Person. in&id. Difl Vol. I I. No. Il. pp. 1147-1152. Pnnted in Great Bntain. All rights reserved
1990 Copyrtght
0191~8869/90 53.00 + 0.00 C 1990 Pcrgamon Press plc
EFFECTS OF AGE AND AGE-RELATED DIFFERENCES IN AUDITORY INFORMATION PROCESSING ON FLUID AND CRYSTALLIZED INTELLIGENCE WAFTALI
‘Department
RAZ,‘.* PAUL J. MOBERG? and DARYL MILLMAN*
of Psychology, Memphis State University, Memphis, TN 38152 and *Departmentof Psychology, The Chicago Medical School, North Chicago, IL 60064, U.S.A. (Received 20 March
1990)
Summary-Speed and fidelity of basic information processing and fluid intelligence deteriorate with age, while crystallized intelligence remains stable. There is also evidence that indices of basic information processing correlate with fluid intelligence in a college population. We hypothesized that age-related changes in simple auditory information processing may explain a significant proportion of age-related decline in fluid, but not crystallized intelligence. To test this hypothesis, we obtained two-tone frequency discrimination thresholds as well as scores on fluid and crystallized intelligence tests in a group of healthy adults spanning ages from I8 to 83. We then compared linear models specifying the relations among those measures. The model postulating age effects on fluid intelligence mediated by differences in frequency discrimination fitted the data better than the model restricted to direct effects of age on fluid intelligence. No effects of information-processing index or age on crystallized intelligence were observed.
INTRODUCTION
The process of growing old is accompanied by deterioration of peripheral sensory functions (Corso, 1981), progressive slowing of central information processing (Cerella, 1985), and decline of important aspects of intellectual aptitude (Horn, 1985). A subset of abilities, known as fluid intelligence and concerned with abstract reasoning, problem-solving, discovery of rules and concept formation, appears to be especially vulnerable to ageing. On the other hand, higher cognitive functions based on culture-bound use of rules, knowledge of facts and use of established concepts comprise crystallized intelligence which remains stable in the course of normal ageing (Horn, 1985). In spite of accumulation of substantial knowledge about all aforementioned facets of cognitive ageing, the links among them are still poorly understood. Cumulative research of the past decade supports the notion that basic information-processing and psychometrically defined intelligence are associated (Nettelbeck & Lally, 1976; Hunt, 1978; Stankov & Horn, 1980; Jensen, 1982; Brand & Deary, 1982; Raz, Willerman, Ingmundson & Hanlon, 1983). Raz, Willerman, and Yama (1987) have reported a significant correlation between a measure of fluid abilities, the Cattell Culture-Fair Intelligence Test (CFIT; Cattell & Cattell, 1973) and a relatively simple index of auditory information processing: frequency discrimination threshold. Stankov (1988) reported that although memory for tones loaded highly on the general auditory factor, it correlated significantly with fluid, but not crystallized intelligence. It appears that performance on a variety of elementary cognitive tasks may explain a significant proportion of variance observed in measures of higher intellectual functions, including, age-related variability. In this study, we attempted to elucidate some of the connections among different domains of cognitive ageing. Namely, we focused our attention on age effects on the relationship between auditory information processing and intelligence. To that objective, we measured performance on a relatively simple auditory information-processing task and representative tests of fluid and crystallized intelligence, and compared linear models specifying the relations among those measures. Since both fluid intelligence and frequency discrimination thresholds deteriorate with age (Horn, 1985; Raz, Millman & Moberg, 1989), and frequency discrimination is related to fluid intelligence in young people (Raz et al., 1987), we hypothesized that age-related elevation of frequency-discrimination thresholds may explain a significant proportion of age-related decline in fluid, but not crystallized intelligence, above and beyond the effect of age alone. *To whom all correspondence PAID Ilill-D
should be addressed. 1147
1148
NAFTALI R\z er al.
