Epprlrpsia. JI(Supp1.
4):S35-S40, 1990
Raven Press. Ltd., New York
0International League Against Epilepsy
Computerized Neuropsychological Assessment of Cognitive
Functioning in Children with Epilepsy
W. C. J. Alpherts and A. P. Aldenkamp
Instituut voor Epilepsiebestrijding,Heemstede, The Netherlands
Summary: The value of a range of computer-aided tests in
the neuropsychological assessment was investigated in
94-177 children with epilepsy, aged 8-18 years, compared
with 68-161 controls in the same age group. Children from
the age of 8 years could cope with rather complex tests in a
wide range of functions: reaction time measurements,
motor speed, information processing, and memory. The
speed of performance tended to increase with age in both
groups, with differences in information processing becom-
The computer is a valuable aid to assessment of
cognitive functioning, both for presenting stimuli and
accepting responses (Golden, 1989). Stimuli may be
simple or complex and either auditory or visual. Re-
sponses can be fed programmed from various devices
such as keyboard, touch screen, or voice key. Stimuli
and responses can be measured with millisecond pre-
cision, providing a highly standardized testing envi-
ronment. Subtle brain abnormalities could affect pro-
cessing speed, which is difficult to detect using
conventional pencil and paper tests, but can be
readily measured by a suitable computer system.
COMPUTER ASSESSMENT
The validity of computerized neuropsychological
assessment has been established in adults but not yet
in children, partly because they can only cope with
verbal material on reaching reading age. This imposes
an age restriction on tapping the verbal memory. On
the other hand, nonverbal processing of complex de-
signs could be comparable to that of adults, with
children only performing at a lower speed (though
not according to Piaget, for instance, who describes
Address correspondence and reprint requests to Mr. W. C. J.
Alpherts at Instituut voor Epilepsiebestrijding, P.O. Box 21, Ach-
terweg 5,2100 AA Heemstede, The Netherlands.
s35
ing apparent from the age of 12 years. The precise control
of stimulus and response required to define the mainly
minor differences between the epilepsy and control groups
can only be fulfilled by computerized testing, which
should undergo further refinement including voice and
language recognition, followed by artificial intelligence.
Key Words: Epilepsy-Seizures-Children-Cognition-
Cognitive disorders-Neuropsychological tests-Comput-
erized psychological tests.
developmental stages). Until now, only a few studies
have paid attention to the age dependency of comput-
erized assessment [Knights et al. (1973), in relation to
mental retardation].
Our work on computerized testing started in 1978
using a Zilog 2-80 microcomputer system called
FePsy. Since this type of computer could only handle
text, a number of questionnaires such as Cattell’s
Anxiety Scale Questionnaire (IPAT) and the Minne-
sota Multiphasic Personality Inventory (MMPI) were
programmed. In 1983, a neuropsychological test bat-
tery for epilepsy was developed using the Apple IIE
microcomputer (Moerland et al., 1986,1988; Al-
pherts, 1987). Over a period of 3 years, many tests
were computerized from different fields of neuropsy-
chology. Ongoing developments in computer tech-
nique led us to convert this system to MS-DOS.
THE FePsy TEST SYSTEM
The “FePsy” system is built around the adminis-
tration of tests and storage of the results in a database.
It is written entirely in Turbo Pascal and uses the
B-Tree Filer routines for the database (Turbo Profes-
sional, 1989). In this way, all of the test and analysis
functions can be performed from the program. Every
patient has a unique identification number, enabling
the system to “recognize” a patient on retesting.
S36 W’.C. J. A L P H E R T S A N D A . P. A L D E N K 4 M P

