Information processing components of substitution test performance

Information processing components of substitution test performance

INTELLIGENCE 9, l 11-136 (1985) Information Processing Components of Substitution Test Performance* LILA F. L A U X AND D A V I D M . L A N E Depart...

1MB Sizes 0 Downloads 64 Views

INTELLIGENCE 9, l 11-136 (1985)

Information Processing Components of Substitution Test Performance* LILA F. L A U X AND D A V I D M . L A N E

Department of Psychology Rtce Umverstty P.O. Box 1892 Houston, TX 77251

Although substitution tests have been included in tests of mtelhgence for years, the underlymg abdmes they measure have stall not been clearly determined This study used componentml analys~s to mvestlgate the mformatton-processmg components underlymg substitution test performance The bases of sex and age differences were also of interest One hundred subJeCts from each of three age groups (9-11. 18-25, and 60-89 years) were tested The componentlal analysis found that subsUtutlon tests measure perceptual speed and, to a lesser extent, memory ablhty and writing speed The component "Sumulus Onentat~on, Response Inltmtlon, and ExecuUon" was related to substitution test performance m the sample of chddren and the sample of older adults but not m the sample of younger adults Verbal abdlty was not slgmficantly related to substitution test performance m the two younger samples but was strongly related to substitution performance tn the oldest sample Although females outperformed males on the Symbol D~glt Test, males &d as well as females on the computerized tasks Apparently, sex differences in substltuaon test performance cannot be explained by the components of the test measured here

INTRODUCTION S u b s t i t u t i o n tests s u c h as the D i g i t S y m b o l S u b t e s t o f the W e c h s l ~ r A d u l t Intelligence Scale ( W A I S ) , the A r m y B e t a S u b s t t t u t t o n T e s t , a n d the C o d i n g Subtest o f the W e c h s l e r I n t e l l i g e n c e S c a l e for C h i l d r e n ( W I S C ) are a m o n g the oldest and best e s t a b l i s h e d o f all p s y c h o l o g i c a l tests a n d are c u r r e n t l y i n c l u d e d m m a n y w i d e l y used i n t e l l i g e n c e m e a s u r e s ( M a t a r a z z o , 1972). T h e s e tests are r e h a b l e , easily a n d q m c k l y a d m i m s t e r e d , r e l a t i v e culture free, a n d s i m p l y and o b j e c t i v e l y scored. T h e Digit S y m b o l S u b t e s t o f the W A I S p a t r s the & g i t s 1 - 9 with nine symbols; the task is to e n t e r the a p p r o p r i a t e s y m b o l into the e m p t y square b e n e a t h the *This research was supported by National Insutute of Health Grant # IR03 MH 36551-01 to David M Lane Requests for reprmts should be addressed to David M Lane at the address hsted above

111

112

LAUX AND LANE

FIG 1

The key from the Symbol Dlg,t Test

digit. This form of the task was adapted by Otis from the Army Beta Substitution Test in which the arrays are reversed such that the symbols are above the digits and the subject enters the appropriate digit into the box below the symbol. Figure 1 shows the key from the Symbol Digit Test which follows the original Army Beta format. A subject's score on these tests is the number of items cqrrectly completed in a short time, typically 90 s. There is convincing evidence that substitution tests measure at least one cognitive, perceptual, or motor ability substantially related to what is traditionally defined as intelligence. First, high correlations are typically found with other measures of intelligence. For example, the Digit Symbol Subtest correlates about .70 with WAIS Full Scale IQ. Second, Digit Symbol Subtest performance is useful in the diagnosis of neurological defects and has been found to be the best of the WAIS subtests for identifying persons with brain damage (Lutey, 1977). Digit Symbol Subtest performance also is sensitive to cognitive impairment due to alcoholism, emotional disturbance, and psychiatric disorders such as schizophrenia (Lutey, 1977; Matarazzo, 1972; Royer, 1971). Third, the Coding subtest of the WISC has been found to be one of the most depressed of the WISC subtests m children with learning disabilmes (Huelsman, 1970; Johnson & Lyle, 1972a; Leton, 1972; Lutey, 1977) and has been useful m predicting reading tmprovement m disabled readers (Huelsman, 1970). Performance on substitution tests has been shown to improve dramatically from childhood to early adulthood and to deteriorate from early to late adulthood In addition, performance on substitution tests has been found to be substantially better for females than for males. Although numerous empirical relationships between substitution tests and important criteria have been estabhshed, there is still considerable ambiguity concerning what substitution tests measure. The followmg factors have been hypothesized by one or more researchers to underly performance on substitution tests: verbal ability, memory ability, perceptual speed, and motor speed. The evidence pertaining to each is summarized below.

Role of Verbal Ability Although the Digit Symbol Test is included on the performance rather than the verbal scales of the WISC and WAIS there is considerable evidence that it involves verbal ability. This is not surprising m light of the fact that the task

SUBSTITUTIONTESTS

113

involves encoding letter-hke stimuli, and encoding speed is related to verbal ability (Hunt, 1978). The normative samples of Wechsler (1955) showed that the Coding and Digit Symbol tests from the WISC and WAIS respectively are correlated with the Vocabulary subtest. Moreover, at every age these tests have shghtly higher correlations with~Verbal IQ than they do with Performance IQ (Lutey, 1977, Matarazzo, 1972; Wechsler, 1949; Wechsler, 1955). Factor analytic studies have generally found that the Digit Symbol Test of the WAIS and the Coding Test of the WISC load primarily on a verbal, factor (Kaufman, 1975; Lutey, 1977, Maxwell, 1960; Silverstein, 1969). Indeed, some researchers have concluded that the Digit Symbol Subtest should be classified as a verbal rather than as a performance test (Lutey, 1977). Royer (1971) was the first to analyze substitution tests m terms of their reformation-processing requirements. Based on his earlier ground-breaking work and some ideas presented by Estes (1974, 1976), Royer (1978) tested the hypothesis that female superiority on the Digit Symbol Test occurs because digit symbol items are recorded verbally and females are more efficient at recoding. Using five arrays which varied in their degree of similarity (high simdarlty items were rotations or rotations and reflections of each other), Royer predicted that high simdanty items would not be as easily recoded verbally and that female superiority would, therefore, be inversely related to item simdarity. He also pre&cted that since high similarity 1terns differ only in sl~atlal orientation, and since males generally are higher in spatial ability than are females, males would do better with the high similarity items. His predictions were confirmed: Females lost their advantage as items became more simdar and were outperformed by males m the high similarity con&tions. These results would have provided even stronger support for Estes's hypothesis if independent measures of verbal and spatial ability had been taken. Further, it is important to note that a female superiority in verbal ability is not universally found, and when it is found it is generally small (Maccoby & Jacklin, 1974). Nonetheless, Royer's findings support the conclusion that verbal abdity plays a role in substitution test performaLnce. Indirect evidence that differences in verbal ability underlie substitution test performance was found by Hulicka and Grossman (1967). Using younger and older adults, they found that older adults made much less spontaneous use of verbal mediation techniques when they attempted to learn paired associates. When instructed to use verbal mediators their learning improved significantly and they showed relatively more improvement than the younger group. Not all stu&es have provided uneqmvocal support for the role of verbal ability in substitution test performance. Storandt (1976), in a study of the effects of aging on Digit Symbol performance, looked at what she called the perceptual motor aspect of performance (writing the symbol) and the cogmtive aspect which she defined as all other components. She found that neither aspect correlated with verbal ability in either young or old adults. More recently Vernon (1983),

