2. Experiment 1: Can Spatial Attention Bias Multistable Motion Perception?
2.2 Experiment 1b: Adding a Manipulation-Check Task to Spatial Attention….49
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2 Experiment 1b: Adding a Manipulation-Check Task to Spatial Attention
In Experiment 1a, participants’ attention was manipulated by instructing them to attend to the four occluders or attend to the four moving lines. However, it is still not clear whether participants obeyed these instructions. Hence, in Experiment 1b, a manipulation-check task of spatial attention was added. Participants had to respond to a probe (the lightening of an occluder or a moving line) presented at the end of each trial as fast as possible. The reaction time to the probe was analyzed to verify whether the manipulation of spatial attention was valid. The prediction is that if participants allocated their attention according to the instructions, RT to the lightening of the occluder and line should be different. The RT should be shorter to for the lightening of attended areas than for the unattended areas.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.2.1 Methods
Participants
Another nine participants were recruited with the same standards as described for Experiment 1a.
Design
This experiment is a one factor (attention) within-participant design.
Participants’ attention was manipulated by instructing them to attend to the four occluders and attend to the four moving lines in different blocks. Simultaneously, all participants were instructed to hold coherent motion perception through the experiment. In addition, at the end of each trial, participants had to respond to a probe—the lightening of an occluder or a moving line—by pressing a key as fast as possible. The response time (RT) to the probe was measured as a dependent variable.
In addition, participants had to report their motion perception (coherent or separate) by key-pressing until a probe presented. The average percentage of time in perceiving coherent motion was measured as another dependent variable.
‧
The stimuli used in this experiment were similar to that in Experiment 1a, except that the four gray occluders were changed into hollow occluders to make the two probe conditions similar (that is, the light of an occluder outline or the light of a line).
Apparatus
The apparatus used here was identical to that in Experiment 1a.
Procedures
The procedure was similar to Experiment 1a, except that participants had to additionally respond to a probe that was shown 3, 4, 5, 6, 7, or 8 seconds after the trial began by pressing the numeral “4” key with the left hand. The six probe-showing times were the same in all blocks, but the sequence was different (by randomizing) so participants could not predict when the probe would present itself. After responding to the probe, the trial ended and a new trial began. Since the recording of motion perception was terminated during probe presenting, the lightening of the occluder or line would not influence the multistable motion perception process.
This experiment had four blocks—two attending-to-occluders blocks and two
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
attending-to-moving lines blocks. The order of the four blocks was counterbalanced within participants. Each block had six trials corresponding to the six probe-showing times. In total, each participant has to do 4 × 6 = 24 experimental trials.
‧
The trials in which the RT to the probe was greater than 1500 ms were deleted and not included in further analysis. One trial was deleted for two participants, and another participant’s entire data were deleted because the deleted trials were too much (six trials of total 24 trials). Thus, the data for further analysis contained only eight participants’ non-deleted data.
The mean percentage of time perceiving coherent motion and RT to the lightening of the occluder or line under the two attention conditions are plotted in Figure 10. One-way ANOVA shows that the effect of attention is significant. The percentage of time perceiving coherent motion is higher in the attending-to-occluders condition (67.28%) than in the attending-to-moving lines (49.17%) (F(1,7) = 24.87 , p
< 0.01, partial η2 = 0.780), which is consistent with the hypothesis that attention can bias multistable motion perception. The type of bias was also consistent with the prediction that the attended areas look closer. On the other hand, a two-way ANOVA of RT shows that the interaction of the RT to the probe under the two attention conditions and the two probe conditions is significant (F(1,7) = 6.35 , p < 0.05, partial η2 = 0.476), and the pattern of interaction in Figure 10 implies that the participants
allocated their attention on demanded areas. That is, the value of subtracting the RT of
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
lightening line from the RT of occluder lightening under the attending-to-lines condition is larger than that under the attending-to-occluders condition. The main effect of the two attention conditions (F(1,7) = 0.91 , p = 0.371) and the two probe conditions (F(1,7) = 1.38 , p = 0.278) are not significant.
Figure 10. Results of Experiment 1b. Mean percentage of time perceiving coherent
motion and RT to the lightening of an occluder or a line under the two attention conditions.
