize stimulus detection. Thus in the neutral condition the oculomotor program amounts to nothing more than postural control, whereas in the cued condition, it is a part of active exploratory behaviour. The consequence of this is that, although the eyes are aligned with the fixation point in both conditions, in the cued condition a running motor program has to be stopped and substituted for another one. whereas in the neutral condition the programming sector of the ocular system is not “occupied”, since posture can be maintained at a low motor level. The diffuse mode of processing information corresponds therefore to the readiness of premotor system to be programmed, whereas the cost paid when the fixation point is cued and the stimulus appears in a peripheral location reflects the time necessary to change motor programs. A final point which deserves some comments concerns the attentional advantage that near stimuli have in respect to far stimuli. Our starting hypothesis was that stimuli closer to the body should be favored in respect to those far from the body because more neuronal circuits control attention in peripersonal space [19,20]. The global advantage observed for the near position in Experiment 4 and for the invalid responses in Experiment 1 appears to confirm this hypothesis. However, neither the valid nor the neutral condition of Experiment 1 show any difference between these conditions. Thus. if a greater representation of the peripersonal space facilitates near stimuli, its influence is not very strong. An alternative interpretation of the “near effect” is that when subjects deploy attention to a certain point in space waiting for stimuli in the sagittal dimension, an attentional space is dynamically created which goes from the fixation point, maximal locus of attention, towards the body of the observer. This hypothesis readily explains why unexpected near stimuli are responded to faster than unexpected far stimuli when attention is focused on a central or a peripheral point. On the other hand, since this hypothesis does not postulate any intrinsic, neuronal basis favoring near stimuli, it has no difficulty, in contrast to the structural hypothesis, in explaining the absence of the “near” effect in the valid and in neutral trials. When attention is focused on a near or a far point its effect is the same. Similarly, when attention is diffuse, as it is on neutral
ATTENTIONMOVEMENTS IN THE THREESPATIAL l>IMENSIONS
29
trials, no advantage for a near stimulus is to be expected. Although this is clearly a post-hoc interpretation of the findings, the idea that attention extends towards the observer from a chosen point in the sagittal plane, can easily explain the experimental findings. Finally, it is interesting to note that in case of vergence eye movement, convergence is easier than divergence [l], a further finding in favor of close link between attention and eye movements. Acknowledgements-This research was supported by funds from the Consiglio Nazionale delle Ricerche and the Minister0 della Pubblica Istuzione to G.R. and C.U. L.C. was postdoctoral fellow of CNPq-Brasil (proc. 200.049/82). The final version of the paper was written while C.U. was a fellow of the Institute for Advanced Studies of the Hebrew University ofJerusalem. The authors wish to thank K. Kirsner for a critical reading of the manuscript and G. lanelli for the computer programs.
REFERENCES I. CARPENTER, R. Movements oj the Eyes. Pion, London, 1977. 2. DOWNING, C. J. and PINKER, S. The spatial structure of visual attention. In Arrention and Performance XI, M. 1. POSNER and 0. S. M. MARIN (Editors), pp. 171-187. Lawrence Erlbaum Associates, Hillsdale, N. J., 1985. 3. EGLY, R. and HOMA, D. Sensitization of the visual field. J. exp. Psycho!.. Hum. Percept. Perform. 10, 778%793. 1984. 4. ERIKSEN, C. W. and HOFFMAN, J. E. The extent of processing of noise elements during selective coding from visual displays. Percept. Psychophys. 14, 155-160, 1973. noise suppression or signal enhancement? Bull. 5. ERIKSEN, C. W. and HOFFMAN, J. E. Selective attention: Psychonom.
Sot. 4, 587-589,
1974.
6. ERIKSEN, C. W. and YEH, Y. Allocation of attention in the visual field. J. exp. Psycho/., Hum. Percepf. Perform. 11, 583-597, 1985. underlying the unilateral neglect syndrome. In Hemi7. HEILMAN. K. M. and WATSON, R. T. Mechanisms inattention and Hemispheric Specialization (Advances in Neurology Series) 18, E. A. WEINSTEIN and R. F. FRIEDLAND (Editors), pp. 93-106. Raven Press, New York, 1977. stages in the processing of visual information. Percept. Psychophys. 18, 348-354, 8. HOFFMAN, J. E. Hierarchical 1975.
following commissural section. In 9. HOLTZMAN, J. D., VOLPE, B. T. and GAZZANIGA, M. S. Spatial orientation Varieties ofAttention, R. PARASURAMANand D. R. DAVIES (Editors), pp. 375-~394. Academic Press, Orlando, FL, 1984. 10. JOHNSTON, W. A. and DARK, V. J. Selective attention. A. Rev. Psycho/. 37, 43-75, 1986. 11. JONIDES, J. Further toward a model of the mind’s eye’s movement. Bull. Psychonom. Sot. 21, 247-250, 1983. Englewood Cliffs, N. J.. 1973. 12. KAHNEMAN, D. Attention and Efort. Prentice-Hall. the Edinburgh Inventory. Neuropsychologia 9, 13. OLDFIELD, R. C. The assessment and analysis of handedness: 97.113,197l.
