VCRPostcr

Door

Figure 4-1. An overhead view of the rectangular testing room.

The overhead image of the room and the participants' responses were recorded with a VCR, which took input from a small video camera mounted on the ceiling just above the swivel chair. A Gateway PC4200 desktop computer randomized the order of the targets for each participant.

Participants were randomly assigned to two groups. One group studied the object array while facing the Computer (studied view), and was tested while physically facing the Closet (updated view). The other group studied the objects while facing the Closet (studied view), and was tested while facing the Computer (updated view). Thus, the studied view and updated view were counterbalanced across the groups.

Participants were blindfolded and sound-masked before they were led into the testing room and sat in the swivel chair. Then they were turned to face the pre-determined orientation according to the group they were assigned to. Then they were instructed to remain stationary, and the blindfold was removed. The experimenter then named and pointed to the five targets one by one, asking the participants to remember them. The participants were allowed to turn their head (and head only) to observe the targets, but not to move their body or feet. The participants were allowed to study the objects for as long as they desired. Then they were blindfolded and put on an earphone with constant white noise as a sound mask, and were asked to point to the targets in a random sequence. If they made a mistake (> 20° from the correct direction), the blindfold was removed and they were asked to study again. The sequence continued until they could point to all targets correctly. The learning period typically took less than 5 minutes without the need to re-learn the targets.

The directions of the targets and the pointing responses were measured from the TV monitor after the testing was completed, by superimposing a transparent radial grid on the monitor. The reaction time was measured from the ending of the target name to the completion of the pointing response, which was indicated by the stabilization of the hand. The angular error for each response was calculated as the small angle (i.e., < 180°, unsigned) between the correct direction and the pointing direction. Since the objects were 60 ± 4° away from their neighbors (123° between Computer & Closet), the AD (angular disparity) between targets were considered multiples of 60° (0°, 60°, 120°, and 180°) and combined in the analysis.

The question of interest is which view(s) do people show advantage at after they study a scene from one perspective and then physically turn to take a different perspective during testing. Performance (RT and angular error) was analyzed as a function of the angular disparity between the imagined heading and the study heading (AD_Study), and the angular disparity between the imagined heading and the current heading (AD_update), using linear regression. Both RT and angular errors showed the angular disparity effect from the updated view and increased as the imagined heading deviated from the current heading, ts(38) > 2.1, ps < .04, suggesting that participants had a representation of the updated view (see Figure 4-2, top panels). In contrast, there was no significant angular disparity effect for the studied view, ts(38) < 1.6, ps > .11. Thus, people showed little evidence that they remembered the scene from the heading they studied it after they made a small turn to face a different orientation, but at the same time showed significant advantage at making judgments from the novel, updated view. This study provides initial evidence that the studied view and the updated view representations may be independent of each other.

One concern with this design is that the AD_study and AD_update were not completely independent of each other; in fact, there is a moderate negative correlation between these two variables. Thus, it is possible that the benefit of the studied view was obscured by the strong effect of the updated view and thus difficult to detect (Mou et al, 2004). To address this issue, we selected two test perspectives that had the same AD_update (120°) but different AD_study (0° & 120°) to de-couple the two factors. Even when AD_update was held constant, there was still no significant effect of AD_study (ts(14) < 1.2, ps > .26), suggesting there is little evidence of the studied view representation after spatial updating occurred.

To further examine the studied view representation during the updating process, we conducted a second study that varied the AD_study and AD_update independently (c.f. Mou et al., 2004). This study was identical to the first one except for the following. Fourteen participants were randomly assigned to two groups. One group studied the targets while facing the Computer, and the other group studied the targets while facing the Closet. Then all participants turned to face the Poster while being tested. In both cases, participants made a 120-degree turn while blindfolded. In this design, the angular distance between the test perspective and the studied view (AD_study) and between the test perspective and people's actual perspective (AD_update) were varied independently. That is, AD_study was 60° and 180° when AD_update was 60°, and AD_study was 0° and 120° when AD_update was 120°. If participants had any memory of the studied view available after they turned, they should show an angular disparity effect of the studied view while AD_updated was held constant.

