Joumal of Expcrimmtal Rychology:
Humil Pereption and Ferfomme
19E3,Vol. 9, No. 4, 607-621
Copyright 1983by the
Amerien Psychologi€l Asiation, lnc.
Analysis of a PerceptualSkill
Timothy A. Salthouseand Kenneth Prill
University of Missouri-Columbia
Four experimentswereconductedwith a trajectory-intersection(videogame)task
to identify the information-processingmechanismsresponsiblefor performance
differencesassociatedwith initial ability and practice. We concludedthat proficiency differencesassociatedwith initial ability are largely attributable to differencesin the revisionofprocessingoperationsand, to a lesserextent,to differences
in the effectiveness
of some component operations.Practice-relatedproficiency
differences were less associatedwith component revision differences, and there
was no evidencethat the performance improvement causedby practice was acof individual components.
companiedby an increasein the effectiveness
The purposeof the presentstudy is to examine the,natureof skill on a relatively simple perceptualtask. Both the performanceof
elementary component operations and the
pattern of component execution are examined in subjectsdiffering in level of performancein their first encounterwith the task,
and in comparisonsbeforeand after practice
on the task. to determine whether the same
processingcharacteristicsare associatedwith
each type of performanceproficiency.
ExperimentalTask and ProcessModel
A fairly novel experimentaltask-judging
the temporal intersection of two trajectories-served asthe activity in which skill differences were examined. A target moved
along a left-to-right trajectory at a variable
speedand angle,and the subjectwirsto control the initiation time of the vertical trajectory of a projectile with the goal of making
the projectile and target trajectoriesintersect
in spaceat the samepoint in time. This task
is somewhatsimilar to an automobile driver
attempting to judge the gap (target)between
cars that is sufficient to allow one to merge
into the flow of traffic. It also resemblesthe
activity of judging the future position of a
ball (target)in order to catch, kick, or deflect
it, and the activity of leadinga moving target
in attempting to shoot it. Becauseof the
Requestsfor reprints should be sent to Timothy A.
Universityof MisSalthouse,Departmentof Psychology,
souri.Columbia.Missouri6521l.
mode ofpresentation(i.e.,on a video screen),
the current task is probably most similar to
certain video arcade games, and indeed this
resemblancewas maximized with sound effectsand an exotictask name("Photon Phantasy") to capitalizeon the interestsofcollege
studentsin such amusements.
The first step in the current investigation
consistedof the development of an initial
model of how the trajectory-intersection task
is performed in order to provide a basis for
deriving hypothesesabout the factors contributing to superior performance. Several
other studieshave used a similar task (e.9.,
Gerhard, 1959; Gottsdanker & Edwards,
1957;Runeson,1975),but the major finding,
that performance was better with increased
time to observethe trajectories,wasnot very
helpful in characterizing the specific processesinvolved in the task., The current
model,generatedfrom intuitive speculations,
is expressedin the flow chart of Figure l.t
The initial component in the model involvesdetermining the position of an imaginary line extendingabovethe launching ap
paratus where the target is vulnerable for intersection (the vertical vulnerability area;
VVA). The second component consists of
estimating the time that the target will arrive
in the VVA above the launching apparatus
I The componentsin this diagram,aswith most such
abstract organizational charts, refer to functionally, but
not necessarilyphysically, distinct processes,and the sequential arrangement portrayed suggestsa typical, but
not necessarilythe only, sequenceofcomponents.
607
608
TIMCIHY A. SALTHOUSE AND KENNETH PRILL
Modelof Processesin
TrajectoryIntersectionTask
Determinelaunchpositionand
verlicalvulnerabilityarea (WA)
Sample target angle and velocity and compute
estimated time of arrival (ETA) to VVA
Computecriticaltarget position(CTP)in WA
(lntersectionol horizontaltargettrajectory
with verlicallaunchtrajectory)
Computelaunchlag time (LLT)between
launchoositionand CTP
ls ETA > (LLT + FeactionTime)?
Figure I.Initial modelFffthe major componentsinvolved in the task of initiating a vertical trajectory to
intercepta target moving in a horizontal trajectory.
(the estimatedtime of arrival; ETA). We assumedthat both the velocityor speedand the
angle or direction of the target are used to
maketheseETAsbecauseboth variablescontribute to the actual time. Next, on the basis
of information about the initial position and
angle of the trajectory the subject is postulated to determine the location in the VVA
at which the target will intersectthe launch
trajectory (the critical target posilion;CTP).
Another component involves estimatingthe
time required by the launched projectile to
traversethe distancefrom the launching position to the CTP (the launch lag time; LlJl).
The final component in the model is a
decisionphasein which the subjectevaluates
whetherthe ETA of the target is equal to the
sum of the LLT and one'sown reaction time
to push the fire button. If the two times
match, the subjectpressesthe button, and if
the arrival time exceedsthe sum of the LLT
and reaction time, the subject waits. Two
strategiesare possibleif the subject reaches
the decisionstagewith ETA lessthan the sum
of the LLT and reaction time: The subject
could simply withhold the response,or the
fire button could be pressedin what is recognizedas a probabl-vfutile effort.
Althoughwe do not claim that this model
accuratelyand completelydescribeshow subjects actually perform the trajectory-intersection task, it doeshave heuristic value in suggestingseveralpossibilitiesfor the nature of
skill differencesin this task. For example,
becausethe accuracyof the final decisionis
dependenton the accuracy of each of the
prior components,skilled individualsmight
perform better than unskilled ones simply
becausethey are more effective or efficient
in the performanceof one or more of the
components.This hypothesisis exploredin
Experimentsl, 3, and 4 by assessing
discrimination accuracy of vertical trajectories
(Component A), left-to-right trajectories
(ComponentC), ETAs (ComponentB), and
LLTs (ComponentD).
Another possible re:$on for proficiency
is that individualsof differingskill
differences
levels might use alternative sequencesof
components.As an example,although the
flow chart-ofFigure I indicatesthat subjects
merely wait (i.e., cycle back only to the decision component), if the arrival time is
greater than the sum of LLT and reaction
time, it is conceivablethat somesubjectsuse
thiSadditionaltime to revisetheir initial trajectory and time estimatesby cycling back
to earlier componentsin the sequence.This
hypothesisis explored by examining task
performance (hit percentage)as a function
ofthe time the targetpath wasdisplayedprior
to the launch location.Ifsubjectsdo not revise their initial estimates,the function relating hit percentageto path observationtime
should be relatively flat becausethe time beyond that neededfor the initial estimate is
merely spent waiting. On the other hand, if
subjectsdo revisetheir estimatesby cycling
back to earlier components,the accuracyof
their estimatesis likely to improve with more
opportunities for revision.This would result
in hit percentageincreasingas a function of
path observationtime.
An interesting question, therefore, is
whether the performance difference across
individualsof rarv
ficiencyis evidenrl
or the initial ter,cl
function relating hi
servailontime. A C
skilled subjecr ino
greater path obnerr:
cate a different pat
ponentsbetweenski
.;ects.An interceprdr
suggestthat the skill
attributabte to differ
reuslons, and com
nlsmswould haveto
lor the performanr
skilled and unskilled
Categm
Although skiil bre
on specificactivity
.a
tually distinct €te8pl
trfied (cf. Nobte. liTt
mediatedskill or erg
comparisonsare berr
parable)individuals
b
amountsof exffieno
rty. However,even at
periencethere are oftr
terencesin task profici
caregoryof skill can b
rhe tevelof initial atili
parisonsare between
IgTmtlg near the lop r
rnDutronand thoseir
near the bottom of ilr
tlon aftercomparablea
on the task.A third nr
ent skill levelsmighr br
respectto demographi
rarn segmentsof tlre
I
young children or old
Known to be deficient
groups (e.g., young
ad
perceptual,cognitirc.
a.
pectedto be relerranl
ro
task of interest,and thu
pect task-proficiencr.
dif
to thesedemographicct
r he examinationof :
egoriesof individuet
<
comparison of the infi
mechanismsassociated
PERCEPTUAL SKILL
individuals of varying levels of overall proficiency is evident in the rate ofgain (slope)
or the initial level (x-axis intercept) of the
function relating hit percentageto path observationtime. A slopedifference,with only
skilled subjectsincreasingin accuracy with
greaterpath observationtime, would implicate a different pattern of processingcomponentsbetweenskilled and lessskilled subjects.An interceptdifference,however,would
suggestthat the skill variationsare not simply
attributable to different amounts of estimate
revisions, and consequently,other mechanismswould haveto be postulatedto account
for the performance differences between
skilled and unskilled individuals.