14. POSNER, M. I. Chronometric E~yplorations of Mind. Lawrence Erlbaum Associates, Hillsdale, N. J., 1978. 15. POSNER, M. I., SNYDER, C. R. R. and DAVIUSON, B. J. Attention and the detection of signals. J. exp. Psychol., Gen. 109, 160-174, 1980. 16. POSNER, M. I., Walker, J. A., Friedrich, F. J. and RAFAL, R. D. Effects of parietal injury on covert orienting of attention. J. Neurosci. 4, 1863%1874, 1984. 17. SHAW, M. L. A capacity allocation model for reaction time. J. exp. Psycho/., Hum. Percept. Perform. 4,58&598, 1978.
18. SHAW, M. L. and SHAW, P. Optimal allocation ofcognitive resources to spatial locations. J. exp. Psycho/., Hum. Percept. Perform. 3,201- 21 I, 1977. 19. RIZZOLATTI, G. Mechanisms ofselective attention in mammals. In Advances in Vertebrate Neuroethology, J. P. EWERT, R. R. CAPRANICA and D. J. INC;LE(Editors), pp. 261-297. Plenum Publishing Corporation, New York 1983. attention and unilateral neglect. In 20. RIZZOLATTI, G. and CAMARDA, R. Neural circuits for spatial Neurophysiological and Neuropsychological Aspecfs ofSpatial Neglect, M. JEANNEROD(Editor). North Holland, Amsterdam (in press). and 21. RIZZOLATTI, G., RIGGIO, L., DASCOLA, 1. and UMILTA, C. Reorienting attention across the horizontal vertical meridians: evidence in favor of a premotor theory of attention. Neuropsychologia 25, 3140, 1987. 22. UMILTA, C. and NICOLETTI, R. Attention and coding effects in S-R compatibility due to irrelevant spatial cues. In Attention and Performance XI, M. I. POSNER and 0. S. M. MARIN (Editors), pp. 457471. Lawrence Erlbaum Associates, Hillsdale, N. J., 1985.
Vol. 25. No. IA. pp. 19-29, Bntain
MOVEMENTS DIMENSIONS
0028-3932187 $3.00+0.00 Pergamon Journals Ltd.
1987
OF ATTENTION IN THE THREE SPATIAL AND THE MEANING OF “NEUTRAL” CUES
Lutz DE GONZAGA GAWRYSZEWSKI,*
LUCIA
CARLO Istituto
di Fisiologia
RIGGIO,
GIACOMO
RIZZOLATTI~
and
UMILTA
Umana,
Universita
di Parma,
Italy
Abstract-Six experiments were conducted to examine the effect of various attentional manipulations on reaction time to visual stimuli. The first three experiments compared the responses to stimuli presented in the depth (Experiment 1 ), along the horizontal (Experiment 2), and vertical (Experiment 3) meridians in a valid condition (stimulus presented in the cued position), an invalid condition (stimulus presented in the alternative position to the cued position) and a neutral condition (no information on stimulus position). The most interesting result was the demonstration that attention can be moved along the sagittal plane in the absence of vergence eye movements and that when attention is focused on a certain point, unattended points between this point and the observer (i.e. near points) are responded faster than unattended points beyond it (i.e. far points). In the frontal plane no asymmetry was found between the responses to unattended points above or below the fixation, whereas a certain, albeit non-constant, advantage was present for unattended stimuli on the right of the fixation point in respect to those on the left of it. The second series of experiments was similar to the first one, except that a new situation was introduced in which the fixation point was cued and stimuli could appear either in correspondence to it or in a peripheral position (invalid condition with attention at the fixation point). The results showed that in this new situation the responses to unattended stimuli are much longer than they are under neutral conditions, and as long as they are under conventional invalid condition. It is suggested that the so called neutral condition is a condition of diffuse attention and an attempt is made to explain it in terms of a premotor theory of attention.
INTRODUCTION THERE IS clear evidence that attention can be allocated to different points in the visual field in the absence of saccadic eye movements [4, 5, 8, 15; see also 141. The present research addresses two questions related to this finding. The first question concerns a fundamental human skill. Can people selectively attend to points located at different depths? The second question concerns the effect of cueing the fixation point. Does this procedure impair reaction time to peripheral stimuli relative to a condition (neutral condition) where, supposedly, attention is on the fixation point, but stimuli in correspondence to it are never presented? The first issue is important because visual exploration normally involves continuous fixation changes along the sagittal dimension. Thus, if covert orientation of attention is a normal way of exploring space, this capacity should not be confined to the frontal plane, but should also include the depth plane. Furthermore, physiological data show that the external
* Author’s present address: Universidade Federal Fluminense, RJ, Brasil. t Correspondence to be addressed to: G. Rizzolatti, Istituto Gramsci 14, 43 100 Parma, Italy. 19
Departamento di Fisiologia
de Neurobiologia, Umana,
Universita
Niteroi,
24000
di Parma,
via
space is not a unitary structure but its perception is mediated by different neural circuits [19, 201. In particular, the space around the observer (peripersonal space) is partly controlled by circuits responsible for the organization of reaching mouth and arm movements and of vergence eye movements, and partly by circuits controlling saccadic eye movements. By contrast, space that is distant from the observer (far space) is under the control of circuits which are responsible for conjugate eye movements. It is therefore interesting to see if there is a directional bias in the depth plane congruent with this anatomofunctional organization. The second aim of this paper is to clarify the location of attention under the so-called “neutral” condition. According to POSNER [14], a “neutral” condition is one in which subjects expect stimuli with equal probability in two locations equidistant from the fixation point. Posner argued that since there is no advantage to orient attention towards one or the other location, attention should remain in register with the eyes and therefore at the fixation point. This point of view has been generally accepted in practice and the advantages (benefits) and the disadvantages (costs) of attending or not attending to a given stimulus location have usually been