CALIFORNIA STATE UNIVERSITY, NORTHRIDGE
THE EFFECTS OF SHAPE-CODING AND
MIRROR-IMAGING ON THE PERFORMANCE OF A
SIMULATED ROOF BOLTING TASK
A thesis submitted in partial satisfaction of the
requirements for the degree of Master of Arts in
Psychology
by
Jayne M. Schurick
May, 1982
The Thesis o
ayne M. Schurick is approved:
Tyler Blake
California State University, Northridge
ii
Acknowledgement
The author would like to acknowledge the U.S. Bureau of Mines for
its support of this research.
The work was performed under Contract
#H0292007 with Mr. Richard Curtin acting as Contract Monitor.
iii
Q .
Table of Contents
Page
Introduction
1
Arrangement of Controls
Previous Research
Shape-Coding of Controls
1
8
13
Method
17
Subjects
Apparatus
Procedure
17
17
21
Results
28
Questionnaire Results
Control Errors and Time to Complete
the Task
Discussion
28
31
44
Arrangement of Controls
Shape-Coding of Controls
Trial
Recommendations
Replication and Further Research
47
48
49
50
52
References
54
Appendix A
57
Description of Mine Simulator
Appendix B
61
Subject Instructions
Appendix C
66
Questionnaire Administered after
Experiment and Summary of Responses
iv
9
List of Figures
Number
Title
Page
1
Two principles of control arrangement showing relative sequences
of controls.
3
2
Illustration of a dual boom roof
bolter.
5
3
Control-burner arrangements used
in stovetop experiments.
10
4
Illustration of the roof bolter
simulator with boom fully extended and raised.
18
5
Shape-coded control knobs used
in the experiment.
20
6
Drilling/bolting pattern subjects
were instructed to use.
24
7
Mean number of substitution errors
by code, control arrangement, and
trial.
35
8
Mean number of substitution errors
by control coding and trial.
38
9
Average time to complete each cycle
by code, control arrangement,
and trial.
41
Top view of underground coal mine
simulator.
59
10
v
'
List of Tables
Number
Title
Page
1
Sequence of Control Activations
25
2
Reported Expectation of Control
Arrangement
29
3
Mean Number of Errors for each
Control Arrangement with
Shape and No Shape-Coded
Controls
33
4
Analysis of Variance for Control
Activation Errors
36
5
Significant Differences among
Ordered Means using a NewmanKeul s Procedure
37
6
Significant Differences among
Ordered Means using a NewmanKeu l s Procedure
42
vi
ABSTRACT
THE EFFECTS OF SHAPE-CODING AND
MIRROR-IMAGING ON THE PERFORMANCE OF A
SIMULATED ROOF BOLTING TASK
by
Jayne M. Schurick
Master of Arts in Psychology
Certain underground mining machines have two identical control
stations located on opposite sides of the machine.
Often, operators
who are accustomed to working on one side of the machine, switch to
the control station on the other side, and confuse controls, a situation that can result in serious accidents and injury.
This thesis
addresses the question of how the controls should be arranged to simplify transfer from one station to the other.
The controls can be
either mirror-imaged or arranged in a left-to-right sequence.
The use
of shape-coded control knobs was also investigated to determine their
relative advantage with each control arrangement.
a 2 X 2 X 8 mixed design.
The experiment used
The first factor was arrangement of con-
vii
trols (mirror-image left-right), the second factor was code (shape or
no shape-coded controls), and the third factor was trial (four practice and four transfer).
Subjects were randomly assigned to one of the four treatment (2 X
2) conditions and learned a manual control task using either shapecoded control knobs or standard round knobs.
They then transferred to
the other side of the machine and performed the same task with the
controls positioned in either a mirror-imaged or left-right arrangement of the first set.
Time and errors were recorded during the prac-
tice and transfer trials.
Performance did not significantly differ
between the arrangement and code groups, a.lthough there was a slight
performance advantage for left-right arrangement and shape-coding.
The effect of trials was significant; training reduced both the number
of errors· and time to complete the task.
In order to investigate subjects• expectations for a particular
arrangement, errors were analyzed in terms of the control that was
used and which should have been used.
(Symmetrical errors might in-
dicate an expectation whereas nonsymmetrical errors would result from
other confusions.)
There were no differences by group, but overall,
significantly more nonsymmetrical errors were made.
In another anal-
ysis, performance was assessed based on subjects' reported expectation
on a post-experiment questionnaire.
Even though there was a general
tendency to report a left-right expectation, there were no differences
in performance between those who reported mirror-image or left-right.
viii
Introduction
The development of many modern machines has become increasingly
specialized and task specific.
This specialization reduces much of
the operator's workload by allowing him to develop repetitive sequences of control actions which in turn make performance of the task
somewhat automatic.
Automaticity results in increased proficiency as
reflected by several performance measures, some of which are reduced
task completion times, improved accuracy, reduction of fatigue, and
increased freedom to attend to other matters (Brebner &Sandow, 1976).
Certain types of machines have more than one operator control
station.
These types of machinery include aircraft, some utility
vehicles, and some underground mining equipment.
On such dual control
station equipment, there is a problem of how the manual controls
should be arranged so that the learned skills are best transferred
from one workstation to another.
One objective of the present study
was to investigate this problem.
A second objective was to determine
whether shape-coded control knobs are advantageous for the learning
and performance of manual control tasks.
Each of these purposes will
be discussed separately.
Arrangement of Controls
The problem of control arrangement is illustrated by cockpit controls for pilots who may sit in the left seat during one flight and in
the right seat during another.
If a pilot who normally flies in the
left seat becomes accustomed to locating a particular control to his
right and closest to the pedestal (center), where then would he expect
to find the duplicate control when flying from the right seat- to his
2
right or closest to the pedestal? For this and many other tasks there
are two basic principles for the arrangement of controls:
mirror-
imaging and left-right sequencing.
Mirror-imaging refers to two sets of controls being placed symmetrically about the same axis through the machine.
Thereby, a specific
control that is located closest to some machine point at one station
would be placed closest to that point at the other station.
The al-
ternative principle, left-right arrangement, implies that the controls
are arranged in sequence, left to right, and maintain the same relative positions at both control stations.
Therefore, a control located
to the operator•s left at one station would remain on his left at the
other.
Both principles are illustrated in Figure 1.
It is likely that the operator possesses certain expectations of
how controls will be arranged.
These expectations may be based on
repeated exposure to situations in which some standard principle has
been adopted.
The
11
C1
ockwi se for increase .. stereotype is such an
expectation (Brebner & Sando\'/, 1976).
However, such stereotypes tend
to vary across groups of people (e.g., different nationalities).
Other expectations are more fundamental and thus applicable across
populations.
These are described by Welford {1968) as inherent ten-
dencies to code and store information in economical ways.
Sandow (1976) identify the
category.
11
Brebner and
Warrick Principle .. as falling into this
Briefly, the Warrick Principle states that an indicator
will move in the same direction as the nearest point on the control
knob.
Unfortunately, as Brebner and Sandow {1976) note, there are difficulties in identifying stereotypes and once identified, there can be
0
0
01
3
A
Figure 1.
B
Two principles of control arrangement showing relative sequences of
controls. Operator positions are indicated by the circles. The controls are machine referenced on machine A (mirror-imaging) and body
referenced on machine B (left-right arrangement).
w
4
p '
problems of interpretation.
The cockpit problem is a good example.
It is likely that both the body (left-right) and the spatial (mirrorimage) reference cues are important to the pilot.
Which principle
is
.
easier to learn? Which would the operator revert to in an emergency?
The answers to these questions and determination of true population
stereotypes for many pieces of equipment would provide valuable information to equipment and instructional designers.
This paper addresses the problem of identifying the best principle
of control arrangement for roof bolter machines that have two workstations.
A roof bolter is used to drill holes and insert bolts into the
roof of mine entries after coal has been extracted.
An illustration
of a typical dual workstation roof bolter is shown in Figure 2.
On
each boom there is an operator workstation that contains five to ten
levers.
At a minimum these levers control boom positioning, machine
stabilizing, and drilling functions.
Some machines also have tramming
controls (to move the machine), electrical controls (such as main
power and lights), and additional boom positioning controls (such as
boom tilt).
A recent analysis of the causes of roof bolter accidents in underground coal mines found that approximately 60% of all reported accidents might be attributed to poor design of the machine (Helander,
Conway, Elliot, &Curtin, 1980).
It has further been estimated that
at least 10% of the accidents are caused by deficiencies in control
design that result in control activation errors (Miller & Mclellan,
1975).
It is therefore suggested that a human factors design of con-
tro1s should considerably reduce the number of accidents in underground coal mining.
5
~~Operator
Dril
Figure 2.
.....--Control panel
""
Illustration
of a dual boom roof bolter.
6
Roof bolting requires frequent manipulation of the drill steel
with one hand, leaving the other hand free to operate the controls.
Therefore, an operator working at the left side of the bolter uses his
right hand at the controls whereas the operator on the right side uses
his left hand.
Most of the time, operators work in pairs on estab-
lished sides of the machine.
Occasionally, however, operators may
switch sides or fill in for someone on the nonregular side.
This
presents the problem of how the functionally identical sets of controls should be spatially arranged:
Should they maintain their posi-
tions relative to the operator•s body in a left-to-right arrangement
or relative to the roof bolter in a mirror-imaged arrangement? There
have been no studies of accident statistics nor empirical research to
investigate this problem.