I 14

LAUX AND LANE

who tested 100 college students, found no relationship between scores on the Vocabulary subtest or full scale Verbal IQ scores and Digit Symbol test scores In a factor analysis by Berger, Bernstem, Klein, Cohen, and Lucas (1964) which utlhzed the full adult age range, a strong correlation between Digit Symbol performance and a verbal comprehension factor was found only in a young adult sample (18-19 years old) but not in any other group (25-34, 45-54, or 60-70 years old). Johnson and Lyle (1972b) reasoned that if a major factor underlying subsmmon test performance were the degree to which verbal codes are employed, then pretramlng in the use of verbal codes would reduce the difference between high and low scorers on substitution tests. Using 6- to 9-year-old children as subjects, Johnson and Lyle found that pretralning improved the performance of both high- and low-scoring subjects, but there was no evidence that the pretrainmg affected them differentially. In summary, the preponderance of studies supports the view that substitution test performance is related to verbal ablhty The findings are most conclusive for young adults, although Storandt's and Vernon's results are inconsistent. There is good but not qmte as consistent evidence for the relationship in children. In the normative sample, Wechsler found that the correlation between Digit Symbol Test performance and WAIS Verbal IQ scores was higher than the correlation between Digit Symbol Test performance and WAIS Performance IQ scores in all adult groups, but there is no other evidence that verbal ability is related to substitution test performance in the elderly.

Role of Perceptual Speed Several researchers have concluded from factor analytic studies that Digit Symbol performance involves a perceptual speed factor. ~ a study by Davis (1956), the Digit Symbol Test loaded on two factors: perceptual speed and numerical ability. Wechsler (1955) also found that the Digit Symbol loaded on a perceptual factor, although it loaded on a factor he called memory and freedom from distractibility as well. Berger et al. (1964) found that for elderly subjects the Digit Symbol loaded highest on a perceptual factor and a memory factor, although for young adults its primary loading was on a verbal factor, as reported above. On the basis of factor-analytic studies, Guilford (1967) concluded that the Digit Symbol Test measures Evaluation of Flgural Units (which is essentially equivalent to perceptual speed) and Memory for Symbolic Implications about equally.

Role of Memory A number of studies have sought to investigate the role of memory in substitution test performance. In an early study, Burik (1950) tested 50 young women on the Digit Symbol Test. Immediately after completing the test, subjects were asked to

SUBSTITUTIONTESTS

I 15

recall as many digit-symbol pairs as they could. No correlation between Digit Symbol scores and this incidental learning measure was obtained. Neither was there a correlation between Digit Symbol scores and scores on the Associate Learning Subtest of the Wechsler Memory Scale. Finally, Burtlk found that instructing subjects to try to learn the digit-symbol pairs actually lowered their performance. Thus, Burik's study suggests that memory is not an important component of Digit-Symbol performance. Contrary evidence was obtained by both Wechsler (1958) and Berger et al (1964) who found that the Digit Symbol Test loaded on a memory factor. Consistent with these findings, Silverstein (1969) found that the Digit Symbol had its highest loading on a verbal/memory factor. Johnson and Lyle (1972b), using a sample of children, found a relationship between performance on the Coding Subtest of the WISC and the number of stimulus pairs that could be recalled. They also found a relationship between coding performance and memory for shapes. Johnson and Lyle (1973) classified children as either good or poor coders on the basis of the WISC Coding Subtest. Using an alternate form of the test, the researchers had both groups memorize the altered symbol-digit pairs to a criterion of two consecutive perfect trials. During testing, half of each group was provided with the code while the other half was not. Although subjects who had the code present performed better than those who did not, having the code present was equally beneficial to the two coding ability groups. Thus, for both good and poor coders, referral to the code is quicker than remembenng the pairs. In addition, good coders needed fewer trials to memorize the pairs to criterion and continued to perform significantly better than poor coders when both groups had memorized the pairing before testing, although their superiority was slightly reduced. Erber, Botwlnick, and Storandt (1981) employed essentially the same methods as Johnson et al. with groups of young and old adults. As is typically found, young adults performed considerably better on the substitution test than did older adults Interestingly, although memorizing the pairs improved the performance of the young adults, it did not do so for the older adults Thus, the age difference in substitution test performace increased following memory training which insured that both groups could recall all digit-symbol pairs before testing. Erber et al. concluded that differences in memory for pairs does not explain the difference in substitution test performance between young and old adults since equating the pair knowledge of the two groups increased the advantage of the younger group. This conclusion should be interpreted cautiously, however, since although the age groups were equated in terms of knowledge of the pairs, no attempt was made to equate them in terms of the speed with which the pairs could be recalled. Erber et al. also found that young adult subjects learned the pairs sigmficantly faster and made significnatly fewer mistakes during acquisition than did older adults, a finding consistent with the well-documented deficit in older adults for paired associate learning (Hulicka et al., 1967). It appears therefore, that al-

I 16

LAUX AND LANE

though memory for pairs itself may not be directly related to substitution test performance, rate of paired-associate learning is correlated with substitution-test performance. The role of memory in substitution test performance is unclear. Correlational studies provide moderately strong support for the hypothesis that memory is important In these tests. Other studies have found that rate of acquisition differentmtes between good and poor coders. Since instructing subjects to attempt to memorize the pairs actually lowers performance on substitution tests It appears that good performers are not actively attempting to learn the pairs. Nevertheless, some studies have found differences between high and low scorers in the actual number of pairs recalled after testing. Only a beginning has been made in testing the role of memory in substitution test performance. There is as yet no direct ewdence that memory plays a role in substitution test performance.

Role of Motor Speed The first evidence that motor speed plays an important role m substitution-test performance was obtained by Burik (1950). In Burik's study, Digit Symbol Test performance correlated from .44 to .52 with various tests of motor speed and thus loaded heavily on a motor speed factor in a factor analysis. Consistent with Burik's results, Lyle and Johnson (1973) found that above average performers on the WISC coding subtest could write Xs faster than could below average performers. There was also a difference in the speed with which the symbols could be copied, but this difference seemed to be due solely to writing speed since an analysis of covariance with coding ability (good versus poor coders) as the independent variable, speed of writing Xs as the covariate, and speed of copying symbols as the dependent variable, failed to find a significant effect of coding ability. Storandt (1976) examined the relationship of copying speed to Digit Symbol Test performance m young and elderly adults and found that the proportion of time spent copying was about 50% in both groups. Older people were slower on the motor as well as on the other aspects of the task. Thus, although it is clear that writing speed is related to substitution test performance, it is not the only factor involved.