Categoriesof Skill
609
skill. Just as it rnay be unlikely that a given
categoryofskill can be explainedby a single
information-processingmechanism, so also
might it be unreasonableto expect that all
skill categoriescould be explained by the
samecombination of mechanisms.
Two of the three skill categoriesdescribed
here were investigatedin the following manner: Experiments l, 2, and 3 contrastedindividuals in the upper and lower quartiles of
the testedpopulation; and Experiment 4 investigatedthe effectsoffairly extensivepractice in a sample of young adults. An additional experiment, which was conducted to
examine the performanceof young and old
adultsat moderatelevelsof experience,is also
briefly discussed.Experimentsl, 2, and 3 are
described together becausethey were formally similar and differedonly in the specific
analytical tests administered after the standard trajectory-intersectiontask. The tests
were designedeither to assessthe effectivenessof component operation or to explore
the between-taskgenerality of the observed
skill differences.The various tests were designedto be assimilar aspossibleto the standard trajectory-intersectiontask with the
minimum number of modificationsnecessary for carrying out the relevant manipulation.
Although skill broadly refersto proficiency
on a specificactivity, at least three conceptually distinct categoriesof skill can be identified (cf. Noble, 1978).The first is practicemediatedskill or expertisein which the skill
comparisonsare betweenthe same(or comparable)individuals before and after varying
amounts of experienceperforming the activity. However,even at the same level of experiencethere are often large individual differencesin task proficiencyand thus another
categoryofskill can be establishedbasedon
the level of initial ability. Here the skill comExperiments1,2, and 3
parisons are betweenthose individuals perMethod
forming near the top of the population distribution and those individuals performing Subiects
near the bottom of the population distribubetween the ages
tion after comparableamounts of experience A total of96 college undergraduates participated
in a
17 and 30 (32 in each experiment)
on the task. A third manner in which differ- of
single 45-min. session.
ent skill levelsmight be distinguishedis with
respectto demographiccharacteristics.Certain segmentsof the population (e.g.,very Apparatus
young children or older adults) might be
Stimuli were displayed on a Hewlett.Packard Model
known to be deficient to other population l3l lA Display Monitor controlled by a PDP I l/03 labwere
groups (e.g.,young adults) in a variety of oratory computer. Two l0-key telephone keyboards
the computer to register responses.
perceptual,cognitive,and motor abilitiessus- also connected to
pectedto be relevant to performanceon the
task of interest,and thus one might also ex- Procedure
pect task-proficiencydifferencesto be related
The standardtrajectory-intersectiontask involvedthe
(a 1.0" X 2.0" rhombus)movingin a linearpath
target
characteristics.
demographic
to these
to right acrossthe screen.On different trials,
from
left
The examination of severaldifferent catthe initial position of the target varied from bottom to
egories of individual differencesallows a top alongthe left verticalaxisofthe screen,the trajectory
comparison of the information-processing anglevariedfrom - I 3.0oto 34.0" relativeto horizontal'
mechanismsassociatedwith each type of and the trajectory speedvaried from 22.5" to 45.0" per
610
TIMOTHY A. SALTHOUSE AND KENNETH PRILL
sec. The position of the launching apparatus along the
bottom horizontal axis of the screen also varied from
trial to trial, but the speed of the projectile remained
constant at 60" per sec.Across trials, the time from the
initiation ofthe left-to-righttrajectoryto the VVA ranged
from 220to 1,090msec,and the LLT to reachthe critical
target position ranged from 0 to 420 msec.2Different
sound effectswere associatedwith the target motion' the
projectile motion, and the eiplosion createdby the simultaneousinters€ctionof targetand projectile.
Five blocks of the trajectory-intersectiontask, each
consistingof 50 trials, were administered as the first task
to subjectsin Experimentsl, 2, and 3. Subsequenttasks
differed acrossthe three experiments,but were presented
in the same counterbalanced order for all subjects. In
Experiment l, yes/notestsof vertical trajectory left-toright trajectory and LLT were administered along with
a modified (blanked trajectory) trajectory-intersection
task.Subjectsin Experiment2 receiveda choicereaction
tim€ task and three modified trajectory-intersection tasks
(constant target, moving launch, and detonation).
Method-of-adjustmentprocedureswere used in Experiment 3 to nssessac\curacyof left-to-right and verticaltrajectory alignment, and precision of target- and projectile-time
estimation.
Vertical-trajectory discrimination (yes/no procedure).
Trials in this task consistedofa displayofthe launching
apparatus and the projectile at various distances above
the launch site. On half of the randomly selectedtrials
the projectile was directly above the launch site (i.e.,
verticallyaligned),and on the other half of the trials the
projectilewasdisplayed.6o to the left or right of vertical
alignment.The subjectwas instructedto pressa key on
vertically
the right keyboard when the projectile was
"yes") and to
alignedwith the launchingapparatus(i.e.,
pressa key on the left keyboard when the projectile and
"no"). Two
launchapparatuswereout ofalignment (i.e',
with
a rangeof
trial blocks,each consistingof 50 trials
projectileJaunchsite distances,were administered.
Left-to-right-tmjectory discrimination (WS/no proce'
'lhe
display in this task consistedof a 500-msec
dure).
presentation of the first zoqoof lhe target trajectory a
500-msecblank period, and finally a 500-msecpresentation of a singledot directly abovethe launching ap'
paratus.On half of the randomly selectedtrials the dot
wasan extrapolationofthe targettrajectory and on the
other halfofthe trials the dot wasdisplacedl.2o above
or below the true intersectionpoint. The subject was
the
instructed to pressa key on the right keyboard when
"yes")'
dot was aligned with the initial trajectory (i.e.'
and to pressa key on the left keyboard when the dot and
"no"). Two
initial trajectory wereout of alignment(i.e.,
trial blocks,eachconsistingof 50 trials with a rangeof
trajectory-dot distances,were administered.
LLT discrimination (yes/no procedure). In the standard trajectory-intersection task an auditory signal was
presentedfor the entire duration that the projectile rras
in motion on the display.The presenttask capitalized
on the correlationbetweensoundduration and projectile
distance by presenting, for 500 msec, a display of the
projectile displacedabovethe launch site by a randomly
varied distance,followed immediately by a sound. On
a randomlyselectedhalfofthe trials, the duration ofthe
sound correspondedto the time it would havetaken the
projectileto traversethat distance,and on the other half
ofthe trials the sound duration was approximately 125
msec too long or too short for the distance displayed.
The subject was instructed to press a key on the right
keyboard when the sound duration and projectile distance matched, and to press a key on the left keyboard
when they did not match. Two trial blocks' each consisting of 50 trials with a range of launch-projectile dis
tances,were administeted.
tasks. All four of the tasksusMethod-of-adjustment
ing the method of adjustment followed the samegeneral
procedure. First, the stimulus configuration was dilplayed and the subject could either increase or decrease
the value of the relevant parameter.The stimulus could
then be displayed and the prior steps repeated as frequently as desired,and finatly a decisionwasregistered.
