A neuropsychological test battery for the Apple ll-E

A neuropsychological test battery for the Apple ll-E

Int. J. Man-Machine Studies (1986) 25, 453-467 A neuropsychological test battery for the Apple II-E M. C. MOERLAND, A. P. ALDENKAMP AND W. C. J. ALPH...

2MB Sizes 6 Downloads 54 Views

Int. J. Man-Machine Studies (1986) 25, 453-467

A neuropsychological test battery for the Apple II-E M. C. MOERLAND, A. P. ALDENKAMP AND W. C. J. ALPHERTS

Instituut voor Epilepsiebestrijding, Meer en Bosch/De Cruquiushoeve, Achterweg 5, 2103 S W Heemstede, The Netherlands (Received 20 March 1986 and in revised form 22 July 1986) The computerized part of a neuropsychological test battery for the assessment of cognitive functions in epilepsy is described. The test module forms the central part of the software. The report and research modules work on the data generated by the test module. The upper layer of the test module is formed by the Test Operating System (TOS), which enables flexible and fast shifting of tests and permits optimum use of memory and assembler routines. Tests currently in use are described and quantitative as well as qualitative improvements of assessment procedures are illustrated. Accurate timing of stimulus-and-response events is seminal to several of these improvements and may be of special importance for research in epilepsy. Several problems which may arise and accumulate from the current lack of standardization of hard- and software design in the research on automated psychological testing are summarized.

Introduction Experiments with automated psychological testing have been going on since the early 1960s and are not bound to the use of (micro) computers per se (Space, 1981; Thompson & Wilson, 1982). Early work on computerized testing involved the (off line) scoring and interpretation of personality inventories like the MMPI (Pearson, Swenson, Rome, Mataya & Brannick, 1965), followed by experiments which also included the (on line) administration of tests in the personality and cognitive domain (Elithorn & Telford, 1969; Space, 1975). At the end of the 1970s microcomputer technology gave a new impetus to these testing practices. Especially in the U.S.A. a number of automated psychological tests are now commercially available, although many of these tests concern the scoring and interpretation of tests in the personality domain. In our Institute we have been using automated forms of several personality inventories (including MMPI, EPPS and Cattell's IPAT) since 1978. Programs for the administration and scoring of these tests have been written in Fortran and run on Zilog-Z80 microcomputers. Among other limitations, these systems are bound to textual displays and can therefore only handle a narrow range of psychometric tests, i.e. questionnaires and inventories. In 1984 a Research Grant from the Dutch Health Organization TNO (Grant: CLEO-A52) enabled us to explore the wider applicability of computerized testing within the context of neuropsychological assessment in epilepsy. The theoretical underpinnings of this project have been outlined by Aldenkamp et al. (in prep.). This article describes the general features of our testing system and gives more specific details on software development. 453 0020-7373/86/100453+ 15503.00/0

9 1986 Academic Press Inc. (London) Limited

454

M. C. M O E R L A N D

ET AL.

Choice of hardware The choice of a particular hardware configuration is a doubtful enterprise as long as the psychological community has not arrived at national or international standards for hard- and software development. Admitted that such standards could only be provisional, while "revolutionary" new machines are announced every month, it would take the burden from institutes missing expertise and might prevent the waste of programming effort caused by developing the same tests for different machines. Besides, general hardware experts often fail to acknowledge the specific requirements a psychological test system needs, which may be indeed of minor or no importance to those concerned with the automation of a salary administration or a steel rolling mill. Massive internal and external storage capacity for instance is often overweighted against the facility to control stimulus exposure and response latency times within millisecond precision; the latter being a conditio sine qua non for an automated psychometric test system. Several of the criteria for hardware selection, important for our neuropsychological test system, are summarized in Beaumont (1981). Some of these are quite obvious and pertain to features which are available for almost any current microcomputer system. Among these are facilities to display highresolution color graphics and to interface different kinds of peripherals. Commercially available peripherals like light-pen, graphics tablet and joystick or self-designed pushbutton devices will often support or replace the keyboard as the primary response medium. The quality and costs of these features, however, vary considerably from system to system. Accompanying software may become a time saver and program enhancer. An important criterion not mentioned by Beaumont is the facility to control stimulus exposure and response latency times. Reaction time measurements form the keydependent variable in cognitive research and modem technology permits these techniques to supply classical psychometric assessment (Sternberg, 1981). More specifically, in the area of epilepsy research, knowledge of temporal relationships between physical events and clinical phenomena can be of utmost importance in understanding these phenomena (Aarts, Binnie, Smit & Wilkins, 1983). The specific problems of timing with raster scan displays have been well documented (Reed, 1979; Lincoln & Lane, 1980; Cavanagh & Anstis, 1980). Although some inherent limitations do remain in controlling visual exposure times with raster scan techniques, some others can be resolved by a specific hardware design or can be put under programmer's software control (Diener & Smee, 1984). Timing of response latencies can be handled by using a hardware clock card with ms precision or by specific software techniques with less than 50 p, (!) s precision (Femano & Pfaff, 1983). Together with local requirements of multiple and easy transportable systems and its widespread use in psychological labs all over the world, the Apple II-e microcomputer seemed our best choice at that time (see Fig. 1). Each test system consists of an Apple II-e microcomputer, a Taxan Extended 80 columns card, two disk-drives, a Taxan RGB Vision II color monitor, the Apple Graphics tablet, the Gibson light-pen system and the Apple Imagewriter matrix printer. A special push-button device has been interfaced to the game I/O. Two slots have been reserved to interface a tachistoscope and audio recorder. The Taxan 80 columns card was chosen because it additionally supplies an interface with an RGB color

