Sei sulla pagina 1di 11

BRAIN AND COGNITION 32, 405415 (1996)


Effects of Stimulus Number and Target-to-Distractor

Ratio on the Performance of Random Array
Letter Cancellation Tasks

Department of Neurology, University of Medicine and Dentistry of New JerseyRobert
Wood Johnson Medical School, New Brunswick, New Jersey, and Center for Head
Injuries, JFKJohnson Rehabilitation Institute, John F. Kennedy Medical Center,
Edison, New Jersey

Cancellation tests are commonly used in the clinical assessment of visuospatial

function, but there has been little study of task characteristics influencing perfor-
mance. This study was designed to assess factors which affect cancellation perfor-
mance. Sixteen healthy subjects sequentially performed four random-array letter
cancellation tasks. The forms contained 50 and 100 stimuli and target:distractor (T/
D) ratios of 1: 4 and 1: 9 with target letter A and randomly selected letter dis-
tractors. The primary performance measure was calculated as the number of cor-
rectly cancelled targets divided by the time to complete the task, corrected for accu-
racy. This measure revealed a strong effect of T/D ratio ( p , .0001), with
performance adversely affected by higher proportion of distractors. There was no
effect of stimulus number. This suggests that T/D ratio should be considered in
cancellation test design and interpretation. 1996 Academic Press, Inc.

Random array cancellation tests are commonly used in the clinical assess-
ment of neglect and other disorders of visuospatial function. Unfortunately,
there has been little standardization of cancellation paradigms and few re-
ports regarding the stimulus characteristics which may influence task perfor-
mance. All cancellation testing involves the identification and marking of
specific target stimuli, but the instruments reported in the literature vary
widely on many stimulus dimensions. It has therefore been difficult to gener-

This work was presented in part to the International Neuropsychological Society, (Cincin-
nati, OH), February 1994. The contributions of Everett Hills, MD, Henry Kong, Ruth Archer,
and two anonymous reviewers are gratefully acknowledged. Address reprint requests to David
S. Geldmacher, M.D., at Alzheimer Center, University Hospitals of Cleveland, 11100 Euclid
Avenue, Cleveland, OH 44106.
0278-2626/96 $18.00
Copyright 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.

alize findings across disease states or even among reports within a well-
defined syndrome such as neglect.
Albert (1973) popularized one common version of cancellation testing that
employs short line segments distributed pseudorandomly on a sheet of paper.
This task contains no stimuli which require the withholding of a response
(distractors) and therefore probably does not draw significantly on the re-
sources of selective attention. It would, however, be expected to require spa-
tially directed attention. Control subjects made no errors in Alberts original
study, suggesting that a ceiling effect limits the effectiveness of the test
in more subtle cases of visuospatial impairment. Villardita, Smirni, and
Zappala (1983) used an expanded version of the task, with 90 rather than
40 targets, in an attempt to increase the instruments sensitivity to subtle
visuospatial dysfunction associated with Parkinsons disease, but did not
compare performance between their task and the original. Chatterjee, Men-
nemeier, and Heilman (1992) subsequently demonstrated that increasing the
number of targets revealed neglect not evident on cancellation displays with
fewer stimuli. Kaplan, Verfaellie, Meadows, Caplan, Pessin, and DeWitt
(1991) reported that increased distractors on a cancellation task resulted in
more omission errors in neglect patients. Since they kept the target number
constant, it is unclear whether the effect was based on more stimuli to
consider as potential targets or on a more adverse target to distractor ratio
It is not clear how a greater number of stimuli affect performance of the
task. Although more stimuli may make the task more difficult by prolonging
the overall duration of performance or increasing the likelihood of habitua-
tion, a higher number of stimuli might also ease the demands on directed
attention by requiring shorter interstimulus shifts of attention. Since these
factors are not mutually exclusive, the net effect of stimulus number on can-
cellation performance by normals is unpredictable.
The degree to which stimulus filtering or selective attention is engaged
also influences cancellation performance. Gauthier, Dehaut, and Joanette
(1989) concluded from their study of cancellation with easily verbalizable
silhouette targets and distractors that the presence of distractors increases
the potency of the cancellation format for detecting visuospatial dysfunction.
Similarly, Rapcsak, Verfaellie, Fleet, and Heilman (1989) found that greater
similarity between targets and distractors enhanced the sensitivity of cancel-
lation tasks for identifying neglect. The difficulties of assessing the impact
of distraction on cancellation performance are exemplified in the work of
Stone, Halligan, Wilson, Breenwood, and Marshall (1991), who reported the
performance of a control population on a visually complex cancellation task
involving stars as targets and different sized stars, letters, and words as dis-
tractors. Despite the complexity of the stimulus instrument, their study did
not address the potential effects of the distractor type and number on perfor-
mance. Negative effects of increased target and distractor number have been