The '7" and *9" keys on the riSht keyboard decreased
and increasedthe parameter value, respectively,and the
"8" key causedthe alteredstimulus configurationto be
displayed.Any key on the left keyboard resulted in the
registration of that parameter setting and initiated the
subsequenttrial. Each task involved 30 trials with a range
of (where relevant) heighs, angles, and speedsof the
target,and horizontal positionsofthe launch apparatus.
Be"ausethe four tasks were designedto assessthe efficiency ofthe componentsproposedin Figure l, they
were termed the VVA, ETA, CTB and LLT tests. In the
VVA test, the stimulus configurationwas the launching
apparatus and a single dot, which was to be adusted in
the horizontal dimension to be in alignment with the
launchingapparatus.In the CTP test, the stimulus configuration was 150 msec of th€ targot trajectory and-a
single dot, which was to be adjusted in the vertical dimensionto b€ in alignmenl with the initial trajeclory.
The ETA and LLT testsinvolved adjustmentsof the
time delaybetweentwo stimulus events.The fint stimulus event in the ETA test was the initial 150 msec of
the targpt trajectory, and the secondevent was a vertical
line displaced across trials at rious distances to the
right of the trajectory. In the LLT test, the first stimulus
eventwasthe displayofthe launchingapparatusand the
secondevent was a horizontal line displaced acrosstrials
at various distancesabove the launch site. In both tests
the subject adjusted the duration betu€en the offset of
the first stimulus evgnt and the onrt ofthe secondevent
to correspond to the time required by the target (ETA
test) or the projectile (LLT test) to traversethe indicated
distance.
Choicereaclion time. A choicereaction time task was
administered to determine whether performance on such
an elementary task would be correlated with performance on the more complicated trajectory-intersection
activity. Either an X or an O wasdisplayedafter a fixation
stimulusconsistingof four dots at the cornersof a 1.2" Y^
1.8" rectangle.Subjectswere instructedto pressthe left
key for X and the right key for O. A minimum of 75
triats, the first 25 of which were practice, composed a
trial block. The mean reaction time for the last 50 trials
with an accuracyof 9O%or greaterservedasthe primary
dependentvariable.
2 This combinationof times resultedin sometrials in
which it wasphysically impossible to achievea trajectory
intersection without pressing the fire button before the
app€aftrnceof the target.
Blankcd tralercn .
trajectory_inrcrro
!11d
zt
/ msiec(approxinf
was displayed.Thc rarl
and the subjecr stil ri
ro mtersectthe target
h
on an exfadatim
bcr
rne target r4iectory q
blocks of 50 trials ;ch
constant targa. Th
rraJectory_int€rsectiootr
verslon except that thc
r
not change from trial ro
and the horizontal pcit
vaned randomly acrq
I
angle remained conir
::.d . per sec,and 0", rtr
tu jlu! each rrcre prc
twovmglaunch. Tirdl
was that the launchi4
|
ten acrossthe screen,al
I
movrng from Ieft to rLt
moment of actiratim L
tra;ectory wasverticd fr;
apparatusat the tirr
d
taunctungapparatusah
task Involved longcr arcr
the_standardtash bur el'
marned unchanged.h,o
were presented.
. Detonalion. This,rr d
Jecrory-lnteNectionus& ir
ttne on the screeningced
o
rrne was describedas a mir
ated at the exact momcnr
d
rhe standard task, $c ddol
becausethe entire WA
br
the button press.Tr|o trid
presented.
Ra
. The high_and low.r
unguishedon the basi!
centagefor Btocks2_5
dard trajectory-inter$
btock of trials was cq
the data were not en l
top quartile of the diJ
each experiment serlt
group, and thosein
lhe
srrtutedthe low_ability
for thehig
rr::n?ces
J3.9Vo
for Exffimenr
2:
InenJ and37.6%for I
l9r Jhelow-abilitygroup
fn_Telt t, t9.O%fq
zU.5To
lbr Exffimenr
abthtydifferences
were
each experimenl:Erp
1
PERCEPTUAL SKILL
i
t
Blanked trajecto42 This task was similar to the stan_
dard trajectory-intersection task except that only the first
217 m*c (approximately 20%) of the target trajectory
was displayed.The target continued on its original patir
and the subject still attempted to launch the projectile
to intersect the target, but the judgments had to bebased
on an extrapolation becauseonly the initial portion of
the target trajectory was actually displayed. Two trial
blocks of 50 trials each were presented.
Constant target. The constant-target version of the
trajectory-intersection task was identical to the standarc
version except that the speedand angle ofthe target did
not change from trial to trial. The height of the target
and the horizontal position of the launching apparatus
varied randomly across trials, but the target speed and
angle remained constant at intermediate values (i.e..
33.8' per sec,and 0o, respectively).3
Two trial blocks of
50 trials each were presented.
Moving launch. The distinguishingfeatureof this task
was that the launching apparatus moved from right to
left acrossthe screen,at I 5o per sec,while the target was
moving from left to right. Launching occurred at the
moment of activation by the key press,and thus the
trajectory wasvertical from the position ofthe launching
apparatus at the time of the key press. Because the
launching apparatus always began at the far right, this
task involved longer averagepath observationtimes than
the standard task, but all other aspectsof the task remained unchanged. Two trial blocks of 50 trials each
were presented.
Detonation. This task differed from the standard trajectory-intersectiontask in havinga continuous vertical
line on the screeninsteadofa launchingapparatus.The
line was describedas a mine field that was to be detonated at the exact moment the target intersectedit. Unlike
the standard task, the detonation task contains no LLT
becausethe entire VVA is simultaneouslyacti ted with
the button press.Two trial blocksof 50 trials eachwere
presented.
Results
The high- and low-ability groupswere dis_
tinguishedon the basisofthe averagehit percentagefor Blocks 2-5 (200tnals) ofthe standard trajectory-intersectiontask. (The first
block of trials was consideredpractice and
the data were not analyzed.)Subjectsin the
top quartile of the distribution of scoresin
each experiment served as the high-ability
group, and thosein the bottom quartile constituted the low-ability group. The mean hit
percentagesfor the high-ability groups were
33,9Vofor Experiment l, 34.9Vofor Experiment 2, and 37.6Vofor Experiment 3; means
for the low-ability groups were 19.4Vo
for Experiment l, l9.0%ofor Experiment 2. and
2O.5Vofor Experiment 3. As expected,the
ability differenceswere highly significant in
each experiment: Experiment l, t(14):
6ll
12.39,p< .001;Experiment2, t(14): 10.55,
p <.001; Experiment 3, (14) : 12.82,
p < .001.
The high-ability groups were predominantly male (8 out of 8 in ExperimentsI and
3, 6 out of 8 in Experiment 2), whereasthe
low-ability groups were predominantly female (7 out of 8 in Experiments I and 2, 6
out of8 in Experiment3). The reportednumbersof minutes per weekspentplaying video
gamesoverthe last6 months were68.1 versus
24.1,t(14) = 2.66,p < .05,for the high_and
low-ability groups in Experiment l, respectively;67.8versus39.0,t(14): l.l9 (ns),for
the comparablegroupsin Experiment ?; and
133.0versus6.8, t(14) : 2.73,p < .05, for
the comparablegroups in Experiment 3. lt
is not clear whether the video game experience was a cause,or merely another consequence,of what are termed initial ability differences.For the purposes of this projecl,
however,initial ability can be operationally
defined as level of performanceon the first
encounter with the current experimental
task.
Performanceon the vertical-trajectory leftto-right-trajectory and LLT discrimination
taskswas expressedas percentagecorrect in
the yes/no decision. These data were analyzed with 2 X 3 (Ability X Task Difficulty)
analysesof variance(nNovns).The task-difficulty variablewascreatedby dividing a relevant distancedimension into equal thirds.