NEUROPSYCHOLOGICAL TEST BATTERY

FIG. 1. Tr

455

configuration.

monitor and color display facilities in text mode. The RGB color display is markedly superior to the standard PAL (NTSC for North America) composite video display.

Programming language Programs run under the Disk Operating System 3.3 and are written in Applesoft Basic. Real time applications have been programmed in 6502 assembly language. Although Basic is not the most versatile programming language, it is most widely used in microcomputer programming and a Basic interpreter is part of the Apple's firmware. Source codes of both the interpreter and the DOS 3.3 operating system have been published and well documented in the past years, which allows unconventional use of this software for experienced programmers. Moreover, most of the software which comes with the peripherals is tailored to its use from within Basic and it might take a hard time or even be impossible to use them from a less well-documented Apple Pascal or other language. Vital shortcomings, such as the slowness of program load procedures under DOS 3.3, have all been overcome so far with reasonable programming efforts. Apart from the system software and the use of a program editor, assembler and debugger, the Higher Text Program (Aldrich & Aldrich, 1981) is used in several tests. This program has been designed to be built in ones own software and is able to generate numerous types and sizes of characters and figures on the high graphic screens.

456

M . C . MOERLAND E T A L .

Program modules The software consists of three separate modules: 1. The TEST MODULE handles the administration, scoring and saving of the data in a so-called random-access file on floppy disk. Our main efforts center on this module, which will be described in detail below. 2. The REPORT MODULE reads the datafile created on disk by the test module and generates a summary of the test results of an individual testee, optionally on screen or printer. The program calculates meaningful statistics on raw scores, provides available norms and meaningful subdivisions of scores, e.g. left v s right-hand performance or performance on the first half v s the second half of the test. 3. The RESEARCH MODULE can selectively read out the datafile and store these "group data" in another file for general research purposes or for local norm development. Using a terminal emulation program, our Apple II-e is converted into an asynchronous terminal which is connected by means of a modem and telephone line to the Cyber Mainframes of the University Computer Centre Amsterdam (SARA). Data are read out and sent to the host computer where powerful data manipulating and statistical programs (including SPSS and BMDP) are available for further processing. THE TEST MODULE

Programming a single test is different from incorporating the same test in an integrated battery. With a single test one can be satisfied if after the boot and progra m load procedures the system shows up with instructions to input identifications or start the test. Questions of program design, memory use, documentation, instructions, datafile format, etc. can be tailored to the specific test involved; while time-consuming start-up and load procedures are often taken for granted. Within the concept of a test battery, however, these questions require more precise consideration. Structural similarities in program design, outer appearance and start procedures of different tests may enhance programming efforts and speed up the process of acquaintance by testee and test technician. Common use of assembler routines and drivers by different tests can also be efficient, especially if this software has not to be reloaded from disk on every occasion. Problems arise when programs have to run in different parts of memory, depending on the all or non-use of (graphic) screens and peripherals. This makes simple "chaining" of programs impossible. Complications also occur when programs have to be loaded from physically different diskettes. Provisions should also be taken to make keyboard input of testee's identifications superfluous on different test runs. Finally the speed of DOS 3.3 load procedures is known to be slow, mainly due to some internal shortcomings of the file manager. To meet these problems and shortcomings adequately a vital piece of software has been written, which we might denote as the Test Operating System (TOS). TEST OPERATING SYSTEM (TOS)