Depending on the manufacturer, control panels of dual-boom roof
bolters may vary in number of controls, orientation of levers, and
location on the machine.
An arrangement of controls according to
subtask recently suggested by Helander et al. {1980} was adopted for
the present investigation.
The results should, however, generalize to
all types of control panels since the question being asked involves
relative placement of controls and not particular controls.
Six con-
trol functions were chosen:
1.
Boom sump, a repositioning control used to extend and retract
the drill head.
2.
Boom slide, a repositioning control used to change the lateral
position of the drill head.
3.
Boom raise, a drill control which is used to raise or lower
the drill head.
7
4.
Drill rotate, a drilling control used to rotate the drill head
in both forward (normal drilling) and reverse directions.
5.
Drill clamp, used to hold the drill into place.
6.
Stab jack, used to stabilize the machine once it has been
correctly positioned.
The controls are vertically oriented levers that operate in a
push-pull direction.
Control knobs are round and about one and one
half in. (3.81 em) in diameter.
The panels are typically mounted
parallel to the boom.
One might assume that equipment operators would choose to perform
the most critical task with their preferred hand.
The critical task
of the roof bolter operator might be either manipulation of the controls or of the drill steel.
There is evidence, however, that this
assumption is not necessarily true.
It has ·been documented that a
task learned with one hand can be performed with the other.
This
ability is called bilateral transfer of training.
Bilateral transfer was first noted by Weber in 1844 (Woodworth,
1938) when he observed that children trained to write with their right
hands could successfully produce mirror-writing with their left hands.
Since that time, bilateral transfer of training has been studied in
various experimental situations ranging from perceptual skills (e.g.,
Volkman, 1858) to motor skills (e.g., Bray, 1928; Cook, 1936).
In-
deed, textbooks in experimental psychology (e.g., Woodworth, 1938;
Stevens, 1951, & McGeoch, 1952) treat the concept of bilateral transfer as an accepted principle.
The question, therefore, does not seem to be which hand should
perform which task but how the controls should be arranged so that
8
either hand can efficiently use them.
When an operator changes sides
of the machine, and thus, the functions controlled by each hand, bilateral transfer of training should occur.
Assuming that operators
can successfully operate the controls with either hand, there still
remains the question of how they transfer the highly practiced control
sequences - with reference to the machine or to themselves.
Previous Research.
Inspired by a study of cockpit control errors
performed by Fitts and Jones {1947), Pigg {1954) investigated bilateral transfer of training under two experimental conditions:
mirror-
image and left-right sequence of controls.
In order to determine the best arrangement, Pigg constructed control panels of four switches each.
Sixty-four subjects were grouped
so that one-half learned a perceptual motor task with their left hands
and the other half with their
ri~ht.
The task consisted of turning
off one of four colored lights by activating the correct switch.
Subjects then performed
other control panel.
11
the same task .. with the opposite hand at the
The arrangement of the second set of controls
was either mirror-imaged or identical to the first set.
The group that performed on mirror-imaged controls required less
time to learn the correct switch for each of the lights; however, this
effect was significant
(~
< .05) only for the first eight trials.
Even though there were no significant differences between the groups
for the remainder of the task {trials 9 - 32), the mirror-image group
consistently made fewer control errors and had a lower percentage of
subjects making errors.
Subjects' performance and verbal reports
indicated that the mirror-image arrangement was expected over the
left-right arrangement by a ratio of approximately two to one.
Addi-
9
tionally, subjects preferred the mirror arrangement by approximately a
three to one ratio.
Two other studies that dealt with mirror-imaged controls were
intended to investigate a common source of frustration - the arrangement of stovetop controls (Chapanis & Lindenbaum, 1959; Ray & Day,
1979).
Both Chapanis and Lindenbaum, and Ray and Day constructed
mock-up stove tops on which four burners located on a horizontal surface could be controlled by different combinations of controls located
on a vertical surface.
Mirror-imaging was defined in these studies as
the arrangement of controls such that the leftmost and rightmost controls corresponded to both back burners or both front burners.
In
this way, the two left controls were a mirror-image of the two right
controls.
In essence, an implied center line on the stove served as a
machine reference for the two columns of burners.
The control-burner
configurations used in each study are shown in Figure 3.
The task in the first experiment (Chapanis & Lindenbaum) required
subjects to locate a stimulus light in the middle of the burner by
activating the correct control.
Subjects were randomly divided into
four groups, each group being exposed to one of the configurations.
Eighty stimuli were randomly presented on one of the four burners.
Of the control-burner relationships, the mirror-image arrangement
(stove 2) produced the fewest errors (although not significantly different from stoves 3 and 4) and the lowest reaction time scores (significant for the last 40 trials).
In the Ray and Day experiment, subjects were instructed to turn
off a burner light by operating the associated control.
Errors were
recorded for each burner configuration and subjects reported their
10
p
Q
Q
Q
0
®0 00
0
0
G
0 0 0 0
(1)
(2)
Mirror
Mirror
Q
Q
G
0
0@ 0®
{3)
Left-Right
Figure 3.
0
0
G
G
0
0 0 0 0
( 4)
Left-Right
Control-burner arrangements used in stovetop
experiments. Arrangements (1) and (2) are
mirror imaged. Chapanis & Lindenbaum used only
arrangements (2), (3), and (4); Ray and Day
used all four configurations.
'
ll
preferences.
The mirror-image configuration was preferred by a slight
majority of subjects and produced fewer errors.
There are differences between these experiments and the.current
investigation.
The stovetop experiments involved the arrangement of
controls on a single panel.
Subjects performed with their preferred
hands and bilateral transfer was not an issue.
These experiments may
not be generalizable to either cockpit or roof bolter control panels
and may indicate nothing more than a preference for symmetrically
arranged controls on a single panel.
Pigg•s experiment involved the use of two panels and bilateral
transfer, and therefore, has some similarity to cockpit and roof bolter controls.
An important difference between roof bolters and cock-
pits, however, is the locations of the control panels.
Whereas the
control panel(s) is placed directly in front of the operator in the
cockpit, duplicate roof bolter controls are on opposite sides of the
machine.
When the roof bolter operator changes workstations, he also
turns his body.
It was expected that this would provide strong orien-
tational cues that would favor the mirror-image arrangement.
A more relevant example of mirror-imaged controls is found on some
utility vehicles used by Southern California Edison.
These single
boom vehicles have control stations on both sides so that the operator
always has maximum visibility of the bucket and worker.
at each workstation are mirror-imaged.
The 10 levers
Several operators of these
vehicles stated that they preferred this arrangement.
Unfortunately
there have been no experimental investigations to verify the best
arrangement.
12
The utility vehicles provide a similar set of circumstances to
those of roof bolter machines.
Functionally identical controls are
required on opposite sides of the machine and the operator must turn
his body from one side of the vehicle to the other.
The main differ-
ence between this example and roof bolters, however, is that it is not
mandatory for the utility truck operators to change hands on particular controls.
Because the operator can use both hands, a mirror-image
arrangement would force him to change hands and a left-right arrangement would not.
Roof bolter operators, on the other hand, must change
hands.
The critical variables for determining the best arrangement of two
sets of functionally identical controls, therefore, might be:
1.
One vs. two-handed operation.
2.
Whether the operator can use the same hand or must switch
hands (that is, whether bilateral transfer is mandated by the
task).
3.
The number of workstations and their locations on the machine.
Whether controls should be mirror-imaged may also depend on several other characteristics of the machine and the task.
The type and
location of feedback may be one important consideration.
Since the
roof bolter operator receives immediate visual feedback from the front
of the machine, he may form direct associations between a particular
control function and the machine itself.
For example, the operator
may learn that the drilling control is the closest one to the boom.
If this association holds for both sides of the machine, machine referenced arrangements (mirror-imaged) would be indicated.
This may
also be true of aircraft in which there is a locus of control, for
13
instance, to make the plane move in a forward direction.
Many control rooms are laid out in such a way that the operator
cannot directly see what he is controlling.
There may be an indicator
display but this usually does not take the form of a locus of control,
and therefore may not be a strong enough feedback cue for forming
associations between controls and their locations.
Indeed, evalua-
tions of nuclear power plant control layouts have cited mirror-imaging
of controls as a source of confusion to operators (e.g., Swain &
Guttman, 1980).
Still there is another more practical consideration.
For example,
many manufacturers of single boom roof bolters place the drill and
boom raise controls closest to the drill steel.
An operator can then
easily align the drill steel while simultaneously manipulating the
control levers with his other hand.
Hence, a left-right arrangement
of controls on a dual boom bolter might be impractical, since it would
require the operator (at one of the workstations) to reach over other
controls, thereby increasing the risk of him inadvertently actuating
one or more of the other functions.
Shape-Coding of Controls
It is generally recognized that the coding of controls by shape
improves their discriminability.
Several military investigators have
identified and experimentally validated shapes for visual and/or tactile discrimination (Bradley, 1959; Hunt & Craig, 1954; Jenkins,
1947).
The basic design of these experiments was to have subjects
compare knobs and report whether they were identical or different.
The comparisons were made two at a time and either both knobs were
presented tactually or one knob was to be felt and compared to a vis-
14
ual representation.
While Hunt and Craig, and Jenkins were concerned
strictly with discriminability, Bradley was also interested in the
manipulability of knobs and the precision with which they could be
set.
The latter investigation identified classes of discriminable
shapes only for cylindrical knobs.
This work emphasizes the impor-
tance of choosing shapes for a specific application, such as the precision with which knobs need to be set and the size and function of
the controls.