As shown in Figure 4-2, bottom panels, performance deteriorated as the test perspective deviated further from their current heading (non-significant for RT, t(67) = 1.7, p = .09; significant for error, t (67) = 5.0, p < .001). There was still little evidence for memory of the studied view, however. Neither RT nor angular error increased significantly as a function of angular disparity from the studied view, ts(67) < 1.7, ps > .09. Thus, there was again little evidence of the effect of the studied view after people made a small turn, even when AD_update was held constant.

Figure 4-2. Object localization performance when people made a turn after studying targets from one perspective. The top panels showed RT (left panel) and angular errors (right panel), plotted as a function of AD_Update, The bottom panels showed performance as a function of angular disparity relative to the studied-view, grouped by AD_Update (i.e., A-I=0°, 60°, 120°).

Figure 4-2. Object localization performance when people made a turn after studying targets from one perspective. The top panels showed RT (left panel) and angular errors (right panel), plotted as a function of AD_Update, The bottom panels showed performance as a function of angular disparity relative to the studied-view, grouped by AD_Update (i.e., A-I=0°, 60°, 120°).

2.2.2 View-dependent representations after disorientation

There is another possibility on the failure to show evidence of the studied view. It is possible that the studied view was never really represented under these experimental conditions. For example, because the study orientations were mis-aligned with the walls of the room, they may be very difficult to learn. Shelton and McNamara (2001) showed that when people study an array of objects from two perspectives, first from an angle mis-aligned with the room's axes and followed by one aligned with the room's axes, people ended up not remembering the scene from the initial studied perspective. Instead, they remembered the object layout from the second, canonical view. Furthermore, Mou & McNamara (2002) showed that when specifically instructed to memorize the scene from a canonical perspective while physically studying it from a non-canonical perspective, people are capable of ignoring their true study perspective and show advantage of the canonical perspective rather than the actual studied perspective. Although in the current study people never "studied" the objects from a second perspective, nor were they instructed to pay attention to the canonical view, it is still possible that people somehow never represented the initial studied view and chose to represent the canonical view instead, leading to the benefit in the updated view, which happened to coincide with the room's axis.

To further examine whether the advantage was truly a result of the representation of the updated view, or a result of the canonical view representation, we asked people to study the targets exactly the same way as before, and then disoriented them right after the studying period. Disorientation destroys the updating process by definition, because updating requires one's perception of self-movement, and cannot occur when one loses track of one's heading. Thus, the disorientation procedure should minimize the effect of updating. If people tend to represent the canonical view instead of the studied view in our environment, then they should continue to show an advantage of the canonical view but no advantage of the studied view even after disorientation. On the other hand, if the lack of studied-view effect was due to the subsequent updating process, then there should be a stronger effect of the studied view when the influence of the updating process was eliminated.

We tested ten participants who were randomly assigned to five groups, two in each group. Each group studied the objects while facing one of the target objects, i.e., each group had a different studied-view. Then immediately after the studying period, they were blindfolded and sound-masked, and were asked to turn themselves in the swivel chair for one minute, changing directions from time to time to induce disorientation. At the end of the disorientation period, they were asked to stop, facing whatever direction they happened to be facing, and sat there for 30 seconds to recover from any physical disturbance. Then they were tested in 20 trials as before, imagining facing each of the objects and pointing to each of the remaining ones. Again, this design allowed the two independent variables, AD_study and AD_canonical, be manipulated independently across subjects.

Performance (RT and error) was analyzed as a function of the angular disparity from the studied view and from the canonical view corresponding to the Poster. As shown in Figure 4-3, top panels, performance decreased significantly as the test perspective deviates further from the studied view, for both RT and angular errors, ts(43) > 2.7, ps <.01. In contrast, performance was not affected by the angular disparity between the test perspective and the canonical direction, ts(43) < 1.2, ps > .25.