measured relative to “neutral” conditions of this type. The assumption however that in “neutral” condition attention is aligned with the eyes is by no means proved. For example it is possible that under neutral condition subjects switch attention randomly or systematically between the two locations where stimuli can appear. If this is true, the results obtained under “neutral” conditions are nothing else but the arithmetic average of the responses given to the attended and non-attended stimuli. A second possibility is that under “neutral” condition subjects do not focus attention on any particular location but distribute attention over a large region of the visual field using the diffuse mode of attention postulated by JONIDES [l I]. Finally, a third possibility is that under “neutral” conditions attention is “disengaged” from its alignment with gaze and ready to move to a new position. According to this point of view deployment of attention involves a multistage process [16]. One stage could be involved in releasing attention from the stimulus (disengagement), a second stage could control the actual movement of attention, and a third stage could be responsible for capturing the stimulus (engagement). In order to test whether attention is focused on the fixation point under neutral conditions we forced the subjects to keep their attention on the fixation point by presenting stimuli with high probability of occurrence at that point, and we compared the responses to peripheral stimuli in this condition with those obtained under “neutral”conditions. The results showed that responses to peripheral stimuli were much longer in the new condition, than under “neutral” conditions. Before reporting the experiments, an additional theoretical point requires clarification. Orienting of attention is usually described as the movement of a sort of an internal eye, which can be directed to different parts of the visual space, thus enhancing information processing efficiency at the selected position [IO]. As explained in the companion paper [21] we believe however that the notion that attention moves is a crude metaphor to describe the fact that attention can be allocated to different parts of the space and that it should be avoided. Our proposal is that redeployment of attention depends on motor programs which compute the next target position for the eyes or for other effecters in the external space. Orienting of attention is a consequence of motor programming. For convenience however we will observe convention and we will use the term “movements of attention”, but in an exclusively descriptive way and without any implication of an internal eye or any type of mechanisms of this kind.
ATTENTION
MOVEMENTS
IN THE THREE
EXPERIMENTS
SPATIAL
DIMENSIONS
21
1-3
Methods In these experiments we studied the capacity of normal subjects to shift attention in the three spatial dimensions. Since a different set-up was used in each experiment, they are described separately. Male students of the University of Parma served as paid subjects in each study. They were right-handed according to Edinburgh Inventory [13], had normal or corrected vision and were nai’ve as to the purpose of the experiment. Also eye dominance for sighting was assessed. Experiment 1 (Depth). Eight subjects were tested. They sat in a dimly illuminated cubicle in front of a horizontal platform having three vertical bars inserted in it. These bars (15 cm high) were aligned along a sagittal axis and were located at 19, 38 and 57 cm from the subject’s eyes (Fig. I). An adjustable head- and chin-rest assured that this distance remained constant. On the upper extremity of each bar there was a green LED. The LED on the far bar (57 cm) and on the near bar (19 cm) were used for generating the stimuli, the LED on the intermediate bar (38 cm) for producing the cues and the fixation point. The LEDs used for generating the stimuli were screened except for a central hole that differed in diameter for the two LEDs. The resulting stimuli were circles with 0.1 mm dia. (near stimulus) and 0.3 dia. mm (far stimulus). Because of their different linear dimensions the two stimuli covered retinal regions of the same dimension (circles of about 0.3” dia. The vertical location of the stimuli and of the fixation point were adjusted for each subject and were positioned at the eye level. Stimulus duration was 10 msec. On the intermediate bar, beside the upper LED, that served as the fixation point, there were other two LEDs. They were partially screened and, according to the way they were turned on, the following cues appeared: a cross (arms 0.5” long, 0.1” wide), a cross plus an arrowhead pointing up or a cross plus an identical arrowhead pointing down. The arrowhead was 0.4” high and 0.2” wide. The subjects were instructed to press a microswitch as quickly as possible at the appearance of the stimuli, using the right thumb. Each trial began with the presentation of one of the three cues. The cross indicated that the stimulus could occur with equal probability at the near or far position. The cross with the arrowhead pointing up indicated a high probability (80%) of occurrence of the far stimulus, the cross with the arrowhead pointing down indicated a high probability (80%) for the near stimulus. The three cues occurred at random and had the same probability of occurrence. The subject was invited to be ready to respond to either stimulus in the case of the cross and to shift his attention (but not his eyes) to the stimulus position indicated by the arrow in the case of the other two cues. The interval between the onset of the cue and the stimuli varied randomly between 1500 and 2000 msec. The cue remained on until the subject responded. Stimulus presentation was controlled by a microcomputer which also FAR
67cm)
FP
(38cm)
NEAR
(19cm)
-
+
Y-l& -%_LJ_Jd 0
0
_d
FIG. 1. Diagram showing the position of the stimuli with respect to the eyes. The two circles indicate the location of the two LEDs used for generating the stimuli, the square the location of the fixation point (FP). Below the square the cues used during the experiment are represented. The inset shows, enlarged, the retinal points stimulated by the two LEDs: near LED 4.5” temporal retina (T); far LED 1.5” nasal retina (N).