Summary It is reasonable to suppose that each of the abilities discussed previously bears some relationship to substitution test performance. Based on the information now available, however, it is not possible to draw even moderately firm conclusions concerning the relative importance of these factors. Part of the problem is that factor analysis, the method used in the vast majority of studies, is mcapable of providing an unambiguous solution to this problem. Although factor-analytic studies have provided a great deal of important information, they have probably

SUBSTITUTIONTESTS

117

reached the limits of their usefulness. Fortunately, a new and promising methodology for approaching this problem exists, and it is described below. COMPONENTIAL ANALYSIS Robert Steinberg (1977), in an influential book, presented a method designed to identify the underlying components of cogmtive tasks. This method, which is called Componentlal Analysis, involves (a) developing an information-processing model of the task, (b) using this model to define and measure the components or elementary reformation processes involved in the task, and (c) determimng the relative xmportance of indiwdual differences in the components for explaimng indwidual differences in the task itself. The first step in a Componential Analysis is to break the task into a series of subtasks, each of which requires less reformation processing than the preceding task. The general strategy ~s to give subjects successively greater amounts of prior reformation (precueing) so that successively less processing is required in the remaining parts of the task. For example, in an A:B::C:D type of true/false analogy problem, a task with one precue presents the subjects with part A of the analogy before presenting the whole problem for solution. A task with two precues presents A:B before presenting the rest of the problem. Componential analys~s assumes that subtasks reqmring less processing are "contained in" those that require more processing--that is, the reformation processing in a precued subtask is a subset of the information processing required in a subtask with less precuemg In the example given above, the reformation processing required to solve the analogy after being presented with A:B ~s a subset of the information processing required to solve the analogy problem when only A has been presented. For this assumption to be true, precueing must not alter the basic nature of the task. This assumption of additivity can be tested by examining the correlations among the subtasks. If the subtasks are ordered according to amount of precueing, the correlations between tasks should decrease as a function of their distance from the dmgonal, resulting in a simplex matrix. Scores on the task and subtasks are called interval scores. They are used m conjunction with the model of the task to estimate the durations or difficulties of the components of the task. These parameter estimates are correlated with reference ability test scores in order to validate parameter estimates as well as to identify components of task performance related to individual differences in other tasks. The present study applied Componential Analysis to the Symbol-Digit Substltution Test in which symbols are presented and digits serve as responses. We chose to use the Symbol-Digit Test rather than the Digit-Symbol Test because the former test is more amenable to componential analysis and to the development of

118

LAUX AND LANE

a computerized analog. Although Royer (1971) and Royer, Gdmore, and Gruhn (1981) showed that performance on these two tests is influenced m the same manner by a variety of variables, it cannot be safely assumed that all aspects of them are identical. Therefore, it should be kept in mind that our results do not necessarily generalize to other substitution tests. The information processing sequence for each item on this test can be viewed as consisting of (a) detecting and encoding the symbol, (b) finding the symbol in an array of nine symbols, (c) encoding the digit paired with that symbol, (d) selecting the proper response, and (e) initiating and executing that response. In order to break the task into subtasks, a computerized analog of this test was developed. On each trial of the analog, an array of nine symbols is presented, each with either a " 0 " or a " 1 " below it. Simultaneously, a single randomly chosen element of the array (the target stimulus) is presented below the array. The task is to respond as quickly as possible by pressing a key corresponding to a " 0 " or a " 1 " to in&cate whether a " 0 " or a " 1 " was paired with the symbol m the array. The pmrings of the symbols and digits remains fixed over all trials. A series of subtasks was developed to estimate the time needed to complete the processing of each of the components. The subtasks are. 1. Cue 0: This subtask is identical to the Symbol Digit Analog described above except that the order of the symbols in the array and the symboldigit pairings vary randomly over trials. 2 Cue 1: The target symbol is presented alone until the subject m&cates by pressing a button that he or she is ready to continue. At that point, the array of symbols with " 0 " s and '" 1 "s appears and the subject responds as in the Symbol Digit Analog. 3 Choice Reaction Time: Subjects are presented with either a " 0 " or a " 1 " and respond by pressmg a corresponding key as quickly as possible. 4. Simple Reaction Time: Subjects press a designated key as quickly as they can when a lighted square appears in the middle of the screen Since the symbols m the symbol-&glt pairings of the Symbol Digit Test are letterhke but are not letters, two versions of the first three tasks described above were developed. In one version of each of these tasks, symbols in the pairs were the letters B, S, V, A, J, M, Y, X, and Z In the second version of each task, symbols in the pairings were patterns: These patterns are shown in Figure 2. Since there were two versions of each of the first three tasks (Symbol Digit Analog, Cue 0, and Cue l) and two reaction time tasks, there was a total of eight computerized tasks.

FIG 2

Symbolsused m the "pattern" version of the Symbol Dlg~tanalog

SUBSTITUTIONTESTS

119

TABLE 1 Computerized Subtasks and the Components Hypothesmed to Make Up Each Subtask Components

Subtasks Analog Cue 0 Cue 1 Choice RT Simple RT

Stimulus Orientation, Response Inmatlon and ExecuUon x x x x x

Dlg~t Encoding, Response Selection

Array Search Match Decision

x x x x

Symbol Encoding

Memory Faohtatlon

x x x

Table 1 presents the analog tasks and subtasks and indicates the components hypothesized to make up each subtask. The presence of an " X " in a column headed by a component indicates that the component is hypothesized to be contained in the task. Thus, all five components are hypothesized to be included in the Analog task, all but memory facilitzataon are hypothesized to be included m the "Cue 0 " subtask, etc. From the scores on these subtasks, estamates of the component times can be made. The goals of this study were to obtain mformation relevant to the following questions: (a) What are the component abihties underlying substitution test performance and what is their relauve importance? (b) What components are responsible for age and sex differences in subsntution test performance? and (c) Does the relatwe Importance of the underlying components change as a function of age or sex?

METHOD Subjects There were three groups of subjects: children, young adults, and elderly adults. Subjects m the children's group were 50 male and 50 female school children ranging m age from 105 months to 149 months. The children attended a Catholic sctiool m Houston. The subjects m the young adult group were college students ranging in age from 17 to 25 years of age. Twenty-five males and 25 females were students at Rice University; an additional 25 males and 25 females were students at the University of Houston. These subjects were paid $5 for their partlcapatlon. Most of the subjects in the older adult group were recruited from neighborhood senior citazen's centers and the Jewish Commumty Center Others were contacted individually, often based on suggestions from subjects obtained from the groups mentioned above A total of 22 males and 78 females were included. These subjects were also paid $5 for their partlopation.

120

LAUX AND LANE

Apparatus and Tasks A TRS80/Model 1 microcomputer was used to control the computerized tasks. The main program was written in Basic and a machine language program was used to present the stimuli and to record the response times. Subjects were given 50 trials on each task. Although subjects were instructed to respond as quickly as they could, it was stressed that accuracy was more important than speed. In addition to the eight computerized tasks, each subject completed the paper and pencil version of the Symbol Digit Test (symbols are located in the array and numbers are written below them) and a speed of copying digits test. The Digit Copying Test resembles the Symbol Digit Test but no coding is required. Subjects simply copy the digit from the upper squares into the empty squares beneath them as rapidly as possible. For both of these tests, subjects were given 90 s and their score was the number of correct responses made in that time. The Vocabulary Subtests of the WISC and WAIS were chosen as the reference tests for verbal ability. In factor analytic studies these tests consistently show a single high loading on the verbal factor. They are quickly administered and very reliable. Two tests of perceptual speed from the French Kit of Reference Tests for Cognitive Factors (French, 1972) served as reference tests of this ability. They are the "Finding A ' s " and "Number Comparison" Tests. The Finding A's Test was adapted for use with all ages to ehminate the possible effect of differences m reading ability among children, young, and old adults. In the French Kit version of Finding A's, subjects search through columns of words and mark out any word containing the letter " A " . In our version, words were replaced with randomly generated nonwords. The letter A never appeared more than once in a nonword. Of course, altering the "Finding A ' s " test in this way may have changed its factor composition somewhat. We felt, however, that it could not be used validly in a developmental study without eliminating the possible confounding of reading ablhty and perceptual speed. Scores on Finding A's represent the number of A's correctly marked in 120 s. The Number Comparison Test requires subjects to compare two numbers of equal length (three to nine digits) and decide if they are identical. If they differ, an X is placed m the space between them. The score is the number of Xs correctly marked m 90 s.