Based on the above and similar experiments, recommendations for
the use of shape-coded controls can be found in several human factors
handbooks (Van Cott &Kinkade, 1972; Woodson & Conover, 1964) and
textbooks (McCormick & Sanders, 1982).
In general, these texts re-
commend the use of shapes that are easily distinguishable and that are
associated with or resemble (when feasible) the function of the control.
In addition the Military Standardization Handbook (MIL-HDBK-759,
proposed, 1981) proposes 16 shapes for use on knobs that must be recognized by touch alone.
The control shapes used in the present study
were selected from this handbook.
Implied by the research and recommendations for the use of shapecoded controls is the assumption that the learning and performance of
a task will be improved as long as the shapes are discriminable.
This
assumption may be unwarranted without consideration of the task.
For
example, just because subjects can distinguish between differently
shaped knobs does not necessarily mean that such knobs will increase
productivity or reduce accidents.
There is a lack of empirical re-
search of the practical advantages of shape-coded controls.
15
One exception to the paucity of practical research is a study of
differently shaped controls in underwater applications (Carter, 1978}.
Divers adjusted 15 knobs of different shapes and sizes in three locations on their diving suits.
Speed and accuracy of adjustment were
assessed and discriminability was not a factor.
The results indicated
that large diameter knobs were adjusted more rapidly and accurately
than small knobs and that tactile markers (for example, pointers)
generally enhanced the accuracy of knob adjustment.
Because discrim-
inability was not a factor in this study, it is difficult to generalize the results to the present experiment.
This study of controls
used in underwater environments does, however, illustrate the need to
investigate the particular task and environment before choosing knobs.
There are at least two other factors that need to be considered
before shapes are recommended.
shapes to use.
The first involves the choice of which
Several investigators (mainly in the aircraft indus-
try) have attempted to describe shapes that have associations with the
function to be controlled.
This is not an easy task.
Except for a
few controls on a few machines, there is little agreement among users
as to the appropriate matching of shapes and functions.
Nevertheless,
suggested shapes can be found in several design guides as previously
mentioned.
A second factor to be considered is cost.
controls are expensive to manufacture.
Shape-coded
The replacement of missing or
broken knobs would also be expensive and time consuming.
The reason for including shape-coded knobs in the present experiment was to determine whether they would enhance performance, as measured by a reduction in errors and in the time needed to complete the
task.
Also of interest was whether any differences between the two
16
arrangements would be reduced or eliminated by using shape-coded controls.
Method
Subjects
Thirty-four male subjects served in the present experiment, howc
ever, two were eliminated due to extremely poor performance.
All but
five were undergraduate students at California State University,
Northridge; the other five subjects were human factors professionals
employed by Canyon Research Group, Inc.
Subjects were paid $15.00 for
their participation, which lasted from two and one half to three
hours.
All subjects were tested individually.
Apparatus
A simulated roof bolter (see Figure 4) was constructed to provide
a task similar to actual roof bolting work.
The simulator measured
4o.5 in. (115.57 em) in length with the boom fully extended, 28 in.
(71.12 em) in width, and 28 in. (71.12 em) in height with the boom
raised.
Six levers controlled the following functions:
1.
Boom sump
2.
Boom lateral slide
3.
Stab jack
4.
Boom raise
5.
Drill
6.
Dri 11 clamp
During testing of the equipment, the drill clamp failed to function
properly.
The clamp was therefore removed from the machine and the
control was disconnected, although it remained on the control panel.
The controls were vertically mounted on a 16 in. {40.64 em) by 6 in.
{15.24 em) control panel and operated in a forward {push) - backward
{pull) manner.
Each control therefore had two directions of movement.
17
18
Recording equjpment
Control
Dri 11
'\
Control panel
Stab jack
Figure 4.
Illustration of the roof bolter simulator with boom
fully extended and raised. The control panel could
be positioned as shown or on opposite side of the
machine.
19
The levers were 6 in. {15.24 em) long and were separated from each
other at the base by 1.75 in. {4.45 em).
The control functions were
attached to the panel by means of a contact so that even though the
levers were fixed, the functions they controlled were interchangeable.
Additionally, a toggle switch mounted in back of each lever allowed
the direction of the function to be reversed.
A separate switch con-
trolled power to the machine.
The control panel was attached to a 25 in. {63.5 em) arm that
extended at a 45 degree angle from the roof bolter and was placed
about 6 in. {15.24 em) above the floor.
The control panel could be
detached from the arm and repositioned on an identical arm on the
opposite side of the machine.
transfer performance.)
(This was done for each subject to test
Because the functions were interchangeable, a
mirror-image arrangement of the controls could be achieved.
The only
direction reversal that was made for the transfer task was boom slide
so that pushing the lever always resulted in the boom moving away from
the subject.
The control knob on each lever was removable to allow exchange
between the shape-code and no shape-code conditions.
Standard round
black knobs, 1 1/2 in. {3.81 em) in diameter were used for the no
shape-code condition.
Shape-coded knobs were chosen in accordance
with MIL-HDBK-759 which identifies discriminable shapes for heavy
equipment control levers.
Shapes that had some resemblance to the
lever functions were selected.
The knobs, approximately 1 l/2 in.
{3.81 em) in diameter are shown with their associated functions in
Figure 5.
(The actual order of the controls used was:
drill clamp,
boom sump, boom lateral slide, stab jack, boom raise, and drill.)
20
Figure 5.
Shape-coded control knobs used in the experiment.
From left to right, the knobs represent the
following functions: boom lateral slide, boom
sump, stab jack, boom raise, drill clamp, and
drill.
21
The machine functions were powered by hand (Skill) drills that
turned threaded rod.
was:
The range of movement for each boom function
sump, 11 in. (2/.94 em); slide, 9 in. (22.86 em); and raise, 14
in. {35.56 em).
All functions were electrically connected to an event
recorder and to the control panel.
The recording equipment included an Esterline Angus event recorder
and a digital clock that recorded elapsed time in seconds.
The event
recorder was installed in a 14 in. (J5.56 em) by 28 in. (71.12 em)
cabinet located at the back of the roof bolter.
Each of 12 channels
of the event recorder was connected to one of the control functions
(six controls by two directions).
The experiment was conducted in a simulated underground coal mine
(described in Appendix A) with a 48 in. (121.92 em) ceiling height.
Two wooden beams (4 ft. (12.2 m) long 2 X 4s) were attached to the
ceiling.
Subjects drilled and inserted one-half inch bolts into the
beams which were clamped into place to allow replacement with undri ll ed beams.
Procedure
The independent variables under investigation were shape-coding
versus no shape-coding of control knobs and mirror-image versus leftright arrangement of the controls.
two-phase experiment.
These variables were tested in a
The first was a learning phase which consisted
of learning a roof bolting task on one side of the machine.
During
the second, transfer phase, the subject performed the roof bolting
task on the other side of the machine.
The variable, shape versus no shape-coding of controls was consistent from the first phase to the second; that is, subjects who used
22
shape-coded knobs during learning also used the same shapes on the
transfer task, and similarly, subjects in the no shape condition used
standard round knobs for both phases.
The mirror-image/left-right
variable, however, was only applicable for the transfer trials.
independent variables formed a 2 X 2 X 8 mixed design.
The
Subjects were
randomly assigned to one of the four groups and were alternately
assigned to start on the right or left sipe of the roof bolter.
Two dependent variables were used.
One was time to complete the
task, measured by a digital stop watch from the time the subject
turned on the power of the roof bolter to the time he completed a
specified bolting cycle and turned off the power.
The other dependent
variable was the number of errors made by the subject.
Four different
types of errors were identified:
1.
Direction - subject activated the correct control but in the
wrong direction, for example, pushing instead of pulling.
2.
Sequence - subject operated one or more controls out of order,
either by forgetting the proper sequence or temporarily forgetting to activate a particular control and then going back
to it.
3.
Omission- subject forgot to operate a required control.
The
stab jack control was the only possible error of omission that
could have been made in this study since all other controls
were necessary to perform the task.
(Incidentally, roof bol-
ter operators frequently forget or intentionally disregard the
stab jack, Helander et al., 1980.)
4.
Substitution - subject confused one control for another.
Substitution errors were relatively easy to recognize; typi-
23
cally the subject would either make a comment or quickly reverse the incorrectly activated control and activate the correct one (or make a series of these rapid adjustments if he
could not find the correct control).
Substitution errors were of particular interest on the transfer task,
especially those for which the substituted control was symmetrical to
the intended function, indicating an expectation for a certain control
arrangement (mirror-image or left-to-right).
Two experimenters independently observed the performance of each
subject.
They recorded the type of error, the actual control chosen,
and correct control (for substitution errors only) for each experimental cycle.
Additionally, the event recorder continuously produced a
permanent record of the controls that were being activated.
After testing each subject, the experimenters recreated the sequence of control activations made by the subject and analyzed all
errors by comparing the correct sequence of control activations to the
permanent record.
The experimental task consisted of positioning the boom, setting
the stab jack, drilling a hole, and inserting a bolt into each of four
specified locations into the wooden beams.
This task was repeated
four times during the learning phase and four times during transfer.
Each subject therefore inserted a total of 32 bolts.
pattern is shown in Figure 6.
The bolting
The subject had to turn on the machine
power at the beginning of each cycle then center the boom and turn off
the power to end the cycle.
The correct sequence of control activa-
tions to complete one cycle is given in Table 1.
24
I
3
1
~
I
2
Figure 6.