LUIZ DE GONZAGA GAWRYSZEWSKI et al.
22
recorded RTs to the nearest msec. Visual feedback about speed and accuracy was shown on a visual display located below the fixation point after each response. RTs shorter than 140 msec or longer than 1000 msec were discarded and additional trials were conducted at the end of each trial block to complete the sequence. Each subject attended two experimental sessions on consecutive days. Each session consisted of 3 blocks of trials separated by a 3 min rest period and was preceded by a series of practice trials. During each block (consisting of 86 trials) the stimuli appeared according to a quasi-random sequence with a maximum of three consecutive presentations at each position. The sessions and the blocks were identical except for the randomized sequence of stimuli. Experiment 2. (Horizontal meridicln). Eight subjects (seven new ones and one from the previous experiment) were tested. Each subject sat in front of a cathode-ray tube controlled by an Apple II computer with his head positioned in an adjustable head-and-chin rest. Each trial began with the appearance of three squares (boxes) and a fixation point. The three boxes (1.5” x 1.5”) were horizontally aligned with their centers at a distance of lo” from each other. The fixation point was a dot positioned at the center of the central box. 500 msec after the appearance of the boxes a cue was shown above the central square. The cue could be a cross, an arrow pointing to the left or an arrow pointing to the right. The cross indicated that the probability of the occurrence of a stimulus in the peripheral boxes was equal, the arrows indicated a high probability of occurrence (75%) in the designated position. No stimuli were presented in the central box. The cues were presented in random order and occurred with the same probability. The interval between the onset of the cue and the stimuli varied between 800 and 1200 msec. The instructions to the subjects were the same as in Experiment 1. The stimulus was a “rectilinear spiral” of 1” x I”. Stimulus duration was 100 msec. The responses were executed by pressing a key of the computer keyboard (letter B) using the right index finger. Visual feedback about speed and accuracy was presented on the screen after each response. Reaction time was measured to the nearest msec by the computer. For other details see Experiment I. Experiment 3. (Vertical meridian). Eight new subjects participated in the experiment. The apparatus and the experimental procedure were identical to those of the previous experiment except that the screen was rotated by 90’ Thus the stimulus configuration that was horizontal in Experiment 2 was vertical in Experiment 3. The cues were presented to the left of central square in the first session and to the right of it in the second session for four subjects. This was reversed for the remaining subjects. Data analysis. Six median RTs were obtained for each subject in each experiment. The variables which defined the treatments were stimulus position (near far or right left or upperlower) and condition: valid (stimulus in the position indicated by the cue), invalid (stimulus in the position opposite to the one indicated by the cue), neutral (no indication about the stimulus location). The six medians were used for statistical analysis (see Results).
Results Experiment 1. (Depth). Figure 2 (left side) shows the mean RT (average of the medians across the subjects) to near and far stimuli in the three basic experimental conditions. An analysis of variance with stimulus position (near and far) and condition (valid, neutral and
Horizontal
Depth stimulus 260 _
1
near 3 position
far
Valid
Neutral
meridian
stimulus
position
right
.
.
Invalid
Valid
Vertical stimulus
II
Neutral
meridian position
upper
Invalid
Valid
Neutral
Invalid
FIG. 2. The results of Experiment 1 (Depth), 2 (Horizontal meridian) and 3 (Vertical meridian) are shown respectively in the three panels. In each panel mean RTs to peripheral stimuli in the three basic conditions (valid, neutral and invalid) are presented. Empty bars indicate the responses to near, right and upper stimuli. Filled bars indicate the responses to far, left and lower stimuli.
ATTENTION
MOVEMENTS
IN THE THREE SPATIAL
23
DIMENSIONS
invalid)
as within-subjects factors showed a significant effect of stimulus position, P~0.03 and condition, F (2, 14)= 15.20, P
Table 1.
Post-hoccomparisons
involving
type of condition
and stimulus
Comparison Valid Invalid
position
for Experiments
Neutral
Invalid
Analysis Condition
Analysis Stimulus position
Valid
Neutral
Overall Near Far
205 205 204
223* 223t 224t
205 205 204
239t 225t 252t
223 223 224
239* 225 252t
Overall Valid Invalid Neutral
2.-Horizontal meridian
Overall Left Right
283 284 282
294* 290 299
283 284 282
326t 335t 317t
294 290 299
326t 335t 317*
Overall Valid Invalid Neutral
3.-Vertical meridian
Overall Upper Lower
281 282 279
290 291 288
281 282 279
290 291 288
304* 308; 300
Overall Valid Invalid neutral
Experiment
1 .-Depth
* Significant
at alpha =0.05;
t significant
304t 308t
3w
at alpha =O.Ol. Other comparisons
l-3
Comparison Near Far 218 205 225 223 Lef 303 284 335 290 Upper 294 282 308 291
221* 204 252t 224 Right 299 282 317* 299 Lower 289 279 300 288
not significant.