These data suggest that people did remember the objects from the perspective they studied them, even though the room is relatively small and these studied perspectives were mis-aligned with the room's axes. These data further suggested that the lack of an effect of the studied view when people took a small turn but not disoriented was not due to their failure to represent the studied view to begin with, because the learning stage was the same in both studies, and people did show evidence of the studied-view representation if they were disoriented. Thus, it was clear that the studied view was what people initially memorized, but the subsequent updating process eliminated its effect.

There are two possible reasons that the effect of the studied view may be eliminated after updating occurred. First, the updated-view representation may replace, or wipe out the studied view representation itself, and therefore the representation is gone forever. Shelton and McNamara (2001) suggested that canonical views may replace, or override an initial "studied view" representation. Thus, it is also possible that a similar "replacement" process might occur between updated view and the studied view.

Second, the updated view may dominate, or mask the studied view and thus the studied-view representation is temporarily un-accessible when updating is in operation. To examine these possibilities, we asked people to study targets in a room from a given perspective, then make a turn while blindfolded (i.e., update), and then get disoriented for testing. If updating makes the studied-view representation temporarily un-accessible, then disorientation, which disables the updating process, should unmask the effect of the studied view. In contrast, if updating abolishes the studied-view representation once it occurs, then there shouldn't be any recovery of the studied view even after disorientation.

We tested twelve participants who were randomly assigned to two groups, one facing the Computer during studying and one facing the Closet. They were then blindfolded and turned to face the Closet and Computer, respectively, and were asked to sit there for one minute, and think about where the objects were from their current heading to ensure that updating was completed. Then they were disoriented and tested on their spatial

Figure 4-3. Object localization performance when people were disoriented. The top panels showed RT (left panel) and angular errors (right panel) as a function of AD_Study, grouped by angular disparity relative to the canonical direction (C-I=0°, 60°, 120°, 180°). The bottom panels showed performance as a function of AD_Study.

Figure 4-3. Object localization performance when people were disoriented. The top panels showed RT (left panel) and angular errors (right panel) as a function of AD_Study, grouped by angular disparity relative to the canonical direction (C-I=0°, 60°, 120°, 180°). The bottom panels showed performance as a function of AD_Study.

judgments as before. In this procedure, people should encode the studied view first, then generate an updated-view representation after they made a small turn while maintaining their sense of direction. After that, the updating process was disrupted before testing. If the updating process temporarily obscures the studied-view representation, then people should show advantage of the studied view. If the studied-view representation was replaced by the updated-view representation, then there should be no recovery of the studied view even when participants were disoriented during testing.

The results were shown in Figure 4-3, bottom panels. Performance was analyzed as a function of the angular disparity between the imagined heading and the studied view, and between the imagined heading and the updated view (i.e., the orientation they were facing right before the disorientation procedure, not their actual heading after disorientation). People's judgments of the target direction were significantly impaired when the test perspective deviated further away from the studied view (significant for RT, ¿(53) = 2.1, p = .05, non-significant for errors, t(53) = 1.7, p = .10)2. In contrast, there was no angular disparity effect relative to the updated view after people were disoriented (ts(53) < 1). A further analysis selecting test perspectives that had AD_study constant (120°) while AD_update varied (0° and 120°) again showed little evidence of the updated-view representation (ts(22) < 1.5, ps > .16).

Several conclusions can be drawn based on these findings. First, studied-view representations are preserved when people change perspectives physically. Second, the spatial updating process limits one's access to the studied view representation while it is in operation, but does not abolish the representation itself, so that the studied view can be retrieved again if people are disoriented after updating occurs. Third, the studied-view and the updated-view representations are independent of each other, updating can occur without the studied view representation. Fourth, the lack of AD effect for the updated view after people are disoriented suggested that the updated-view representation is relatively transient and is lost once the updating process is disrupted.

0 0

Post a comment