When patients are tested repeatedly or with multiple
instruments, the results are merged. Each patient has
two basic identification files, plus a file with results
for every test. Figure 1 shows an example: patient I is
tested twice (two separate sessions) with slightly dif-
ferent tests: patient 2 is tested three times on one day
(three versions within one session).
No computer peripherals are needed. All of the
responses have to be made using the keyboard. In
case of “yes” and “no” responses, a group of keys can
be used on the left or right side of the keyboard, so no
typewriting ability is required.
Output can be passed to the printer or an ASCII file
for subsequent processing by a statistical package like
SPSS,and online to the electroencephalogram (EEG)
by means of a Brainbox. This apparatus writes ASCII
text on one channel of the EEG, such as stimulus
onset, response latency, and correct or incorrect
responses, allowing the test to be correlated with
the EEG.
Retesting usually leads to a learning or practice
effect. For tests in which every item has the same level
of difficulty, we use an item pool. In every trial. the
item pool is randomly searched and the trial built up
in a balanced way. This means that the learning effect
is not due to “known items.” In other tasks (the
CVST. for instance), we are interested in the time
profile of the individual items as well. Here, a semi-
random procedure is applied, so fixed retests must be
made. The database is flexible enough to enable fu-
ture developments like tailored testing and online
statistical analysis to be included in much the same
way as an expert system.
Binary choice reaction time
This is a measure of the reaction time that includes
a decision component. In this task, there are two
stimuli: a red square on the left or a green square on
the right side of the screen. The reaction time is mea-
sured with the left hand on the left stimulus. and with
the right hand on the right stimulus.
Tappiog task
This task is an adaptation of the similar task in the
Halstead-Reitan battery. It provides a measure of
motor speed. Tapping speed is measured for each
hand.
Computerized visual searching task (CVSr)
This task deals with visual information processing
and is an adaptation of Goldstein’s visual searching
task (Goldstein et al., 1973; DeMita and Johnson.
1981). A centered grid pattern has to be compared
with 24 surrounding patterns. only one of which is
identical (Fig. 2). The test consists of 24 trials (the 24
patterns change after I2 trials).
Recognition
In the recognition task. four conditions are possi-
ble: simultaneous or serial presentation, with verbal
or nonverbal stimuli. An array with six words or four
figures, the “memory set.” is presented (the learning
phase). ARer a delay of 2 s. the screen display changes
to show one of these words (or figures) hctwcen
others. This word then has to be identified. In the
serial presentation. recall of the order of the stimuli is
required. A recognition memory paradigm is onen
applicable in patients with memory loss (Lcvin.
1986).

DESCRIPTION OF THE TASKS

Simple reaction time measurements
Basic perceptuomotor performance is assessed.
Stimuli are either auditory (800 Hz tone) or visual (a
white square on the screen). The subject is instructed
to react as quickly as possible.

File 1 File 2 Separate test files

Identification Test date Test results test 1

Patient 11.58,r-.8k- Readion times

Memory
Attention
03.03.89 Attention
- IMotor S p e e d

05.24.891 -Attention
05.24.892 -
Pab’ent -E 05.24.894
Attention
-Attention
FIG. 1. Structure of the database.
FIG. 2. Example of a computerized vlsual searching task
(CVST). (Arrow in center of gid indicates me blodc to be matched
with one of the sunarnding patterns.)