(See Fig. 2)

The TOS consists of three parts: 1. A speed-load utility (Bongers & Schouten, 1985) which is able to load Applesoft and binary files 12 to 15 times faster than under standard DOS 3.3 load procedures,

NEUROPSYCHOLOGICAL TEST BATTERY

457

MainmenuI

Speed loader

L Assembler routines $ 8 0 0 - $ C 0 0

FIG. 2. Test operating system (T.O.S.)

where less than 1 k per second can be loaded. The loader is fully relocatable, occupies 3 k RAM and is put into memory by a "bootstrap" routine at the startup. Speedload files (including DOS 3.3) are stored in contiguous blocks on a fast boot diskette by means of the speedload master utility program. Fast boot diskettes are fully compatible with normal DOS 3.3 procedures. Apart from loading the different test programs within a few seconds or less, this offers the opportunity to take specific advantage of the Apple's bank switched and auxiliary memory areas. Looking at the Apple II-e's memory map (Watson, 1982, p. 72), 16 kbyte of bank switched RAM has been mapped in the area $D000-$FFFF and is available for data or program storage by setting the correct soft switches. The 12 k ROM in this address space contains the System Monitor and the Applesoft Interpreter. Another 64 kbytes of memory have been added by installing the extended 80 columns card in the auxiliary slot. In combination with the speedloader, this gives a neat opportunity to speedload all kinds of software needed at different test occasions in auxiliary or bank switched RAM at the very start-up of the system and within less time than a normal start-up and program load procedure would take. Programs are speed-loaded and moved to auxiliary RAM for future use by the built-in auxmove routine at $C311. The bank switched area can be made available by setting the correct soft switches. The bank switched area has been reserved by us for test-specific assembler routines like timing of push-buttons, etc. The auxiliary memory part is used so far to store recurrent menu programs, drivers for light-pen and graphics tablet, the speedloader program and the Higher Text Program (Aldrich & Aldrich, 1981).

458

M . C . MOERLAND ET AL.

2. A nucleus assembler program which should be permanently available to all programs. The memory area $800-$C00 has been reserved for this purpose by putting lomem to $C00 at the very beginning of the start-up procedures. The program takes care of setting and resetting soft switches when bank switched memory needs to be called from within a test program. It also holds address tables and subroutines to move either assembler or Applesoft programs from auxiliary to main memory and optionally can move control to these programs. In the same area a range of memory locations has also been reserved to pass variables between programs or assembler routines. Applesoft programs can " p o k e " here important information such as test number and memory configuration needed, while reaction times, for example, on individual trials can be " p e e k e d " from other fixed locations. 3. An Applesoft general menu program. This program is loaded and stored in auxiliary memory during start-up. The programs' main menu shows the different test areas available, while moving on to the submenus gives the opportunity to speed-load and run individual tests. Before loading a test the program can check in a split second whether the correct disk is in drive 1 and undertake actions to correct it if this is not the case. In some cases it is also necessary to modify certain Applesoft pointers before running the program. Running a test will automatically overwrite the general menu program, but from within every test it is always possible to return to the main menu or one of the submenus. It should be stressed that moving through menu's or loading tests only takes a few seconds or less. THE T U R N KEY SYSTEM

The software has been developed as a so-called "Turn key system", which means the test technician only has to put the diskette with the tests in drive one and turn on electricity to get into an interactive conversation with the system. First testee's name, "patient n u m b e r " and test date are requested. Mathematical checks are performed on the validity of the digit strings of date and patient number and the name is concatenated to a m a x i m u m of 20 characters. These three data are then put aside as program variables and protected for use at different program runs. From this point the conversation is limited to "two key" choices from a multiple-choice menu. The top of the screen shows the alternatives from which to choose, while the status lines at the bottom show the different test parameters currently in use. A choice from the menu is taken by typing a letter f r o m the keyboard. Where keyboard input is requested all "invalid" choices are locked off by the software. The program "highlights" a valid choice immediately, but only undertakes action when the choice is confirmed by typing a R E T U R N . F r o m the main menu one can choose the psychological domain from which to take a test. Memory, attention/concentration and perceptual motor tests have been constructed and software provisions have been made to incorporate tests for problem-solving activities and language functions. Each of these domains physically corresponds with a diskette which has to be put in drive 1 of the test system. From the main menu one can additionally choose to get instructions on the screen or input a new patient name, number and test date. The submenus on the second level show a list of tests available in this domain. After a choice of a particular test is confirmed, a speedload and run of this test is done at a particular place in memory. In the first line of every test a call is made which takes care for a particular memory configuration. Standard choices in every test are instructions, practice trials, back to main menu, back to submenu and