METHOD
Subjects
Forty-nine volunteers were recruited from Northern Illinois communities as well as from staff and students of The Chicago Medical School. Prospective participants were interviewed and screened for a history of audiological problems, environmental and occupational noise exposure, use of ototoxic medication, history of alcoholism, drug abuse, major psychiatric disorders, diabetes, and hypertension-conditions known to be associated with various degrees of cognitive impairment (White, Cartwright, Cornoni-Huntley & Brock, 1986). All Ss scored above 29 (out of 30) on the Mini-Mental State Examination (MMSE) (Folstein, Folstein & McHugh, 1975) and all were community-dwelling and self-supported individuals. Pure-tone air-conduction audiograms were obtained using a Tracer audiometer (Model RA214). At 750 Hz, Ss were required to have unilateral thresholds below 20 dB HL. Because five elderly Ss could not qualify for the study due to grossly impaired hearing, the final sample included 44 Ss (14 males). The mean age in that sample was 39.8 (SD = 19.5) and mean duration of formal education was 15.9 yr (SD = 2.8). The Spearman rank-order correlation between age and education was p = -0.11, NS. Tests
The normal number of correct answers on the Cattell Culture-Fair Intelligence Test (CFIT)Scale 2, Form A was used as a measure of fluid intelligence. Initially, the Vocabulary subtest of the (WAIS-R) was administered to measure crystallized intelligence. However, we found that 12 Ss (all students) have previously taken this test, and for them Extended Vocabulary (V3) from the Educational Testing Services Factor-Referenced Tests Kit (Ekstrom, French, Hat-man & Derman, 1976) was administered as a substitute. Raw scores of WAIS-R vocabulary, and total number of correct answers on V3 corrected for guessing was used as mesures of vocabulary. A total of 30 Ss were administered both vocabulary tests, the correlation between which was quite high: r = 0.88. For those who had only V3 scores, values predicted by a regression of WAIS-R vocabulary on V3 were assigned. Apparatus and procedure Ss were run singly, in a dimly-lit sound-attenuated chamber. Before each session, the task was explained, and several practice trials were administered to verify comprehension of the instructions. Ss were instructed to execute the task at their own pace by initiating a new trial only when they felt comfortable and were ready to listen. Stimuli were digitally synthesized sine-wave bursts with total duration of 40 msec, fundamental frequency range of 770-870 Hz, and peak signal level of 75 dB SL. Stimuli were presented monotically in a two-interval forced choice (21FC) paradigm with a visual warning message, and visual feedback. The order of the ears was randomized. The frequency discrimination task was administered in three 1-hr sessions. Six to seven blocks of trials (40-60 trials each) were administered in each session. Frequency discrimination thresholds (6F), were computed for 70.7% correct using a transformed adaptive staircase procedure (Levitt, 1971). The performance index (6F) was computed as an average of the best four of the last six blocks. The test-retest reliability of this procedure is about 0.90 (Raz, 1985; Johnson, Watson & Jensen, 1987). Further details of the apparatus and procedure employed for measuring frequency discrimination thresholds can be found in Raz et al. (1989).
RESULTS
The zero-order correlations among age, frequency discrimination thresholds, and the measures of fluid and crystallized intelligence are presented in Table 1. Since the distributions of 6F was skewed, a logarithmically-transformed index (log dF) was used. The correlations indicate a moderate association between age, auditory information processing, and fluid intelligence. The scatter plots in Fig. 1 show that those correlations do not reflect undue influence of outliers.
Ageing. information processing, and intelligence
1149
To explore the relationships among the variables in Table 1 further, two linear models were fitted to the data. These models are depicted in path diagrams in Fig. 2. Age was assumed to be a source variable that is influenced neither by intelligence, nor by auditory information processing, whereas frequency discrimination was postulated to influence fluid intelligence or to be unrelated to it. Based on the evidence reviewed in the introduction, the reversed effect, i.e. frequency discrimination thresholds being influenced by intelligence, was considered highly improbable. Absolute auditory accuity measure (pure-tone threshold) was not included in this analysis because it was used as a screening device, and, by design, had a severely truncated range. Model I is a fully-recursive three-variable model reflecting the hypothesis that fluid intelligence is directly affected by ageing per se as well as by age-related changes in auditory information processing. Model II, nested within Model I, postulates independent age effects on fluid intelligence Table I. Zero-order correlations among the variables included in the models Log dF
CFIT
Vocabularly
0.46
-0.77 -0.52
0.24 -0.00 0.09
Age Log dF CFIT
Correlations I >0.29 are significant at P < 0.05 level, 2-tailed.
6-
101 0
-4
’
’
20
’
’
40 Age
’
’
’
60
’
’
60
II
1 100
0
I 20
I
I
I
40
lyr)
Age
60
I
I 60
(yr)
40 r
36 t 30 c 24
16 t q
12
-3
'
I -2
I -1
I 0
I 1
I 2
I 3
I 4
log dF Fig. I. Scatter plots and regression lines illustrating relationships among age, frequency discrimination and fluid intelligence.
I II Fig. 2. Path diagrams of models I and II with standardized path coefficients indicated on the paths.