/
4
I
Wooden beams
I
Drilling/bolting pattern subjects were instructed
to use. The numbers represent approximate locations an·d sequence for drilling. The roof
bolter would be located in the foreground.
25
Table 1
Sequence of Control Activations
1.
Turn on main power
2.
Extend/retract boom and slide boom left/right to drilling
position
3.
Lower stab jack
4.
Raise boom (with drill in place) to roof
5.
Continue raising boom while rotating drill
6.
Lower boom while reversing drill
7.
Continue lowering boom
8.
Raise boom (with bolt in place) to roof
9.
Continue raising boom while rotating drill
10.
Lower boom
11.
Raise stab jack
12.
Extend/retract boom and slide boom left/right to next
drilling position
13.
Repeat steps 3 through 12 for three remaining holes
14.
Lower drill as far as it will go
15.
Center boom
16.
Turn off main power
26
Upon reporting to the experiment, each subject was given a description of coal mining and of the experimental task (subject instructions are provided in Appendix B).
personal protective equipment.
The subject was then given miner's
Mandatory safety equipment consisted
of a helmet and cap lamp, safety belt with the lamp battery, knee
pads, and gloves; optional equipment included coveralls, boots, safety
glasses, and ear protection.
The subject practiced turning the cap lamp on and off several
times before the interior lights were turned off and before the subject and experimenters entered the mine simulator.
The function of
each lever on the roof bolter and the power switch were explained and
demonstrated by one experimenter.
For subjects in the shape-code
condition, the shapes were emphasized and subjects asked to feel the
knob as it was explained.
Subjects were told that the shape-coded
knobs were to help them distinguish among the controls and to learn
the task.
time.
Subjects were also told to operate only one control at a
The experimenter went through one bolting cycle (four bolts)
without actually inserting any bolts and then the subject performed
his first cycle with guidance from the experimenter.
No data were
recorded for this cycle and the subject was free to ask questions.
Following this introduction, the recording equipment was turned on
and the subject performed four task cycles.
The stop watch was man-
ually initiated and terminated at the beginning and end of each cycle.
In between each cycle, the experimenter repositioned the wooden beams
in the ceiling to provide undrilled areas for drilling and bolting.
After the four bolting cycles, the subject was sent out of the
mine simulator for an approximately 20 minute break while the experi-
27
menter moved the control panel to the opposite side of the machine.
For the mirror-image
group~
the control functions were interchanged
accordingly as were the shape-coded knobs for that condition.
The
control panel was simply repositioned for the left-right groups.
Before reentering the
mine~
subjects were informed that the con-
trol panel had been moved and that they were to perform the same task
they had learned.
The instructions explicitly stated that the exper-
imenter had the capability to exchange and/or reverse the controls but
that the purpose was not to trick the subject - only to test his performance on identical controls placed on the opposite side of the roof
bolter, see Appendix B.
Once back inside the mine simulator, the subject performed four
more bolting cycles while error and time measurements were recorded.
The bolting pattern remained the same as it had been for the learning
task.
After completing the experiment, a short questionnaire was administered (presented in Appendix C).
Subjects were questioned on their
knowledge of the purpose of the experiment, specific controls that
they consistently
confused~
the tactile and/or visual benefits of the
shape-coded controls, profession or major in school, handedness, and
experience with heavy equipment.
Subjects were then told the purpose
of the experiment and were paid for their participation.
Results
Two subjects who were tested fairly early in the experiment performed so poorly that it was decided to eliminate their data.
the subjects were in the mirror/no shape-code condition.
Both of
Their data
was replaced by two additional subjects, therefore providing a total
sample of 32 to be used in the analyses.
Questionnaire Results
Responses to the questionnaire administered after the experiment
are presented with the questionnaire in Appendix C.
When reviewing
these responses it is important to remember that they were given following the task and that subjects may have been influenced by their
particular experimental treatment.
A fairly liberal policy \'las adopted in classifying responses to
the question about whether a subject had knowledge of the experiment•s
purpose.
included:
Responses that were included in the affirmative category
11
determining the best placement of controls, ..
11
identifica-
tion of problems between the right and left sides of the machine,"
11
Viewing controls from a different perspective, .... evaluation of con-
trol configurations, .... testing transfer of knowledge, .. and .. evaluation
of stimulus-response compatibility in a control reversal study...
Only
one subject stated the exact purpose of testing machine versus body
referenced control configurations, and only two of the subjects who
had shape-coded control knobs mentioned shape-coding as a purpose.
The second question asked subjects how they expected the controls
to be arranged on the transfer task.
presented in Table 2.
A summary of the responses is
Overall, 66% of the subjects reported an expec-
tation for a left-to-right arrangement with only 31% saying they ex-
28
29
Table 2
Reported Expectation of Control Arrangement
Reported Expectation
Mirror
Left-Right
Don't Know
Mirror/Shape
5
2
1
Mirror/No Shape
2
6
0
Left-Right/Shape
0
8
0
Left-Right/No Shape
3
5
0
10
21
1
Group
Total
30
pected mirror-imaged controls.
Of those who did say mirror-imaged,
however, half were in the mirror-image, shape-code condition and none
were in the left-right, shape code condition.
This seems to indicate
that the shapes where helpful in determing arrangement.
In'general
the groups that used shape-coded knobs expected the controls to be
arranged according to the configuration they were given.
There may be
a preference for left-to-right arrangement but the shape-coded knobs
influenced the reported expectation.
(Four subjects also mentioned
that they had assumed that the control panel had simply been moved to
the opposite side of the roof bolter and did not realize that the
experimenter could or would have the time to interchange controls.)
The controls causing the most problems were boom sump, boom slide,
and stab jack.
The problems represented confusions between controls,
forgetting to operate a control (mainly the stab jack), and forgetting
the direction of movement for a particular function.
The only differ-
ence in errors between groups was on the stab jack control for the
shape versus no shape-code conditions.
Six subjects in the shape-code
group reported having problems with the stab jack while 12 subjects in
the no shape-code group had problems, possibly indicating that the
shape of the stab jack helped somewhat.
Also many subjects reported
that the direction of the stab jack seemed to be reversed (they
thought a forward-push activation should have lowered the stab jack
when in fact it raised it).
The most frequently confused controls for all groups were the boom
sump and slide.
In addition to subjects attesting to this on the
questionnaire, while performing the task, many commented on the
trouble they were having remembering which of these two controls acti-
31
vated each function.
Very few subjects reported that the shape-codes helped them to
distinguish between controls.
There were 13 accounts (7 subjects) of
the shapes helping visually (mainly boom sump, boom slide, and stab
jack) and 9 reports {4 subjects) of tactile advantages (boom sump,
stab jack, and the drill control).
Based on the responses to this
question and question #2 (expectation for mirror-image or left-right
arrangement), it appears that the shapes helped subjects determine the
arrangement of controls but did not.help in distinguishing between
individual controls.
Subjects' profession or major in school is not discussed since
there were many different categories and none were more highly represented in an experimental condition than the others.
Most subjects
had no previous experience working with heavy equipment.
Those re-
porting experience had mostly used forklifts and light construction
equipment.
There was no differential experience among groups.
Fin-
ally, of the 29 subjects who reported their handedness, most were
right handed {23), and these were fairly
ev~nly
divided among the
groups.
Control Errors and Time to Complete the Task
Separate analyses of variance (BMDP 2V, Dixon & Brown, 1979) were
performed on the four different types of errors and on the time data.
In all of these analyses, trial (four learning and four transfer) was
a within subject variable, and arrangement (mirror-image versus leftright arrangement) and code (shape versus no shape) were grouping
variables.
32
The mean number of errors made by each group is presented in Table
3.
Since errors of direction, sequence, and omission were not of pri-
mary interest to the study, they were not subjected to further interpretation.
Errors of sequence and omission mainly seem to be indica-
tive of task knowledge and although direction errors might be interesting (to determine compatibility between direction of movement
and function), direction was not manipulated in this study and no firm
conclusions can be drawn.
The only significant difference for direction and sequence errors
was trial, f (7, 196) = 8.53,
respectively.
~
< .01 and f (7, 196) = 2.67,
~
< .05,
A Newman-Keuls test of ordered pairs of means revealed
a significant difference between learning trial 1 and transfer trial
4, f (8,196) = 2.57,
~
< .05 for direction errors. Using the same a
posteriori procedure, there were no significant differences between
the means for sequence errors.
significance, f {1, 28)
= 3.03,
Coding of control knobs approached
~
< .10 for sequence errors. The mean
number of errors for shape-coded controls was .266 and for no shapecoded controls, it was .461, indicating a slight advantage of shapecoding.
~
Arrangement approached significance, f (1, 28) = 3.32,
< .10 for errors of omission. Examination of the means showed that
the mean number of errors was .281 for mirror-imaged controls and .148
for left-right arranged controls.
This indicates a slight advantage
for left-right arrangement.
The errors that are of main interest to the study are substitution
errors, those for which the wrong control was activated.
Substitution
errors have practical significance for both the coding of control
knobs and the arrangement of controls on each side of the roof bolter.
/
33
Table 3
Mean Number of Errors for each Control Arrangement
with Shape and No Shape-Coded Controls
Control Arrangement
Mirror Image
Left-Right
Shape
No Shape
Shape
No Shape
Direction
2.89
2.33
2.14
2.66
Sequence
.25
.56
.28
. 36
Omission
.27
.30
. 78
.22
1.45
1.56
.81
1.63
Type of Error
Substitution
34
The mean number of substitution errors is shown graphically by code,
control arrangement, and trial in Figure 7.