In both the vertical-trajectorydiscrimination
and LLT discrimination tasks the distance
wasthat betweenthe launch site and the projectile position. The distancein the horizontal-trajectory-discrimination task was that
betweenthe end ofthe initial trajectory and
the dot whose alignment was to be judged.
In all cases,we assumedthat the shorter the
relevantdistance,the easierthe task.
The aNove on the data from the verticaltrajectory-discrimination task revealed a
nonsignificanteffectof ability (F < 1.0),but
a significant difficulty effect, F(2, 28\:
I13.06,MS": 29.20,p < .0001,and a sig-
3The target is not stationary
in this task, but the speed
and direction of targetmotion remainedconstantacross
trials insteadofvarying from trial to trial as in the standard task.
612
TIMOTHY A. SALIHOUSE AND KENNETH PRILL
nificant Ability X Difficulty interaction, F(2,
28): 4.35,MS.: 29.20,p < .05.Bonferroni
t testsat eachlevel of difrculty indicatedthat
the high-ability subjectswere more accurate
than the low-ability subjectsonly at the easiest difficulty level: At the easylevel (84.67o
v s . 7 7 . l T o()1, 4 ) : 2 . 1 8 , p< . 0 1 ; a t t h em o d t(14) : -1.20
eratelevel (67.5Vovs. 7l.U%o),
(ns); and at the difficult level (52.37o vs.
t(14) : -.05 (ns).
52.4vo),
eNovn
on the left-to-right-trajectoryThe
discrimination data indicated significant effectsof ability,F(1, l4) : 6.95,MS":70.21,
p < .05; difficulty,F(2,28) : 25.02,MS, =
73.45,p < .0001;and Ability x Diftculty,
F(2,28) : 3.74,MS, : 73.45,P < .05' Bonferroni t tests at each level of difficulty revealedthat ability differencesweresignificant
only at the easiestlevel of difficulty: at the
(14) : 3.06,
easylevel (88.49ovs. 75.3Vo),
p<.01; at the moderatelevel (66.97ovs.
t(14): 2.07 (ns);and at the difficult
58.07o),
t(14) : -.67 (ns).
level(62.8Vovs. 65.6Vo),
No effectswere significant in the eNovn
on the data from the LLT discrimination
task:ability,F(1, 14) : 1.48,MS" = 253.23,
p > .20; difficulty,F(2,28) : 2'91,MS. =
147.86,p > .05; and Ability x Difficulty, F(2,
28): 1.2r,MS" : 147.86,
P>'30. The overall mean level of performancewas 65.17o.
Performance on the method-of-adjustment taskswasexpressedin terms of the constant error (mean error of adjustment) and
the variable error (standarddeviation ofthe
adjustments)for each of three levelsof task
difficulty (close,middle, or far distancesof
the relevantparameter).
The eNova on the constant error for the
ETA task revealedthat the high-ability subjects produced significantly longer time intervals,and greaterelrors, than the low-ability subjects-236 msec vs. 57 msec, F(1,
< .01-but
l4) : 19.30,MS" : 35,438.74,P
neither the task diftculty nor Ability x Difficulty Effects approached significance (p >
.10). Only the task-difficulty effect was significant (p < .05) with the variable error
measure(close : 130 msec, middle : 169
m s e c ,f a r : 1 8 0 m s e c ) ,F ( 2 , l 4 ) : 6 . 3 3 ,
MS":l'118.72'P<.01'
There was a significant ability difference
on the constant error measurefor the LI-T
task with the high-ability subjectsproducing
longertime intervals,but smallererrors,than
the low-ability subjects(-97 msec vs. -293
msec),f'(1, 14) : 11.82,MS.: 38,916.74,
p < .005. The task-difficulty effect was also
significant(close: -153 msec, middle:
-208 msec, far : -223 msec),F(2,28) =
p < .01,but not the
5.84,MS": 3735.97,
Ability x Difficulty interaction (F < 1.0).No
effectsweresignificantin the analysisof variable error (all ps > .10).
The analysesof the measuresfrom the
CTP task indicated that only the difficulty
effectfor the constanterror measurewas significant (close = .30o, middle : 0o, far =
-'41"), F(2,28): 8.66,MS, : 41.00,P <
.005. The main effectof ability and the Ability x Difficulty interaction did not approach
significancewith either constant or variable
error (ps > .30), and the difficulty effectwith
the variable error measurealso fell far short
of significance(p > .30).
The difficulty effect was significant in the
vvA task for both constant error (close =
.17o,middle= .32o,far: .40"),F(2,28):
31.61,MS" = l l8, p < .0001,and variable
error (close: .16o, middle : .24", fat:
.28"\,F(2, 28): l2.l l, MS" : .84,p < .0005.
Neither the main effect of ability nor the
Ability x Difficulty interaction was significant for either variable (ps > .20).
Although the high-ability group had slightly
fasterchoicereaction time (394 msecat 957o
accuracyvs.427 msecat 949oaccuracy),this
difference was not statistically significant,
t(14): l-26,P> -20.
Performanceon each modified trajectoryintersection task was initially assessedwith
a Group (hieh-ability vs. low-ability) x Task
(standardvs. modified) lNovn on overall hit
percentage.The task manipulation in this
analysisis contaminatedwith order and practice effectsbecausethe standard task was always presentedearlier and for more trials
than the modified task, but the interaction
term indicates whether the ability differences
are significantly altered with the modified
task. The Group X Task interaction was significant for the blanked-trajectory.F(1, l4) :
6.72,MS.: 22.16,p < .05,and detonation,
F(1, 14): 8.33,MS": 39.03,P < .05,tasks,
but not for the constant-target, F(1,
14)= 2.33,MS. =
launch,F(t, lCj =
.20, tasks.
The constant_u
tasks both led to ir
rugn_and low_abili
percentagesin the
!t.4% for the higbr
ror the low_abilitys
.01. Mean hit per
launch task were 3
subjects and 27.6%
Jects,,(14) = 4.06.
performance
on
and detonation rert
tne.standardtask,
f
subjects.
Hit 1
?9rrlg
l6.t%o
and 10.3%.n
and low-ability grou
Jecrorytask, (t4) =
mancein the detona
lor the high_abilirvsr
tow-ability subjoas ,
ro summarize th
pearsthat the subjec
tne standard trajecro
arso somewhatmone
perrormrng subjecs
about the alignrneafc
tglnght trajectories ,
difficulty is not roo g
cedure is used. Ttrerr
rable differencesin tt
lnatlng the time roqu
or target to traversea
thouglt high_abitirysr
ouce longer time in&
subJects.