Procedure Subjects were tested individually. The computerized tasks were presented first for all subjects. The order in which these eight tasks were presented was randomized individually for each subject. Subsequently, each subject completed the Symbol Digit and the Digit Copying Tests which were also presented in random order. Finally, the two perceptual speed reference tests and the vocabulary test were administered with vocabulary always being admimstered last.

121

SUBSTITUTION TESTS TABLE 2 Means and Standard Devtauons of the Symbol Digit, the Digit Copying, the Number Comparison, the Finding A's, and the Vocabulary Tests as a Function of Age and Sex Children M

Symbol Digit All Males Females Digit Copy All Males Females Number Comp All Males Females Finding A's All Males Females Vocabulary All Males Females

Young Adults SD

M

SD

Elderly Adults M

SD

41 91 39.56 44 20

(8 36) (6.51) (9.36)

68 42 64 84 71 78

(9 56) (7 71) (9 97)

33 82 36 09 33 14

(12 88) (14.61) (12 23)

105 23 99 25 111 08

(19 81) (16 82) (20 91)

176 04 171 91 179.92

(20 95) (19 91) (21 36)

119 92 125 27 118 30

(33 81) (40 69) (31 59)

7 41 6 83 7 97

(2 16) (1 93) (2 26)

14 43 13 91 14 92

(3.03) (2 85) (3.14)

8 69 9 69 8 40

(3 11) (2 68) (3 18)

12 12 10 88 13 35

(4 74) (4 03) (5 10)

31 63 29 71 33 47

(9 41) (9 31) (9 22)

18 84 18 27 19 00

(7 13) (7 88) (6 97)

35 05 35 27 34 84

(4 63) (4 61) (4 69)

61 54 61 30 61 77

(11 65) (9 90) (13 21)

57 03 64 87 54 93

(14 03) (9 93) (14 26)

RESULTS AND DISCUSSION Basic Data, Paper and Pencil Tests Table 2 presents the means and standard deviations of the Symbol Digit Test, the Digit Copying Test, the Finding As Test, the Number Comparison Test, and the Vocabulary Tests as a function of age and sex. The expected marked improvement m performance from childhood to early adulthood as well as the even more remarkable dechne in performance from young to older adulthood was found on the Symbol Digit Test. Young adults completed more than twice as many items as older adults, and children completed about 25% more items than the older group. Similarly, performance on the Digit Copying Test and the two perceptual speed tests (Finding As and Number Comparison) improved from chddhood to early adulthood and dechned again with old age, although performance in old age did not fall below the level of the chtldren as did performance on the Symbol Digit Test.

122

LAUX AND LANE

On the Vocabulary Test there was a large difference between the performance of the children and the performance of young adults. The young adults also scored shghtly higher than did the older adults (about .33 standard deviations). Since vocabulary does not typically decline with age, it is likely that the difference obtained here was due to our sampling procedure. All of our young adults were enrolled m college, with half of them attending a highly selective university. For both the children and the young adults, females showed the typicallyfound large advantage over males in substitution test performance. Females m these two groups also performed substantially better on the Digit Copying Test and the two perceptual speed tests (Finding A's and Number Comparison). In the older group, however, the female advantage not only disappeared, but a slight male superiority was obtained on the Symbol Digit Test, the Digit Copying Test, and the Number Comparison Test. For the children and the young adults, there were virtually no sex differences on the Vocabulary Test. In the older adult group, however, males showed a large advantage over females. We beheve that the male subjects m our older adult sample were more educated and perhaps more intelligent than the females in our older adult sample due to our method of samphng. Males were generally much more reluctant to participate than females, and many of those who were willing to participate had professional training and several were still revolved to some degree in professional activities. This group of older men actually scored h~gher on the Vocabulary Test than did the college students who served as our young adult group! Table 3 shows the correlations among the Symbol Digit, the Digit Copying, the Finding As, the Number Comparison, and the Vocabulary Tests as a function of sex and age. For the children and young adults, the correlations among all the paper and pencil tests were generally higher for females than for males. In both these groups vocabulary did not correlate with Symbol Digit Test performance nor generally with any of the other tests. These findings are consistent with Storandt's (1976) findings but contrary to the Wechsler normative groups where there was a substantial correlation between Digit Symbol Test performance and Vocabulary scores at all ages. It should be noted, however, that we used the Symbol Digit Test rather than the Digit Symbol Test used by Wechsler. In the older adult group, correlations among the paper and pencil measures were umformly high for both sexes. The Vocabulary Test correlated highly with the Symbol Digit Test, and correlated moderately h~ghly w~th the Digit Copying Test and with the perceptual speed measures as well. These results are not consistent with Storandt's, who found no correlation between Digit Symbol Test scores and Vocabulary Test scores in either the young adult group or the older adult group. The finding of a substantial correlation between the Vocabulary Test and the Symbol Digit Test for older adults in conjunction w~th the finding of practically

123

SUBSTITUTIONTESTS

TABLE 3 Correlations Among the Symbol Digit, Digit Copying, Finding A's, NumberComparison, and VocabularyTests Symbol Digit

Symbol Digit Digit Copying Finding A's Number Comp Vocabulary

1 00 40** 29* 50** 00

Symbol Digit Digit Copymg Finding A's Number Comp Vocabulary

1 00 05 18 46** 28

Symbol Digit Digit Copying Finding A's Number Comp Vocabulary

1 00 78** 67** 70** 52*

Dxglt Copying

Finding A's

Children 66** 69** 1 00 44** 32* l 00 52** 35* - 02 20 Young Adults 39** 30* 1 O0 36* - 26 1 O0 04 44** - 04 38** Elderly 70** 56** 1 00 44** 65** 1 00 51" 72** 36 67**

Number Comparison

62** 59** 60** 1 O0 - 01 64** 51 ** 18 1.00 16 58** 55** 44** 1 00 33

Vocabulary

12 34* 01 04 1 00 - 03 - 20 - 06 07 1 00 50** 38** 36** 34** 1 O0

Note Correlations for females are above the diagonal, correlations for males are below the diagonal *p < 05 **p < 01

no correlation between these tests for the younger samples is particularly interesting. First, ~t is consistent w~th the dedifferentiat~on hypothesis that the factorial structure of intelligence becomes simpler with age (Cunningham, 1980; Remert, 1970). Second, it suggests that people who have good vocabularies when they are young do not perform particularly well on the Symbol Digit Test, but, as they grow older, their performance on the Symbol Digit Test improves relative to that of their cohorts. Since, in general, Symbol Digit performance dechnes w~th age, this means that people with good vocabularies are less susceptible to this decline. Also, people with good vocabularies tend to be better educated and, therefore, probably remain more mentally active as they grow older than do less well educated people. It thus appears that being mentally actwe helps to prevent the sharp decline with age that typically occurs on the Symbol Digit Test and other speeded tests. Naturally, this hypothesis must be considered to be only speculative at the present time; however, ff it should prove to be correct, then it would mean that age-related declines in mental functioning can be lessened or prevented by a proper environmental intervention.