NEUROPSYCHOLOGICAL TEST BATTERY

459

start of test, using the same letter from test to test. Other choices are test-specific and involve for instance numbers of trials to be used, exposure time of .test stimuli, time delay between test items, kind of stimuli to be used and self-paced or machine-paced test taking. Test-specific choices often lead to submenus from which a choice is to be made in exactly the same manner (see Fig. 3).

FIG. 3. Test menus may lead to submenus: two successive displays. After confirmation of the choice by the return key, the value o f the test parameter is instantly updated in the bottom status lines and the program branches back to the test main menu. The keyboard's escape key is an invisible choice, which can be used at almost any point during run time. It breaks the current action of the program, optionally lets you save the (incomplete) data and returns you to the menu of the particular test. Although a little expertise is sufficient to break and restart a running program, the escape sequence fits more elegantly into our turn key concept and besides is able to save incomplete data. The data are saved in random access files. Although these files are somewhat less economical in memory use c o m p a r e d with normal sequential text files, they offer more convenience when selectively reading out data. The program always requires you to put the data diskette in the second drive. Again error handling routines forestall the program losing control when I / O errors occur; e.g. after putting in the wrong diskette. The line number and type of error are sent to the screen with a short note on correction measures to be undertaken.

Description of tests DIGIT SPAN TEST (see Fig. 4) The regular forward and backward sequences of the Wais Digit Span can be administered as either auditory or visual. I m p r o v e d standardization for the auditory form has been gained by a moving arrow on the screen which determines the sequence and tempo of the digit string spoken out by the test technician. Testee's response

460

M. C. MOERLAND E T A L .

FIG. 4. Digit Span Test in response mode.

sequences are fed into the computer by pointing the light-pen to the digits in a square matrix on the right half of the screen. The response sequence is instantly displayed at the bottom of the screen and can be corrected easily until the light-pen hits " e n d of response". The program does take care of the correct criteria for test break. SIMPLE REACTION-TIME TEST

For a chosen number o f trials, reaction times for simple auditory or visual stimuli are measured. Stimulus exposure endures until a push-button response is given. The interstimulus interval is randomly varied from 3- to 5 s. BINARY-CHOICE REACTION-TIME TEST

Either a red- or a green square-inch block is displayed in a random sequence in either the left or right half of the C R T display. Testee is asked to push one of two buttons corresponding to the position of the colored block on the screen. In the self-paced form a response is instantly followed by substitution o f another block in either the same or the opposite position. Speed and accuracy of responses are measured and the test breaks after a pre-chosen number of trials. In the machine-paced form, speed of exposure is a function of the ratio of correct and incorrect responses. The test breaks when the total number of correct responses falls below 50%. A non-computerized form of this test has been used in our institute as a test for mental strength for m a n y years and the computerized form resembles the old one in almost every respect.

