Journal of Gerontology 1978, Vol. 33, No. 4, 541-545
Fatigue Effects on Intelligence Test Performance in the Elderly1 Walter R. Cunningham, PhD, Carol M. Sepkoski, BA, and Mary R. Opel, BA 2
N issue of central importance in interpreting research on intellectual functioning A in the elderly concerns fatigue effects. Fatigue effects represent a particularly serious issue for researchers administering long batteries of tests and seeking to make generalizations regarding abilities and age. Kamin (1957) suggested that decline of various subtest scores on the Primary Mental Abilities battery (PMA) may be an artifact of the order in which the subtests are administered (i.e., fatigue or boredom may influence different subtest scores). He administered the PMA subtest in a Latin square design to young and old subjects. Results indicated that there were no significant order effects, and Kamin suggested that age decrements were indeed due to changing cognitive abilities rather than to performance related variables. Furry and Baltes (1973) proposed that elderly subjects are sensitive to situational variables in test-taking such as fatigue-producing conditions and that these may result in performance related age differences. The investigators administered the PMA to subjects of three age groups; subtest position and pretestfatigue conditions were varied. Results indicated that, among the elderly subjects there
'This research was supported by AOA Grant #90-646-01, which is gratefully acknowledged. The authors would like to express their sincere appreciation to Jo Anne Judd for assistance in participant recruitment and scheduling, test scoring, and keypunching. The cooperation of Ms. Emily Barefield and Reverend Summers of AARP is also gratefully acknowledged. 2 Psychology Dept., Univ. of Florida, Gainesville 32611.
were no position effects, but the fatigue condition significantly influenced performance on the Word Fluency, Verbal Meaning, and Reasoning Subtests. Furry and Baltes emphasized interpretation of age by fatigue interactions and concluded that the intelligence score differences among the elderly, as contrasted to the young, may be influenced by situational determinants. Therefore, it may be inappropriate to relate age differences solely to ability specific factors. Furry and Baltes discussed fatigue effects as one example of ability extraneous influence and suggested that fatigue influences, together with other performance variables, might complicate the interpretation of the existing abilities literature. They used a specific fatigue-producing condition in their battery, a modified version of the Finding A's Test (P-l in the Educational Testing Service (ETS) Kit of Reference Tests for Cognitive Factors; French et al., 1963). The modification consisted of lengthening a letter cancellation task to 20 min. They also noted that this fatigue producing treatment was excessively fatiguing. They acknowledged that the procedure might be more fatiguing than simply taking other tests. The present study attempted to determine whether this specific task yields similar effects in contrast to actually taking a long battery of ability tests. It was hypothesized that the Finding A's task would be more fatiguing than a long battery of subtests and that fatigue effects 541
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
The purpose of the study was to investigate effects of fatigue on intelligence test performance in the elderly. Dependent variables were Verbal Comprehension, Numerical Facility, Perceptual Speed, and Word Fluency tests. Fatigue effects were investigated by varying the number of previous tests administered, by introducing breaks between tests in some conditions, and by using a pre-test fatigue-producing condition, a modified form of the Finding A's test. Subjects' ages were between 57 to 91-years. It was hypothesized that the Finding A's test would be more fatiguing than a long battery of tests and that introducing a break condition between the Finding A's test and the main battery would alleviate fatigue effects. Analyses of Variance resulted in a main effect due to a pre-test condition for the Perceptual Speed test only, and only when the main battery was preceded by the Finding A's task (p < .001). It appears that the elderly are not as susceptible to test fatigue as previous results seemed to suggest.
