2.3.1 Introduction
The same experimental procedure used in Experiment 1 was repeated in
Experiment 2, however an alternative method for manipulated visual perspectives was
applied. Instead of manipulating the gaze of the agent, during so-called Invalid trials,
the visibility method was used (i.e., Cole et al. 2016). Specifically, the agent held a
newspaper, resting on the table, and looked down towards it. This therefore controlled
for where the agent was attending. As many authors have pointed out (e.g., Cole et al, 2016), the dot perspective task and gaze cueing paradigm confound the agent’s view
with the location of visual attention. Consequently, in the present experiment, the
agent always looks to the same position, i.e., down towards the table/newspaper. To
reiterate, Experiment 2 specifically assessed the possible confounds of diverting the agent’s attention to see whether this manipulation is driving the discrepancies in
spontaneous visual perspective taking literature.
All other aspects of the procedure, in terms of blocking of trials and
ambiguous numbers, were kept as a direct replication of Experiment 1. It is predicted
that if the suggested spontaneous perspective taking theory is indeed due to the
representation of another’s visual perspective, obscuring the view of the ambiguous
number for the agent will eradicate any spontaneous perspective taking-like effects.
However, if similar patterns in the data are found, irrespective of whether the agent
can see the number or not, the results would suggest that an alternative cognitive
process could be driving spontaneous perspective taking effects in this experiment
2.3.2 Method
All aspects of the Method were as described for Experiment 1, with the
following three exceptions. One, a new sample of participants was gathered. Two,
visibility was manipulated instead of validity using an occluding barrier rather than
the gaze direction of the agent and three, participants were informed that the agent
could not see the target number for half of the trials (they were not told how this was
achieved).
In terms of the sample, 38 participants were recruited with a mean age of
20.00 (SD = 1.86, range = 18-27), with 27 of the sample being female (11 male). Six
participants identified themselves as left-handed whilst the remaining 32 identified
themselves as right-handed. All 38 participants reported normal, or corrected to
normal vision, and were recruited through the University of Essex online volunteer portal known as ‘SONA’, with participants being reimbursed for their time.
Additionally, instead of manipulating validity Experiment 2 manipulated
visibility, in which the agent was depicted as holding and gazing towards a newspaper
resting on the table directly in front of the ambiguous number. Consequently, the agent’s ability to perceive the number was directly obstructed, hereafter known as
Non-visible. See Figure 2.6 for an example of the different stimuli that was used for
Figure 2.6: The two top images represent the two critical conditions in the basic ambiguous number paradigm. If the ambiguous number effect is due to a representation of the agent’s visual perspective, then no such effect should occur when the agents view is obstructed by the
occluding barrier in the form of a newspaper in the two bottom images.
2.3.3 Results and Discussion
Data Preparation. The data were collated and reorganised using Microsoft
Excel and collapsed into a single data file. The data were then transformed to produce
mean RTs, excluding any outliners that were 2 standard deviations above or below the
mean, for the two levels of each manipulated factor of visibility and ambiguity.
Additionally, a percentage for correct self-perspective responses for each condition
Reaction Time. Figure 2.7 depicts the mean RT for the four conditions. A
repeated measures ANOVA with ambiguity and visibility as within-participant factors
found a significant main effect of ambiguity, F(1, 37) = 25.73, p < .0001, np2 = .41,
but no significant main effect of visibility was found, F(1, 37) = .8, p = .38, np2 = .02.
A significant interaction was also found, F(1, 37) = 4.13, p < .05, np2 = .1.
