Distance Estimation Experiment 2
Distance Estimation Experiment #2
Method
Participants. 24 undergraduate students (13 women) with a
mean
age of 21.74 (SD = 4.31) were recruited from the University of Washington
department of Psychology human subject pool. All participants were given
extra credit toward their Psychology class for their participation except
for two who were paid $10 per hour.
Stimuli and apparatus. With few exceptions, the stimuli
and apparatus used were exactly the same as reported in
experiment one. On
each trial, the computer randomly selected one of three geometric fields
of view (50, 80, or 100). At participants’ viewing distance of 38 cm, the
50 degree GFOV presented identical physical and geometric fields of view.
The 80 degree GFOV was the same as that used in experiment one. Figure
DE2-1 illustrates the appearance of the three GFOV conditions. Unlike
experiment one, the grid was present on every trial, and the environment
provided feedback only to people who were assigned to the feedback
condition. The VE also recorded the amount of time taken to complete each
trial.
Figure DE2-1
All participants used the desktop system described in experiment one,
although the resolution was changed to 640 x 480 (75 Hz. refresh) and a
different graphics accelerator was used (Diamond Fire GL 3000). The frame
rate for this system was approximately 24.18 frames per second.
Procedure. Participants were trained on the rudiments of VE
navigation with a mouse. They were then required to complete a ‘virtual
obstacle course’ in less than four minutes before continuing
participation. Participants were then randomly assigned to either receive
feedback or not (subject to the constraint that equal numbers of
participants appear in each feedback condition). As in experiment one,
each trial began with a red and a green block being randomly placed inside
the environment. Participants were instructed not to spend too much time
on any one trial, but were also told that they should feel free to view
the blocks from whatever perspectives they felt necessary to arrive at an
accurate distance estimation. Participants made as many distance
estimations as they could in the remaining time, taking breaks whenever
they wished.
Results
Participants completed an average of 87.46 trials (SD = 23.19), with
roughly one-third at each FOV setting. Accuracy (percent distance error),
estimated parameters, and model fits were calculated separately for each
participant and were subsequently used as dependent variables in separate
3 (FOV) x 2 (feedback) x 2 (gender) repeated measures ANOVA’s.
Percent overestimation Accuracy was clearly affected by both
feedback and GFOV, but not by gender. In general, participants were more
accurate when given feedback and when viewing the VE with the moderate
GFOV’s as illustrated in figure DE2-2. An ANOVA confirms that main effects
of
GFOV (F(2, 40) = 35.61, p < .001) and feedback (F(1, 20) = 9.50, p = .006)
are highly significant. No other main effect or interaction approached
significance. Effect sizes (h2) for FOV and feedback were estimated at
.64 and .32 respectively.
Participants tended to overestimate distances much more and to be more
variable in their responses without feedback (M = 24.49, SD = 26.55) than
with (M = 0.80, SD = 6.99). Figure DE2-3 illustrates the percent of
overestimation for participants in both feedback conditions across the
first ten trials. By the seventh trial, the difference between the
percent of overestimation for participants who received feedback (M =
6.38, SD = 38.42) was significantly lower than for those who did not
receive feedback (M = 41.06, SD = 35.61) (t(22) = 2.29, p = .023).
Figure DE2-2
Specific effects of GFOV were examined with pairwise contrasts. These
revealed that accuracy was significantly different between the 50 degree
FOV condition (Mfeedback = -11.19; SD = 9.73; Mnone = 14.09, SD = 26.13)
and the 80 degree FOV condition (Mfeedback = 0.82; SD = 7.44; Mnone =
23.45, SD = 26.42) (F(1,20) = 38.115, p < .001). In fact, for the
participants who received feedback, the underestimation of 11.19% in
trials with a 50 degree FOV was significantly less than zero (t(11) =
3.98, p = .002). Overestimation was also significantly reduced between
the 80 and the 100 degree GFOV trials (Mfeedback = 6.47; SD = 10.62; Mnone
= 34.97, SD = 29.58) (F(1,20) = 11.53, p = .003).
Figure DE2-3
Steven’s exponents and power model fit The simple power model
was fit separately to each participant’s data for each GFOV condition.
Across the 24 participants, exponent estimates from these model fits
ranged from 0.54 to 1.03 (M = 0.80, SD = 0.12) and were significantly less
than one (t(23) = 8.30, p < .001). For 11 (46%) of the participants, the
95% confidence interval of the exponent estimate did not contain one. R^2
values from these models ranged between .32 and .72 (M = .56, SD = 0.11).
Neither exponent estimates nor R^2 values were significantly affected by
GFOV, feedback, gender, or any of their interactions.
[ return to top ]