542
CUNNINGHAM, SEPKOSKI, AND OPEL
for the long battery conditions, as described below. Test materials were selected from the ETS battery rather than the PMA because the primary purpose of the study was to determine the degree of seriousness of fatigue effects in the ETS battery, since the first author was in the process of planning a large-sample, large-variable study of factor structure of abiliMETHOD ties in late life. The current experiment involved no variation of position for any of Subjects Letters were mailed to all members (approxi- the tests, since Furry and Baltes (1973) and mately 300) of the local chapter of the Ameri- Kamin (1957) found no order effects for elcan Association of Retired Persons in Gaines- derly subjects. ville, Florida. The letter explained the nature Fatigue Battery 1 consisted of Advanced of the study and invited participation; a $5.00 Vocabulary (V-4), Number Comparison (P-2), gratuity was also offered. Thirty males and 75 Division Test (N-2), and Word Fluencyfemales volunteered, and of these 24 males and Suffixes (Fw-1) administered in that order. 56 females were scheduled for testing sessions. Fatigue Battery 2 consisted of Vocabulary Twenty-seven subjects canceled for various (V-2), Finding A's (P-l), Subtraction and reasons. The final sample consisted of 17 Multiplication (N-3), and Word Fluencymales and 36 females. The range of age was Beginnings and Ending (Fw-3) in that order. 57 to 91 years, with a mean of 72 years. The Word Fluency-Suffixes was constructed for range of educational level was 7 to 20 years, use in the subsequent state-wide study and with an overall mean level of approximately was intended to be similar to the ETS test 14 years, with the male average being higher Fw-1. The test consists of two parts. First, than the female average by about 1 year. The the participant is given 3 min to write down as sex ratio in the sample approximates that of many words as possible ending in the suffix the full membership of the local chapter. "TION"; then the participant is given 3 min to write as many words as possible ending in the suffix "AL." The Word Fluency test Beginnings and Endings also consists of two MATERIALS Three subtests of the ETS battery were used parts. The participant is given 3 min to write as dependent variables and were presented as many words as possible beginning with the in the following order: Extended Range Vocab- letter " S " and ending with the letter "R" and ulary Test (Verbal Comprehension, V-3),3 then is asked to write as many words as posIdentical Pictures (Perceptual Speed, P-3), sible beginning with the letter " T " and ending and the Addition Test (Numerical Facility, with the letter " D . " N-l). A Word Fluency task, Prefixes (Fw-2, designed by the first author to be similar to Design the ETS version of this test) was also included Subjects were randomly assigned to six as the fourth dependent variable. This test testing conditions. The design was a 2 (Sex) consists of two parts. First, the participant by 6 (Condition) arrangement. Since Furry is given 3 min to write as many words as pos- and Baltes had found several sex interactions sible beginning with the prefix "PRE." In and main effects, it seemed appropriate to the second part, the participant is asked to include sex as a factor in the design. Condiwrite as many words as possible beginning tions varied in number of tests, testing breaks, with the prefix "BE" in 3 min. These four and pre-test fatigue as follows: Condition 1 — subtests (V3, P3, Nl, Fw2) will be referred Control: Main battery of 4 tests administered to as the main battery. straight through without previous tests (45 Other tests from the ETS battery were used min); Condition 2 — Four preliminary tests as potentially fatigue producing instruments (Fatigue Battery 1), a rest interval of 10 min, then the main battery (85 min total); Condition 3 — Preliminary tests (Fatigue Battery ••Letter-number abbreviations follow the ETS system used in French 1), main battery (75 min) with no rest interval; el al. (1963). would be alleviated by introducing break conditions during the session. It was expected that subtest scores for Word Fluency and Perceptual Speed would be most effected by fatigue, due to their highly speeded nature.
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
FATIGUE EFFECTS ON INTELLIGENCE
Procedure All sessions were scheduled in the morning in such a manner that the main battery was administered beginning at 11 a.m. Subjects were tested in groups of 8 to 10. Instructions were given by the same test proctor throughout the study. Subjects were first given a data sheet to fill out concerning their age and educational level. They were then instructed by the test proctor as to the procedure for taking the tests and recording their answers. Individual subtests were distributed by the test proctor and by an aide. For the Finding A's test (in Conditions 5 and 6), every 2 min the subjects were instructed to circle the word at which they were looking. This last procedure follows Furry and Baltes. RESULTS
Effects of testing condition and sex were analyzed by an overall least squares multivariate analysis of variance. Significant main
effects were followed up by univariate analysis for each subtest of the main battery. Since Condition 6 had been found to yield fatigue effects previously, a planned comparison was carried out contrasting Condition 6 with the other conditions when univariate Ftests were significant. A preliminary test of the homogeneity of the 12 cells' variance-covariance matrices, using the procedure indicated by Morrison (1967) with Box's (1950) F-distribution approximation appropriate for small cell Ns, indicated that the null hypothesis was plausible (Fiio,2i2 = .466,/? > .05). The nonorthogonality problem (unequal numbers of subjects in various cells) was handled using the elimination or model comparison approach of Applebaum and Cramer (1974). The overall multivariate analysis of variance Fatigue x Sex interaction was nonsignificant (F20,i27 = 1-24, p > .05). The multivariate test for Fatigue effects, eliminating Sex effects, was significant (F20,i27 = 2.60, p < .001) as was the test for Sex effects, eliminating Fatigue effects (F4,38 = 2.69, p < .05). The results for the univariate tests for V-3, P-3, N-l, and Fw-2, are presented in Table 1. No interactions were statistically significant. The only significant univariate F-ratio for the Fatigue conditions was for P-3 (F5,4i = 5.41, p < .001). The planned comparison for P-3 in Conditions 1 through 5 against 6 yielded significant results (F5,47 = 5.27, p < .001, mean for Condition 6 = 26.33; mean for five other conditions = 41.23). A one-way analysis of variance of the same scores in Conditions 1 through 5 was not significant (F4,37 = 1.44,