reported in individuals with spatial neglect (Kaplan et al., 1991; Chatterjee

et al., 1992), but the specific impact of the T/D ratio on cancellation perfor-
mance among healthy control subjects has not been fully explored.
The effects of stimulus number and T/D ratio have, however, been exten-
sively studied by many investigators on noncancellation visual search tasks
employing reaction time (RT) as the dependent variable. On tasks consisting
of targets and distractors sharing perceptual elements, such as letters, reac-
tion time increases in relationship to the number of potential targets. Orderly
array cancellation tasks like the 2-and-7 test are also reported to demon-
strate this effect (Ruff, Evans, & Light, 1986). Several nomenclatures have
evolved for the description and interpretation of this common finding includ-
ing serial, effortful, are controlled processing. It is unknown if sim-
ilar effects are true for cancellation of randomly arrayed letters.
The effects of total stimulus number and T/D ratio can be studied using
the cancellation paradigm in clinical settings with a minimum of specialized
equipment or training. The purpose of this study was to determine the impact
of factors known to affect visual exploration in laboratory based RT methods
on random-array letter cancellation tasks. Understanding these factors is im-
portant in developing bedside tasks which are sensitive to subtle visuospatial
dysfunction, as well as in the interpretation of results from the pathologic
populations in whom the test is currently applied.
A second purpose of the study was to further explore the feasibility of
combining both accuracy and speed in a single performance measure for
cancellation. Results of this measure have been reported for the effects of
healthy aging (Geldmacher, Rowland, & Riedel, 1995) and traumatic brain
injury (Geldmacher and Hills, in press). Orderly array cancellation tasks are
typically scored by counting the errors in a predefined time period, but sub-
jects in such tasks are directed to use a specific spatial search pattern. Ran-
dom array tasks are used to better understand spatial search patterns and
spatially directed attention, but are rarely timed. Nonetheless the facility and
efficiency of visual search performance has ecological validity for daily ac-
tivities like driving. Additional, clinically applicable, tasks to evaluate such
abilities have potential uses for measuring recovery after brain lesions or
gauging the impact of rehabilitative or pharmacologic interventions.

Subjects. Sixteen healthy subjects, 10 men and 6 women, were enrolled. Their mean age
was 29.7 years (range 1941) and mean education was 15.9 years (range 1220). Fourteen
subjects self-reported as right handed. The protocol was reviewed and approved by the institu-
tional review board of the Robert Wood Johnson University Hospital. Informed consent was
obtained from all subjects prior to enrollment.
Instruments. Four cancellation forms were prepared on white 8.5 3 11-inch paper. The four
forms were as follows: (1) 5 targets, 45 distractors; (2) 10 targets, 40 distractors; (3) 10 targets,
90 distractors; and (4) 20 targets, 80 distractors. The target letter for each was A; distractors
were chosen at random from other upper case letters (see Fig 1).

FIG. 1. The four test forms. Top left, 10/40; top right, 5/45; bottom left, 20/80; bottom
right, 10/90. Actual size for all forms is 8.5 3 11 inches.

Because previous reports have shown different error patterns along nearfar and leftright
axes, as well as in central regions, of cancellation forms (Geldmacher, Doty, & Heilman, 1994,
1995), equal numbers of stimuli were arranged in five spatial zones, each covering an area
of 18.7 inches. The diamond shaped central zone was 4.3 inches on each side and centered
on the page. Left distant, right distant, left near, and right near zones were irregular pentagons
defined by the subjects midline, the midline of the cancellation form in the frontal axis, and
border of the central zone. Figure 2 depicts the layout of the zones.
Procedure. Subjects were confortably seated at a desk or table and were instructed to use
their preferred hand to place a pen mark through every occurrence of the target letter on the
cancellation form. Speed and accuracy of performance were equally stressed, but unlike or-
derly array paradigms, no specific scanning path was suggested or required. Following the
instructions, cancellation forms were sequentially placed on the desk, centered approximately
12 from the edge and at the subjects midline. The examiner ensured no lateral or rotational
movement of the sheet or the subjects trunk; subjects had free movement of head and eyes.
Scanning with the finger or pen tip was allowed. The examiner used a stopwatch to time the
interval from the subjects first mark until the completion signal for each sheet. Subjects sig-
nalled completion for each cancellation form either verbally or by placing the pen down,
whichever occurred first. The order of stimuli was counterbalanced across subjects.
Data collection and analysis. The time for completion as well as number and location of
all unmarked targets was recorded for each cancellation form. For purposes of analysis, the
primary performance was defined as the number of correct responses divided by the time to
completion, corrected for the accuracy of the target identification. This can be represented as