There is a trq
Jects to have faster rer
ablllty subjects.buL p
smattsamplesizeand/r
this differencefailed r
nlncance. The abilitv
to be rather task specif
srons of the trajectory
to much smailer,and ir
rcant, ability ditrerence
. As noted earlier, rrial
.lectory-intersectienrd
ot the taryet trajectofv I
fs:n9n ofthe taunchiq
lnal.
thesevaluesntre
PERCEPTUAL SKILL
6t3
14): 2.3t, MS,:9.05, p > .10,or moving- of time (by dividing distance by speed),and
launch,F(1, 14) = 1.46,MS" = 20.58,p > hit percentagewas then examined as a func.20. tasks.
tion of target path observation time. Six
The constant-targetand moving-launch groupings of path observation times were
tasksboih led to improved performancefor formed on the basis of roughly comparable
high- and low-ability subjects.The mean hit intervalswith a minimum of 15trials in each
percentagesin the constant-target task were category.The mean hit percentagesfor each
4l.4%ofor the high-ability subjectsand,28.8%oobservation time grouping are displayed in
for the low-abilitysubjects,t(14) : 4.65,p < Figure 2.4
.01. Mean hit percentagesin the movingWe conducted 2 X 6 (Initial Ability X
launch task were 39.6Vofor the high-ability Time) ANovAs on the data from each expersubjects and 27.6Vofor the low-ability sub- iment summaized in Figure 2. The same
jects,(t+; : 4.06,p < .01.
pattern of results emerged with each set of
Performance on the blanked-trajectory data as all effectswere statistically significant:
and detonation taskswas reducedrelative to ability, F(1, 14) > 58.2;MS" < 143.4,p <
the standardtask, particularly for the high- .0001;time, F(5, 70\ > 17.6,M,S"< 130.0,
ability subjects.Hit percentageaveragedonly p <..0001;Ability X Time, F'(5,70) > 2.7,
l6.lVo and l0.3%o,respectively,for the high- M S " < 1 3 0 . 0 , p < . 0 5 .
and low-ability groups with the blanked-traThe patternsportrayedin Figure 2, in conjectory task, (14) : 2.16,p < .O5.Perfor- junction with the significantAbility X Time
mancein the detonationtask averaged22.3Eo interactions, indicate that the high-ability
for the high-ability subjectsand 19.l7ofor the subjectsincreasedtheir hit percentagewith
low-abilitysubjects,t(14\ : .73, ns.
additional path observation time, whereas
To summarize the results thus far, it ap- the low-ability subjectseither did not, or did
pearsthat the subjectswho perform well on so to a much lower extent. On the basis of
the standardtrajectory-intersectiontask are the argument presentedearler, this finding
also somewhatmore proficient than poorer can be interpreted as indicating that only the
performing subjects at making judgments high-ability subjectswere revisingtheir judgabout the alignment of both vertical and left- ments by recycling through the component
to-right trajectories, at least when the task sequencein the time betweenthe initial esdifficulty is not too great and a yes/no pro- timate and the arrival of the target at the
cedure is used. There is no hint of compa- launch position. The functions of Figure 2
rable differencesin the accuracyof discrim- are remarkable,not only with respectto how
inating the time required for the projectile clearly this trend is representedbut also in
or target to traversea specifieddistance,al- the degree to which the three independent
though high-ability subjectstended to pro- experimentsyielded very similar results.
duce longer time intervals than low-ability
Further confirmation of the hypothesis
subjects.There is a trend for high-ability sub- that high-ability subjectsdiffer from low-abiljects to have faster reaction times than low- ity subjectsin engagingin more extensive
ability subjects,but, perhapsbecauseof the updatingof the decisionestimatesis available
small samplesizeand/or the largevariability, from the data of the two modified trajectorythis differencefailed to reach statistical sig- intersectiontasksin which ability differences
nificance. The ability differencesalso seem were significantly reduced. First consider the
to be rather task specificastwo modified ver- blanked-trajectorytask in which the target
sions of the trajectory-intersectiontask led
to much smaller,and in one case,nonsignifa We conducted two additional analyseswith (a) path
icant, ability differences.
observation times corrected for variations in LLT due
As noted earlier,trials in the standardtra- to targets at varying heights and (b) averagemiss disjectory-intersectiontask varied in the speed tances substituted for hit percentages.The direction of
misses shifted from too late to too early with inofthe target trajectory and in the horizontal the
creasedpath observation time, but otherwise the results
position ofthe launchingapparatus.For each ofthese analyses
werenearly identical to thosepreviously
trial, thesevalueswere converted into units described.
TIMOTHY A. SALTHOUSE AND KENNETH PRILL
6r4
'l
/'ExP
Exg.3
.t--o
'it:4
2
t7/*'r.-Ex1
rol 25"k
ttt
-=
r
-.1''rF
o
@
CD
.U
c
o
f 4 0
o
o
o)
(E
-c
o
o
@
o
(D
(L
Exp.3
20r
30
y<Exp. 1
6
(L
Exp.2
Bottom25"/.
<350
413
539
Path ObservationTime (msec)
task as a function oftime to observe
Figure 2. Percentageofhits in the standardtrajectory-intersection
j. (tni aott.O lines are the data from subjectsin the top
2,
and
l,
the targettrajectory ln nxpeiiments
quartiles of the
quartiles of the sample, and the sotd fnes are the data from subjectsin the bottom
samples.)
was also completely
path is visible for only the first 217 msec of path continuation time
resultsare consisthese
of
Both
eliminated.
sublethe
ihe total trajectory. Blanking out
that a major facquent firget trajectory should eliminate the tent with the interpretation
differencesin
ability
the
for
opportunity for revisionsbeyond the initial tor responsible
decifrequent
more
is
the
task
standard
the
eitimate, and consequently,the performance
subhigh-ability
among the
differences between high- and low-ability sion updating
jects.
s,rouDsshould be reduced or eliminated if The detonation version of the trajectoryirucir of the superiority of high-ability subtask provides another opportuintersection
jects
-of is due to their more frequent rwisiolt
the correspondence
examining
for
nity
this
examined
We
co-ponent estimates'
differencesand the monotonimplication by analyzing the data lop tfe betweenability
functions relating hit perincreasing
blanked-trajeitory task in terms of the six ically
time' The relevant
path
observation
to
centage
goupings of path observation(continualion)
4. An n'uove'
Figure
in
illustrated
are
data
ability
two
for
the
Iime. Vfe"n hit percentages
was signifeffect
time
the
revealedthat only
gxoupsare displayedin Figure 3. AnnNovn
:
:
143'55,
MS,
P<
5'3t,
70)
F(5,
i-ndicited a small but significant effect -of icant,
relationcomplex
(other
The
<
1.0).
:
Fs
.001
=
l:l'63'
4.62, MS:
ability, f'(1, 14)
percentageand path oqs"Tp < .05, but nonsignificant effectsof time, ship between hit
to explain, but it is
difficult
F(5, 70): 2.32,MS. = 78.41,Pl, .0-5,-3nd vation time is
ofthe trajectoryversion
this
that
:1.26,
noteworthy
MS":
fime x AbilitY, F'(5, 70)
the elimination
in
resulted
task
intersection
78.41,p > .25.
and the linear
difference
ability
the
both
of
indicates
3
2
and
A comparison of Figures
percentage
and path obhii
that not only was the ability difference re- trend between
subjects'
high-ability
for
time
duced but the trend for high-ability subjects servation
and movingconstant-target
the
Because
to improve hit percentage with increased
Figure 3. hrccnrl
traJectoryin Exg
Note that al$orrsh
samepath for rhe r
Iaunch versionsof tt
task allow normal u
tory information. I
posedherewould lea
ability differencesrx
same manner as ob
40g,
fo * o
ct)
g
5 aoc
I
l0-
I't4tre4.FerccntaXro
rn trpennrnt
l. tTh
615
PERCEPTUAL SKILL
o
I
o
C')
(!
JU
o
Y
o
(L20
Toozs'k P
Bottom25%
<350
413
539
664
>850
789
Path Obseruation Time (msec)
ofhits in the blanked-trajectorytask as a function ofcontinuation time ofthe target
Figure 3. Percentage
trajectory in Experiment l. (The labelsrefer to the top and bottom quartiles ofthe sampleof subjects.
Note that although the target trajectory was only visible for the first 217 msec, it continued along the
samepath for the times indicated.)
launch versionsof the trajectory-intersection
task allow normal updating of target-trajectory information, the interpretation proposedherewould leadto the expectationthat
ability differenceswould be exhibited in the
same manner as observedin the standard
versionof the task. As noted earlier,this was
indeed the case,and thus the results from
thesetasks are also consistentwith the suggestionthat at leastpart ofthe ability differencesare attributable to more frequent revision of decisionestimates.
o
t 3 0
(!