Table 1. Univariate Analyses of Variance for Four Dependent Variables. Wf2
Nl
P3
V3
df
MS
F
MS
F
MS
F
125.88 491.18 22.24
1.54 6.01*
79.68 1451.22 285.20 152.21
9.53** 1.87
81.76
423.45 7.66 121.98 78.27
5.41***
Interaction Residual
5 1 5 41
Total
52
88.15
114.31
Source Main Effects Fatigue
Sex
.27
.10 1.56
183.00
.52
MS
F
75.20 40.90 92.16
1.06
.58 1.30
70.76 279.02
Note: V3 is the Extended Range Vocabulary Test (Verbal Comprehension); P3 is Identical Pictures (Perceptual Speedl); Nl is an Addition Test (Numerical Facility); and Wf2 is Prefixes (Word Fluency). * p < .05 **p < .01 ***p < .001
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
Condition 4 — Two sets of four preliminary tests (Fatigue Batteries 1 and 2) with a rest interval between them followed 'by another rest interval and the main battery (150 min); Condition 5 — Modified Finding A's test, rest interval, main battery (75 min); Condition 6 — Modified Finding A's test and the main battery without a rest interval (65 min). It is important to note that the Finding A's (P-l) test in Fatigue Battery 2 was administered in the usual way as a short, speeded test, not as the lengthy modified version used by Furry and Baltes.
543
544
CUNNINGHAM, SEPK0SK1, AND OPEL
III
IV
V
VI
CONDITIONS
Fig. 1. Mean Perceptual Speed Scores (P-3) with Standard Errors across six Experimental Conditions.
p > .05) (see Fig. 1). Significant sex effects were found for V-3 (F M | = 6.01,/? < .05, male mean = 35.00, female mean = 30.44) and N-l (F,.4i = 9.53, p < .01, male mean = 47.12, female mean = 37.06). Results indicated that there was a deficit for scores on the P-3 subtest among the elderly only when the main battery of tests was preceded without a break by a fatigue-producing condition (Condition 6). There was no significant deficit for Vocabulary scores in Condition 6 as found by Furry and Baltes (1973). This, as well as other differing results, may have been due to different test materials, scheduling arrangements, or the increased educational level of the present sample in contrast to the elderly sample of Furry and Baltes. Lack of power due to the modest sample size employed was also a possible source of differing results. On the other hand, none of the nonsignificant main effect fatigue F-ratios are remotely close to statistical significance. DISCUSSION
Although Furry and Baltes' (1973) recent work was primarily focused on age comparisons, their results emphasized the issue of effects of fatigue on intelligence test performance. In particular, the results with the Modified Finding A's task raised the possibility that lengthy test-taking procedures may yield biased results. The present findings indicate that fatigue effects may not have as strong an impact as Furry and Baltes' results seem to suggest. Fatigue effects were found only in. Condition 6, as indicated by its contrast
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
II
EXPERIMENTAL
with the other five conditions (p < .001). No significant differences were found among the other five conditions (p > .05). Fatigue effects were alleviated simply by introducing a break between the Finding A's test and the main battery (Condition 5). Even the longest battery of tests, Condition 4 (150 min) showed no significant deficit of scores. This suggests that Furry and Baltes' results may have been less general than was thought previously. It is possible that fatigue effects for moderate length batteries are specific to pretest tasks such as the Finding A's test and that the majority of the research work showing age differences of scores on intelligence tests among the elderly cannot reasonably be attributed to fatigue effects. However, because of the differences in conditions (particularly time of administration; subjects were tested at various times of day in the Furry & Baltes study), the educational level of the samples (mean educational level for the Furry-Baltes sample was 11.5 years) and since different tests were employed in the two studies, these conclusions are quite tentative. The present study did not find a deficit of Word Fluency scores under the Finding A's Condition as did Furry and Baltes. This may have been due to the fact that the present Word Fluency subtest was more restrictive than that of the PMA. The present test required subjects to list words beginning with "PRE" and "BE," in separate timings. The PMA Word Fluency test requires subjects to list words beginning with a single consonant. The present test may have been measuring vocabulary to a greater extent than the PMA Word Fluency. It should be noted that conceptually a Word Fluency task involves both the amount of vocabulary stored and efficiency (speed/accuracy) of retrieval from that store. Whether the task is more heavily involved with the amount of words stored or efficiency presumably depends at least in part on the restrictiveness of the stimulus materials. Thus, using a letter for a stimulus would probably put greater emphasis on retrieval, whereas a relatively more restrictive stimulus such as a prefix would put greater task emphasis on the amount of words in the store. Thus, it is speculated that a possible de-emphasis on retrieval may have made this task less sensitive to fatigue, in contrast to the PMA fluency task.