Figure 2.7: Mean RTs for each the four conditions, with standard error bars included
To examine the significant interaction three planned comparison t-tests were
conducted. Firstly, there was a significant difference in RT for Visible, Unambiguous
(M=459.93, SD=55.57) and Visible, Ambiguous (M=483.43, SD=60.36) conditions;
t(37) = -3.89, p > .001, d = 0.63, BF10 = 68.31. Thus, showing that RT for Visible
Unambiguous conditions was significantly shorter than Visible Ambiguous
conditions. This is further supported by a Bayesian t-test which found the data to be
68 times more likely under the alternative hypothesis. Additionally, there was a
significant difference between Non-Visible, Unambiguous (M=462.16, SD=65.77)
and Non-Visible, Ambiguous (M=470.93, SD=64.49) conditions; t(37) = -2.79, p >
440 445 450 455 460 465 470 475 480 485 490 495 Visible Non-Visible R ea ct io n T im e (m s) Unambiguous Ambiguous
.008, d = 0.45, BF10 = 4.979. In other words, these results show that within the Non-
Visible condition, there was a significant difference between ambiguity conditions
with Unambiguous being significantly shorter than Ambiguous conditions. This is
further supported by a Bayesian t-test which found that the data to be 5 times more
likely under the alternative hypothesis. Alternatively, there was no significant
difference in RT for Visible, Ambiguous (M=483.43, SD=60.36) and Non-Visible,
Ambiguous (M=470.93, SD=64.49) conditions; t(37) = 1.85, p = .07, d = 0.3, BF10 =
0.81, and is further supported by a Bayesian t-test which found the data to be 1.2
times more likely under the null than the alternative hypothesis.
As can be seen participants were significantly faster to respond during
Unambiguous conditions, irrespective of the occluding barrier. Plus, the significant
interaction demonstrates that participants are significantly faster to respond during
Unambiguous trials in both Visible and Non-Visible conditions. However, most
importantly there was no significant difference in RT between Ambiguous Visible
and Ambiguous Non-Visible conditions. These results therefore dispute the
spontaneous visual perspective taking claim. The claim is clear, if this phenomenon is
in fact the spontaneous assumption of an alternative visual perspective, the effect
should not have been identifiable during Non-Visible conditions at all, where the
agent was unable to view the ambiguous number.
Error Rate. A repeated measures ANOVA with ambiguity and visibility as
within-participant factors did not find a significant main effect of ambiguity, F(1, 37)
= .24, p = .63, np2 = .006, or visibility, F(1, 37) = 1.83, p = .18, np2 = .05. There was
also no significant interaction found, F(1, 37) = 3.23, p = .08, np2 = .08. Consequently,
Overall, in isolation the significant ambiguity finding suggests that
participants were assuming the visual perspective of the agent as previous literature
has suggested. In other words, if we take the non-barrier conditions only, the data
suggests that participants were spontaneously assuming the agent’s visual perspective
during Ambiguous conditions, leading to the longer RTs in comparison to
Unambiguous conditions. However, importantly, this ambiguity effect was found
irrespective of whether the agent could see the ambiguous number of not. To reiterate
the rationale for the present experiment, if the assumption of the agent’s visual
perspective is indeed driving the results of this experiment, then no such effect should
have occurred when the alternative perspective cannot view the ambiguous number.
Therefore, Experiment 2 supports the findings of Experiment 1, in that alternative
cognitive processes must be driving the spontaneous visual perspective taking-like
effects observed. This again coincides with the findings of Cole, et al. (2015) whom
also found perspective taking-like effects during conditions in which the alternative
perspective was unable to view the target.
Further Analysis
As with Experiment 1, additional analyses were undertaken to ensure that the
results found overall were exhibited in both ambiguous number pairs. ‘68’-‘69’ Ambiguous Number Pair
Reaction Time. Figure 2.8 depicts the mean RT for the four conditions using
the ‘68’-‘69’ ambiguous number pair. A repeated measures ANOVA with ambiguity
and visibility as within-participant factors found a significant main effect of
ambiguity, F(1, 37) = 28.27, p < .0001, np2 = .43, but no significant main effect of
visibility, F(1, 37) = .22, p = .65, np2 = .006. There was also no significant interaction,
faster to respond during trials with the ‘69’ ambiguous number, in comparison with
‘68’. In addition, the presence of an occluding barrier did not affect this data trend.