Experiment 2. (Horizontal meridian). The results of the experiment are presented in the middle part of Fig. 2. An analysis of variance with stimulus position (right and left) and condition (valid, neutral and invalid) as within-subjects factors showed that only condition was significant, F (2, 14)=46.37, Pt0.001. The difference between responses to right (299 msec) and left (303 msec) stimuli were not significant, whereas the interaction between stimulus position and condition reached significance, F (2, 14) = 4.3 1, P < 0.05. The post-hoc comparisons are presented in Table 1 (2-Horizontal meridian). The most interesting result is the asymmetry of the responses to left and right under invalid condition. No asymmetry is present under valid and neutral treatments. Experiment 3. (Vertical meridian). Figure 2 (right side) shows the results of this experiment. An analysis of variance with stimulus position (upper and lower) and condition (valid, neutral and invalid) as main factors showed that only condition was significant, F (2, 14)= 7.15, P~0.01. The post-hoc comparisons are shown in Table 1 (3-Vertical meridian). Significant differences were found between the invalid and the two other conditions but not between the valid and the neutral condition. The advantage of valid over neutral conditions was of 9 msec. Since it was present in 7 out of 8 subjects, it is possible that the absence of significance between the valid and neutral conditions represents a type 2 error. The difference between the responses to upper and lower stimuli (294 vs 289 msec) was not significant.
24
LUIZ IX
GONZAGA
GAWRYSZEWSKI
EXPERIMENTS
et 01.
4-6
These experiments were aimed at clarifying the concept of neutral condition. The main difference with the previous three experiments was that stimuli could appear at the fixation point. Right-handed 1131 male students of the University of Parma served as paid subjects. They had a normal or corrected vision and were naive as to the purpose of the experiment. Euperimmt 4. (Dephl. Eleven subjects participated in the experiment. The same experimental apparatus as m Experiment I was used. The only change consisted of a modification of the LEDs system on the central bar which allow,ed presentation of a stimulus (a vertical bar 0. I wide. 0.5 high) at the fixation point. The near and far stimuli were the same as in Experiment I. The duration of all stimuli was 10 msec. Each trial began with the appearance of one of the following cues: an arrowhead pointing up (see for its dimension Experiment l), an arrowhead pointing down or two arrowheads pointing in opposite directions. The arrowhead pointing up indicated a high probability of stimulus occurrence (SO”%) in the far position and a low probability of a stimulus occurrence (20%) in the near position. The arrowhead pointing down indicated the opposite (80”/0 for the near position .20% for the far positionl. Presentation of two arrowheads indicated that the probability of occurrence of a stimulus at the lixation point was 60”/0 and that the probability of occurrence of a stimulus at the far and near positions was 20% for each position. In seven subjects eye position was controlled using an infrared eye movement monitor (Mode1 200 Eye Movement Monitor, G. W.. Applied Science Laboratories). The operating principle consists in detecting the change in reflected infrared light between the sclera and the iris. The resolution of the system was better than 0.5 The output of the monitor as well as a mark corresponding to the onset ofthe stimulus was recorded on a polygraph. At theend ofeach experimental session, the records were analysed and the trials in which eye movements occurred identified and discarded. The number of trials in which the subjects did not maintain fixation was negligible (less than 3%). E.qwrimrnt 5. (Horizontalmrridian). Eight subjects (six new ones and two from Experiment 4) participated in the experiment. The stimulus arrangement followed Experiment 2. The only difference consisted in the cue instructions. Presentation ofa cross indicated that the probability ofoccurrence ofa stimulus at the fixation point was 500/u, while the probability of occurrence of a stimulus to the right or to the left of fixation point was 25% for each position. Presentation of an arrow pointing to the right or to the left indicated a high probability (75%) of occurrence of a stimulus in the designated position and a low probability (25%) in the opposite one. Other instructions to the subjects and data analysis were the same as in Experiment 2. Exuperimenr 6. (C’wtical mrridim). Eight new subjects participated in Experiment 6. The stimulus arrangement followed Experiment 3. The cue instructions were identical to those of Experiment 5. Presentation of a cross indicated that the probability of occurrence of a stimulus at the fixation point was 5O”/u, while the probability of occurrence of a stimulus above or below the fixation point was 25”/; for each position. Presentation of an arrow pointing up or down indicated a high probability (75”/“) ofoccurrence of a stimulus in the designated position and a low probability (25%) in the opposite one. Other instructions and data analysis were the same as in Experiment 3.
Results Experiment 4. (Depth). Figure 3 (left side) shows the results of the experiment. The RTs were slowest when the subject paid attention to the central location and the stimulus appeared in a peripheral location (invalid condition with attention at the fixation point, IC). They were slow, but to a lesser degree, when the subject paid attention to a peripheral location (near or far) and the stimulus appeared in the opposite. peripheral location (invalid condition with attention at a peripheral point, IP) and fastest when the cue and the stimulus corresponded (valid condition). The responses to near stimuli were faster than those to far stimuli. An analysis of variance of RTs with stimulus position (near and far) and condition (valid, IP and IC) as within-subjects factors showed significant main effects of stimulus position, F (1, 10) = 5.06, P < 0.05 and condition, F (2, 20) = 17.60. P < 0.001. The post-hoc comparisons are shown in Table 2 (4Depth). In this and in the following experiments there were two types of valid trials, those in response to an attended peripheral stimulus (traditional valid trials) and those in response to an attended central stimulus. For compatibility with the previous experiments we decided to use the traditional valid trials in the analysis. In the present experiment RT for the two types of valid trials was identical
ATTENTION
MOVEMENTS
Horizontal
Depth stimulus 290 ii $ 6
IN THE THREE SPATIAL
stimulus
position
near
r:
far
a
right
3407
Vertical
meridian position a
25
DIMENSIONS
meridian
stimulus 350-I
position
upper
r
lower
a
270
; ._ 5 ._ ;; : L
250
230
Valid
IP
IC
Valid
IP
IC
Valid
IP
IC
FIG. 3. The results of Experiments 4,5 and 6 are shown respectively in the three panels. In each panel mean reaction times to peripheral stimuli in the three basic conditions are presented. The basic conditions were: valid, invalid with attention at a peripheral point, IP; invalid with attention at the fixation point, IC. All conventions as in Fig. 2.