FATIGUE EFFECTS ON INTELLIGENCE
SUMMARY
Previous research raised the possibility that lengthy intelligence test sessions might yield biased results due to fatigue effects. The purpose of the present study was to investigate effects of fatigue on intelligence test performance in the elderly. Dependent variables were Vocabulary, Numerical Facility, Perceptual Speed, and Word Fluency tests. Fatigue effects were investigated by varying the number of previous tests administered, by introducing breaks between tests in some conditions, and by using a specific pretest fatigue-producing condition, the modified Finding A's test. Subjects were between the ages of 57 to 91 years with an average age of 72 years. Mean educational level was 14 years. It was hypothesized that the modified version of the Finding A's test utilized to simulate fatigue effects in previous work would be more fatiguing than a long battery of tests and that introducing a break condition between Finding A's test and the main battery would alleviate fatigue effects. Analyses of Variance yielded a main effect due to a pre-test condition for the Perceptual Speed test only when the main battery was preceded by the Finding A's task (p < .001). Preceding the main battery by various numbers of tests (up to a total session of 150 min) yielded no statistically significant effects. It appears that the elderly are not as susceptible to test fatigue as previous results seemed to suggest.
REFERENCES
Applebaum, M. I., & Cramer, E. M. Some problems in the nonorthogonal analysis of variance. Psychological Bulletin, 1974, 81, 335-343. Box, G. E. P. A General distribution theory for a class of likelihood criteria. Biometrika, 1950,56, 317-346. Cunningham, W. R. & Birren, J. E. Age changes in human abilities: A 28-year longitudinal study. Developmental Psychology, 1976, 12, 82-83. French, J. W., Ekstrom, R. B., & Price, L. A. Manual for kit of reference tests for cognitive factors. Educational Testing Service, Princeton, 1963. Furry, C. A. & Baltes, P. B. The effect of age differences in ability-extraneous performance variables on the assessment of intelligence in children, adults, and the elderly. Journal of Gerontology, 1973,28, 73-80. Kamin, L. J. Differential changes in mental ability in old age. Journal of Gerontology, 1957, 12, 66-70. Morrison, D. F. Multivariate statistical methods. McGraw-Hill, New York, 1967.
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
Differences in results between the current study and those of the Furry and Baltes study may be related to the time of day of test administration. In the current study, all participants began the dependent variable battery at 11 a.m. This was due in part to the traditional psychometric preference for morning ability testing and also due to the subjective observation of the authors from pilot studies that participants appeared to be more fatigued at trie end of afternoon sessions. In contrast, Furry and Baltes tested participants at various times, including afternoon and evening sessions. It is possible that such differences in scheduling may have resulted in higher fatiguability in the Furry and Baltes' sample. A main effect due to sex was found for Nl (p < .01). Past research has often found that males of older generations score higher than females on mathematical tests (e.g., Cunningham & Birren, 1976). It is also possible that these effects were related to sex differences in educational attainment in the current sample. In the same way, the significant sex effect in favor of males on V3 may be due to educational differences in these samples. Condition 4 (4 preliminary tests, rest interval, 4 preliminary tests, rest interval, main battery) was included to determine whether or not elderly subjects could take a long battery of tests and not be seriously affected by fatigue or boredom. With the present kind of sample, these results suggest that it is safe to administer batteries of tests in sessions of up to 150 min, if break periods are allowed. However, this conclusion should be tempered by consideration of the modest sample size employed in the current study. It is always possible that a larger sample would yield more significant effects. It is surprising that only three studies have been carried out focusing on this issue, considering the large number of descriptive studies in the literature which assume such effects are negligible. Clearly, far more research needs to be done with regard to these issues. It would be interesting to attend to the question of the maximum length of test batteries possible among the elderly before obtaining decrements attributable to fatigue effects. Possible fatigue interactions with time of day, educational level, and age within the elderly range would also be of interest for further study.