Epilepsia. Vol 31. Suppl 4. 1990

 COMPUTERIZED ASSESSMENT OF COGNITIVE FUNCTION
TABLE I. Hand preference by group
Right Left Ambidextrous
Control group 9 1.7% 6.9% I .4%
Epilepsy group 86.0% 12.6% 1.5%
Other tasks not further described here include a
vigilance task and the Seashore rhythm. Develop-
ment work has recently started on several other tests,
including problem solving and assessing hemispheric
dominance.
RESULTS
Data have been gathered from a broad range of
people with epilepsy compared with “normal” con-
trols. Here, we present data for children aged 8-18
years. The groups have been tested with an Olivetti
M24. Data on about 200 children also exist for the
Apple IIE testing environment but are not presented
here. The number of subjects tested with various
tasks were 68- 161 controls and 94- 177 children with
epilepsy.
The control group was selected from regular
schools in The Netherlands and Sweden. The epi-
lepsy group is composed of children from different
epilepsy centers and schools for special education.
Mean I.Q. levels are above 85.
Reaction times and motor speed
As the distribution of reaction times is always
skewed, a cutoff score is first calculated and an upper
limit set at two standard deviations above the mean
in every subject. Second, a lower limit is set at 130 ms
in the auditory mode, and at 150 ms in the visual
mode. Reaction times below these values are consid-
ered to be outliers. Differences are calculated for the
dominant and the nondominant hand, preferences in
each groups being divided as shown in Table 1. Left-
handedness occurs nearly twice as often in the epi-
lepsy group, consistent with other findings (Van der
Vlugt and Bakker, 1980).
Simple reaction times do not show significant dif-
ferences between the groups under comparable con-
ditions (Table 2). As expected, there is an agedepen-
TABLE 2. Simple reaction times (in ms)
~~
Auditory
Dominant
Control group 304 (95)
Epilepsy group 290 (62)
SD in parentheses.
s3 7
TABLE 3. Binary choice reaction times (in ms)
and correct responses
Reaction time Correct response
Control group 404(126) 53.5 (7.1)
Epilepsy group 446 (227) 52.6 (8.5)
SD in parentheses.
dent effect in the direction of lower reaction times
with rising age.
The binary choice reaction time tends to be longer
over the whole age range in the epilepsy group, but
the difference between the groups is not significant
(p = 0.15). There is a significant difference (p <
0.002), however, in the number of correct responses
(Table 3).
In the tapping task, the mean number of taps is not
only lower than found previously (Dodrill, 1979), but
the dominant/nondominant difference is also greater
(Table 4). Overall, for thl: dominant hand, there is a
small but significant difference between the epilepsy
and control groups. Figure 3 shows little difference
between the groups up to the age of 11 years, the
dominant hand in the control group performing con-
sistently better thereafter.
The CYST shows a more marked agedependent
effect (Fig. 4), with control subjects performing over
twice as fast at 18 years as at 8 years. In the epllepsy
group, this decrease is much less from 11 years on-
wards, resulting in a clear and significant difference
throughout adolescence. This differenceis not easy to
explain. The groups have not been studied for epi-
lepsy-related factors such as severity, type, and medi-
cation. Should these be responsible, it seems strange
that their effect is not detectable until the age of
1 1- 12 years. There could, however, be an interaction
between brain maturity and an epilepsy-related factor
such as “mental slowness” (Aldenkamp et al., 1990).
Below the age of 12 years this would not be easy to
measure accurately, but above this age the perfor-
mance of controls increases steadily, whereas some
children with epilepsy seem to have already reached
their “top” performance.
The way children with epilepsy solve the visual
searching task could differ from those in the control
Visual
Nondominant Dominant Nondominant
281 (73) 306 (72) 314 (78)
279 (57) 31 I(54) 319(61)
S38 W. C. J. ALPHERTS AND A. P. ALDENKAMP
TABLE 4. Tapping [esl scores in seconds
Mean number of taps, 5 X 10 s (SD)
Dominant
____ ~
Control group 49.6' (9.0)
Epilepsy group 47.2' (7.1)
Significant difference: p < 0.03.
group, but there is no significant difference between
the mean errors made in each group (2.21 f 2.8 for