FIG. 2. Layout of the five spatial zones superimposed on each of the stimulus forms.

Performance 5 (correct responses/total targets) 3 (correct responses/total time).

Secondary analyses were conducted for other potentially pertinent measures of performance
such as error rate, completion time, errors/time, and errors/target. The spatial distribution of
errors was also considered as a secondary measure.

Data for the effects of subjects, errors, completion time, and combined
performance for each form are shown in Table 1. ANOVA demonstrated
that these differ significantly (F[3,60] 5 12.42, p , .0001). Therefore post hoc
analysis with t-tests for paired samples was conducted to analyze the effect
of stimulus number and T/D ratio on the performance measure. For the anal-
ysis of stimulus number, performance on both 50-stimulus forms was com-
pared with both 100-stimulus forms. Similarly, for analysis of target to dis-
tractor ratio, both 1 : 4 forms were compared to both 1: 9 forms. Table 2
depicts the data when the forms are combined for stimulus number and
T/D ratio. There was no significant effect of stimulus number [t(31) 5 20.189;
p 5 .85]. There was a strong effect of T/D ratio [t(31) 5 9.34; p , .0001].
For secondary measures, ANOVA demonstrated a trend toward significant
differences in mean errors per target when 50 and 100 stimuli and 1: 4 and 1:
9 T/D ratios were grouped (F[3,124] 5 2.02, p 5 .11). ANOVA did demonstrate
Errors, Completion Time, and Performance Score for Each Stimulus Form

Subjects Mean Mean Maximum Completion Mean Mean Mean

Test with errors per errors per errors per time completion errors/ performance
form errors subject target subject a range time second score

5/45 1 .06 .012 1 5.616.7 8.3 6 2.6 0.01 6 0.04 0.63 6 0.14
10/40 12 1.25 .125 4 6.511.5 9.0 6 1.5 0.16 6 0.16 0.87 6 0.19
10/90 9 1.06 .106 4 9.222.9 15.5 6 3.8 0.08 6 0.10 0.54 6 0.17
20/80 12 2.0 .100 6 10.923.7 17.4 6 3.64 0.13 6 0.12 0.96 6 0.24

Minimum errors 5 0 for all forms.
Errors, Completion Time, and Performance Scores for Different Task Factors

Mean Mean Mean Mean Performance Mean

Task errors completion time/ errors/ score performance
factor per target time stimulus second range score

50-Stimuli 0.07 6 0.10 8.6 6 2.1a 0.17 6 .04 0.08 6 0.14 0.291.14 0.75 6 0.21
100-Stimuli 0.10 6 0.10 16.5 6 3.8a 0.16 6 .04 0.10 6 0.11 0.261.54 0.75 6 0.29
1:4 T/D 0.11 6 0.10 13.2 6 5.1b 0.18 6 .03 0.14 6 0.4c 0.551.54 0.92 6 0.22d
1:9 T/D 0.06 6 .10 11.9 6 4.9b 0.16 6 .05 0.05 6 0.08c 0.260.89 0.59 6 0.16d
t 5 213.26, df 5 31, p , 1026.
t 5 2.6, df 5 31, p , .02.
t 5 3.31, df 5 31, p , .002.
t 5 9.34, df 5 31, p , 1026.