545
Fatigue Effects on Intelligence Test Performance in the Elderly1 Walter R. Cunningham, PhD, Carol M. Sepkoski, BA, and Mary R. Opel, BA 2
N issue of central importance in interpreting research on intellectual functioning A in the elderly concerns fatigue effects. Fatigue effects represent a particularly serious issue for researchers administering long batteries of tests and seeking to make generalizations regarding abilities and age. Kamin (1957) suggested that decline of various subtest scores on the Primary Mental Abilities battery (PMA) may be an artifact of the order in which the subtests are administered (i.e., fatigue or boredom may influence different subtest scores). He administered the PMA subtest in a Latin square design to young and old subjects. Results indicated that there were no significant order effects, and Kamin suggested that age decrements were indeed due to changing cognitive abilities rather than to performance related variables. Furry and Baltes (1973) proposed that elderly subjects are sensitive to situational variables in test-taking such as fatigue-producing conditions and that these may result in performance related age differences. The investigators administered the PMA to subjects of three age groups; subtest position and pretestfatigue conditions were varied. Results indicated that, among the elderly subjects there
'This research was supported by AOA Grant #90-646-01, which is gratefully acknowledged. The authors would like to express their sincere appreciation to Jo Anne Judd for assistance in participant recruitment and scheduling, test scoring, and keypunching. The cooperation of Ms. Emily Barefield and Reverend Summers of AARP is also gratefully acknowledged. 2 Psychology Dept., Univ. of Florida, Gainesville 32611.
were no position effects, but the fatigue condition significantly influenced performance on the Word Fluency, Verbal Meaning, and Reasoning Subtests. Furry and Baltes emphasized interpretation of age by fatigue interactions and concluded that the intelligence score differences among the elderly, as contrasted to the young, may be influenced by situational determinants. Therefore, it may be inappropriate to relate age differences solely to ability specific factors. Furry and Baltes discussed fatigue effects as one example of ability extraneous influence and suggested that fatigue influences, together with other performance variables, might complicate the interpretation of the existing abilities literature. They used a specific fatigue-producing condition in their battery, a modified version of the Finding A's Test (P-l in the Educational Testing Service (ETS) Kit of Reference Tests for Cognitive Factors; French et al., 1963). The modification consisted of lengthening a letter cancellation task to 20 min. They also noted that this fatigue producing treatment was excessively fatiguing. They acknowledged that the procedure might be more fatiguing than simply taking other tests. The present study attempted to determine whether this specific task yields similar effects in contrast to actually taking a long battery of ability tests. It was hypothesized that the Finding A's task would be more fatiguing than a long battery of subtests and that fatigue effects 541
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
The purpose of the study was to investigate effects of fatigue on intelligence test performance in the elderly. Dependent variables were Verbal Comprehension, Numerical Facility, Perceptual Speed, and Word Fluency tests. Fatigue effects were investigated by varying the number of previous tests administered, by introducing breaks between tests in some conditions, and by using a pre-test fatigue-producing condition, a modified form of the Finding A's test. Subjects' ages were between 57 to 91-years. It was hypothesized that the Finding A's test would be more fatiguing than a long battery of tests and that introducing a break condition between the Finding A's test and the main battery would alleviate fatigue effects. Analyses of Variance resulted in a main effect due to a pre-test condition for the Perceptual Speed test only, and only when the main battery was preceded by the Finding A's task (p < .001). It appears that the elderly are not as susceptible to test fatigue as previous results seemed to suggest.