Experiment 5. (Horizontal meridian). The results of this experiment are illustrated in the central part of Fig. 3. The analysis of variance with stimulus position (right and left) and condition (valid, IP and IC) as main factors, showed that only condition was significant, F (2, 14) = 59.67, P
DISCUSSION The purpose of the present experiments was two-fold: first, to study how attention is oriented in the three dimensions ofspace (Experiments l-3), and second, to examine how the alignment of attention with fixation affects the speed of reaction time to stimuli in nonattended positions (Experiments 4-6). The results of the first experiments showed that attention can be moved in depth as well as in the other spatial dimensions. When a subject expects a stimulus at a certain location in the sagittal plane his responses to it are faster than to stimuli presented in other locations in the
LUIZ DE GONZAGA GAWRYSZEWSKIet ul.
26
Table 2. Posr-hoc Experiment
comparisons
S.-Horizontal meridian
6.-Vertical meridian
type of condition
Analysis Stimulus
4.-Depth
involving
position
and stimulus
position
Comparison
for Experiments Analysis Condition
Near
Far
245 222 242 271
255* 228 256* 282
Valid
IC
Valid
IP
IC
IP
Overall Near Far
225 222 228
277t 271t 282T
225 222 228
249t 242+ 256i.
217 271 2x2
249t 242t 256t
Overall Valid IP IC
Overall Left Right
280 281 219
3237 320t 326T
280
327t 327t 3277
323 320 326
321 321 327
Overall Valid IP IC
Overall
289 292 286
324t 316T 331t
289 292 286
317t
324 316 331
317 321 314t
Overall Valid IP IC
Upper Lower
* Significant at alpha=0.05; tsignificant attention at a peripheral point; IC=invalid
281 279
at alpha=O.Ol. with attention
3217 314t
446
Other comparisons not significant. at the fixation point.
Comparison
LfJff 309 281 327 320 upper 310 292 321 316 IP=
Right 311 219 321 326 LOWW 310 286 314 331t
Invalid with
same plane. Similarly, if the subject is not informed where in depth the stimulus will appear his responses are slower than when he knows the exact stimulus location. That changes in eye vergence are not responsible for the effect has been already shown by DOWNING and PINKER [2] and Experiment 4 clearly indicates that attentional effects with stimuli presented at different distances from the observer can be obtained in the absence of eye movements. Thus there is no doubt that attention can be dissociated from vergence eye movements. An interesting aspect of the depth experiments (Experiments 1 and 4) is the asymmetry between the capacity to detect near stimuli when attention is allocated to a far point and that of detecting far stimuli when attention is allocated to a near point. When a subject expects a near stimulus, and a far stimulus is presented, the subject needs 27 msec more to respond to it than when he expects a far stimulus, and a near stimulus is presented. If this phenomenon is measured with reference to the neutral condition, the cost for moving attention towards the near stimulus is negligible (2 msec), but the cost for moving it to the far stimulus amounts to 28 msec. DOWNING and PINKER [2] reported a qualitatively similar effect but of smaller magnitude than that reported here. This could be due to methodological differences between the experiments. For example binocular depth cues were present in our experiments but not in those of Downing and Pinker. An attempt to provide an explanation of the observed asymmetry is presented at the end of the discussion. There is another point concerning depth experiments that deserves some comments. The above discussion rests on the assumption that the subjects did use depth for discriminating stimulus position. One might argue, however, that, due to suppression of the input from the non-dominant eye, the stimuli were in fact aligned along the horizontal meridian. For example, if the input from the left eye were to have been suppressed, the near stimulus would have been to the left of the right eye fovea and the far stimulus to the right of it. This interpretation seems to be unlikely because (a) the ambient light was strong enough to provide the subjects with other depth cues, besides retinal disparity and (b) the suppression of one eye is not a common phenomenon among normal subjects. Moreover, one feature of our data argues against the eye-suppression explanation. One prediction from this explanation is
ATTENTION
MOVEMENTS
IN THE THREE
SPATIAL
DIMENSIONS
27
that any right-left bias seen along the horizontal meridian should manifest itself as a near-far bias in the depth situation. Since most of our subjects in Experiments 1 and 4 were right-eye dominant (10 out of 19 subjects showed right-eye dominance; 3 were left-eye dominant and for the other 6, eye dominance was uncertain), one would have predicted an advantage for the far stimulus, given the right-side advantage observed in horizontal dimension (see below). On the contrary we found a clear advantage for the near stimulus. Movements of attention along the vertical meridian did not show any bias in favor of upper or lower stimuli. In contrast, an advantage was found for right stimuli in the case of attentional shifts along the horizontal meridian. The cost, measured with reference to the neutral condition, was 18 msec for shifting attention to the right, and 45 msec for shifting attention to the left. The advantage for right stimuli, although reported by others [7, see other ref. in 221, is not consistently observed. For example it was absent in Experiments 5 of the present paper and in the experiments reported in the companion paper [21]. If this right-left asymmetry is a reliable effect, one may speculate that it is a reflection of right hemisphere dominance in spatial attention. According to this interpretation, the right hemispace is controlled by both hemispheres, while the left hemispace is controlled exclusively by the right hemisphere. The consequence of this functional arrangement should be a greater salience of right stimuli than left stimuli. The second series of experiments (Experiments 46) dealt with the problem of the neutral condition. The question we asked was the following: if the probability of stimulus occurrence renders the fixation point attentionally important, will the RTs to peripheral stimuli differ from those obtained under neutral condition? In other words is the conventional neutral treatment genuinely neutral in both spatial and functional terms with attention focused at the fixation point? Our results showed that when there is an advantage to focus attention on the fixation, RTs to the non-attended stimuli are as long as those obtained when attention is oriented to a peripheral stimulus and much longer that thby are under neutral condition. These findings support the following conclusion. First the difference between the conventional invalid condition (with attention on