Because control arrange-
ment could not have an effect on performance until the transfer
trials, the two shape-code conditions (mirror-image and left-right)
should produce similar curves on the four learning trials and the no
shape-code conditions (mirror and left-right) should show similar
curves.
As can be seen in Figure 7, all groups except the left-right,
shape group produced similar patterns of results.
This group had
superior performance on all learning trials, an unexplained advantage
that seemed to carry over into the transfer trials.
According to the analysis of variance (see Table 4), trial was the
only variable to-reach statistical significance, F (7, 196) = 16.84,
~
< .01. A Newman-Keuls procedure revealed several significant dif-
ferences among the ordered means.
The obtained differences and levels
of significance are shown in Table 5.
Trials 1 and 5 (learning trial
1 and transfer trial 1) had significantly more errors than learning
trials 3 and 4, and transfer trials 2, 3, and 4.
A trial by code interaction approached significance,£.. (7, 196) =
1.84,
~
< .10. These data are-plotted in Figure 8. The interaction
indicates that the shape and no shape-code groups performed differentially across trials.
As can be seen on the graph, the two groups
performed initially at about the same level.
The group that used
shape-coded knobs then made fewer control activation errors (especially evident on the first transfer trial) on every subsequent trial
except the third transfer trial on which they made slightly more
errors.
35
Mirror/Shape.
•o-o• Mirror/No
Shape
6.0
VI
s....
•
•
Left-Right/Shape
,[J
0
Left-Right/No Shape
0
s....
s....
LLJ
5.0
s::::
0
+l
ttl
>
.,....
+l
u
4.0
c:r:
r-
0
s....
+l
s::::
0
u
3.0
s....
Q)
..c
E
:::::5
z:
s::::
2.0
ttl
Q)
~
1.0
L1
L2
L3
Tl
L4
T2
T3
Trial
Figure 7.
Mean number of substitution errors by code, control
arrangement, and trial.
T4
36
Table 4
Analysis of Variance for Control Activation Errors
Source
df
MS
F
Arrangement (A)
l
5.35
. 51
Code (C)
l
13.60
1.29
c
l
7. 91
.75
Error
28
10.54
Trial (T)
7
28.72
16.84**
T
X
A
7
2.55
1.50
T
X
c
7
3.14
1.84
T
X
AX
7
.58
.34
196
1.71
AX
c
Error
** £ < . 01
37
Table 5
Significant Differences among Ordered Means
using a Newman-Keuls Procedure
Ordered Means a
Ordered Means
. 500
. 688
. 781
.781
(8)
(6)
(7)
(4)
.969 1.56 2.75
(3)
(2)
(l)
2.88
(5)
.500
2.25** 2.38**
.688
2. 06*
2. 19*
. 781
1. 97*
2.1 0*
.781
1. 97*
2.1 0*
.969
1. 78*
l. 91 *
1.56
2.75
2.88
a Numbers in parentheses indicate trial number:
1 - 4 represent
learning trials 1 - 4, and 5 - 8 represent transfer trials 1 - 4.
*£
< • 05
** £
< • 01
38
•
6.0
•
o--o
Shape
No Shape
Ul
s....
s....
s....
0
LLJ
5.0
s:::
0
.,....
+->
It!
>
.,....
+->
u
c:t::
4.0
,.....
0
s....
s:::
+->
0
u
3.0
s....
QJ
..0
E
:::l
z
s:::
It!
2.0
QJ
:::;:::
1.0
L1
L2
L3
Tl
L4
T2
T3
Trial
Figure 8.
Mean number of substitution errors by control coding
and trial.
T4
39
In addition to the mixed analysis of variance, two other analyses
were performed using substitution errors on the transfer trials as the
dependent variable.
These were performed to investigate whether there
was an expectation for either mirror-imaged or left-right arranged
controls.
The first analysis of variance was a three factor mixed design
with repeated measures on one factor (see Bruning & Kintz, 1968 for
details of the analysis).
The dependent variable was the number of
symmetrical and nonsymmetrical errors made on the four transfer trials
(a within subjects variable).
A symmetrical error might indicate an
expectation for one control arrangement.
If a subject had mirror-
imaged controls, for example, and he expected left-right arrangement,
he would confuse symmetrical controls:
drill and extra {disconnected
drill clamp), boom sump and raise, and slide and stab jack.
The same
would be true for a subject who had left-right arranged controls but
expected a mirror-imaged arrangement.
The analysis indicated an over-
all difference between symmetrical and nonsymmetrical errors, f. {1,28)
= 25.79,
~
< .001, however, there were no differences by group. This
may mean that there was no expectation for either left-right or mirror-imaged controls.
As expected, however, there were significantly
more nonsymmetrical errors (mean of 3.81) than symmetrical errors
(mean of 1.03).
This simply means that there were fewer symmetrical
confusions.
The second analysis was a t-test between the mirror-image and
left-right groups using only subjects who had reported a correct
expectation on the questionnaire (question #2).
The t-test was not
significant, indicating that there were no performance differences
40
between groups of subjects who reported a correct expectation for one
or the other arrangement.
Finally, an analysis of covariance was performed in an attempt to
partial out some of the variability due to differences in learning.
The mean number of substitution errors on learning trials 3 and 4 was
the covariate and the mean number of errors on the first transfer
trial was the dependent variable.
f (1, 27) = 5.82,
~
The covariate was significant,
< .05 implying that it was a useful predictor of
the number of errors made during the first trial of transfer.
Arrangement of controls was the only variable that approached significance, I {1, 27),
~
< .06. Subjects in the mirror-image group made
an average of 3.56 errors and those in the left-right arrangement
group averaged 2.19 errors.
This indicates a slight advantage of
1eft-right arrangement of contra1s.
The average time to complete each cycle is graphed by group in
Figure 9.
As with the error data, similar graphs should be produced
by the two shape-code groups and the two no shape-code groups on the
learning trials.
As shown, all groups except the mirror/no shape
group had very simi 1ar time scores.
In the analysis of variance, the
only significant effect was trial, I (7, 168)
=
31.84,
~
< .01. The
significant differences between the ordered means according to a Newman-Keuls test are shown in Table 6.
Learning trial 1 differed from
all other trials except transfer trial 1, and the first transfer trial
differed from learning trials 3 and 4 and transfer trials 2, 3, and 4.
In an attempt to partial out some of the subject variance, time to
complete learning trials 3 and 4 was averaged and used as a covariate
with time on transfer trial 1 as the dependent variable.
This covar-
41
450
•
•
Mirror/Shape
Mirror/No Shape
B
B Left-Right/Shape
.
D----0 Left-Right/No Shape
o----o
-;400
"'0
s:::
u
0
Q)
(/)
s:::
•r-
.:.t:.
(/)
ltl
1Q)
+l
Q)
r-
0.
5 350
u
0
+l
Q)
E
•r-
1-
300
L1
L2
L3
Tl
L4
T2
T3 - T4
Trial
Figure 9.
Average time to complete each cycle by code, control
arrangement, and trial.
Table 6
Significant Differences among Ordered Means using a Newman-Keuls Procedure
Ordered ~1eans a
Ordered Means
316.00
325.46
327.29
336.32
343.36
355.61
371.11
(8)
(7)
(4)
(6)
(3)
(2)
(5)
388.39
(1)
55.11**
72.39**
325.46
45.64**
62.93**
327.29
43.82**
61. 11**
336.32
34.79*
52.07**
343.36
27.75*
45.04**
316.00
355.61
39.61**
32.79*
371.11
388.39
a Numbers in parentheses indicate trial number:
1 - 4 represent learning trials 1 - 4, and .5 - 8
represent transfer trials 1 - 4.
* Q < .05
** £ < .01
-1=:>
N
43
9 •
iance analysis did not yield any significant differences.
A t-test was performed using subjects who correctly reported an
expectation for mirror-image or left-right controls (questionnaire
item #2).
.
The analysis was performed between the mirror-image and
left-right groups using time to complete the first learning trial as
the dependent variable.
The !-test was not significant, an indication
that the reported expectation was not reflected in time to complete
the transfer task.
Discussion
The objective of the present experiment was twofold; the first was
to determine how two functionally identical sets of controls should be
spatially arranged on a roof bolter that has two control stations
located on opposite sides of the machine.
The task of drilling holes
and inserting roof bolts, whether controlled at either workstation is
performed at the head or front of the roof bolter.
Additionally, the
operator•s noncontrolling hand guides the drill and bolts into place.
Therefore an operator working at one workstation would switch the
functions of each hand at the other workstation.
tical principles of arranging the controls:
right arrangement.
There are two prac-
mirror-image and left-
The past research that has been done regarding the
arrangement of controls has questionnable applicability to the present
investigation (for example, Chapanis & Lindenbaum, 1959; Pigg, 1954;
Ray & Day, 1979).
The second objective of the experiment was to determine whether
coding the control knobs by shape would enhance their discriminability
and the learning and performance of the roof bolting task.
Previous
investigators have shown that shape-coded controls do enhance discriminability and have implied that they therefore improve performance
(Bradley, 1959; Hunt & Craig, 1954; Jenkins, 1947).
There have been
few investigations of the use of shape-coded controls under actual
working conditions (an exception is Carter, 1978) and no studies performed in underground mining settings.
The impetus for studying shape-coding and arrangement of controls
was the number of serious accidents involving roof bolter operators
and their helpers.
Although it remains unknown how many of these
44
45
accidents result from operators confusing controls, it has been suggested by Helander et al. (1980) that a human factors design of controls would be the first step to enhance safety.