542
CUNNINGHAM, SEPKOSKI, AND OPEL
for the long battery conditions, as described below. Test materials were selected from the ETS battery rather than the PMA because the primary purpose of the study was to determine the degree of seriousness of fatigue effects in the ETS battery, since the first author was in the process of planning a large-sample, large-variable study of factor structure of abiliMETHOD ties in late life. The current experiment involved no variation of position for any of Subjects Letters were mailed to all members (approxi- the tests, since Furry and Baltes (1973) and mately 300) of the local chapter of the Ameri- Kamin (1957) found no order effects for elcan Association of Retired Persons in Gaines- derly subjects. ville, Florida. The letter explained the nature Fatigue Battery 1 consisted of Advanced of the study and invited participation; a $5.00 Vocabulary (V-4), Number Comparison (P-2), gratuity was also offered. Thirty males and 75 Division Test (N-2), and Word Fluencyfemales volunteered, and of these 24 males and Suffixes (Fw-1) administered in that order. 56 females were scheduled for testing sessions. Fatigue Battery 2 consisted of Vocabulary Twenty-seven subjects canceled for various (V-2), Finding A's (P-l), Subtraction and reasons. The final sample consisted of 17 Multiplication (N-3), and Word Fluencymales and 36 females. The range of age was Beginnings and Ending (Fw-3) in that order. 57 to 91 years, with a mean of 72 years. The Word Fluency-Suffixes was constructed for range of educational level was 7 to 20 years, use in the subsequent state-wide study and with an overall mean level of approximately was intended to be similar to the ETS test 14 years, with the male average being higher Fw-1. The test consists of two parts. First, than the female average by about 1 year. The the participant is given 3 min to write down as sex ratio in the sample approximates that of many words as possible ending in the suffix the full membership of the local chapter. "TION"; then the participant is given 3 min to write as many words as possible ending in the suffix "AL." The Word Fluency test Beginnings and Endings also consists of two MATERIALS Three subtests of the ETS battery were used parts. The participant is given 3 min to write as dependent variables and were presented as many words as possible beginning with the in the following order: Extended Range Vocab- letter " S " and ending with the letter "R" and ulary Test (Verbal Comprehension, V-3),3 then is asked to write as many words as posIdentical Pictures (Perceptual Speed, P-3), sible beginning with the letter " T " and ending and the Addition Test (Numerical Facility, with the letter " D . " N-l). A Word Fluency task, Prefixes (Fw-2, designed by the first author to be similar to Design the ETS version of this test) was also included Subjects were randomly assigned to six as the fourth dependent variable. This test testing conditions. The design was a 2 (Sex) consists of two parts. First, the participant by 6 (Condition) arrangement. Since Furry is given 3 min to write as many words as pos- and Baltes had found several sex interactions sible beginning with the prefix "PRE." In and main effects, it seemed appropriate to the second part, the participant is asked to include sex as a factor in the design. Condiwrite as many words as possible beginning tions varied in number of tests, testing breaks, with the prefix "BE" in 3 min. These four and pre-test fatigue as follows: Condition 1 — subtests (V3, P3, Nl, Fw2) will be referred Control: Main battery of 4 tests administered to as the main battery. straight through without previous tests (45 Other tests from the ETS battery were used min); Condition 2 — Four preliminary tests as potentially fatigue producing instruments (Fatigue Battery 1), a rest interval of 10 min, then the main battery (85 min total); Condition 3 — Preliminary tests (Fatigue Battery ••Letter-number abbreviations follow the ETS system used in French 1), main battery (75 min) with no rest interval; el al. (1963). would be alleviated by introducing break conditions during the session. It was expected that subtest scores for Word Fluency and Perceptual Speed would be most effected by fatigue, due to their highly speeded nature.
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
FATIGUE EFFECTS ON INTELLIGENCE
Procedure All sessions were scheduled in the morning in such a manner that the main battery was administered beginning at 11 a.m. Subjects were tested in groups of 8 to 10. Instructions were given by the same test proctor throughout the study. Subjects were first given a data sheet to fill out concerning their age and educational level. They were then instructed by the test proctor as to the procedure for taking the tests and recording their answers. Individual subtests were distributed by the test proctor and by an aide. For the Finding A's test (in Conditions 5 and 6), every 2 min the subjects were instructed to circle the word at which they were looking. This last procedure follows Furry and Baltes. RESULTS
Effects of testing condition and sex were analyzed by an overall least squares multivariate analysis of variance. Significant main
effects were followed up by univariate analysis for each subtest of the main battery. Since Condition 6 had been found to yield fatigue effects previously, a planned comparison was carried out contrasting Condition 6 with the other conditions when univariate Ftests were significant. A preliminary test of the homogeneity of the 12 cells' variance-covariance matrices, using the procedure indicated by Morrison (1967) with Box's (1950) F-distribution approximation appropriate for small cell Ns, indicated that the null hypothesis was plausible (Fiio,2i2 = .466,/? > .05). The nonorthogonality problem (unequal numbers of subjects in various cells) was handled using the elimination or model comparison approach of Applebaum and Cramer (1974). The overall multivariate analysis of variance Fatigue x Sex interaction was nonsignificant (F20,i27 = 1-24, p > .05). The multivariate test for Fatigue effects, eliminating Sex effects, was significant (F20,i27 = 2.60, p < .001) as was the test for Sex effects, eliminating Fatigue effects (F4,38 = 2.69, p < .05). The results for the univariate tests for V-3, P-3, N-l, and Fw-2, are presented in Table 1. No interactions were statistically significant. The only significant univariate F-ratio for the Fatigue conditions was for P-3 (F5,4i = 5.41, p < .001). The planned comparison for P-3 in Conditions 1 through 5 against 6 yielded significant results (F5,47 = 5.27, p < .001, mean for Condition 6 = 26.33; mean for five other conditions = 41.23). A one-way analysis of variance of the same scores in Conditions 1 through 5 was not significant (F4,37 = 1.44,