N E U R O P S Y C H O L O G I C A L TEST BATTERY

461

VIGILANCE TASK A string of eight characters is displayed in the center of the screen for a pre-determined number of trials. Exposure time and interstimulus interval can also be chosen in advance. The testee has to decide on the appearance of a character "A" at a random position in the stimulus string, which consists half of the trials exclusively of the letter X. Responses are given by pushing the yes or no button on the response medium. In this case we chose the game paddles as primary-response medium. The paddles can easily be held while the testee sits back and performs the task, which might take half an hour or more. The test administration is machine-paced; failing to respond during the inter-stimulus interval means a trial is counted as missing. During a period of 100 ms after stimulus exposure, no responses are accepted to avoid confusing a late response on the former trial with the new one. Speed and accuracy of the responses are recorded; sensitivity and bias are computed according to the signal detection model (Davies & Parasuraman, 1982). CORSI BLOCK TAPPING TEST The test has been adapted from a computerized form of Corsi's Block Tapping Test developed by the EEG department of our Institute (Aarts et al., 1983). The specific aims of the latter study induced several deviations from the original Corsi Block Test (Milner, 1971), which are not pertinent to the act of computerizing this test itself. So the test was presented as a continuous performance task and the block configuration had been constrained to seven instead of nine blocks. For our purposes it seemed reasonable to stay more close to the original design. We maintained the presentation of the block sequences by flashing them on and off, while immediate recall was also required by tipping with the light-pen to the correct blocks. However, we re-established the original number and pattern of blocks, while the test was also restricted to two fixed parts. First, the memory span for block sequences is determined by adding one block to every successfully recalled sequence. Second, the superspan (normal span +1) performance with a recurrent pattern on every third trial is determined for a total of 24 trials. Block sequences have been adapted from Smirni, Villardita & Zappala (1983), who found that memory performance on the Block Tapping Test not only depends on sequence length but also on other path characteristics. Moreover these findings forestall careless use of random functions in test development. COMPUTERIZED VISUAL SEARCHING TASK (CVST) (see Fig. 5) This test has been described by Goldstein, Welch, Rennick, & Shelly (1973) as a successful instrument for discriminating brain damaged from normal and psychiatric populations. DeMita & Johnson (1981) developed a computerized version of the same task for the DEC LSI-11/03 computer and also report successful discriminating potential of this test for brain-damaged and non-brain-damaged groups. The task consists of finding a grid pattern out of 24 which matches the one in the centre of the screen. Grid patterns are displayed in checkerboard fashion and are numbered from 1 to 24. The target pattern is denoted by an arrow on the right side and is selected from each of the four quadrants to balance the location of the matching grid. Two displays of 24 different grid patterns can be provided. Responses are recorded by typing the correct number from the keyboard.

462

M.C. MOERLAND

Illll HI I IllU I IIIII IIIII IIIII IIII

Illll l II IIIII IIIII IIIII

Z IlUl IIIII

IIIII 6

IIIII

IlUl

UlII

III I

I II

l Ill lllll IIUl II II I III 3 IIIII 4 IIIII III I IlUl lUll lUll UlU III I IllU 7 IIIII ~ IIII II II U II IIIII

IIIII

ET AL.

U II Illll IIIII S

II II

IIIII 9

I III IIIII I I I I I tO

III I

UlII

II II I III I III UlII IIIII IIIII IIIII II II lUlltt I IIIlz IIIII <_ 1111113 1111114 UBD ilUl HI HI II II II II IIIII lUll Illll IIIII lUll lUll IIIII IIIII III I IIIII IIIII IIIII I Ul IIIII IIII U Ils I 1116 1111117 1111118 I I I I I t 9 IIIII IIIII IIIII I III IIII IIIII IIIII I III IIIII IIIII IIIII IIII! IIIII III I IIIII I III IIII lUll IIIII IIIII IllU20 I I 1121 I I I I z z I 11123 I I I 1 2 4 IIIII IIIII UlII IIIII IIIII II II IllU I III IIIII III I UlU II II

[--

BLOK IS GELIJ~ RRH HUMMER:

FzG. 5. Computerized visual searching task: Display 1.

FINGER-TAPPING TEST The test has been adapted from the original Halstead Reitan Battery. Speed of fingertapping is measured for the index finger of the right- and left hand separately; five times for a period of 10 s each. Feedback on performance is given by a counter in the center of the screen, while the number of taps on different 10 second trials is shown at the b o t t o m of the screen. The tapping device can either be the two push-button keys on the keyboard or another device connected to the PB0 and PB1 of the game I / O . ICONICS TEST The test has been designed to assess iconic memory, which refers to "the persistence of visual impressions and their brief availability for further processing" (Solso, 1979). The task has been adapted from Averbach & Coriell (1961) who used tachistoscopic presentation of letter tableaux with partial recall as reproduction method. The testee is asked to fixate a white pixel in the center of the screen and expect an auditory signaling cue. This cue precedes the display of a four-letter matrix in the center of the screen for a m a x i m u m of 200 ms. Forty milliseconds later an underlining character is displayed in one and only one of the letter positions. The testee is now asked to name this letter. Again speed and accuracy measures are recorded. VISUAL HALF FIELD TEST (VHT) The V H T is known as one of the techniques to study functional differences between the left and the right cerebral hemisphere. A variety of stimuli and tasks has been used with this test, but they all require controlled stimulus presentation in one or both of the visual hemifields. Our test combines many of these tasks and stimuli; while test