Figure 2.8: Mean RTs for each of the four conditions for the ‘68’ and ‘69’ ambiguous number pair, with standard error bars included
Error Rate. A repeated measures ANOVA with ambiguity and visibility as
within-participant factors did not find a significant main effect of ambiguity, F(1, 37)
= .71, p = .4, np2 = .02, or of visibility, F(1, 37)= 2.82, p = .1, np2 = .07. There was
also no significant interaction, F(1, 37)= .14, p = .71, np2 = .004. As can be seen there
was no significant difference in error rate for trials using the ‘68’ and ‘69’ ambiguous
number pair.
‘88’-‘89’ Ambiguous Number Pair
Reaction Time. Figure 2.9 depicts the mean RT for the four conditions using
the ‘88’-‘89’ ambiguous number pair. A repeated measures ANOVA with ambiguity
and visibility as within-participant factors did not find a significant main effect of
ambiguity, F(1, 37) = .008, p = .93, np2 = .0001, or of visibility, F(1, 37) = 1.26, p =
.27, np2 = .03. There was also no significant interaction, F(1, 37) = 3.05, p = .09, np2 =
440 450 460 470 480 490 500 510 Visible Non-Visible R ea ct io n T im e (m s) Non-Ambiguous Ambiguous
.08. Thus, there was no significant finding in terms of RT when isolating the ‘88’ and ‘89’ ambiguous number pair.
Figure 2.9: Mean RTs for each of the four conditions for the ‘88’ and ‘89’ ambiguous number pair, with standard error bars included
Error Rate. A repeated measures ANOVA with ambiguity and visibility as
within-participant factors did not find a significant main effect of ambiguity, F(1, 37)
= 2.69, p = .11, np2 = .07, or of visibility, F(1, 37) = .88, p = .35, np2 = .02. There was
also no significant interaction, F(1, 37) = 5.37, p = .03, np2 = .13. Thus, there were no
significant finding reported in terms of error rate when isolating the ‘88’ and ‘89’
ambiguous number pair.
Overall, these data again reveal an ambiguous number effect; RTs were longer
when the participant and agent can interpret the number differently. This on its own is
consistent with past reports of visual perspective taking. Furthermore, the significant
interaction identified that this effect is reduced when the model cannot see the
number, which again is suggestive of an allocentric visual perspective computation.
However, the visual perspective taking notion is clear in what it predicts when a
420 430 440 450 460 470 480 490 Visible Non-Visible R eac ti o n Ti m e (m s) Unambiguous Ambiguous
alternative perspective cannot see the ambiguous number; there should be no effect at
all. Yet, in Experiment 2 a visual perspective taking-like effect was found when the
agent was unable to see the ambiguous number; i.e., when the newspaper obstructed
the view. Therefore, this effect cannot be concluded to be the rapid and spontaneous computation of the agent’s allocentric visual perspective. Other cognitive processes
must be driving the results of this experiment. However, contrary to Experiment 1, a
significant interaction between ambiguity and visibility was also found in Experiment
2. Yet, further analysis identified that the ambiguity effect between Unambiguous and
Ambiguous conditions was found in both of the visibility conditions, and no
significant difference between Ambiguous Visible and Ambiguous Non-Visible
conditions were found. Thus, although the effect was reduced in the Non-Visible
condition, again the visual perspective taking theory is clear that the effect should not
have been identifiable in this condition. Consequently, it can be concluded that
Experiment 2 again does not support the spontaneous visual perspective taking theory,
as the same effect was identified during condition in which the allocentric visual
perspective could not see the ambiguous number.
In Experiment 3, the importance of the occluding barrier will be revisited.
Specifically, the occluding barrier manipulation and variations to the avatar’s stance
will be applied within the dot perspective task. This will assess whether alternative
cognitive processes, such as gaze following, or mental rotation can be used to
2.4 Experiment 3 – Avatar Stance, Occluding Barriers, and the Dot Perspective