In a further analysis of the data of RT to the probe, the eight participants can be divided into two groups depending on whether their individual patterns of interaction imply that the participants allocated their attention on demanded areas, which is
‧
defined by “the index of attention.” That is, subtract the RT of line lighting from the RT of occluder lighting under the attending-to-lines condition and under the attending-to-occluders condition, then subtract the later from the former; this value is the index of attention. If the index of attention is positive, the participant is assigned to the "positive interaction" group, implying he/she does allocate his/her attention on demanded areas. Conversely, participants are assigned to the "negative interaction"
group if the index of attention is negative.
The mean percentage of time perceiving coherent motion under the 2 (group) × 2 (attention) conditions are plotted in Figure 11. There are seven participants in the
“positive interaction” group and only one participant in the “negative interaction”
group. Two-way ANOVA shows that the effect of attention (F(1,6) = 6.26, p < 0.05, partial η2 = 0.510) and interaction is significant (F(1,6) = 6.67, p < 0.05, partial η2 = 0.527). Simple main effect shows that the effect of attention is significant only in the
“positive interaction” group (F(1,6) = 51.72, p < 0.001) but not significant at the
“negative interaction” group (F(1,6) = 0, p = 0.965). This evidence implies that the effect of attention will show, as predicted, only when participants allocate their attention on demanded areas (in the “positive interaction” group). Hence, the index of attention can be used in later experiments to filter out the data that does not meet the criterion of allocating attention on demanded areas.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 11. The mean percentage of time perceiving coherent motion under the 2
(group) × 2 (attention) conditions.
‧
This experiment repetitively verified the main result of Experiment 1a that the percentage of time for perceiving coherent motion was higher in the attending-to-occluders than in the attending-to-moving-lines condition. This result is consistent with the prediction that spatial attention can bias multistable motion perception by making attended areas look closer. Furthermore, after reducing Experiment 1a’s trial duration from 1 minute to 3 to 8 seconds, the effect of attention increased dramatically. The difference of the percentage of time perceiving coherent motion under the two attention conditions increased from 5.89% to 18.11%. This implies that participants can keep their attention on demanded areas better during a shorter duration trial than in a longer one. The manipulation check task of attention also showed that most participants allocated their attention according to the instructions. On the other hand, the increasing effect of attention brings another interesting issue worth investigating in the future: is the effect of intention still more powerful than the effect of attention after reducing the trial duration? Because intention was not manipulated in this experiment, this question cannot be answered here.
However, even if the results of Experiments 1a and 1b were consistent with the
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
hypothesis that attention can bias multistable motion perception, it is still not certain whether the effect of attention is actually affected by the depth perception mechanism, making attended areas look closer. Accordingly, in Experiment 2, this question will be verified directly.
‧
3. Experiment 2: The Effect of Attention: Can Spatial Attention Alter Perceived Depth?
Experiment 2 is a depth-judgment task designed to investigate whether spatial attention can affect depth perception, making attended areas look closer. This experiment has nothing to do with multistable motion perception, but the diamond stimulus is still used here in order to determine whether spatial attention affects the stimulus used in Experiment 1b. In other words, the test will indicate whether attending to the four occluders can make the occluders look closer and whether attending to the four lines can make the lines look closer.
In the depth-judgment task, nine different levels of binocular disparity of the occluders and the lines are manipulated. Participants have to judge which one is farther in depth, the four occluders or the four lines, after the diamond stimulus is briefly shown. Participants’ attention is also manipulated in different blocks, as in Experiment 1b. The point of subjective equality (PSE, on which the participants have a 50% probability of reporting that the lines are more behind the occluders) will be measured as the dependent variable, and it is predicted that the PSE under the attending-to-occluders condition will be different from that under the attending-to-lines condition. If attention can make the attended areas look closer, then
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
when participants attend to the four occluders, the occluders will look closer. Thus, participants will perceive occluders and lines in the same plane only when the lines are defined by binocular disparity as in front of occluders, so the PSE should be smaller than zero. In the same way, the four lines will look closer when attending to the four lines. Thus, participants will perceive occluders and lines in the same plane only when the occluders are defined by binocular disparity as in front of lines, so the PSE should be larger than zero. Therefore, the PSE under the condition of attending to the four lines will be larger than under the condition of attending to the four occluders.