the epilepsy vs. 2.36 f 2.3 for the control group), and
their profiles of the trial searching time are strikingly
alike (Fig. 5). This suggests a similar strategy for
problem solving. Consequently, the difference be-
tween the groups can only be expressed in terms of
slower mental speed.
In the recognition task with simultaneously pre-
sented stimuli, there is a clear difference between the
"words" and the "figures". (Fig. 6). The figures are
presented as nonsense to prevent verbal processing,
and this could explain some of the differences. There
seems to be no agedependent effect for the figures,
the proportion of correct responses staying around
50%, whereas the percentage of correct word re-
sponses increases from about 60%at 8 years to 80%at
18 years, the control group generally performing a
little better than the epilepsy group.
In the serial presentation of recognition material,
an age effect is present for both words and figures. but
there is no difference between the epilepsy and the
control groups (Fig. 7).
Simultaneous and serial recognition should repre-
sent slightly different memory processes. the former
being a genuine recognition task requiring storage of
memory and subsequent recall. Moreover, the serial
presentation introduces an extra difficulty: the set has
- 1
8 9 10 11 12 13 14 15
AGE (years)
FIG. 3. Tapping task, number of dommant-hand taps by age
(group means for epilepsy patients and controls).
Epilepsia. V d . 31. Suppl. 4. 1990
Nondominant
-
42.9 (7.7)
41.4 (7.6) YF 15
*5
8 9 10 11 12 13 14 15 16 17 18
AGE (years)
FIG. 4. Computerized visual searching task (CVST); man
s~archinstime fagoups eged 8-18 m.shomng ~ o ~ e s w e
separation of epirepsy and contrd ~ o u p after
s 11 yeers of age.
to be stored in order, a process more like "recall." In
younger children, this extra load is seen in the verbal
form, but by about the age of 16 years the difference
has disappeared.
CONCLUSIONS
The focus of this study lies in showing how com-
puter-aided testing can contribute to the neuropsy-
chological assessment of children with epilepsy. If the
computer can be used from an age of about 8 years.
performance can be accurately compared in epilepsy
and control groups. Results of computerized tests in-
dicate that children can perform the kind of tasks
employed. As expected, age effects are seen. in the
controls as well as the epilepsy group, especially with
more complicated tasks that require quick decision
making (binary choice), complex information pro-
cessing (CVST), and memory.
Children have no difficulty in handling the key-
board in a simple. specified way-no real typing is
required. Performances at the age of 8 yearsarc some-
times much slower than at I8 years (CVST), but still
I
acceptable. Motor speed (tapping) increases with age.
as expected. A smaller improvement is found in the
recognition tasks, except for the recognition of simul-
taneously presented words. Automated reading will
probably lead to enhanced performance. and research
in our department should enable forms to be adapted
for the younger and more handicapped children.
Time-based tasks, such as the simple reaction time
measurement, binary choice reaction time. and-in
more complex form-the CVST all rely on the spc.c.d
of information processing. It has been found that
16 17 18 raising the complexity of the decision and search
processes reveals an interaction with mental sped
and age. Except for binary choice reaction time. this
isapparent fromabout 11-IZyearsofapc. I f i t i \ t r u c
 COMPUTERIZED ASSESSMENT OF COGNITIVE FUNCTION
TRIALS
that no epilepsy-specific mental dysfunctioning exists
(Seidenberg, 1989), this is hard to explain. Children
of 8 years of age in both groups can perform the tasks
adequately, but intergroup differences in speed of in-
formation processing begin to emerge at about the
age of 1 1-12 years.
Since the tasks presented here are not typically
“frontal sensitive,” the results cannot be explained in
terms of maturation of the frontal lobe. An epilepsy-
related pattern of neuropsychological dysfunctioning
on the level of attention/arousal and information
processing might be postulated (Stores et al., 1978;
Aldenkamp et al., 1990).
Finally, establishingthat computerized assessment
can contribute to neuropsychologicaltesting suggests
that the method could be extended. Its combination