Spatial Distribution of Errors by Zone
Spatial zone

Left Right Left Right

Form distant distant Central near near Total

5/45 0 0 1 0 0 1
10/40 2 7 5 3 3 20
10/90 2 4 3 3 5 17
20/80 6 4 6 10 6 32
Total 10 15 15 16 14 70

significant differences in mean errors/second with the same task groupings

(F[3,124] 5 3.63, p , .02). Post hoc analysis with a two-tailed t test for paired
samples revealed no effect of stimulus number (t[31] 5 20.80, p 5 .43).
There was a large effect of T/D ratio, with more errors/sec occurring in the
1/4 condition (t[31] 5 3.31, p , .003).
The spatial distribution of errors pooled across all forms is depicted in
Table 3. The low number of errors per form precluded useful statistical analy-
sis of individual forms. 2 analysis demonstrated that the distribution of er-
rors on the page was not significantly different from a random pattern
( 2 5 .869, p 5 .35).

The results demonstrate that increasing the number of stimuli to be
searched on a cancellation task does not independently affect combined
search accuracy and search time per potential target, though obviously search
duration as a whole increases. In addition, a higher target number resulted
in lower accuracy, as measured by the rate of errors per subject. In addition,
the amount of distracting information strongly influenced the cancellation
performance score on these random array tasks. Forms with a lower propor-
tion of distractors, though having higher performance scores, also had a
higher rate of errors per second, suggesting a speedaccuracy tradeoff. Fur-
thermore, lower performance scores with the lower T/D ratio suggest that
random array letter cancellation tasks require effortful or controlled
information processing approaches.
Although random array cancellation tests have proved useful in the re-
search and clinical evaluation of disorders of visuospatial function, their
overall utility and application has been limited by the lack of broad objective
measures of performance. Accuracy, or the number of correctly cancelled
potential targets, has been the most common scoring method, but has not
accounted for the time required to perform the task. Alternatively, some in-
vestigators have used time as a measure and analyzed error number and pat-

tern separately (Weintraub & Mesulam, 1987). Structured array cancellation

tasks, as used in neuropsychological approaches to selective attention (Ruff
et al., 1986), employ both time constraints and measures of accuracy. These
tasks do not, however, assess the spatial aspects of attention (i.e., directed
attention) since subjects are instructed to use a specific search strategy.
The combined time and accuracy performance measure used in this study
discerns a quality of cancellation different than accuracy alone, since the
highest proportion of errors (2.0 per subject) and the highest individual num-
ber of errors (6) occurred on the cancellation form with the best combined
performance score (20/80). This measure also does not duplicate the effect
of time alone, as the longest mean completion time (17.4 sec) occurred on
the form with the best overall performance score. The rate of errors per target
did not distinguish between the task factors, but the number of errors per
second provided results similar to the combined performance score. Al-
though the number of omission errors in this small healthy population was
low, the ability to assess directed attention by analyzing the spatial distribu-
tion of errors remains available as a complementary scoring system.
Several investigators have reported that, in neglect syndromes, the number
of stimuli available for cancellation influences the number of correctly can-
celled targets. Chatterjee et al. (1992) found that increasing stimulus number
resulted both in more errors and in more correctly cancelled targets. Kaplan
and colleagues (1992) reported that increasing the number of distractors
without changing target number resulted in fewer correct cancellations.
Taken together, these reports suggest that T/D ratio is more pertinent than
stimulus number in neglect. Differential effects on these tasks might also
be expected with different neglect subtypes. Neglectful patients with poor
attentional capacity would be expected to perform most poorly on tasks with
greater intrinsic distraction since these depend on more selective attention.
In contrast, individuals with intentional or hypokinetic neglect would likely
experience more difficulty with the larger shifts in attentional focus required
between potential targets on forms with fewer stimuli. Similarly, hypokinesia
in parkinsonism may have influenced the error rate in the Villardita, Smirni,
and Zappala (1983) study, since patients were faced with the competing de-
mands of increased stimuli and smaller interstimulus distances. Impairments
in visuomotor scanning are major contributors to the cognitive deficits fol-
lowing traumatic brain injury and are more sensitive to T/D ratio than stimu-
lus number on the paradigm used in this study (Geldmacher and Hills, in
press). The effect of T/D ratio observed in this study should therefore be
attributable to the speed and efficiency of redirections of attention in the
search field, or the speed of information processing when the attentional
focus is appropriately redirected, or a combination of both.
The combined measure of cancellation performance appears to capture
the more subjective, qualitative aspect of performance that might be consid-
ered the quality of the search. It was designed to incorporate the extremes