A manipulation-check task of spatial attention (as in Experiment 1b) will also be executed to verify whether the manipulation of attention is valid.
‧
Another 14 participants were recruited with similar standards as described for Experiment 1a. After finishing this experiment, they received either one bonus point in the “Methods of Psychological Experiment” course or 100 NT dollars.
Design
The experiment is a 2 (attention) × 9 (binocular disparity) completely within- participant design. Participants’ attention was manipulated by instructing them to attend to the four occluders or to the four lines in different blocks, as in Experiment 1b. Binocular disparity was manipulated using a stereoscope at nine different levels (-4, -3, -2, -1, 0, 1, 2, 3, and 4 pixels). A greater positive disparity means that the lines were more behind the occluders; the greater negative disparity means that the lines were more in front of the occluders. Zero disparity means that the occluders and the lines were at the same depth. Participants had to judge which were behind (rather than in front), the occluders or the lines. Thus, the possibility that participants’ judgments were biased toward the instructions of attention can be ruled out. The point of subjective equality (PSE, on which the participants had a 50% probability of reporting
‧
that the lines were behind the occluders) was calculated as a dependent variable.
The experiment contained eight blocks split over two days; each day contained four blocks: two attending-to-occluders blocks and two attending-to-lines blocks. The order of the two attention conditions was counterbalanced within participants. Each block contained 22 trials in which the nine levels of binocular disparity were repeated two times, respectively, in a random sequence. The main task was to judge the depth relationship between occluders and lines. The remaining four trials were probe-detecting trials, including two occluder-lightening trials and two line-lightening trials (all were in zero disparity). In these trials, participants had to respond to the probe as fast as possible and did not need to do depth judgment. In total, each participant had to complete 2 × 4 × 22 = 176 experimental trials in two days. The PSE was calculated as a dependent variable.
Materials
The diamond stimulus used here was identical to that in Experiment 1b, except that it was static (the four lines were not moving).
Apparatus
The apparatus used here was similar to that described for Experiment 1b, except
‧
that the View Sonic CRT computer monitor was changed from a 17-inch to a 19-inch and a stereoscope was added. With the help of the stereoscope, participants can easily fuse a stereogram with binocular disparity to form the perception of the three-dimensional diamond stimulus.
Procedures
The procedures were similar to that described for Experiment 1b, except that participants did not have to respond to their motion perception of the diamond stimulus because it was static. Instead, they had to judge the depth relationship between occluders and lines.
At the beginning of each block, participants were instructed to allocate their attention on the four occluders or the four lines throughout the whole block.
Participants pressed the numeral key “4” with their left hand to start each trial by themselves when they were well prepared (fixated on the central cross and concentrating). After 150 ms, 300 ms, or 450 ms latency—applied at random to eliminate the expectancy effect, increasing the level of attention—the static diamond stimulus appeared for 600 ms and then was covered by a random-dot mask to prevent the afterimage from interfering. Participants had to report their depth judgment—occluders or lines were behind—by key-pressing after the diamond
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
stimulus was shown. Pressing the numeral key “1” with the right hand indicated that they perceived the occluders behind the lines, and pressing the right-hand numeral key “3” indicated that the lines were seen as behind the occluders. In probe-detecting trials, the lightening of an occluder or a line was presented using 150 ms, 200 ms, or 250 ms latency after the diamond stimulus was shown. Participants had to press the left-hand numeral key “6” as fast as possible when seeing the probe. The next trial would not begin until participants pressed the key to report their depth judgment or probe detection. The flow chart of the procedure in Experiment 2 is shown in Figure 12.
Diamond stimulus show 600 ms 150/300/450 ms delay
Show instruction Press “0” to start
Mask until response or
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 12. The flow chart of the depth-judgment trials in Experiment 2. By using a
stereoscope, participants’ left eye and right eye see a left picture and right picture, respectively, which is then fused into a three-dimensional diamond stimulus.