a peripheral point) and neutral condition is not due to the distance that attention has to travel in order to reach the non-attended location. Secondly, and more importantly, the results indicate that attention is not oriented to the fixation point under neutral condition or, if it is, only part of its capacity is allocated to that point. If attention is not focused on the fixation point under the neutral treatment, what has happened to it? One reasonable possibility is that subjects allocate attention randomly to one or the other of the two equiprobable positions where the stimulus can occur [see for a similar hypothesis 93. If this were the case however the S.D. should be higher under the neutral condition than under the other conditions since it involves an average of valid and invalid trials. Our data show that this is not true. The S.D. in the neutral condition was slightly higher (55 msec) than in the valid condition (47 msec) and slightly lower than in the invalid condition (64 msec). The same pattern of results was obtained in the horizontal, vertical and sagittal dimensions. Thus a “gambling” interpretation of neutral condition can be ruled out. Another possibility is that subjects allocate the attention simultaneously to both of the possible stimulus locations under neutral treatment, but, given the limited capacity of the attentional resources [12], the responses are slower than with a single attended stimulus location, Although there is some evidence concerning the division and distribution of attention to various locations in the visual field [3, 17, 181, it is unlikely that this is the case
2x
LUIZ m G~NZAC;.~ GAWRY~KE~~KI
et al
under the experimental conditions used here. Under conditions similar to ours POWER et al. [15] found that subjects were unable to allocate attention to two points lying on each side of the fixation point. More recently JONIDES [ 1 l] and ERIKSEN and YEH [6] reached the same conclusion. The idea that subjects pay simultaneous and selective attention to two stimulus locations under neutral treatment appears to be rather unlikely. A plausible and more attractive interpretation of the neutral condition is that attention is not focused at all but is diffused. As suggested by JONIDES [ 1 l] there are two modes in which subjects can attend to a visual display: (a) by allocating attentional resources evenly across the display and, (b) by concentrating them in response to a cue on one display location. Our argument is that the way in which subjects are instructed to deal with the stimuli during neutral trials induces them to switch from the focused mode of processing information adopted during cued trials to a diffuse mode. In the companion paper [Zl] we proposed that in visual detection tasks the time necessary to “move” attention corresponds to time required to program a motor act, or, if this has been programmed, to modify it. How can this interpretation account for the diffuse mode of processing information? Let us consider the motor programming point of view. If one compares neutral conditions to conditions in which the stimulus is expected at the fixation point, it is clear that they are not identical. In the first condition a precise outer alignment is not necessary since no stimuli are going to effect at the fixation point. In contrast, in the second condition a continuous precise adjustment of ocular axes is essential if the subject is to optimize stimulus detection. Thus in the neutral condition the oculomotor program amounts to nothing more than postural control, whereas in the cued condition, it is a part of active exploratory behaviour. The consequence of this is that, although the eyes are aligned with the fixation point in both conditions, in the cued condition a running motor program has to be stopped and substituted for another one. whereas in the neutral condition the programming sector of the ocular system is not “occupied”, since posture can be maintained at a low motor level. The diffuse mode of processing information corresponds therefore to the readiness of premotor system to be programmed, whereas the cost paid when the fixation point is cued and the stimulus appears in a peripheral location reflects the time necessary to change motor programs. A final point which deserves some comments concerns the attentional advantage that near stimuli have in respect to far stimuli. Our starting hypothesis was that stimuli closer to the body should be favored in respect to those far from the body because more neuronal circuits control attention in peripersonal space [19,20]. The global advantage observed for the near position in Experiment 4 and for the invalid responses in Experiment 1 appears to confirm this hypothesis. However, neither the valid nor the neutral condition of Experiment 1 show any difference between these conditions. Thus. if a greater representation of the peripersonal space facilitates near stimuli, its influence is not very strong. An alternative interpretation of the “near effect” is that when subjects deploy attention to a certain point in space waiting for stimuli in the sagittal dimension, an attentional space is dynamically created which goes from the fixation point, maximal locus of attention, towards the body of the observer. This hypothesis readily explains why unexpected near stimuli are responded to faster than unexpected far stimuli when attention is focused on a central or a peripheral point. On the other hand, since this hypothesis does not postulate any intrinsic, neuronal basis favoring near stimuli, it has no difficulty, in contrast to the structural hypothesis, in explaining the absence of the “near” effect in the valid and in neutral trials. When attention is focused on a near or a far point its effect is the same. Similarly, when attention is diffuse, as it is on neutral
ATTENTIONMOVEMENTS IN THE THREESPATIAL l>IMENSIONS
29
trials, no advantage for a near stimulus is to be expected. Although this is clearly a post-hoc interpretation of the findings, the idea that attention extends towards the observer from a chosen point in the sagittal plane, can easily explain the experimental findings. Finally, it is interesting to note that in case of vergence eye movement, convergence is easier than divergence [l], a further finding in favor of close link between attention and eye movements. Acknowledgements-This research was supported by funds from the Consiglio Nazionale delle Ricerche and the Minister0 della Pubblica Istuzione to G.R. and C.U. L.C. was postdoctoral fellow of CNPq-Brasil (proc. 200.049/82). The final version of the paper was written while C.U. was a fellow of the Institute for Advanced Studies of the Hebrew University ofJerusalem. The authors wish to thank K. Kirsner for a critical reading of the manuscript and G. lanelli for the computer programs.