The use of mirror-image or left-right arrangement has been recognized by the mining industry as a problem (Halopoff, March 1981) and
as yet, it remains unresolved.
As a consequence there are no stan-
dards, and the manufacturers of roof bolters construct their machines
in different ways which further magnifies the problem.
has been considered but never tried on roof bolters.
Shape-coding
Many machine
operators, however, code individual controls by applying layers of
electrical tape around certain controls, welding extensions onto controls, or bending the controls in different directions (Helander et
al., l9HO).
The present study was designed to investigate both shape-coding
and control arrangements with a high fidelity roof bolting task.
It
was felt that the similarity of the experimental task and equipment to
actual roof bolting conditions would increase its applicability, although not necessarily its generalizability to other situations such
as control arrangements in cockpits or nuclear control rooms.
Subjects learned a highly realistic roof bolting task using a
simulated roof bolter in a simulated underground coal mine.
Shape-
coding and arrangement of controls were manipulated in a 2 X 2 X 8
mixed design and time to perform the task and number of errors were
the dependent variables.
It was hypothesized that both mirror-imaging
and shape-coding would improve performance.
In general, the data collected were highly variable.
The large
amount of variability in the error data may be a reflection of the
46
subject's ability to perform the task or understand instructions or of
his motivation.
If this is true of the subjects used in the experi-
ment, one might wonder whether it is also true of actual roof bolter
operators.
Of course, miners receive greater extrinsic, and possibly
intrinsic, rewards for their work but it is likely that some grasp the
task and/or instructions better than others.
It is thus very likely
that the highly variable performance seen in the experimental situation can also be found among roof bolter operators.
This is further
substantiated by the limited selection and training procedures found
in most coal mines.
Equipment operators often substitute for one
another on machines that they have had little training and experience
with.
The amount of time necessary to complete one drilling and bolting
cycle was also highly variable.
The reasons for this variability were
more obvious, however, than for the error data.
The time measure
seemed to reflect such extraneous factors as eye-hand coordination in
placing a bolt into the bolt extension, ability to perform a secondary
task (that is, preparing the drill and bolt while positioning the
boom), and differences in the degree of automaticity achieved by subjects.
Some subjects appeared to have the sequence so well planned
and timed that they smoothly and rapidly activated the controls in the
correct order.
Others paused between control activations as if they
needed time to calculate the next move.
Generally the slow subjects
were clumsy throughout the task, dropping bolts, showing unsureness
with each activation, and were slow to recognize and react to their
errors.
47
The main problem with the amount of variability in the data is
that it concealed differences, if any existed, between the shape/no
shape-code and mirror-image/left-right arrangement conditions.
The
only variable that had a strong effect, despite the variability, was
practice over the eight trials.
(This is discussed in detail below.)
With respect to the error data, three of the four classifications
of errors, direction, sequence, and omission were not directly relevant to the study.
They were analyzed only as indicators of task
knowledge.
Substitution errors were the most important type of error.
Errors
of substitution refer to confusions between controls or inadvertant
activation of a control.
This type of error is relevant to both the
questions of control arrangement and shape-coding.
Arrangement of Controls
The hypothesis that subjects would expect a mirror-image arrangement of controls on the transfer task was not supported by the results.
Arrangement did, however, almost reach significance for errors
of omission (e.< .10) and in the analysis of covariance of substitution errors (£
< .06). In both of these cases, left-right arrangement
obtained a slight advantage.
The lack of a tru·ly significant effect may have been due to the
high amount of variability in the data or it possibly indicates that
there is no true expectation of how the controls should be arranged.
All subjects, regardless of the condition they were in made substantially more errors on the first transfer cycle.
It was as if the
transfer task was an entirely new task for subjects to learn and after
the first cycle, the relationships between controls that had been pre-
48
viously learned once again became useful.
tion of Figure 7.
This can be seen by inspec-
Regardless of condition, subjects made a large num-
ber of substitution errors on the first transfer trial (Tl).
The
error rates then dropped back and stabilized around the pre~transfer
rate (L4).
In addition to analyzing the number of substitution errors, two
analyses were performed to investigate the question of expectation.
The first, an analysis of symmetrical and nonsymmetrical substitution
errors, showed no differences by group, indicating that the subjects•
expectations did not manifest themselves in performance.
There were,
however, significantly more nonsymmetrical errors overall.
This may
mean that the control confusions were due to other factors rather than
an expectation.
A second analysis was based on subjects• reported expectations
(after the experiment).
Even though there was a general tendency to
report left-right, there were no differences in performance between
those who reported mirror-image or left-right.
Shape-Coding of Controls
It was hYpothesized that shape-coding would reduce the number of
errors both in the learning and transfer phases of the experiment and
further, that they would reduce the time needed to perform the task.
The results, however, did not support this hypothesis.
Coding
approached significance for sequence errors (£ < .10) showing some
advantage for shape-coding and a trial by code interaction approached
significance (£ < .10) in the analysis of substitution errors (see
Figure 8).
This indicates that subjects who used shape-coded controls
had a differential learning advantage across trials.
49
As in the case of arrangement of controls, the absence of a large
significant effect of shape-coding may be the result of the highly
variable data or it mqy truly indicate no advantages of shape-coding
in this situation.
Inspection of Figures 7 and 8 shows that the first
transfer trial is the most difficult, regardless of condition, and
that the error rate decreased and stabilized at the second transfer
trial which is equivalent to the pre-transfer rate.
Trial
The within subjects variable of trial was the only one to show
highly significant differences in all of the analyses.
The number of
direction, sequence, and substitution errors significantly decreased
across trials.
Omission errors did not decrease but there were so few
errors of this type that the data were not analyzable.
Additionally,
the only control that could have been omitted was the stab jack and
subjects who had trouble remembering to use it continued to do so
throughout the experiment.
The fact that trial was significant in the analysis of substitution errors has implications for the training of roof bolter operators, especially in light of the non-significant effects of arrangement and coding of controls.
As mentioned previously, the first
transfer trial seemed to give subjects the most trouble after which,
errors rates dropped back to the pre-transfer rates.
This indicates
that training and sufficient practice before actually beginning to
drill and bolt the roof of the mine might be indicated.
Possibly
mandating a 11 dry run 11 or increasing the training programs for roof
bolter operators would accomplish the necessary practice.
Training
programs need to be provided on an individual mine, and individual
50
? '
machine basis, however, because of the many different types of machines available.
This places the responsibility in the hands of mine
owners, mine training personnel, and roof bolter manufacturers.
Trial was also significant in the analysis of time to complete the
task.
According to the Newman-Keuls test on ordered means, the first
learning and first transfer trials took significantly more time to
perform than all other trials (the only exception was between transfer
trial 1 and learning trial 2 where the difference did not reach statistical significance).
The effect of trial on the time data also has implications for the
training of roof bolter operators.
productivity will increase.
If operators can work faster,
The amount of positive transfer from day
to day practice over months of working on one or both sides of the
roof bolter remains unclear and should be investigated, however.
Recommendations
Given the high amount of variability in the data and considering
the practical significance of the reduction of even one error which
might prevent one serious accident or fatality, more consideration
should possibly be given to the variables that approached significance.
It also seems wise to mandate some standard for control
arrangements on dual-boom roof bolters.
Then, at least, all operators
could be trained in the same way and problems associated with working
on different types of machines would be eliminated.
A recommendation is made for the use of shape-coded controls on
roof bolters.
They did help somewhat in both the learning and trans-
fer phases of the experiment.
The use of different sizes of control
knobs, as well as shapes may offer even more advantages although fur-
51
ther experimentation would be necessary before making this
reco~end­
ation.
One of the main disadvantages with the use of shape-coded controls
is the problem of replacement of lost or broken knobs.
Machine opera-
tors would have to assume the responsibility for reporting the need
for new knobs and maintenance personnel would have to assume responsibility for keeping a supply in stock and replacing them when necessary.
With a little extra cost, the lever and control knob could be
cast and molded as a single unit.
This would prevent knobs from com-
ing loose and being lost or stolen.
If one of these units should
break, however, it could result in substantial and costly down-time of
the machine and operator(s) until the control was replaced.
The question of control arrangement is more difficult to analyze
and answer.
There are advantages and disadvantages to both mirror-
imaging and left-right arrangement.
The results of the experiment
indicate a slight advantage in terms of errors for a left-right
arrangement; however, there are practical advantages in placing the
boom raise and drilling controls closest to the drill on both sides of
the machine.
For the common practice of hands-on drilling, in which
the operator holds the drill steel, mirror-imaged controls would minimize the distance he must reach and thus reduce the risk of him inadvertently activating other controls.
Considering the fact that there was no clear expectation for mirror-imaging or left-right arrangement, a recommendation for one or the
other seems premature.
Whether subjects were in the mirror-image or
left-right condition, they formed the correct associations between
controls once they figured out the basic arrangement.
Perhaps this
52
means that there really is no stereotype and that subjects (and
miners) can learn to perform equally well with either arrangement.
Research in this area should continue, however, since a standard needs
to be set by the industry.
Perhaps the research should take the form
of weighing the costs and advantages/disadvantages of each arrangement
rather than trying to determine population expectations.
Replication and Further Research
If a replication of this experiment were to be made, several methodology changes should be considered:
1.
Increase the number of training trials to allow subjects to
reach asymptotic performance.
In the present experiment, sub-
jects were placed in a fairly high pressure task situation in
which they were required to learn the location of controls in
a relatively short amount of time while being evaluated by the
experimenters.