NEUROPSYCHOLOGICAL

TEST BATTERY

463

parameters such as number of trials, exposure time, etc. can be varied as usual. Verbal stimuli can be either single characters or two-, three- and four-letter words. The non-verbal stimuli consist of nonsense figures, which are made by random combination of four simple geometric shapes from a pool of 12. Single stimuli are presented in a pseudo-random sequence in either the left or the right visual field. Double stimuli can also be presented bilaterally, depending on the task to be performed. Tasks requested are simple naming, matching or recognition of the target stimulus from a set of four. In order to control for fixation, the testee is asked to read aloud a random series of digits, appearing at the point of fixation in the center of the screen. The digit series varies in length and the last digit is simultaneously displayed with the target stimulus. Responses can be judged by the test technician by pushing the correct buttons of the push-button display. In case of matching or recognition tasks push-button responses are given by the testee and can be fully processed and judged by the computer. SEASHORE RHYTHM TEST The Rhythm Test is a subtest of Seashore Measures of Musical Talents (Seashore, Lewis & Saetveit, 1960) and is also part of the Halstead Reitan Battery. The testee is required to match 30 pairs of rhythmic patterns which are the same on half of the trials. Written answers have been replaced by push-button responses, which additionally allows accurate timing of response latencies. Instead of using a tape recorder, the patterns are generated by the internal speaker of the microcomputer, following the technical specifications of Seashore et al. (1960). Timing of responses starts at the end of every second rhythmic pattern. "Premature" responses during the stimulus exposure interval had to be ignored, due to the serial processing character of the microcomputer. The trial interval ends after a response has been received or a preset time interval has lapsed. FIFTEEN WORDS TEST The Fifteen Words Test is a Dutch word-learning task (Deelman, 1972). Fifteen words are presented as either auditory or visual on five consecutive trials. The testee's free recall is requested on every trial, while delayed recall is introduced after 15 min. Free recall procedures are well known from experimental psychology and a wide range of computational measures have been developed to study accuracy and organization phenomena (Murphy & Puff, 1982; Pellegrino & Hubert, 1982). These algorithms can undoubtedly be easily implemented in the report module. If on-line response registration could also be realized, the whole task of scoring can be taken away from the test technician and supplementary speed measures can again be provided. However, up till now the automation of the administration part has had little success. We either underestimated the speed of responses given by the testee or overestimated the speed by which response alternatives can be judged by the test technician and fed into the computer by means of the light-pen. Although training may improve performance, tester's speed and accuracy run short when a "burst" of vocal answers is given, which is not uncommon at the beginning of a recall period. We have yet to consider computerizing the administration part by making use of the graphics tablet, where multiple choice and written answers can easily be combined and speed of registration is not dependent on speed of raster scan.

464

M. C. M O E R L A N D E T A L .

RECOGNITION TEST This test combines many tasks in one. From the test menu one can change almost any test parameter which has been described as relevant (Murdock, 1982). The prototypical task involves a study phase where the item material is presented and a test phase where recognition of a study item is requested. The number of study items may be varied up to eight and the items can be digits, numbers, letters, figures or words of different length. Exposure time and recall interval can be varied as usual and push-button responses are either yes/no or forced choice from one or more response alternatives. In case yes/no answers are involved, data can be analysed following Sternberg's additive factor model (Sternberg, 1969). OPERATIONAL EXPERIENCE

The need to develop test software in a continuous interaction with clinical try-outs became apparent from the very start of our project. In fact our first programming efforts resulted in the failure to automate the administration of the Fifteen Words Test. Clinical try-outs and subsequent software modifications have become routine practice in further test developments. Improvements have involved minor and major changes in difficulty level, kinds of test materials, number of practice trials, response medium, etc. Most important however, these clinical try-outs almost guarantee a quick detection of "bugs" which have passed initial program debugging. From a purely psychometric point of view an obvious lack of norms in these initial stages will handicap clinical interpretation of the test results. Therefore care has been taken that in these situations several well-established psychometric tests always formed the major part of the clinical test procedure. By now data are accumulating which will make it possible to compare levels of performance within our epilepsy population. Research on reliability, validity and norming of tests on a non-epileptic population has also been planned. Specific resistance against computerized testing has not been noticed in our patient group. However, in rare instances we suspect that the frequency of the raster scan display or the "flashing" of stimuli with short exposure times may have provoked observed seizures. Additional experience has been obtained in research programs, mainly in the evaluation of side effects of anti-epileptic drugs. These try-outs led to several improvements of our test system. For example, drug evaluation often requires repeated measurements designs for which we had to develop parallel test versions. In our experience, this can be easily fulfilled in computerized testing by random selection of test items from a larger pool or by generating a limited number of fixed parallel test versions. These strategies have been applied in the recognition tasks and in the Fifteen Words Test respectively.