Table 1. Univariate Analyses of Variance for Four Dependent Variables. Wf2
Nl
P3
V3
df
MS
F
MS
F
MS
F
125.88 491.18 22.24
1.54 6.01*
79.68 1451.22 285.20 152.21
9.53** 1.87
81.76
423.45 7.66 121.98 78.27
5.41***
Interaction Residual
5 1 5 41
Total
52
88.15
114.31
Source Main Effects Fatigue
Sex
.27
.10 1.56
183.00
.52
MS
F
75.20 40.90 92.16
1.06
.58 1.30
70.76 279.02
Note: V3 is the Extended Range Vocabulary Test (Verbal Comprehension); P3 is Identical Pictures (Perceptual Speedl); Nl is an Addition Test (Numerical Facility); and Wf2 is Prefixes (Word Fluency). * p < .05 **p < .01 ***p < .001
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
Condition 4 — Two sets of four preliminary tests (Fatigue Batteries 1 and 2) with a rest interval between them followed 'by another rest interval and the main battery (150 min); Condition 5 — Modified Finding A's test, rest interval, main battery (75 min); Condition 6 — Modified Finding A's test and the main battery without a rest interval (65 min). It is important to note that the Finding A's (P-l) test in Fatigue Battery 2 was administered in the usual way as a short, speeded test, not as the lengthy modified version used by Furry and Baltes.
543
544
CUNNINGHAM, SEPK0SK1, AND OPEL
III
IV
V
VI
CONDITIONS
Fig. 1. Mean Perceptual Speed Scores (P-3) with Standard Errors across six Experimental Conditions.
p > .05) (see Fig. 1). Significant sex effects were found for V-3 (F M | = 6.01,/? < .05, male mean = 35.00, female mean = 30.44) and N-l (F,.4i = 9.53, p < .01, male mean = 47.12, female mean = 37.06). Results indicated that there was a deficit for scores on the P-3 subtest among the elderly only when the main battery of tests was preceded without a break by a fatigue-producing condition (Condition 6). There was no significant deficit for Vocabulary scores in Condition 6 as found by Furry and Baltes (1973). This, as well as other differing results, may have been due to different test materials, scheduling arrangements, or the increased educational level of the present sample in contrast to the elderly sample of Furry and Baltes. Lack of power due to the modest sample size employed was also a possible source of differing results. On the other hand, none of the nonsignificant main effect fatigue F-ratios are remotely close to statistical significance. DISCUSSION
Although Furry and Baltes' (1973) recent work was primarily focused on age comparisons, their results emphasized the issue of effects of fatigue on intelligence test performance. In particular, the results with the Modified Finding A's task raised the possibility that lengthy test-taking procedures may yield biased results. The present findings indicate that fatigue effects may not have as strong an impact as Furry and Baltes' results seem to suggest. Fatigue effects were found only in. Condition 6, as indicated by its contrast
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
II
EXPERIMENTAL
with the other five conditions (p < .001). No significant differences were found among the other five conditions (p > .05). Fatigue effects were alleviated simply by introducing a break between the Finding A's test and the main battery (Condition 5). Even the longest battery of tests, Condition 4 (150 min) showed no significant deficit of scores. This suggests that Furry and Baltes' results may have been less general than was thought previously. It is possible that fatigue effects for moderate length batteries are specific to pretest tasks such as the Finding A's test and that the majority of the research work showing age differences of scores on intelligence tests among the elderly cannot reasonably be attributed to fatigue effects. However, because of the differences in conditions (particularly time of administration; subjects were tested at various times of day in the Furry & Baltes study), the educational level of the samples (mean educational level for the Furry-Baltes sample was 11.5 years) and since different tests were employed in the two studies, these conclusions are quite tentative. The present study did not find a deficit of Word Fluency scores under the Finding A's Condition as did Furry and Baltes. This may have been due to the fact that the present Word Fluency subtest was more restrictive than that of the PMA. The present test required subjects to list words beginning with "PRE" and "BE," in separate timings. The PMA Word Fluency test requires subjects to list words beginning with a single consonant. The present test may have been measuring vocabulary to a greater extent than the PMA Word Fluency. It should be noted that conceptually a Word Fluency task involves both the amount of vocabulary stored and efficiency (speed/accuracy) of retrieval from that store. Whether the task is more heavily involved with the amount of words stored or efficiency presumably depends at least in part on the restrictiveness of the stimulus materials. Thus, using a letter for a stimulus would probably put greater emphasis on retrieval, whereas a relatively more restrictive stimulus such as a prefix would put greater task emphasis on the amount of words in the store. Thus, it is speculated that a possible de-emphasis on retrieval may have made this task less sensitive to fatigue, in contrast to the PMA fluency task.
