Assembly Training Experiment #1
Assembly Training Experiment #1
A summary of this experiment was published in the CHI ’98
Summary (pp. 339-340). It is available for downloading as a
postscript file here . Or view it as a .pdf
file here:
Method
Participants.
A total of 40 people ( 17 men) recruited from the University of Washington
Psychology department’s Human Subject Pool participated in the experiment
in return for extra credit in their introductory psychology course.
Materials and apparatus.
The real world tasks on which all participants were tested involved
correctly placing approximately ten blocks in a particular configuration
in one of three ‘circuit boards.’ The circuit boards were flat structures
(appx. 125 cm^2) with a pattern of holes cut into them. To solve each
puzzle correctly, a participant had to insert each block into a specific
location. Successful completion of this task required thinking about the
spatial properties of the puzzle pieces.
Three training environments were developed to teach people how to
perform this task: a paper tutorial, a video tutorial, and a VE trainer.
A great deal of effort was devoted to making the paper and video tutorials
effective and complete. These tutorials were constructed using modern
principles of tutorial design (Price & Korman, 1993).
The circuit boards and blocks were modeled in a VE using World Up by
SENSE8 Corporation on a Pentium Pro 200 with a Diamond Fire GL 3000
graphics accelerator card. Participants who trained in the VE sat
approximately 58 cm from a 40 cm x 30 cm (21 inch) flat-panel color
monitor. The virtual interface (see figure below) allowed the user to
select between three modes of interaction: 1) ‘Navigation’ mode allowed
users to change the position and orientation of their viewpoint, 2)
‘Pick’ mode enabled users to translate the virtual blocks in the plane of
the displayed scene, and 3) ‘Rotation’ mode enabled users to rotate the
blocks. User interaction with the VE was accomplished solely through the
mouse and keyboard. Mouse movements controlled navigation and translation
of the blocks, while block rotations were accomplished through key
strokes.
Procedure.
Participants were randomly assigned to one of the three training
conditions in which they learned how to solve three circuit board
problems. All groups were given practice drills and quizzes to verify
that they had learned the correct placement of the blocks. Participants
who trained from the paper tutorial and the VE were given as much time as
they needed to complete the training. Training time for the video group
was constant. Immediately after completing the training for each
circuit board, participants were timed on their ability to assemble it in
the real world. Participants returned a week later and attempted to solve
the same puzzles with no further training. They were also timed on this
task.
Results
Training time.
People took more time to train in the VE trainer (M = 18.18 min., SD =
6.64) than either the video group (6.72 min.) or the paper tutorial group
(M = 5.47 min., SD = 1.66). The difference between the VE training time
and that of each other group is significant by Tukey HSD tests.
Immediate transfer.
Mean solution times immediately after training were relatively fast for
each training condition (M = 0.95 min., SD = 0.83) indicating that each
method was clearly effective at training the task. Differences in the
effectiveness of training medium were tested using an analysis of
covariance (ANCOVA) in which three levels of training and two levels of
gender were entered as independent variables and the amount of time spent
training was entered as a covariate. The mean time to solve each of the
three puzzles immediately after training was the dependent measure. The
analysis revealed no significant effect of any of the independent
measures.
Long-term transfer (retention).
The same ANCOVA was performed on the solution times of participants’
performance a week later. This analysis yielded a significant main effect
for training condition and no other significant effects. Contrasts showed
that the solution times for people who had trained in the VE were
significantly faster than either the paper tutorial group (t(37) = 2.87, p
= .007) or the video group (t(37) = 2.40, p = .021) The figure above
illustrates the effect of training condition on people’s long-term
performance.
Training condition was the most powerful predictor of retention,
accounting uniquely for 18% of the variance in solution times. Training
time accounted uniquely for 5% and gender accounted uniquely for 5%. In
total, the model fit by the ANCOVA accounted for 36% of the variance in
solution times.
[ return to top ]