Simple RT M SD Errors (%) Choice RT M SD Errors (%) Cue 1 (Letters) M SD Errors (%) Cue 1 (Patterns) M

338 103 4 0

670 130 30

2061 302 55

2621

678 152 27

2090 429 44

2734

Males

362 111 3 5

All

Children

2844

2118 527 32

685 172 32

385 115 30

Females

1693

1254 211 28

391 70 15

273 70 2.5

All

1639

1256 202 3.0

379 50 22

256 63 3.0

Males

Young Adults

1744

1252 220 28

402 83 10

289 72 1.8

Females

3143

2210 502 25

675 364 18

446 244 15

All

2939

2176 979 25

582 194 0.8

423 256 2 2

Males

Elderly

3205

2220 884 25

703 398 20

453 241 12

Females

TABLE 4 Means and Standard Dewatlons of Response Latencles (m mdhseconds) and Percent Errors for Each of the Computerized Tasks as a Function of Age and Sex

SD Errors (%) Cue 0 (Letters) M SD Errors (%) Cue 0 (Patterns) M SD Errors (%) SD Analog (Letters) M SD Errors (%) SD Analog (Patterns) M SD Errors (%) Sample Size

1288 40 2654 784 2.8 3467 1319 38 2532 947 22 3143 996 32 73

1290 20 2580 711 18 3830 1265 3O 2246 822 12 3124 1316 18 22

1287 3.8 2637 765 25 3435 1302 2.8 2466 923 2O 3139 1071 30 95

279 3.2 1493 233 1.0 2013 323 32 1191 273 20 1617 376 32 49

281 3.0 1512 219 20 1997 283 18 1230 267 2.2 1570 281 25 46

283 3.0 1502

226 15 2005 303 20 1210 270 22 1594 332 28 95

655 38 2589 625 10 3207 681 18 2166 618 12 2713 798 25 49

523 70 2555 586 28 3053 560 38 2064 529 35 2604 733 50 48

601 54

2573 603 19

3131 626 27

2116 575 2.2

2659 764 38 97

126

LAUXAND LANE

Basic Data, Computerized Tasks Table 4 presents the means and standard deviations of the response times and percent errors of the eight computerized tasks as a function of age and sex. Subjects were dropped from these analyses if their error rate on at least one task approached 50%, indicating that they were performing randomly, or if their total error rate was more than three standard deviations above the mean of the remainlng group members. This resulted in a total of 13 subjects being dropped leaving 97 children, 95 young adults, and 95 older adults for a total sample of 287. In computing the mean response times for each subject, the first l0 trials on any task were considered practice trials and were discarded. Means were then computed on the remaining 40 trials. Since means can be very sensitive to outhers and therefore inefficient estimators of central tendency, we also computed means trimmed 50% (the mean of the scores from the 25th to the 75th percentile). Since the analyses with the trimmed means were almost identical to the analyses with the untrimmed means, we report only the analysis on untrimmed means. In order to test the assumption of additlvity, the correlation matrices among the subtasks were examined. If the assumption is valid, then the correlation matrices should show the simplex form. That is, correlations should decrease as a function of the number of different components involved in the subtasks. The correlation matrices among these subtasks for the three age groups are shown in Table 5. Correlations generally decreased as a function of the number of components on which the tasks differed supporting the assumption of additlvlty. The most important comparison involves the correlations between the Symbol Digit Analog and the four types of subtasks. These correlanons (averaging over the letter and pattern stimuli) increased systematically as a function of the number of common components. Of the 24 comparisons between adjacent subtasks that can be made for each age group, 5 were contrary to the expected direction for the children, 5 were contrary to the expected d~rection for the young adults, and 9 were contrary to the expected direcnon for the older adults. However, none of these devianons from a simplex pattern approached significance. Also consistent with the assumption of additivlty is the finding evident m Table 4 that tasks containing more information processing components consistently required substannally more time to complete than tasks contammg fewer components. On the computerized tasks, young adults responded much more rapidly than either children or older adults. The children were faster than the older adults on all computerized tasks with the exception of Choice Reaction Time where there was virtually no difference in the response times of the two groups In the children's groups, males showed a slight-to-moderate advantage over females on all tasks. The males in this group, however, made consistently more errors than the females, indicating that they were sacrificmg accuracy for speed. In the young adult sample, where male and female error rates were similar, males

SUBSTITUTION TESTS

127

TABLE 5 Correlations Among the Computerized Tasks, Computerized Subtasks, and the Symbol Digit Test as a Function of Age SD Analog

SD Analog Cue 0 Cue 1 Choice RT Simple RT Symbol Digit

(78) 70 61 54 30 44

SD Analog Cue 0 Cue 1 Choice RT Simple RT Symbol Digit

(62) 59 50 44 21 41

SD Analog Cue 0 Cue 1 Choice RT Simple RT Symbol Digit

(78) 79 81 50 38 69

Cue 0

Cue 1

Choice RT

Children (n-97) 61 57 75 56 (65) 60 61 36 32 44 41 Young Adults (n-95) 51 57 53 (68) 75 50 68 (54) 58 60 50 31 20 62 57 29 33 Elderly (n-95) 84 85 53 (79) 84 50 86 ( 81) 49 50 56 43 46 40 63 70 45 78 (79) 73 57 38 53

Simple RT

Symbol Digit

34 30 29 32 22

51 51 54 41 22 -

32 27 30 62 00

46 49 52 33 00 -

43 51 47 40 42

77 72 71 45 42 -

Note Correlations among tasks using letters are above the diagonal, correlations among tasks using patterns are below the diagonal On the diagonal are the correlations between the pattern and letter versions of each task where appropriate

r e s p o n d e d f a s t e r t h a n f e m a l e s o n all tasks e x c e p t t h o s e w h i c h utilized letters as s y m b o l s . W h e n p a t t e r n s w e r e s y m b o l s , f e m a l e s n o t o n l y r e s p o n d e d m o r e slowly b u t m a d e m o r e errors t h a n m a l e s . O n the s i m p l e a n d c h o i c e r e a c t i o n t i m e tasks, y o u n g adult m a l e s r e s p o n d e d s u b s t a n t i a l l y faster t h a n f e m a l e s , hut also m a d e n o t i c e a b l y m o r e errors. In the o l d e r adult g r o u p , m a l e s were 2 0 % faster at r e s p o n d i n g to the C h o i c e R e a c t i o n T i m e task t h a n w e r e f e m a l e s a n d they m a d e f e w e r errors. In the o t h e r c o n d i t i o n s , o l d e r m a l e s w e r e g e n e r a l l y s o m e w h a t faster at r e s p o n d i n g t h a n o l d e r f e m a l e s , b u t the d i f f e r e n c e s w e r e not large U n l i k e the two y o u n g e r g r o u p s w h e r e f e m a l e s t e n d e d to s h o w a n a d v a n t a g e w h e n letters were s t l m u h , o l d e r m a l e s w e r e faster t h a n o l d e r f e m a l e s b o t h w h e n letters a n d w h e n p a t t e r n s w e r e u s e d as s y m b o l s . In a d d i t i o n , o l d e r m a l e s m a d e f e w e r errors t h a n f e m a l e s . A s d i s c u s s e d p r e v i o u s l y , the o l d e r m a l e s w e r e a m o r e select g r o u p t h a n were the o l d e r f e m a l e s . In all g r o u p s , s u b j e c t s r e s p o n d e d s u b s t a n t i a l l y faster w h e n d e a l i n g with familiar s y m b o l s (letters) t h a n w h e n d e a l i n g w i t h u n f a m i l i a r o n e s (patterns), a l t h o u g h as s h o w n in T a b l e 5, t h e s e t w o types o f tasks w e r e h i g h l y correlated.