5zo
dt
r
10
<350
413
539
664
789
PathObseruation
Time(msec)
Figure 4. Percentageofhits in the detonationtask as a function oftime to observethe target trajectory
in Experiment2. (The labelsrefer to the top and bottom quartiles ofthe sampleofsubjects.)
616
TIMOTHY A. SALTHOUSE AND KENNETH PRILL
Table I
Summary of ComponenbEfectivenessTbsts
Task
difficulty
Exp. Exp.
Ability
l&3' 4
Component
VVA
Yes/No
Adjustment
constant error
Variableerror
ETA
Adjustment
Constant error
Variableerror
CTP
Yes/No
Adjustment
Constanterror
Variableerror
Yes/No
Adjustment
constant error
Variableerror
x
x
x
x
X
x
"
X
A
x
x
X
X
X
x
x
x
x
Nore.An X indicatesthat the effectulasstatisticallysignificant (p < .05); a dashindicatesthat the information
was not applicable.VVA = vertical vulnerability area;
ETA = estimatedtime of arrival; CTP : critical target
positionl LLT = launch lag time.
" Theseeffectswere significantonly at the easiestlevel
of difficulty.
Discussion
The main purpose of ExPeriments l, 2,
and 3 wasto identify someof the factorscorrelated with performance differences observedin a haphazardsampleof individuals.
As it turned out, the subjectswith extreme
scoresalso differed in sex distribution and
self-reportedexperienceplaying potentially
related video games.These results indicate
that the three skill categoriesidentified in the
introduction are frequently not mutually exclusive becausethe present experimentsattempted to investigateskill as initial ability,
but the individuals in the extreme ability
groups also differed on experienceand demographic(i.e.,proportion of males)dimensions.Although thesedifferencesmake it difficult to determine why the individuals differed in proficiency, the major focus of
Experiments l, 2, and 3 was to identify processingfactors responsiblefor the existing
ability differences,and this goal is not compromised by the characteristics of the samples.
The results of Experiments l, 2, and 3
present an intriguing, but still incomplete,
picture of the nature of skill as initial ability
on the trajectory-intersectiontask. With respect to the model illustrated in Figure l,
there is evidencethat the skilled subjectsmay
be more effective than the unskilled subjects
at performing some of the component operations. It can be seen in Table l, which
summarizesthe major results from the tests
of component effectiveness,that'an ability
effect was obtained in each of the components investigated.Perhaps more remarkable, however,is the small number of significant ability effectsrelativeto the number of
significanttask-difficultyeffects.Sevenof the
I I measureswere found to be sensitiveto the
difficulty level of the task, and all of these
results were replicated in Experiment 4.
Therefore, these seven measures have face
validity in that performance was inversely
related to task difficulty. Howeveq only three
ofthese yielded significant ability effects,and
then only at the easiestdiftculty level for the
WA and CTP components.
It is interestingto note that the high-ability
subjects produced significantly longer time
estimatesthan low-ability subjectsin both the
LLT and ETA tests. Unfortunately, without
further data one can only speculateas to
whetherthe longer subjectivetime estimates
are a cause,or an effect,of the performance
differenceswith.which they are correlated.It
is also noteworthy that both high- and lowability subjectsapparently had illusory perceptions of the trajectory velocities as the
left-to-right velocity was consistently underestimated (positive constant errors), and the
vertical velocity was overestimated(negative
constanterrors). Runeson(1975) reported a
similar misperceptionof velocity,but we have
no satisfactoryexplanation for this illusion,
or the fact that it apparently reversesfrom
vertical to horizontal orientations.
Perhapsthe strongest conclusion possible
at the present time is that the skilled subjects
appear to executethe componentsin a different manner than the lessskilled subjects.
This inference is derived from the functions
portrayedin Figures2 and 3, which are con-
srstentwith rb f
ability sutlanr cy
nenlc
u'ith
dt
target traina!.{
Jeds mcrcll.q,I
tra)€d ia hr:
I
femtoheq
loop- fu&..ir
hrgb{filitv diu
w|rbrbQbct
tbe decid<n caitr
morc prccbe wirh r
Exg
The purpoce of I
v.estigatethe proces
trce-related skill oo
uon task. Some relc
rrom eight yount
aoutts who perform
section tasks for 5Or
the context of a larX
romberg, l9g2). Ua
the time requirernen
in the project, an aw
per task per session
testsof componentef
out. However,comrr
on the first and la* iA
of 288 interveningtri
age(1.e.,your8 = 31
practice (i.e., earlv =
effects,and a significr
vatlon Time interacti<
navlng steeperslom
Practice X path Obs
tlon wassignificantfor
subjects,with the dirc(
tndlcatrngthat the sb
Iatinghit percenr4gp
o{
was steeperwith incrq
Although the age aa
were rnteresting the at
ponent effectiveness
al
tbct of practice(i.e.. 3.
rormatlvenessof the rc
perimenlattemptodro
I
or practtceby providiq
tnals (i.e., 1,800vs. 2d
of component effecdu
tered after practice to
someof the expeced g
6r7
PERCEPTUAL SKILL
sistent with the interpretation that the highability subjects cycle through prior components with additional time to observe the
target trajectory whereasthe low-ability subjects merelywait. In terms of the model portrayed in Figure l, the low-ability subjects
seem to be accuratelycharacterizedby the
loop on the decisioncomponent,whereasthe
high-ability subjects are better represented
with a loop to earlier componentssuch that
the decisionestimatesbecomeprogressively
more precisewith additional time.
Experiment 4
ment was attributable to more accuratespatial and temporal discriminations. A modified matched-groupsdesign was used to capitalize on the availability of data from the
pool of subjectsin Experiments I and 3. A
majority of femaleswere usedas subjectsto
minimize the amount of preexperimental
practicewith related tasks.
Method
Subjects
Eight young adults with a mean age of 22.6 years
(sevenfemalesand one male) participatedin five l-hour
sessionsover a period of 2 weeks.Each received $20 for
participating.
The purpose of Experiment 4 was to investigatethe processescontributing to practice-related skill on the trajectory-intersec- Apparatus
tion task. Some relevant data were available
The apparatuswas identical to that of the preceding
from eight young adults and eight older experiments.
adults who performed two trajectory-intersectiontasksfor 50 experimentalsessionsin Procedure
the context of a larger project (Salthouse&
Ten 50-trial blocks of the standard trajectory-interSomberg, 1982). Unfortunately, becauseof sectiontask wereadministered
on SessionsI through 4.
the time requirementsof the other activities The first and last five trial blocks in the experiment were
in the project, an averageofonly 12.5trials identical, but the remaining 35 trial blocks (i.e.. 5 on
per task per sessionwas presentedand no SessionI and l0 eachon Sessions2,3, and 4) had randomly selectedvaluesof target speed,target
target
testsof componenteffectiveness
werecarried height,and launch location. On Session5, angle,
the subjects
out. However,comparisonsof performance receivedfi ve trajectory-intersection trial blocks followed
on the first and last 188trials (with an average by method-of-adjustmenttestsof horizontal (CTp) and
of 288 interveningtrials) revealedsignificant vertical (VVA) trajectory alignment, horizontal (ETA)
age(i.e., young : 3l.2%o,old = 18.3%)and and vertical (LLT) trajectory timing, and forced-choice
testsof horizontal
vertical trajectory alignment.
practice(i.e., early : 22.9vo,late: 26.7Vo\ The instructionsand
encouragedsubjectsto attempt to
effects,and a significant Age X path Obser- improve their performanceas much as possibleduring
vation Time interaction with young subjects the practiceperiods.As an additional incentive.the inhaving steeperslopesthan old subjects. The dividual exhibiting the geatest improvement from the
Practice X Path Observation Time interac- first to the fifth sessionreceiveda $20 bonus.