‧
The mean probability of reporting that the lines were behind the occluders under 2 (attention) × 9 (binocular disparity) conditions in eight blocks were counted and charted in Figure 13a. The t-test shows that the PSE under the attending-to-occluders condition (PSE = 0.087) and attending-to-lines condition (PSE
= -0.019) is not different (t(13) = 0.413, p = 0.686), which is not consistent with the prediction. In addition, the two-way ANOVA of RT to the probe in eight blocks also shows that the interaction is not significant as predicted (F(1,13) = 2.53, p = 0.136), implicating that participants did not totally follow the instructions to allocate their attention. This might be why the results of the PSE were not consistent with the prediction. Next, only the data in which participants followed instructions to allocate their attention were selected for further analysis. The method of selection is described below.
The eight blocks were divided into four sections (Section 1: Blocks 1 and 2;
Section 2: Blocks 3 and 4, and so on). Each section contained an attending-to-occluders block and an attending-to-lines block to check whether the index of attention (as described in Experiment 1b) in each section is positive. Only the sections in which the index of attention is positive—implying that participants
‧
allocated their attention according to the instructions—are selected for further analysis.
There were 26 sections selected (total: 4 sections × 14 participants = 56 sections;
three participants’ data were not selected), which contained 12 attending-to-occluders-first and attending-to-lines-next sections and 14 attending-to-lines-first and attending-to-occluders-next sections. Thus, the confounding of order effect would be very small. Two-way ANOVA of RT to the probe shows that the interaction is significant (F(1,9) = 9.76, p < 0.05, partial η2 = 0.520), implicating that participants allocated their attention according to the instructions in the selected sections. Scheffe’s post hoc test shows that participants’
response to the lightening of lines was shorter than to the lightening of occluders at attending-to-lines condition (F = 33.92 > F’= 5.12) but not at the attending-to-occluders condition (F = 1.97).
Mean probability of reporting that the lines were behind the occluders under 2 (attention) × 9 (binocular disparity) conditions in the selected sections were counted and are charted in Figure 13b. Two-way ANOVA shows that the main effect of binocular disparity is significant (F(8,72) = 47.28, p < 0.001, partial η2 = 0.840), implicating that the manipulation of binocular disparity is valid. However, neither the main effect of attention nor interaction are significant (F(1,9) = 0.44, p = 0.523;
F(8,72) = 1.61, p = 0.138).
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The PSE under the attending-to-lines condition (PSE = 0.179) is larger than under the attending-to-occluders condition (PSE = -0.045) and consistent with the prediction, but the t-test of PSE is not significant (t(9) = 1.272, p = 0.235). However, the mean probability of reporting lines behind occluders in small binocular disparities (-1, 0, and 1) is significantly higher under the attending-to-occluders condition (49.58%) than that under the attending-to-lines condition (38.33%) (t(9) = 4.045, p <
0.01) (see Figure 13c). This evidence supports the hypothesis that attention can affect depth perception, making attended areas look closer.
(a)
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(b)
(c)
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 13. Results of Experiment 2. (a) Mean probability of reporting that the lines
behind the occluders were plotted under 2 (attention) × 9 (binocular disparity) conditions in eight blocks. (b) Mean probability of reporting that the lines were behind the occluders in the selected sections were plotted under 2 (attention) × 9 (binocular disparity) conditions. (c) The mean probability of reporting lines behind occluders in small binocular disparities (-1, 0, and 1) under the two attention conditions in the selected sections.
‧
The mean probability of reporting lines behind occluders in small binocular disparities (-1, 0, and 1) is significantly higher under the attending-to-occluders condition than that under the attending-to-lines condition, consistent with the hypothesis that attention can affect depth perception, making attended areas look closer. The resulting pattern also shows that PSE under the attending-to-lines condition is larger than under the four occluders condition (although not significant)
and is consistent with the prediction.
There were some problems that may influence the results. One problem is that the probability of reporting lines behind occluders did not decrease with greater negative disparity. The reason might be the conflict between binocular and monocular depth cues at negative disparity. That is, due to the monocular depth cues, such as T-junctions, the lines are easily perceived as completed behind the occluders.
However, the lines should be perceived in front of the occluders under negative disparity conditions. In fact, most participants reported that it was hard to perceive lines in front of occluders (in negative disparity conditions) while it was easy to perceive lines behind occluders (in positive disparity conditions). Notice that this problem may also affect the PSEs measured here because the probability of reporting
‧
國立 政 治 大
立 政 治 大