128

LAUX AND LANE

In the Cue 0 task, which involves searching the array and making a match as well as response selection, initiation, and execution, older adults took 42% more time to respond when the symbols were patterns rather than letters. Younger adults and children, by comparison, needed only 35% and 31% more time respectwely for patterns as opposed to letters. This supports previous research showing that older adults have more difficulty dealing with unfamiliar stimuli than do younger people (Schonfield & Robertson, 1968; Thomas, Waugh, & Fozard, 1978) Since some Symbol Digit Test symbols are not highly familiar, this disadvantage m dealing with unfamiliar stimuli may underlie some of the decrement m substitution test performance which is found in older adult groups.

What Does the Symbol Digit Test Measure? One of the primary goals of this research was to determine the extent to which Symbol Digit Test performance can be explained by the hypothesized underlying components. One way to measure the importance of a given component is to compute the difference between the time taken to perform a subtask which includes the component in question and a subtask identical in all respects except that it does not include the component (this second task will be referred to as the "subsumed task" hereafter). This difference score could then be correlated with Symbol Digit Test performance. There are, however, a few problems with this approach. First, difference scores often have low reliability which might attenuate the correlation. Second, and more important, the derived scores for the various components would not be independent; in fact, many would be highly correlated. We wished to have measures of the various components that were independent so as to more easily determine the relative importance of the components in determining individual differences in Symbol Digit Test performance. In order to obtain independent measures of the underlying components, we correlated the portion of a subtask that is independent (in the sense that it is uncorrelated in hnear regression) of all subsumed subtasks with the Symbol Digit Test In more statisucal terms, this is the semipartial correlation of a subtask with the Symbol D~git Test with the subsumed task(s) partialled out. Although error rates were generally low, it was still important to consider the possibility of speed/accuracy tradeoffs Therefore, we calculated semipartial multiple correlations in which both response time and error rate served as predictors. For example, in calculating the component (actually two components that were intentionally confounded) Digit Encoding and Response Selection with the Symbol D~glt Test, response latency on the choice reaction time task was predicted (using multiple regression) by the response time and error rate on the simple reaction time task. Errors on the choice reaction time task were then predicted by the response time and error rate on the s~mple reaction time task as well as by response time on the choice reaction time task. The residuals from these two regressions were then used to predict Symbol Dig~t Test performance, with the multiple R being the semipartml multiple correlation of interest As a matter of computational convenience, this correlation was computed by taking

129

SUBSTITUTION TESTS

the square root of the difference between the squared multiple correlation in the regression of Symbol Digit Test performance on the simple reaction time task (response time), simple reaction time task (errors), choice reaction time task (response time), and choice reaction time task (errors) and the squared multiple correlation in the regression of Symbol Digit Test performance on the response ttme and error measures for the simple reaction time task. Because some of the stimuli in the Symbol Digit Test are letterlike and some are abstract patterns, the letter and pattern versions of the computerized tasks were both used in regression analyses involving these subtasks. Thus, the semipartial multiple correlation Involved four predictors (latency and errors for each version). Since the distribution of differences between semipartial multiple correlations is not known, significance tests are not reported on the differences between the groups in correlations of components with Symbol Digit Test performance. The Symbol Digit Test is scored in terms of the number of correct responses and the components are scored in terms of latency. Therefore, "positive" relationships would be associated with negative correlations. For the sake of ease of reading, the signs of all the correlations were reversed so that positive relationships were associated wtth positive correlations. Often a componentlal analysis involves collecting data from each subject under a variety of task conditions (not to be confused with subtasks). In these cases it is possible to obtain an index of the goodness of fit of the component model for each subject individually. Since we did not introduce variations in the Symbol Digit Test, this kind of internal validation was not possible. TABLE 6 Correlations of Hypothesized Components with the Symbol Digit Test as a Function of Age and Sex

Children Males Females Young Adults Males Females Elderly Males Females

1

2

3

4

5

6

R

23"* 17 38 ** 03 04 26"* 43"* 51 * 40"*

36"* .16 48"* 49** 32 55 ** 33"* 52" 36 **

45"* 52"* 40 * 40** 46* 47 ** 48"* 48 51 **

22" 40" 17 34** 38 31 ** 16 23 05

20 32 24 18 19 22 27"* 27 24 *

38"* 30" * 23 * 21 ** 11 20" 23"* 04 24" *

79"*

77**

85"*

Component 1 Stimulus orientation, response m~tlat~on and execution Component 2 Digit encodmg and response selection Component 3 Search and match/nomatch decision Component 4 Target encoding Component 5 Memory facdltatlon Component 6 Copying speed R Multiple correlation of Symbol D~glt predicted by all components *p < 05 **p < 01

130

LAUX AND LANE

The correlations of the components with the Symbol Digit Test are shown in Table 6. The multiple correlations of the Symbol Digit Test when regressed on all the components for each of the three age groups are shown also in Table 6. These correlations can be computed either by taking the square root of the sum ot the squared component correlations or by regressmg Symbol Digit scores on all measures.

Stimulus Orientation, Response Initiation and Execution. The correlations of this component, which was measured by response latency and errors on the simple reaction time task, with Symbol Digit Test performance, were .23, .03, and .43 for the children, young, and old adults respecnvely. Thus, individual differences in this very elementary information processing component are an important source of individual differences in the Symbol Digit Test performance of the elderly but play virtually no role in determining individual differences in the performance of young adults and only a modest role m children. To obtain information on the relanonship of th~s component to other standard indexes of mental abihty, the same analyses were done with the reference tests as were done with the Symbol Digit Test. The correlations between the "Stimulus Orientation, Response Selection and Execution" component and the Number Comparison Test were . 10 in children, .02 in young adults, and . 14 in older adults. It appears that indwldual differences in th~s component do not predict individual differences in Number Comparison Test performance at any age. The correlations between this component and the other perceptual speed measure, Findmg As, were .26 for children, . 13 in young adults, and .29 in older adults. Indiwdual differences m this component are only modestly related to individual differences m perceptual speed as measured by the Findmg As Test and, therefore, this component does not appear to be an important component of perceptual speed as defined by our reference tests. The pattern of the correlations of "Stimulus Orientation, Response Selection and Execution" with Vocabulary Test scores m the three groups reveals large differences in the relationship of this component to verbal ability. The correlations were .20 in children, .00 in young adults, and .38 in older adults. Older adults with larger vocabularies have shorter simple reacnon nines than do those with lower vocabularies. That this relationship is not evident m the young adult group offers additional support for the hypothesis that remammg mentally active slows the decline m mental functioning normally found with advancing age Digit Encoding and Response Selection. This component, which is that part of choice reaction time independent of simple reaction rime, is substantially related to Symbol Digit Test performance at all ages. The correlations between this component and the Symbol Digit Test were .36 for the children, .49 for the young adults, and .33 for the older adults. When the relationships of the "Digit Encoding and Response Selection"

SUBSTITUTIONTESTS

131

component with the two perceptual speed tests were examined, the following correlations were obtained. For Number Comparison the correlanons were .44 for the children, .40 for the young adults, and .39 for the older adults. The correlations for the Finding As test were .29, .25, and 00 for children, young and older adults, respectively. The "Dlg~t Encoding and Response Selection" component correlated w~th Vocabulary Test scores. 10 m children, 00 m young adults, and .09 m older adults indicating that speed of encoding digits and selectmg a response did not predict Vocabulary Test scores at any age. Thus, it appears that although this component ~s sigmficantly related to both the Symbol Digit Test and the Number Comparison Test, both of which reqmre subjects to encode d~gits, ~t is not substantially related to performance on tasks which require subjects to encode letters, and ~t does not appear related to verbal ability as measured by vocabulary tests at all.