tion wassignificantfor young but not for old
Results
subjects,with the direction of the interaction
indicating that the 'slopeof the function reThe mean percentageof hits in the trajeclating hit percentageof path observationtime tory-intersectiontask
increased from 26.lTo
was steeperwith increasedpractice.
in Blocks2-5 of SessionI to 36.6Vo
in Blocks
Although the age and practice differences 2-5 of Session5, t(7) : 3.38,p <
.05. The
wereinteresting,the absenceof testsof com_ significanteffectofpractice
indicatesthat the
ponent effectiveness
and the rather small ef_ provisionof 1,800trials betweenthe first and
fect of practice(i.e.,3.BVo\
weakenedthe in- secondtestingsessions
did indeedleadto subformativenessof the results.The current ex- stantial performanceimprovements.
periment attempted to produce greatereffects
Figure 5 illustrates hit percentageas a
of practiceby providing many more practice function of path observationtime
for earlv
trials (i.e., 1,800vs. 288). In addition, tests (Sessionl) and late (Session phases
5)
of
of component effectivenesswere adminis- practice.An eNova indicated that
the practered after practice to determine whether tice,F(1, 7) : 38.0t,MS" : 84.43,p<
.0005,
someof the expectedperformanceimprove- and time f'(5, 35) : 53.76,MS" :
99.99,
TIMCNHY A. SALTHOUSE AND KENNETH PRILL
618
I and 3 contained
p < .0001, effectswere significant, but the and both Experiments
'Practice
mean hit percentThe
subjects.
pools
32
of
X Time interaction was not, F(5,
fro1- lhj two
subjects
control
8
the
of
iges
35\: 2.16,MS": 84.43,P > .05'Thesesta26'l%o,ex'
both
pievious experimentswere
tisiical trends, in conjunction with the patpresent
the
of
mean
the
as
same
tern portrayed in Figure 5, indicate that the ictly the
l
'
Session
on
subjects
lineai relationshipbetweenpath observation e*perimental
The results of the ANovAsconducted on
time and hit percentagewas evident at both
can
of componenteffectiveness
stagesofpractice, and that the additional ex- thejmeasures
signifno
were
There
perienceled to an increasein hit percentage be easilysummarized.
practice (p > .09) or Practice.X
but with only a slight (nonsignificant)steep- icant
(p > .14) effects.There is abperTask-Difficulty
ening of the path observationtime/hit
that the subjects,with
tto
euidence
solutely
centagefunction.
substantially higher
practice
and
additional
of
effects
the
of
To make comparisons
any more accurale
were
performance
the cur- final
practiceon componenteffectiveness,
than subdiscriminations
relevant
ient subjectswere contrasted with subjects at making
but
performance
initial
jects
comparable
of
and
from Experiment I and Experiment 3
Taskpractice'
of
amount
matchedon the basisof overallhit percentage without the same
significant(p < '0-l) for
were
effects
difficulty
a
allowed
This
l.
Session
for Blocks 2-5 of
error (p > ' 16) and
variable
Cfp
ttt"
but
direct examination of the effectsof practice all
(p
>
.87) measures'
error
constant
ETA
subbecausethe
on componenteffectiveness
received
jects from the earlier experiments
Discussion
the componenttestsimmediatelyafter Blocks
from
subjects
the
whereas
1,
Session
of
2-5
The two major findingsof this experiment
the current experiment received 40 addithat practice-relatedskill developswtthare
comthe
tional blocks (2,000 trials) before
increasesin the effectiveponent tests. The matching was quite close out concomitant
temporal components'or
and
spatial
of
t."uu." only 8 subjectshad to be matched ness
(n
=
!
o
c
o
o
o
(L
"
-
<350
413
539
664
789
>850
Time(msec)
PathObservation
task as a function oftime to observe
Figure S.Percentageofhits in the standardtrajectory-intersection
phase of practice.)
the
to
refer
lati
ia
+-liriy
the target trajectory m e*p".ir""it
dramatic dr4l
nent operarq. I
in the find; I
componcd t{
Practrca ca..r {
rn ored hd {
rs. -i66tl lLd
tqcad c
-{
rqIE
tr=-f=l.r
rft-hbx
Crener
An intriguing sp
the results of this :
formation-pro(6su
contributing to \er
sociatedwith ahen
of skill. For eramp
encesfound to acco
ationsassociatcdrrr
to be the sameas rh
ability or age-relarc
One exampleof
ferenceis wident rr
tential differencs ir
nents responsiblc f
processing oryratto
and method-of-ad1u
assessthe accurac\ (
tial discriminations
components in the
task. Somerather sli
were found to be ess
ity level,but the sam
any differencesas a
The present tess
provide information
and not the efficienc
nents, and therefore
of the time, rather rl
component decisio
larger and more con
If the limitation r+e
however,one might
formance difference
section task to be n
time to observethe t
the component oprri
of Figures2 and 5 in
genceratherthan con
path observation tirn
PERCEPTUAL SKILL
t
$
':ll
dramatic changesin the cycling of component operations.The first result is apparent
in the finding that none of the measuresof
componenteffectiveness
exhibitedsignificant
practice effects,despite sizeabledifferences
in overall level of performance (i.e., 26.lVo
vs. 36.6%).The secondresultis an inference
bard on the similar slopesof the functions
relating hit percentageto path observation
time in Figure 5, and the absenceof a significant Practice X Time interaction.
619
pretation has no support at the presenttirne.
It also proved impossible to devise a test to
assessthe effectivenessor efficiency of the
decisioncomponent in the model of Figure
l, and this component is arguably the most
important in the sequence.Despitetheselimitations, it is still surprisingthat component
effectiveness
contributeslittle to the skill variations, particularly those that are practicemediated. Salthouseand Somberg (1982)
havesummarizedthe resultsof many studies,
and added further results of their own. documenting the effectsof practice on elemenGeneral Discussion
tary aspectsof skill, and yet the experienceAn intriguing speculationsuggestedfrom based improvement in performance on the
the results of this study is that different in- trajectory-intersectiontask did not seem to
formation-processingmechanisms may be be accompaniedby increasedaccuracyin the
contributing to variations in proficiency as- relevantcomponents.
sociatedwith alternative conceptualizations
One possiblereasonfor the lack of differof skill. For example, the pattern of differ- encesin component effectivenessis that inencesfound to accountfor performancevari- dividualsat differentskill levelsuseddifferent
ations associatedwith practicedo not appear componentsto perform the task. This sugto be the sameas thoseaccountingfor initial gestioncannot be definitely ruled out; howability or age-relatedperformancevariations. ever, it is difficult to imagine how the task
One example of this configurational dif- could be performed without componentsof
ferenceis evident in the analysisof the po- the type outlined in Figure l. To estimatethe
tential differencesin the individual compo- intersection point of two trajectories, both
nents responsiblefor carrying out specific the spatial and temporal aspectsof eachtraprocessing operations. Both forced-choice jectory must be determined and the inforand method-of-adjustmenttestswere usedto mation from the two trajectoriesmust be inassessthe accuracyof the temporal and spa- tegratedin some manner. The components
tial discriminations postulated to be basic in Figure I thereforeappearto provide a neccomponents in the trajectory-intersection essaryand sufficient set to perform the tratask. Somerather slight accuracydifferences jectory-intenectiontask,althoughalternative
were found to be associatedwith initial abil- levelsof specificationor detail might be more
ity level,but the sametestsfailed to indicate appropriatefor somepurposes.For example,
any differencesas a function ofpractice.