with EEG recording has already been mentioned, but
computerization could also be applied to neuropsy-
”
8 9 10 11 12 13 14 15 16 17
AGE (years)
FIG. 6. Simultaneous recognition of words/figures. aged 8-18
years, showing steady advance in word recognition, compared
with figures (Oh correct).
s39
FIG. 5. Similar profiles for controls and epi-
lepsy patients in computerized visual search-
ing task (CVST).
chological assessment coupled with stimulation of
depth or surface electrodes as in a Penfield-Rasmus-
sen operation. In both cases, the patient is fully con-
scious, and stimulation lasts for no more than 4 s.
The need for precise control of stimulus presentation
and response latency can only be fulfilled by com-
puter control. The same applies to the possibility of
presenting parallel test versions. For some tasks, we
need peripherals like push buttons, graphics tablets,
or even touch screens.
The field of cognitive psychology has developed
many experimental tasks and mathematical para-
digms that can easily be programmed on a computer
and could prove their value in patient settings. Until
recently, responses have had to be given mostly by
pressing buttons or keys, although many tests require
a verbal answer. We are now awaiting further devel-
opment of highly sophisticated voice-and language
18 ”
8 9 10 11 12 13 14 15 16 17 18
AGE (yews)
FIG. 7. Serial recognitionshows no difference between epikPSy
and control groups or with age ( O h correct).
Epilepsia. t‘d. 31. Suppl. 4. 1990
 S40 W. C. J. ALPHERTS AND A . P. ALDENKAMP
-recognition systems. Perhaps, it will not be long
before the incorporation of artificial intelligence
overcomes the rigidity of today's computers.
REFERENCES
Aldenkamp AP. Alpherts WCJ, Dekker MJA, Overweg J. Neuro-
psychological aspects of learning disabilities in epilepsy. €pi-
lepsia 1990;3I(suppl. 41S9-20 (this issue).
Aldenkamp AP. Alpherts WCJ, Moerland MC, Ottevanger N, van
Parijs JAP. Controlled release carbemazepine: cognitive side-
effects in patients with epilepsy. Epilepsia 1987;28:507-14.
Alpherts WCJ. Computers as a technique for neuropsychological
assessment in epilepsy. In: Aldenkamp AP. Alpherts WCJ.
Meinardi H, Storm G, eds. Education and epilepsy. LissefBer-
wyn: Swets & Zeitlinger 1987:lOl-9.
Cull CA, Trimble MR. Automated testing and psychophamacol-
ogy. In: Hineman I, Storier PD. eds. Humanpsychopharmacd-
ogy,Chichester: John Wiley & Sons 1987:I 13-53.
DeMita MA, Johnson JH. The validity of a computerized visual
searching task as an indicator of brain damage. Behav Res
Methods Instrum 1981;13:592-4.
Dodrill CB. Sex differences on the Halstead-Reitan neuropsyche
logical battery and on other neuropsychological measures I
Clin Psycho1 l97935:236-4 I.
Golden CJ. Computers in neuropsychology. In: Butcher JN, ed.
Computerized psychological assessment. New York: Basic
Books 1989344-54.
Epilepsia. V d . 31, Suppl. 4. 1990
Goldstein G. Welch RB, Rcnnick PM,Shelley CH. The validity of
a visual searching task as an indicator of brain damage. J C'tJn-
sult Clin Psqrhol 1973;41:434-7.
Knights RM. Richardson DH. McNarry LR. Automated vsclinical
administration of the Peabody Picture Vocabulary Test and the
Coloured Progressive Matrices. Am J Men! Retard
1973~78223-5.
Levin HS. Learning and memory. In: Hannay HJ. ed. Experimen-
tal techniques in human neurops)achdogy.New York:Oxford
University Pms, 198630942.
Moerland MC, Aldenkamp AP, Alphcrts WCJ. A muropsycholog-
ical testbattery for the Apple WE. Inf I Man-Machine Stud
1986:25:453-67.
Moerland MC, Aldenkamp AP, Alpherts WCJ. Computerized psy-
chological testing in epilepsy. In: Maane FJ. Mulder U M .
Sjouw WPB. Akkennan AE. eds. Computers in ps,wholwy
methuds. instrumentation and psychodiagnostrcs. Lisse/&r-
wyn: Swets & &itlinger, 1988: 157-64.
seidenbmg M. Ncumpsychological functioning of children with
epikpsy. In: Hermann BP. Seidenbq M.eds. Childhnd epi-
lepsies: neuropsychological. psychosocial and intervention
aspecfs.ChiChester John Wiley & Sons. 1989:71-81.
Storm G, Hart J, Piran N. Inattentiveness in school children with
epilepsy. Epilepsia 1978:19169-75.
Turbo RofeSionrl.&Tree Filer5.0. User> manual. Scotu Valley:
TurboPomr Software, 1989.
Van der Vlugt H. Bakker D. Latcralization of brain function in
persons with epikpsy. In: Kulig BM rt al.. cds. Epilepsy and
behavior. LissefBcrwyn: S m & & Znllingn. I9sO..91-7.