REFERENCES I. CARPENTER, R. Movements oj the Eyes. Pion, London, 1977. 2. DOWNING, C. J. and PINKER, S. The spatial structure of visual attention. In Arrention and Performance XI, M. 1. POSNER and 0. S. M. MARIN (Editors), pp. 171-187. Lawrence Erlbaum Associates, Hillsdale, N. J., 1985. 3. EGLY, R. and HOMA, D. Sensitization of the visual field. J. exp. Psycho!.. Hum. Percept. Perform. 10, 778%793. 1984. 4. ERIKSEN, C. W. and HOFFMAN, J. E. The extent of processing of noise elements during selective coding from visual displays. Percept. Psychophys. 14, 155-160, 1973. noise suppression or signal enhancement? Bull. 5. ERIKSEN, C. W. and HOFFMAN, J. E. Selective attention: Psychonom.
Sot. 4, 587-589,
1974.
6. ERIKSEN, C. W. and YEH, Y. Allocation of attention in the visual field. J. exp. Psycho/., Hum. Percepf. Perform. 11, 583-597, 1985. underlying the unilateral neglect syndrome. In Hemi7. HEILMAN. K. M. and WATSON, R. T. Mechanisms inattention and Hemispheric Specialization (Advances in Neurology Series) 18, E. A. WEINSTEIN and R. F. FRIEDLAND (Editors), pp. 93-106. Raven Press, New York, 1977. stages in the processing of visual information. Percept. Psychophys. 18, 348-354, 8. HOFFMAN, J. E. Hierarchical 1975.
following commissural section. In 9. HOLTZMAN, J. D., VOLPE, B. T. and GAZZANIGA, M. S. Spatial orientation Varieties ofAttention, R. PARASURAMANand D. R. DAVIES (Editors), pp. 375-~394. Academic Press, Orlando, FL, 1984. 10. JOHNSTON, W. A. and DARK, V. J. Selective attention. A. Rev. Psycho/. 37, 43-75, 1986. 11. JONIDES, J. Further toward a model of the mind’s eye’s movement. Bull. Psychonom. Sot. 21, 247-250, 1983. Englewood Cliffs, N. J.. 1973. 12. KAHNEMAN, D. Attention and Efort. Prentice-Hall. the Edinburgh Inventory. Neuropsychologia 9, 13. OLDFIELD, R. C. The assessment and analysis of handedness: 97.113,197l.
14. POSNER, M. I. Chronometric E~yplorations of Mind. Lawrence Erlbaum Associates, Hillsdale, N. J., 1978. 15. POSNER, M. I., SNYDER, C. R. R. and DAVIUSON, B. J. Attention and the detection of signals. J. exp. Psychol., Gen. 109, 160-174, 1980. 16. POSNER, M. I., Walker, J. A., Friedrich, F. J. and RAFAL, R. D. Effects of parietal injury on covert orienting of attention. J. Neurosci. 4, 1863%1874, 1984. 17. SHAW, M. L. A capacity allocation model for reaction time. J. exp. Psycho/., Hum. Percept. Perform. 4,58&598, 1978.
18. SHAW, M. L. and SHAW, P. Optimal allocation ofcognitive resources to spatial locations. J. exp. Psycho/., Hum. Percept. Perform. 3,201- 21 I, 1977. 19. RIZZOLATTI, G. Mechanisms ofselective attention in mammals. In Advances in Vertebrate Neuroethology, J. P. EWERT, R. R. CAPRANICA and D. J. INC;LE(Editors), pp. 261-297. Plenum Publishing Corporation, New York 1983. attention and unilateral neglect. In 20. RIZZOLATTI, G. and CAMARDA, R. Neural circuits for spatial Neurophysiological and Neuropsychological Aspecfs ofSpatial Neglect, M. JEANNEROD(Editor). North Holland, Amsterdam (in press). and 21. RIZZOLATTI, G., RIGGIO, L., DASCOLA, 1. and UMILTA, C. Reorienting attention across the horizontal vertical meridians: evidence in favor of a premotor theory of attention. Neuropsychologia 25, 3140, 1987. 22. UMILTA, C. and NICOLETTI, R. Attention and coding effects in S-R compatibility due to irrelevant spatial cues. In Attention and Performance XI, M. I. POSNER and 0. S. M. MARIN (Editors), pp. 457471. Lawrence Erlbaum Associates, Hillsdale, N. J., 1985.