it could be argued that velocity and distance
The present tests were designedonly to areestimatedinsteadof time in the trajectory
provide information about the effectiveness estimates(e.g.,Lappin, Bell, Harm & Kottas,
and not the efficiencyof individual compo- 1975;Rosenbaum,1975),in which casedifnents, and therefore it is possiblethat tests ferent setsof componenttestsmight be more
of the time, rather than the accuracy,of the appropriate than those that were used. It is
component decisions would have vielded clearly necessaryto obtain considerablymore
larger and more consistentskill differences. evidenceon a variety of suspectedcompoIf the limitation were primarily temporal, nents before dismissingthe contribution of
however, one might have expected the per- improved component efficiency to overall
formance differencesin the trajectory-inter- task proficiency,and the present results, alsection task to be reduced with additional though reasonablycleaq can only be considtime to observethe trajectory and complete ered suggestiveat the presenttime.
the component operations.In fact, the data
A difference in the order of the compoof Figures2 and 5 indicate a trend of diver- nents within the processingsequencedoes
genceratherthan convergencewith increased appear to contribute to the skill variations
path obcervationtime, and thus this inter- associatedwith initial ability level and, to a
620
TIMCIHY
A. SALTHOUSE AND KENNETH PRILL
lesserextent,with practiceand adult age.The
skilled subjectsexhibited a steeperfunction
relating hit percentageto path observation
time, and this was interpreted as an indication of more frequent revisionsof the component estimates with additional viewing
time. The less skilled subjects,particularly
those defined in terms of initial abilitv. had
much flatter hit-percentage/path-obiervation-time functions, suggestinglittle or no
revision of the componentestimateswith increasedtime to observethe targetpath. These
differencescan be interpreted as evidence
that the skilled subjectsexecutedthe components repetitively (e.g., nncnE-ABcDEABcDE. .), whereasthe unskilled subiects
completed the entire sequenceonly once
(e.9.,ABcDE-E-E-E.. .).
It would obviously be desirableto obtain
more direct evidencefor the hypothesisthat
skill differencesare related to the manner of
component execution, but all procedures
that were consideredhad severelimitations.
For example, it might seem that eye-movement analyseswould prove relevant to this
issue,but it is difficult to predict which type
of scan pattern would be associatedwith a
given componentsequencebecausemuch of
the information processingcould be carried
out in the absenceof overt eye movements.
A closer examination of Figures 2 and 5
revealsthat there are actually two distinct
segmentsof the function relating hit percentageto path observationtime, and that
the interpretation previously proposed applies best to the segmentfrom 539 to over
850 msec.The first segment,from under 350
to 539 msec,appearsto increasefor subjects
at all skill levelsand may be determinedmore
by factors such as speed of component or
sequencecompletion. The data of Figure 5
also suggestthat the two segmentsare differentially affectedby practice as only the segment from 539 to over 850 msec appearsto
exhibit substantial changeswith additional
experience.
Another result of theoretical significance
in the current experimentsis the demonstration that the sequencewith which elementary
componentsareexecuted(i.e.,nncor-ABcDE
vs. ABCDE-E. . .) may be a more im.
portant determinant of overall skill than the
level of performanceon each separatecomponent.Fleishman(1966)and his colleagues
have demonstratedthat the particular abilities contributing to performance change
acrossstagesof practice and the current resultsare clearlyconsistentwith the suggestion
that skill variations are due in part to the
relianceon different mechanisms.It has also
been argued by many theorists that overall
performance may be more dependent on
strategies(i.e., sequencesof component operation) than on basicabilities or component
proficiency(e.g.,Allport, 1980; Bartlett, 1972:
Edwards,1979;Glaser,1980;Lansman,l98l;
Pellegrino& Glaser,1979;Sternberg,1978;
Welford, 1958,1976).In providingevidence
that certain skill variationsmay be associated
with differencesin the sequenceof component execution,the presentstudy adds substance to the argument that at least some
perceptual skill differences are qualitative
rather than merely quantitative.
References
Allport, D. A. Patternsand actions: Cognitive mechantsmsar€ content specific.In G. Claxton (Ed.), Cognitive psychology:New directions.London: Routledge
& Kegan Paul, 1980.
Bartlett, F. The experimental study of skill. In R. N.
Singer(Ed.),Readingsin motor learning.Philadelphia,
Pa; l*a & Febiger,1972.
Edwards,E. The expert pilot. In W. T. Singleton(Ed.),
The study of real skills: Vol. 2. Complianceand excellence.Baltimore, Md.: Univeniry Park Press,1979.
Fleishman,E. A. Commentson ProfessorJones'paper.
In E. A. Bilodeau(Ed.),lcquisition o/skil/. New York:
AcademicPress,1966.
Gerhard, D. J. The judgment of velocity and prediction
of motion. Ergonomics,1959,2.287-304.
Glaser, R. General discussion:Relationshipsbetween
aptitude,learningand instruction. In R. Snow,P. Federico, & W. Montague(Eds.),Aptitude, learning and
instruction(Vol. 2). Hillsdale,N.J.: Erlbaum, 1980.
Gottsdanker,R. M., & Edwards,R. V. The prediction
of collision.,4rnericanJ ournaI of Psyvhology, | 9 57, 70,
I l0-l 13.
Lansman, M. Ability factors and the speed of information processing.In M. P. Friedman, J. P. Das. &
N. O'Connor (Eds.), Intelligenceand learning. Nevt
York: PlenumPress,198t.
Lappin,J. S., Bell, H. H., Harm, O. J., & Kottas,B. On
the relation betweentime and spacein the visual discrimination of velocity.Journal of Experimental Psychology: Human Perception and Performance, 1975,
I, 383-394.
Noble, C. E. Age, race and sex in the learning and performance of psychomotor skills. In R. T, Osborne,
C.E \.-..- r
b t r t f' , r . , . .
demrc P:r.
Pellcgnnr .' '*
-
C0lllt\Ylc.:.
,.
In R -t \r-r.-.
,2.:.,&: | .. rtl'zi
\
':.^
rr:
'\
Roscnha-n t) \ 1,.
k.rr:r anl a!.cicr.,'
t E * 4-,
1t11271,tr .;
/ _19)--{)-1.
R u n c s o n .S . ! ' r s u a i r and nonnatural i:r
chophl'sics. 19'i
PERCEPTUAL SKILL
The
C. E. Noble, & N. Weyl (Eds.), Human variation:
Acabiopswhology of age, race and sex New York:
demic Press, 1978.
Pellegrino, J. W., & Glaser. R. Cognitive correlates and
.o".pon.tttt in the analysis of individual differences'
In R. J. SternberB. & D K. Detterman (Eds'\, Human
intelligcn(e; Pospectyes t'n its Iheorr and measurernent. Nor*ood. N.J.: Ablex. 1979'
Rosenbaum. D. A. Ferception and extrapolation glveto.irf and acceleration. Jrturnal of Experimental.Psy-1975'
,'lr,,lottr lluman Pocepluvt and Performance'
/ 195--rtll.
Runcson. S Yrsual prediction of collision with natural
and no<rnaturaf morron functions' Perception & Psy'
clrrphlrter l9-5. /,1 16l-166'
n
i
(
1
e
62r
Salthouse,T. A., & Somberg,B. L. Skilled performance:
The effectsofadult ageand experienceon elementary
processes.Journal of Experimental Pslx:hology:General, 1982,l l, 176-207.
Sternberg,R. J. Intelligence researchat the interface
be$e;n differential and cognitive psychology: Prosoectsand proposals.Intelligence' 1978,2' 195-222'
wettorA,A.T. Ageingand humanskl// London:Oxford
Univenity Press,1958.
Welford,A. T. Skilled performance:Perceptialand motor
skl//s.Glenview,Ill.: Scott Foresman,1976'
ReceivedNovember 5, 1982
Revision receivedFebruary 18, 1983 I