of performance that characterize visuospatial attentional impairments de-

scribed in neglect and other cerebrovascular syndromes, neurodegenerative
diseases, traumatic brain injury, and other illness states affecting spatial at-
tention and cognition. The impulsive or impersistent subject who completes
the task quickly, but misses many targets, will obtain a low score on the
basis of accuracy. Conversely, the compulsive or perseverative subject who
laboriously retraces previously searched areas will generate a low score be-
cause of the increased time required to complete the task. In addition, the
ability to draw meaningful conclusions from analysis of the spatial distribu-
tion of errors on random array tasks is preserved. The scoring system further
incorporates the timed performance aspects of orderly array tasks. The intent
of this scoring method was not to supplant more detailed analyses, but rather
to consolidate aspects of previous scoring approaches into a single broad
clinically relevant measure. It was designed to be easily used by rehabilita-
tion therapists and other health care providers as a method of following
visuomotor scanning, directed attention, or visuospatial information pro-
cessing over the course of recovery from or treatment of brain lesions.
The combined performance measure obviously cannot answer all ques-
tions about a patients visuospatial search performance. For instance it will
not distinguish between impulsiveness and impersistence, and experimenter
observations or other measures will continue to be necessary to clarify the
reasons for low performance scores. Nonetheless, as one measure of ability
it can provide a useful tool for monitoring recovery from pathologic extremes
of performance or the success of therapeutic interventions.
Ideally, the combined measure used in this study would allow reproducible
quantitative measurement of performance on sequential testing and facilitate
longitudinal assessments of decline, recovery, or pharmacologic intervention
in disorders of visuospatial function. Its value in comparison to other meth-
ods of assessing cancellation performance such as errors/target or errors/
time has not been demonstrated in this small normal sample. Indeed, the
choice of measure may vary depending on the clinical population of interest.
With the insights into normal performance provided by this study, future
explorations are planned to address the utility of these tasks and measures
of performance in pertinent clinical populations.

Albert, M. L. 1973. A simple test of visual neglect. Neurology 23, 658664.
Chatterjee, A., Mennemeier, M., & Heilman, K. M. 1992. A stimulus response relationship
in unilateral neglect, the power function. Neuropsychologia, 32, 11011108.
Gauthier, L., Dehaut, F., & Joanette, Y. 1989. The Bells test, A quantitative and qualitative
test for visual neglect. International Journal of Clinical Neuropsychology, 11, 4954.
Geldmacher, D. S., & Hills, E. C. Effect of stimulus number, target-to-distractor ratio, and
motor speed on visual spatial search quality following traumatic brain injury. Brain Injury,
in press.

Geldmacher, D. S., Doty, L., & Heilman, K. M. 1994. Spatial performance bias in healthy
elderly subjects on a letter cancellation task. Neuropsychiatry, Neuropsychology, and Be-
havioral Neurology, 7, 275280.
Geldmacher, D. S., Doty, L., & Heilman, K. M. 1995. Letter cancellation performance in
Alzheimers disease. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 8,
Geldmacher, D. S., Rowland, D., & Riedel, T. 1995. Age effects on letter cancellation tasks.
Journal of the International Neuropsychological Society, 1, 129.
Kaplan, R. F., Verfaellie, M., Meadows, M.-E., Caplan, L. R., Pessin, M. S., & DeWitt,
L. D. 1991. Changing attentional demands in left hemispatial neglect. Archives of Neurol-
ogy, 48, 12631266.
Rapcsak, S., Verfaellie, M., Fleet, W. S., & Heilman, K. M. 1989. Selective attention in hemi-
spatial neglect. Archives of Neurology, 46, 172178.
Ruff, R. R., Evans, R. W., & Light H. 1986. Automatic detection vs. controlled search, A
paper and pencil approach. Perceptual Motor Skills, 62, 407416.
Stone, S. P., Halligan, P. W., Wilson, B., Breenwood, R. J., & Marshall, J. C. 1991. Perfor-
mance of age matched controls on a battery of visuospatial neglect tests. Journal of Neu-
rology, Neurosurgery, and Psychiatry, 54, 345350.
Villardita, C., Smirni, P., & Zappala, G. 1983. Visual neglect in Parkinsons Disease. Archives
of Neurology, 40, 737739.
Weintraub, S., & Mesulam, M. M. 1987. Right cerebral dominance in spatial attention, Further
evidence based on ipsilateral neglect. Archives of Neurology, 44, 621625.