FATIGUE EFFECTS ON INTELLIGENCE
SUMMARY
Previous research raised the possibility that lengthy intelligence test sessions might yield biased results due to fatigue effects. The purpose of the present study was to investigate effects of fatigue on intelligence test performance in the elderly. Dependent variables were Vocabulary, Numerical Facility, Perceptual Speed, and Word Fluency tests. Fatigue effects were investigated by varying the number of previous tests administered, by introducing breaks between tests in some conditions, and by using a specific pretest fatigue-producing condition, the modified Finding A's test. Subjects were between the ages of 57 to 91 years with an average age of 72 years. Mean educational level was 14 years. It was hypothesized that the modified version of the Finding A's test utilized to simulate fatigue effects in previous work would be more fatiguing than a long battery of tests and that introducing a break condition between Finding A's test and the main battery would alleviate fatigue effects. Analyses of Variance yielded a main effect due to a pre-test condition for the Perceptual Speed test only when the main battery was preceded by the Finding A's task (p < .001). Preceding the main battery by various numbers of tests (up to a total session of 150 min) yielded no statistically significant effects. It appears that the elderly are not as susceptible to test fatigue as previous results seemed to suggest.
REFERENCES
Applebaum, M. I., & Cramer, E. M. Some problems in the nonorthogonal analysis of variance. Psychological Bulletin, 1974, 81, 335-343. Box, G. E. P. A General distribution theory for a class of likelihood criteria. Biometrika, 1950,56, 317-346. Cunningham, W. R. & Birren, J. E. Age changes in human abilities: A 28-year longitudinal study. Developmental Psychology, 1976, 12, 82-83. French, J. W., Ekstrom, R. B., & Price, L. A. Manual for kit of reference tests for cognitive factors. Educational Testing Service, Princeton, 1963. Furry, C. A. & Baltes, P. B. The effect of age differences in ability-extraneous performance variables on the assessment of intelligence in children, adults, and the elderly. Journal of Gerontology, 1973,28, 73-80. Kamin, L. J. Differential changes in mental ability in old age. Journal of Gerontology, 1957, 12, 66-70. Morrison, D. F. Multivariate statistical methods. McGraw-Hill, New York, 1967.
Downloaded from http://geronj.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 22, 2015
Differences in results between the current study and those of the Furry and Baltes study may be related to the time of day of test administration. In the current study, all participants began the dependent variable battery at 11 a.m. This was due in part to the traditional psychometric preference for morning ability testing and also due to the subjective observation of the authors from pilot studies that participants appeared to be more fatigued at trie end of afternoon sessions. In contrast, Furry and Baltes tested participants at various times, including afternoon and evening sessions. It is possible that such differences in scheduling may have resulted in higher fatiguability in the Furry and Baltes' sample. A main effect due to sex was found for Nl (p < .01). Past research has often found that males of older generations score higher than females on mathematical tests (e.g., Cunningham & Birren, 1976). It is also possible that these effects were related to sex differences in educational attainment in the current sample. In the same way, the significant sex effect in favor of males on V3 may be due to educational differences in these samples. Condition 4 (4 preliminary tests, rest interval, 4 preliminary tests, rest interval, main battery) was included to determine whether or not elderly subjects could take a long battery of tests and not be seriously affected by fatigue or boredom. With the present kind of sample, these results suggest that it is safe to administer batteries of tests in sessions of up to 150 min, if break periods are allowed. However, this conclusion should be tempered by consideration of the modest sample size employed in the current study. It is always possible that a larger sample would yield more significant effects. It is surprising that only three studies have been carried out focusing on this issue, considering the large number of descriptive studies in the literature which assume such effects are negligible. Clearly, far more research needs to be done with regard to these issues. It would be interesting to attend to the question of the maximum length of test batteries possible among the elderly before obtaining decrements attributable to fatigue effects. Possible fatigue interactions with time of day, educational level, and age within the elderly range would also be of interest for further study.
545