The minimal training afforded subjects did not
nearly approximate the long term practice that leads to the
automaticity in control activations observed with actual roof
bolter operators.
Most subjects reported after the experiment
that the easiest and fastest way for them to learn the control
locations was to memorize them left to right.
Therefore it is
not actually known whether there are population stereotypes of
how controls should be arranged.
2.
Construct a roof bolter simulator that has two visible control
panels rather than one that is moved to the other side of the
machine.
Few of the subjects in the experiment assumed that
the experimenter had the capability or the time to change the
controls to a mirror-image arrangement for the transfer phase.
53
Therefore, they simply expected the same left-right arrangement.
This may have supressed any trend in favor of the mir-
ror-image arrangement.
If two control panels are used, the
.
initial instructions to subjects should state that they will
be using both panels during the experiment.
With this know-
ledge, either conscious or unconscious, their performance
would be more revealing and applicable to actual roof bolting.
Additional research is also recommended using shape-coded control
knobs that are grossly different in terms of size and height as well
as shape.
It is felt that this would greatly enhance discriminability
but the question of whether the controls would improve performance
needs to be addressed.
References
Bradley, J.V. Tactual coding of cylindrical knobs (Tech. Rep. No. 59182). Wright-Patterson Air Force Base, OH:
Wright Air Development
Center, Air Research and Development Command, September 1959.
Bray,
c.w.
1928,
Transfer of learning. Journal of Experimental Psychology,
~'
443-467.
Brebner, J. & Sandow, B. Direction-of-turn stereotypes - conflict and
concord. Applied Ergonomics, 1976, 1(1),34-36.
Bruning, J.L. & Kintz, B.L.
Glenview, IL:
Computational handbook of statistics.
Scott, Foresman & Co., 1968.
Chapanis, A. & Lindenbaum, L.E. A reaction time study of four controldisplay linkages. Human Factors, 1959, l(l), 1-7.
Carter, R.C. Knobology underwater. Human Factors, 1978, 20 (6), 641647.
Cook, T.W. Studies in cross-education. Psychological Review, 1936, 43,
149-178.
Dixon, W.J. & Brown, M.B. (Eds.) Biomedical computer programs Pseries.
Berkeley:
University of California Press, 1979.
Fitts, P.M. &Jones, R.E. Analysis of factors contributing to 460
"pilot-error" experiences in operating aircraft controls
(Memorandum Rep. TSEAA-694-12). Wright-Patterson Air Force Base,
OH:
Halopoff,
Aero Medical Laboratory, Air Materiel Command, July 1947.
w.
(Chairman of Working Group 4, SAE Subcommittee 29).
Personal communication, March 1981.
54
55
Helander, M., Conway, E.J., Elliot,
w.w.,
& Curtin, R. Standardization
of controls for roof bolter machines. Phase I:
Human factors/en-
gineering analysis (Tech. Rep. No. 10-3072). Westlake Village, CA:
Canyon Research Group, Inc., September 1980.
Hunt, D.P. & Craig, D.R. The relative discriminability of thirty-one
differently shaped knobs (Tech. Rep. No. 54-108). Wright-Patterson
Air Force Base, OH:
Wright Air Development Center, Air Research
and Development Command, December 1954.
Jenkins, W.O. Tactual discrimination of shapes for coding aircrafttype controls. In P.M. Fitts (Ed.), Psychological research on
equipment design. Washington D.C.:
U.S. Government Printing
Office, 1947.
McCormick, E.J. & Sanders, M.S.
Human factors in engineering and
design (Fifth Edition). New York:
McGraw-Hill, 1982.
McGeoch, J.A. The psychology of human learning (A.L. Irion, Ed.). New
York:
Longmans, Green and Co., 1952.
MIL-HDBK-759 (Proposed). Human factors engineering design for army
materiel. Military Standardization Handbook, 9 January 1981.
Miller, W.K. & Mclellan, R.R. Analysis of disabling injuries related
to roof bolting in underground bituminous coal mines-1973 (Information Rep. 1017). Denver:
Health and Safety Analysis Center,
1975.
Pigg, L.D. Orientation of controls in bilateral transfer of training
(Tech. Rep. No. 54-376). Wright-Patterson Air Force Base, OH:
Air
Development Center, July 1954.
Ray, R.D. & Day, W.D. An analysis of domestic cooker control design.
Ergonomics, 1979,
~(11),
1243-1248.
56
I
Stevens,
s.s.
(Ed.) Handbook of experimental psychology. New York:
Wiley, 1951.
Swain, A.D. & Guttman, H.E. Handbook of human reliability analysis
with emphasis on nuclear power plant applications. Albuquerque,
NM:
Sandia Laboratories, October 1980.
VanCott, H.P. & Kinkade, R.G. (Eds.)
Human engineering guide to
equipment design (Revised edition).
Washington D.C.:
U.S.
Government Printing Office, 1972.
Volkmann, A.W. Uber den Einfluss der Uebung auf das Erkennen
raumlicher Distanzen. Sachs. Akad. Wiss. Ber. Leipzig, 1858, lQ,
38-69. (Taken from Woodworth, R.S. Experimental psychology. New
York:
Holt, 1938.)
Welford, A.T. Fundamentals of skill. London:
Methuen, 1968.
Woodson, W.E. &Conover, D.W. Human engineering guide for equipment
designers (Second Edition). Berkeley:
Press,
University of California
1964~
Woodworth, R.S. Experimental psychology. New York:
Holt, 1938.
'
APPENDIX A
Description of Mine Simulator
57
58
Q '
The mine simulator, constructed in Canoga Park, California was designed to recreate with high fidelity, a low seam coal mining environment.
In achieving such fidelity, consideration was paid to both vis-
ual features and task conditions found in an actual mine environment.
The general configuration of the mine consisted of an 8 ft. by 36
ft. main .. tunnel ...
While one wall of this tunnel was uniformly
straight, simulating a haulage way
11
rib, 11 the other wall contained
three 6 ft. by 8 ft. alcove areas separated by two 3 ft. wide observation chutes.
The general layout and location of the roof bolting
task are shown in Figure 10.
Environmental fidelity features included such structural elements
as roof, floor, sidewalls, and timbers, and ambient features such as
illumination, temperature, and ventilation.
The structural features
were designed to incorporate high visual fidelity with design variability.
The roof consisted of nine 4 ft. by 8 ft. plywood panels and three
6 ft. by 8 ft. panels in the alcoves.
panels were interchangable.
Within each of these sets, the
Each panel was designed to vary ±6 in.
from the average roof height.
Features such as roof bolts, cross beam
timbers, and rock outcroppings were simulated.
designed to be set at two average heights:
The entire roof was
36 in. and 48 in., but
could be, using post inserts, raised to a considerably higher level
without difficulty.
The floor of the simulator, like the roof, was plywood and included dips, bumps, and inclines.
Additionally, irregular wooden
panels were randomly positioned to simulate the irregular floor conditions found in actual mines.
Interchangeable
roof panels
T
8ft.
1
-1
I I~
I
I
I
~
I I I
I I ~)ti~er I I
0
Roof bolter
o
I I
I I I
I
I
I
6 ft.
1
Exhaust-\
fan
Entrance
o
I -+Air
conditioner
I
I
I
36 ft.
Figure 10. Top view of underground coal mine simulator.
tJ1
1.0
60
The walls or ribs of the simulated mine consisted of plywood backing upon which wood blocks of various shapes were randomly placed.
Over these blocks, chicken wire was stretched and a stucco compound
spread.
The result was a highly irregular and visually realistic rib
construction.
The walls tended to attenuate sound within the simula-
tor, thus to some unknown extent simulating a sonic environment.
The ambient features of the simulator, although more difficult to
reproduce accurately, had fair fidelity to actual mine conditions.
The housing unit in which the simulator was located was a masonry
brick structure with a 20 ft. roof and two entrances:
and a garage-type roll-up door.
a regular door
Both entrances were covered to allow
as little outside light as possible to enter the simulator.
Illumin-
ation in the simulator was provided only by cap lamps.
Ventilation was provided by a draw-fan located at the rear of the
s.i mula tor.
The fan drew a draft through the main tunne 1.
Temperature
control was achieved through an air conditioning unit installed at the
front of the simulator.
mately 65-75 degrees.
The temperature was maintained at approxi-
APPENDIX B
Subject Instructions
61
62
Canyon Research Group has been contracted by the Bureau of Mines
to determine the best arrangement of controls on underground mining
machines, in particular roof bolters.
A roof bolter is used to drill
holes and insert long bolts into the roof of underground coai mines
after the coal has been extracted.
The bolts hold the layers of rock
together and help prevent the roof from falling.
Roof bolters vary in size depending on the size of the mine.
Large roof bolters generally have two booms or metal extensions that
are moveable in three directions:
up and down.
forward and aft, side-to-side, and
The drill or bolt is held on at the end of the boom.
Also on the boom are five to ten lever controls that control boom positioning and the drilling functions.
manufacturer.
The actual number depends on the
Currently there are no standards as to how these con-
trols should best be arranged on each boom to make them easy to operate.
Since the introduction of roof bolters into the mining industry,
there have been many serious accidents due to operators confusing controls.
To determine how the controls should be arranged, we constructed a
simulated roof bolter that has the same basic functions as an actual
machine.
You will learn to operate our 11 mini 11 roof bolter by posi-
tioning the boom, drilling holes, and inserting small bolts into the
roof of the mine simulator.