Discussion In recent years interest has been renewed in computerized testing. Single tests and test batteries have been developed at different institutions and for different machines. In our case we started with automated versions of existing personality inventories. Also, in the cognitive domain several of our tests are computerized versions of established

NEUROPSYCHOLOGICAL

TEST BATTERY

465

tests (Digit Span, Seashore Rhythm, Tapping Test, etc.). A step away are those tests which are based on existing tests, but have been adapted for computer administration. Among the latter are the Corsi Block Test and the visual form of the Digit Span Test. Finally several tasks have been added to our test repertoire, which have seldomly been available in clinical settings up till now (e.g. The Visual Half Field and the Iconics Test). The main advantages of computerized test procedures have been described as being convenience, economy and objectivity (Elithorn, Mornington & Stavrou, 1982). Especially when integrated in a flexible TOS, a series of different tests and instructions can be accessed instantaneously. Tests can be administered in a highly standardized fashion and reports of test results can be produced without delay. Some qualitative features seem to be even more promising. Real-time control of stimulus and response events is the critical component in introducing techniques from experimental psychology into the clinical field (Sternberg, 1981) and its specific role in improving psychological assessment in epilepsy has been stressed. Software control of other test parameters, such as test length and item difficulty, also allows more flexible test administration and permits "testing the limits" in a systematic way. The concept of standardized tests and test administration is preserved in the default setting of test parameters at the very start-up of a particular test. Although alternatives have been proposed (Elithorn & Telford, 1969), computerized testing has been developed very much within the traditional psychometric approach (Thompson & Wilson, 1982). In several ways current developments in computerized testing also form a potential threat to classical psychometric requirements. These questions seem to be pertinent even in the case of computerizing existing tests. For instance, Watts, Baddeley & Williams (1982) report systematic differences in level of performance on computerized versions of Raven's Matrices and the Mill Hill Vocabulary Scales. Furthermore, they report minor differences in reliability between computerized and paper and pencil versions of these tests, while the validity of the computerized version could not be challenged at all. In addition performance differences between hand- and computer-administered personality inventories have been reported by Space (1981). Another threat arises from the fact that minor and major differences slip in when computerizing the same test for different hardware configurations. So our Computerized Visual Searching Task for the Apple II-e differs on "minor" points from the one published by DeMita & Johnson (1981), which in ~urn differs in several respects from the original one by Goldstein et al. (1973). In the case of developing experimental tasks, a wide scatter of tasks will almost automatically prevent them from reaching a solid psychometric basis. Therefore some standardization of hard- and software and commitment to tests being worthwhile to be programmed is urgently needed. Even in case one feels little commitment to classical psychometric test requirements, this statement becomes a truism when one starts automating test procedures. Programming tests is time-consuming and the costs of software very soon outgrow the hardware investment. Software experiments outgrow the phase of experimenting with a single test. Cooperation between institutes could improve the quality and increase the number of available tests, while justifying the personal and financial investments which now have to be taken by single institutes. We have started cooperation with some institutes to develop new tests, to collect new data for norms and to convert software from Applesoft Basic to MSDOS Turbo Pascal.

466

M.C. MOERLAND ET AL.