Array Search and Match~No Match Decision. As might be expected this is the most time consuming component of substitunon test performance and individual differences m performance on this component are highly related to performance on the Symbol D~glt Test. The abihty to compare snmuh and to dec~de as to their identity or nonidennty defines both the Evaluanon of F~gural Umts and the Evaluanon of Symbolic Units Factors (Gmlford, 1967) which are conceptually identical to Thurstone's perceptual speed factor Th~s definmon also seems to describe the acnvities which make up the Array Search and Match/No match Decision Component. This component correlated .45, 40, and .48 with Symbol Digit Test performance for the children, young adults, and older adults respectively The correlations of this component with the reference tests support the contention that this component is an aspect of perceptual speed. The correlanons with Number Comparison were .33 for chddren, .36 for young adults, and .50 for older adults. With Finding As the correlanons were .37 for children, 47 for young adults, and .49 for older adults. The correlations with Vocabulary Test scores were .26 for the children, .35 for the young adults, and .43 tbr the older adults. Since the S2¢mbol Digit Test was essentially uncorrelated w~th the Vocabulary Test for the younger two groups ( 12 for the children and - . 0 3 for the young adults), it appears that the part of the Array Search and Match/No Match Decision component that is related to verbal ability is unrelated to substitution test performance for these two groups. Target Encoding. Indwidual differences in the speed with which subjects encoded the target correlated .22 with the Symbol Digit Test for the children, .34 for the young adults, a n d . 16 for the older adults. Speed of encoding the target correlated .24, . 16, and .33 with Number Comparison scores for the

132

LAUX AND LANE

children, young adults, and older adults respectively. For Finding A's, the correlations were. 19, .21, and .37. Target encoding is thus only modestly related to performance on the two perceptual speed tests and appears to predict performance on the two tests about equally well. The relationship of this component to verbal ability, as measured by the Vocabulary Test, is certainly not substantml. These correlations were. 12 for the children, .21 for the young adults, and .28 for the older adults. It is reasonable that this component correlated more highly with the Vocabulary Test than did the Digit Encoding component since the target was a letter and Hunt (1978) found that verbal aptitude and speed of encoding letters are related.

Memory Facilitation. Individual differences in memory for either symbol digit pairs or for the location of the symbols appears to be related, albeit not strongly, to Symbol Digit Test performance. The correlations of this component with the Symbol Digit Test were .20, .18, and .27 for the children, young adults, and older adults respectively. Earlier studies have presented mixed evidence concerning the role of memory in substitution test performance. The present results are perhaps the most direct evidence of the role of memory differences in substitution test performance. It is clear, however, that most of the variance on substitution tests is not due to memory differences. There were no non-trivial correlations between the memory facilitation component and either of the perceptual speed tests or the Vocabulary Test indicating memory facilitation is not related to either perceptual speed or verbal ability. Writing Speed. The writing speed component correlated .38 with the Symbol D~git Test for the children and .21 and .23 for young and older adults respectively. Thus, writing speed played a role in Symbol Digit Test performance at all ages. It is important to note that although the act of copying d~gits must involve encoding the digits, the component as defined m this analysis has the variance due to encoding speed partialled out; what remains should be very nearly just writing speed. Conclusions about What the Symbol Digit Test Measures. There are several conclusions that can be drawn from the data discussed in the previous secuons. First, it is clear that the Symbol Digit Test is, in part, a measure of perceptual speed. For both males and females at all three age levels, the component "Array Search and Match/No match Decision" correlated substantially with the Symbol Digit Test. That this component is a measure of perceptual speed follows not only from the closeness of it to the typical definitions of perceptual speed, but also from the strong correlations between it and the two perceptual speed reference tests. A second conclusion is that the Symbol Digit Test measures a number of

SUBSTITUTION TESTS

133

aspects of information processing ability. All of the components were related to Symbol Digit Test performance at a nontrivial level for at least one age/sex group. It should be noted, however, that several but not all of these components may be construed as components of perceptual speed. A third conclusion is that the Symbol Digit Test measures different abilities in different populations. The component "Stimulus Onentation, Response Initiation and Execution" correlated substantially with Symbol Digit Test performance in the older adult group but not at all in the young adult group In addition, verbal ability was strongly related to Symbol Digit Test performance m the older adult sample, but not in the children's or young adult groups. Fourth, these data provide evidence that individual differences in memory ability are responsible for some of the differences in Symbol Digit Test performance. The degree to which subjects performed better when the symbol-digit pairings were held constant from trial to trial as opposed to varying on each trial was predictive of Symbol Digit Test performance. The relationship was significant, however, only for the elderly subjects. These results do not support the hypothesis that Symbol Digit Test performance is strongly related to verbal ability. The correlations of the components of the Symbol Digit Test with the reference test of verbal ability were not impressive. There was some correlanon of the "Array Search and Match/No match" component with Vocabulary Test scores, but the correlation was not high. Also, the Symbol Digit Test itself did not correlate with the Vocabulary Test in the two younger groups. It is quite interesting, though, that the correlation increased with age. As described previously, this suggests that the decline in Symbol Digit Test performance with age may be reduced by staying mentally active. It ~s clear from these results that the relationship among tasks cannot be explained solely m terms of general intelligence. For example, in the young adult group, the component "Digit Encoding and Response Selection" correlated 49 with the Digit Symbol Test but did not correlate at all with the Vocabulary Test. This is not to say that general intelligence is not involved in these components, but that it is not the whole story. Sex Differences The expected female advantage on the Symbol Digit Test and on the two perceptual speed tests was obtained for both the children and the young adults but not for the older adults. Unfortunately, the analysis of the computerized tasks did not reveal much about the basis of this sex difference: males tended to respond more quickly on the computenzed tasks while performing more slowly on the Symbol Digit Test itself. Apparently either the computenzed tasks do not tap the mformation-processing component responsible for the sex difference or they tap some additional component which males are better at performing and zs not a component of the Symbol Digit Test.

134

LAUX AND LANE

The finding that females performed no better than males on the Vocabulary Test while outperforming males on the Symbol Digit Test suggests that, contrary to the hypotheses of Estes (1974, 1976) and Royer (1978), the sex difference is not due to verbal ability. This conclusion is strengthened by the finding of sex differences on the perceptual speed tests that parallel the sex differences on the Symbol Digit Test. Thus, perceptual speed appears to be the most likely ablhty underlying the sex difference on the Symbol Digit Test.