Do you have any questions so far?
Once we get inside the simulator, I will explain each of the control functions and demonstrate each lever.
I will explain the dril-
ling and bolting sequence and then let you practice by drilling four
holes and inserting a bolt into each while I guide you and answer any
questions you have.
After this first practice cycle, you will perform
63
Q '
several cycles of four bolts each while we time you and record control
activation errors.
I would like you to work as quickly and as accur-
ately as possible.
Your task is to learn the control locations.
Do
you have any questions?
ONCE INSIDE THE MINE SIMULATOR, THE BOOM, CHUCK, DRILL STEEL, BOLT
EXTENSION, AND BOLTS WERE SHOWN AND EACH OF THE FOLLOWING CONTROLS WAS
DEMONSTRATED AND DESCRIBED:
SUMP - moves the boom forward and aft - push to go forward, pull
to come back
SLIDE - moves the boom from side to side - push to move the boom
away from you and pull to come toward you
STAB JACK- stablizes the machine once the boom has been positioned - push to go down (set the stab jack) and pull to go up
(release)
RAISE - raises the boom - push to go up and pull to come down
DRILL - push to drill and pull to reverse; when drilling, the boom
must be raised at the same time
TO SUBJECTS IN SHAPE-CODE CONDITION:
Notice the differently shaped knobs - these are to help you to distinguish between the controls.
TO ALL SUBJECTS:
You will be drilling and bolting cycles of four bolts each.
quence is (DEMONSTRATE).
power.
The se-
At the beginning of each cycle, turn on the
You will then place the drill steel into the chuck and holding
onto the drill steel, sump the machine all the way forward and all the
way (toward/away from) you to the first bolting location.
Next you
64
set the stab jack.
the wood.
You then raise the boom until the drill touches
Now you drill while you continue to raise the boom.
ate the two controls together.
Oper-
Continue until all but about an inch
of the drill is inserted; then lower the boom and reverse the drill at
the same time - until the drill clears the wood.
Then continue lower-
ing the boom until the top of the boom is about even with the top of
the stab jack.
Next you place the bolt extension and bolt into the
drill chuck and raise the boom until the bolt meets the hole you just
drilled.
bolt.
Then drill and raise the boom at the same time to insert the
Once the bolt is in place, lower the
boom~
to leave the
bolt; continue lowering until the top of the boom is even with the
stab jack.
#2.
Raise the stab jack and sump all the way back to position
Drill a hole and insert a bolt then position the boom for hole
#3, drill and insert a bolt; then go to position #4, drill a hole and
insert a bolt. After the fourth bolt is in place, I want you to put
the boom in the neutral starting position:
all the way and sumped all the way back.
the boom should be lowered
The stab jack should be up
. and the boom positioned in about the middle of the machine - use this
welding mark as an approximate position.
Except for the drill and raise controls, operate all other controls singly.
Do you have any questions?
You can now go through one cycle on your own.
guide you through it.
I will sit here and
After that, you will go through several cycles
on your own while we time you and mark down your errors.
AFTER FOUR BOLTING CYCLES:
Ok, now I'm going to give you a 15-20 minute break.
you and explain the next part of the task.
I'll go out with
65
When you go back into the mine simulator, I want you to perform
the exact same task - but the controls will be located on the opposite
side of the machine.
They may be in the same order you learned or
they may be different.
As I indicated earlier, we are looking at
several control arrangements.
TION:
(TO SUBJECTS IN THE SHAPE-CODE CONDI-
The shapes you had before, however, will always be associated
with the same controls.)
Please don't worry about any errors you make
- I want you to perform the task as you would expect the controls to
be arranged.
Pretend you are working at a different machine somewhere
else in the mine.
Is that clear?
WHEN BACK IN THE MINE SIMULATOR:
As you can see, your (left/right) hand now must hold the drill and
your (right/left) hand must work the controls.
is the same (DEMONSTRATE).
start button.
The drilling sequence
We will begin when you press the black
APPENDIX C
Questionnaire Administered after Experiment
and Summary of Responses
66
67
GENERAL QUESTIONNAIRE
1.
What do you think the purpose of this experiment was?
The. Jte6pon6e6 We.Jte. c.l.a.J.:,~.>-L6-Le.d a.6 11 1Je6 11 , hav-ing
:the. pwz.po~.>e. and "no", ~.>how-Lng no k.now.te.dge..
GJtou.p
Ml.JlJi.otL/Shape.
WM.otL/ No Shape.
Le.6:t-Righ:t/Shape.
Le.fi:t-Righ:t/No Shape.
To:ta.t
2.
Yu
lmow.te.dge. o6
No
7
-r
4
5
5
4
3
TI
IT
3
How did you expect the controls to be arranged when you transferred
to the other side of the machine?
·MhuiotL
5
GJtou.p
MUCJioJL/Shape.
M..ifutotL/No Shape.
Le.fi:t-Righ:t/Shape.
Le.fi:t-Righ:t/No Shape.
To:ta.t
3.
~.>orne.
Von':t Know
Le6:t~R-Lgh:t
2
2
6
0
3
8
5
TO
2T
1
0
0
0
-1
When you transferred to the other side, what controls gave you the
most problems?
ManiJ .6u.bje.et6 Jte6ponde.d 6otL bo:th .te.aJtrU.ng and :t1Lan66e.Jt.
aJte. c.l.a.J.:,~.>-L6-Le.d btj eon:tJto.t.
.G
MJJ/j;_§Jt/Shape.
MbutotL/ No Shape.
Le.fi:t-Righ:t/Shape.
Le.fi:t-Righ:t/No Shape.
To :tal
4.
·su.mp · ·sude. · S:tab Jaek.
-r 3
3
5
5
7
5
4
5
IT
H
Rwe.
6
3
6
IT
2
Rupon6e6
VJUU
0
0
1
1
1
0
0
0
-1
Were there any controls you consistently confused? What were they?
Rupan6e..6 aJte. w:te.d
GJtou.p
MIJCJiotL/Shape.
MbutotL/No Shape.
Le.6:t-Righ:t/Shape.
Le.fi:t-R-igh:t/No Shape.
To:ta.t
a.6
:the. pcUM :tha:t We.Jte. Jte.potL:te.d.ttj eonfiu..6e.d.
Stimp/SUde.
SUde./ S:tab Jaek.
3
2
3
5
T3
0
0
1
I
68
5.
Did the shape-coded control knobs help you?
Did they help visually or tactiley?
If. so, which ones?
Onf..y .6ubjec.:t6 who Med .6ha.pe-eoded eonbz.oRA aY1.6WeJted :thi-6 quution.
RupoY1.6U aJte w:ted .6epaJLa.:tei.y 6oft v-Wu.a.i and :ta.ctil.e 6oft rrWvtOft•
.image and ie6:t-JU.gh:t gftoup.6.
VISUAL
Gftoup
Mbul.oft
Le6:t-1Ugh:t
Ta:ta.i
Sump
-22
Slide
S:ta.b .J ad<.
2
1
2
2
4
1
4
3
Ra.-We
1
0
V!Ull.
0
1
1
TACTILE
Gftoup
MlJlJioft
Le6:t-1Ugh:t
To:ta.i
6.
Sump
----r-
Slide
S:ta.b Jac.k.
·Ra..we
1
2
1
2
0
1
1
0
0
2
1
3
2
V!Ull.
2
What is your profession/major?
RupoY1.6U We.Jte no:t c.i.a..6.6-i..6-i..ed by gftoup .6-i..nc.e :the numbe.Jt o6 .6ubjec.:t6
-i..n ea.eh pfto6u.6-i..on/majoft c.a.:tegofty -W .6o .6mali. RupoY1.6U ;[nc.fuded:
n
Au:to body ftepa.inman
V!Lwnme.Jt
Human 6ac.:toM pfto6u.6-i..ona..t
Aeeounting/BM-i..nu.6
Phy.6-i..e.6/Compute.Jt Se-i..enee
Poi-i..Uea.i Se-i..enc.e
T
~~~
2
EaJt.:th Se-i..enee
Eng-i..nee!Ung
Hea.i:th Se-i..enee
Ntdu.Jr..a.l. RuouJtc.u Mana.gemen:t
1
2
1
1
1
1
1
M~h
8-i..oiogy
Speeeh Commun-i..c.a..tion
7.
1
4
5
5
1
N~on
1
Phy.6-i..ea.i Eduea..tion/Ree~tea.Uon
P.6yehoiogy
Human 6ac.:toM (gJtadua:te)
To:ta.i
2
1
1
3Z
Have you ever worked around heavy equipment?
Ail .6ubjec.:t6 who ftUponded yu :to .:thi-6 quution ftepoJt.:ted woftfU.ng
onf..y wUh 6oftlli6:t6 and/ oft Ugh:t c.oM.tw.W.on eqt.Upmen:t.
Me ffi:ted by expe.Jt.imen:ta.i c.onc:Ut.-i_on.
AYI.6WeM
69
ZfuPoJr./Sha.pe
MUvw~r./No
Shape
Le6~-Right/Sha.pe
Le6~-Right/No Shape
To-tal.
8.
Yu
-22
2
4
TO
No
6
6
6
4
22
What is your dominant hand?
GJr.oup
WJiJioJr./ Shape
Mhvl.oJr./No Shape
Le6:t
---r
3
1
Le6~-Right/Sha.pe
· Le6~-Right/No Shape
ToM.
0
6
Right
Mi.-6.6-<..ng Va:ta
0
1
1
1
6
4
6
7
IT
3