References AARTS, J. H. P., BINNIE, C. D., SMIT, A. M. & WILKINS, A. J. (1983). Selective cognitive impairment during focal and generalized epileptiform LEG activity, Brain, 107, 293-308. ALDRICH, D & ALDRICH, R. (1981). Higher Text Plus. Washington: Apple PugetSound Program Library Exchange. AVERBACH, I. & CORIELL, A. S. (1961). Short term memory in vision. Bell System Technical Journal, 40, 309-328. BEAUMONT, J. G. (1981). Microcomputer-aided assessment using standard psychometric procedures. Behavior Research Methods & Instrumentation, 13, 430-433. BONGERS, C. & SCHOUTEN, W. (1985). Speedloader Pro, Een snel-laad utility voor de Apple H familie. Rotterdam: CBWS Productions. CAVANAGH, P. & ANSTIS, S. M. (1980). Visual psychophysics on the Apple II: getting started. Behavior Research Methods & Instrumentation, 12, 614-626. DAVIES, D. R. & PARASURAMAN, R. (1982). The Psychology of Vigilance. London: Academic Press. DEELMAN, B. (3. (1972). Etudes in de Neuropsychologie. Psychologisch r bij patienten met gelocaliseerde cerebrale stoornissen. Groningen: Waiters Noordhof. DEMITA, M. A. & JOHNSON, J. H. (1981). The validity of a computerized visual searching task as an indicator of brain damage. Behavior Research Methods & Instrumentation, 13, 592-594. DIENER, D. & SMEE, W. P. (1984). Apple Tachistoscope. Behavior Research Methods & Instrumentation, 16, 540-544. ELITHORN, A. MORNINGTON, S. & STAVROU, A. (1982). Automated psychological testing: some principles and practice. International Journal of Man-Machine Studies, 17, 247-263. ELITHORN, A. TELFORD, A. (1969). Computer analysis of intellectual skills. International Journal of Man-Machine Studies, 1, 189-209. FEMANO, P. A. & PFAFF, D. W. (1983). Time interval acquisition on a 6502-based microcomputer. Behavior Research Methods & Instrumentation, 15, 521-529. GOLDSTEIN, G., WELCH, R. B., RENNICK, P. M. & SHELLY, C. H. (1973). The validity of a visual searching task as an indicator of brain damage. Journal of Consulting and Clinical Psychology, 41, 434-437. LINCOLN, C. E. & LANE, D. M. (1980). Reaction time measurement errors resulting from the use of CRT displays. Behavior Research Methods & Instrumentation, 12, 55-57. MILNER, B. (1971). Interhemispheric differences in the localization of psychological processes in man. British Medical Bulletin, 27, 272-277. MURDOCK, B. B. (1982) Recognition memory. In PUFF, C. R. Handbook of Research Methods in Human Memory and Cognition. New York: Academic Press. MURPHY, M. D. & PUFF, C. R. (1982). Free recall: basic methodology and analyses. In PUFF, C. R. Handbook of Research Methods in Human Memory and Cognition. New York: Academic Press. PEARSON, J. S., SWENSON, H. P., ROME, H. P., MATAYA, P. & BRANNICK, T. L. (1965). Development of a computer system for scoring and interpretation of Minnesota Multiphasic Personality Inventories in a medical clinic. Annals of the New York Academy of Sciences, 126, 682-692. PELLEGRINO, J. W. & HUBERT L. J. (1982). The analysis of organization and structure in free recall. In PUFF, C. R. Handbook of Research Methods in Human Memory and Cognition. New York: Academic Press. REED, A. V. (1979). Microcomputer display timing: problems and solutions. Behavior Research Methods & Instrumentation, 11, 572-576. SEASHORE, C. E., LEWIS, D. & SAETVEIT, J. G. (1960). Seashore Measures of Musical Talents. Manual. New York: The Psychological Corporation. SMIRNI, P., VILLARDITA, C. & ZAPPALA, G. (1983). Influence of different paths on spatial memory performance in the block-tapping test. Journal of Clinical Neuropsychology, 5, 355-359. SOLSO, R. L. (1979). Cognitive Psychology. New York: Harcourt Brace Jovanovich. SPACE, L. G. (1975). A console for the interactive on-line administration of psychological tests. Behavior Research Methods & Instrumentation, 7, 191-193.

NEUROPSYCHOLOGICAL TEST BATTERY

467

SPACE, L. G. (1981). The computer as psychometrician. Behavior Research Methods & Instrumentation, 13, 595-606.

STERNBERG, R. J. (1981). Testing and cognitive psychology. American Psychologist, 36, 11811189.

STERNBERG, S. (1969). Memory-scanning: mental processes revealed by reaction time experiments. American Scientist, 57, 421-457. THOMPSON, J. A. & WILSON, S. L. (1982). Automated psychological testing. International Journal of Man-Machine Studies, 17, 279-289. WATSON, A. (1982). Apple II Reference Manual. Cupertino: Apple Computer Inc. /r K., BADDELEY, A. & WILLIAMS, M. (1982). Automated tailored testing using Raven's Matrices and the Mill Hill Vocabulary tests: a comparison with manual administration. International Journal of Man-Machine Studies, 17, 331-344.