Age Differences Across all tasks, the young adults outperformed the other two groups by a substantial margin. Children were faster than the elderly subjects on the simple reaction time task, the computerized Symbol Digit Analog (letters), and all the computerized subtasks which have patterns as stimuli, although the differences in the latter cases were not large. Children also were substantially faster on the Symbol Digit Test than older adults (completing 42 items in 90 s as opposed to 34 for the older adult group). The elderly, however, were faster than children at copying digits and on the two perceptual speed reference tests. Thus, the superiority of the children over the older adults can not be due to either writing speed or perceptual speed differences. Subjects in all age groups used memory to some extent when performing the analog tasks as can be seen by comparing their response times in this condition to response times when the pairs and locations varied randomly over trials. When the stimuli were letters, children showed a memory facilitation of 456 ms, young adults showed a memory facilitation of 293 ms, and older adults showed a memory facilitation of 171 ms. When patterns were symbols, the memory facilitations were 472, 312, and 297 for children, young, and older adults respectively. In terms of the percentage improvement as a function of the opportunity to remember the stimulus-response pairings, children improved their mean response times by 18% in the letter analog and by 15% in the pattem analog; young adults improved their response times by about 19% In the letter analog and 15.5% in the pattern analog. In contrast, older adults improved their response times by only 6.5% and 8.6% in the letter and pattern analog conditions respectively. Thus, the well-documented finding that memory dechnes with age may help explain the large decline with age in Symbol Digit Test performance. CONCLUDING REMARKS After a number of factor analytic studies had failed to reveal clearly what the Digit Symbol Substitution Test of the WAIS measured, Wechsler (1958) concluded that this test was factorially ambiguous Our data on the Symbol Digit Test suggest that this ambiguity resulted from the fact that substitution tests are factorially complex and that different underlying abihties are important to substItution test performance in different populations

SUBSTITUTION TESTS

135

It is c l e a r t h a t s u b s t i t u t i o n tests m e a s u r e m e n t a l a b d i t i e s that are related to m e a s u r e d i n t e l l i g e n c e but, b e c a u s e t h e y are f a c t o r i a l l y c o m p l e x , t h e i r u s e f u l n e s s as a d i a g n o s t i c tool is d i m i m s h e d . P o o r p e r f o r m a n c e in c h i l d r e n , for e x a m p l e , m i g h t b e d u e s i m p l y to s l o w w r i t i n g s p e e d or to i m p a i r m e n t in s o m e other, m o r e c o g n i t i v e , ability. W h e n s u b s t i t u t i o n test p e r f o r m a n c e is p o o r , the c a u s e o f the d e f i c i e n c y c a n o n l y b e d e t e r m i n e d b y e x a m i n i n g the i n d w d i u a l ' s p e r f o r m a n c e on o t h e r m e a s u r e s c h o s e n to tap e a c h o f t h e abilities w h i c h u n d e r l i e s u b s t i t u t i o n test p e r f o r m a n c e . T h e c o m p o n e n t i a l a n a l y s i s o f the S y m b o l D i g i t T e s t p e r f o r m e d here s u g g e s t s that it m a y b e p o s s i b l e to d e v i s e s t a n d a r d i z e d tests useful for d i a g n o s i n g specific areas o f w e a k n e s s .

REFERENCES Berger, L , Bemstem, A , Klein, E , Cohen, J , & Lucas, G (1964) Effects of aging and pathology on the factorial structure of lntelhgence Journal of Consulting Psychology, 28, 199-207 Bunk, T E (1950) Relauve roles of the learning and motor factors revolved in the D~glt Symbol Test Journal of Psychology, 30, 33-42 Cunnlngham, W R (1980) Age comparative factor analysis of ability variables in adulthood and old age lntelhgence, 4, 97-132 Davis, P C (1956) A factor analysis of the Wechsler-Bellevue Scale. Educanonal and Psychologzcal Measurement, 16, 127-146 Ether, J T , BotwlnlCk, J & Storandt, M (1981). The impact of memory on age differences in Digit Symbol performance Journal of Gerontology, 36, 586-590 Estes, W K (1974) Learning Theory and Intelligence American Psychologist, 29, 740-749 Estes, W K (1976). Intelhgence and cogmtlve psychology. In L Resnlck (Ed), The nature oJ intelhgenee Hdlsdale, NJ Erlbaum Guilford, J. P (1967) The nature of human mtelhgence New York McGraw-Hill Huelsman, C B., Jr (1970) The W1SC subtest syndrome for disabled readers Perceptual and Motor Skdls, 30, 535-550 Huhcka, J M , & Grossman, V L (1967) Age group comparisons for the use of mediators in pmred-assoclate learning Journal of Gerontology, 22, 46-51 Hunt, E (1978) Mechanics of verbal abdlty Psychological Review, 85, 109-130 Johnson, E G , & Lyle, J G (1972a) Analysis of WISC Coding 1 Flgural reverslblhty Perceptual and Motor Skdls, 34, 195-198 Johnson, E G , & Lyle, J G (1972b) Analysis of WISC Coding. 2 Memory and verbal mediatlon Perceptual and Motor Skills, 34, 659-662 Johnson, E G , & Lyle, J G (1973) Analysis of WISC Coding 4 Paired associate learning and performance strategies Perceptual and Motor Skills, 37, 695-698 Kaufman, A S (1975) Factor analysis of the WISC-R at 11 age levels between 6% and 16V2years Journal of Consulting and Chmcal Psychology, 43, 135-147 Leton, D A (1972) Discriminate analysis of WISC profiles of learning disabled and culturally deprived pupils Psychology in the Schools, 9, 303-308 Lutey, C (1977) Individual intelligence testing Greeley. CO Carol Lutey Pubhshlng Co Lyle, J G , & Johnson, E G (1973) Analysis of WISC Coding. 3 Writing and copying speed, and moUvatlon Perceptual and Motor Skills, 36, 211-214 Maccoby, E E , & Jackhn, C N (1974) The psychology of sex differences Stanford Stanford Umverslty Press, Stanford, CA Matarazzo, J D (1972) Wechsler's measurement and apprmsal of adult intelligence Baltimore Wdhams and Wdkln ,

136

LAUX AND LANE

Maxwell, A E. (1960) Obtammg factor scores on the Wechsler Adult lntelhgence Scale Journalof Mental Sctences, 106, 1060-1062 Remert, G (1970) Comparaave factor analytic studies of mtelhgence throughout the human hfe span In L R Goulet & P B Baltes (Eds.) Lzfe-span developmental psychology. New YorkAcademic Royer, F (1971) Information processing of visual figures In the Digit Symbol substitution test Journal of Expertmental Psychology, 81, 335-342 Royer, F (1978) Sex differences m symbol digit substitution task performance lntelhgence, 2, 145-152 Royer, F L , Gdrnore, G C., & Grnhn, J J (1981) Normative data for the Symbol Digit Substitution Test Journal of Climcal Psychology, 37, 608-614 Scbonfield, M A & Robertson, E A (1968) The coding and sorting of &glts and symbols by an elderly sample Journal of Gerontology, 23, 318-328 Sllverstem, A B (1969) An alternate factor analyttc solutmn for Wechsler's mtelhgence scales Educational and Psychologwal Measurement, 29, 763-767 Sternberg, R (1977). Intelhgence, reformation processmg, and analogwal reasomng The componenttal analysts of human abdtnes Hlllsdale, NJ. Erlbaum Storandt, M (1976) Speed and coding effects m relation to age and ablhty level Developmental Psychology, 12, 177-178 Thomas, J C , Waugh, N C , & Fozard, J L. (1978) Age and famdlanty m memory scanning Journal of Gerontology, 33, 528-533 Vernon, P E (1983). Speed of processing and general mtelhgence lntelhgence, 7, 53-70 Wechsler, D (1949) Manual for the Wechsler lntelhgence Scale for Chddren New York" Psychological Corp Wechsler, D (1955) Manual for the Wechsler Adult lntelhgence Scale New York Psychological Corp Wechsler, D (1958) The measurement and appratsal of adult mtelhgence (4th ed ) New York Psychological Corp