I
J 100
1150
NA~ALI RAN. et al.
and auditory information processing. The models were compared using the method described by Pedhazur (1982) after Specht (1975). According to this approach, a summary index of the amount of variance explained by a fully recursive model (Rk) is computed by subtracting the product of all squared residual path coefficients from 1. Similar index (M) is computed for overidentified models nested within the fully recursive one. The M values for nested models are then compared with Ri using a W statistic, W = -(N - k)ln[( 1 - Rh)/(l - M)], where N is the sample size and k is the number of overidentifying restriction on a nested model. The W statistic is distributed as a x2 variable with k degrees of freedom. A multiple regression model in which age and frequency discrimination served as predictors of fluid intelligence, accounted for 63% of the variance in the fluid intelligence [F(2,41) = 35.17, P < O.OOl]. The goodness of fit of Model I was RL = 0.710. Model II in which the path from frequency discrimination to fluid intelligence was eliminated provided a significantly worse fit to the data: M = 0.679, x2 (1) = 4.34, P < 0.05. Similar analysis for crystallized intelligence was unnecessary, for age, and frequency discrimination explained only a small part of the variance in Vocabulary scores: R2 = 0.07, F(2,41) = 1.65, NS. Multiple correlations produced by the models predicting fluid and crystallized intelligence are equivalent to bivariate correlations between the respective dependent variables and a linear combination of the independent variables. We therefore tested the difference between them using Steiger’s (1980) procedure for comparison of dependent correlation coefficients. The difference between multiple Rs for two full models was significant: Z* = 3.62, P < 0.001.
DISCUSSION The results of this study suggest that age-related decline in fluid intelligence may be mediated in part by individual differences in basic auditory information processing. Previously it was proposed that generalized cognitive slowing could be the cause of cognitive ageing (Birren, 1965; Salthouse, 1985). The relationship between selective intellectual decline and slowing of performance on simple paper-and-pencil tasks has been recently demonstrated by Hertzog (1989). The results reported here suggest an age-related link between fluid intelligence and information processing task that posed no demands on speed akin to those presented in Hertzog’s (1989) tasks. As we suggested elsewhere (Raz et al., 1989), age-related differences on non-speeded information processing tasks, such as frequency discrimination, may stem from differences in trace persistence, short-term working memory capacity and the timing of its reset rather than from primary slowing of information processing, i.e. reduced number of operations per unit of time. It is worth noting that information processing was assessed using an auditory task, while cognitive performance was measured by a test that requires no auditory processing indicates that modality-specific sensory deficits cannot account for the findings. Differential practice effects and differential use of strategies may be a source of concern when individual differences in information processing are studied. Although all Ss showed some improvement on the first five-six blocks (200-300 trials), no substantial gains were observed afterwards, and the Ss presumably reached asymptotic levels of performance. This is in accord with Watson’s (1980) observation that a two-tone discrimination task, unlike procedures involving multiple tones, is rather easy to learn. In any case, there were no indications that perceptual learning during the first blocks depended on explicit strategies. Ss questioned after the session reported no distinct strategies in dealing with the task, although all indicated that they felt increasingly more competent in dealing with the task as the experiment progressed. In spite of their preliminary nature, the findings presented here may be useful in guiding a new approach to an old problem of separating age and cohort effects. Calendar age stands for many things, and cohort membership is one of them. Although no data specifically addressing cohort effects on two-tone frequency discrimination are available, such effects were not found on laboratory measures of auditory discrimination utilizing degraded speech stimuli (Bergman, Blumenfeld, Cascardo, Dash, Levitt & Margulies, 1976). It is unlikely that simplification of the stimuli would induce cohort effects in an auditory discrimination task. Kausler (1982) reviewed the evidence of cohort effects on several basic information processing measures and found those effects
Ageing, information processing, and intelligence
1151
negligible. It may be thus possible to use frequency discrimination or similar information-processinging tasks as vehicles for separating cohort effects from those of ageing per se. Usually, such partitioning of variance requires a cross-sequential design (Buss, 1973). Unfortunately, that approach is time-consuming; it may take years to determine whether a certain cognitive measure is or is not affected by cohort effects. Seeking correlates of cognitive aptitude not contaminated by inter-generational differences in cultural practices and attitudes-relatively simple cognitive markers-may be helpful in testing validity of claims of age-related declines derived from cross-sectional age-group comparisons. The idea is that a cognitive marker task can be used to separate total variance of intelligence measures into cohort-free and cohort-dependent parts. It is highly desirable that such a marker be a performance index associated with IQ and with age, yet showing no cohort effects. To the extent that a cohort-dependent variable (e.g. intelligence) correlates with a cohort-free cognitive index, it exhibits age-related differences; the intelligence variance predicted uniquely by the cognitive marker and jointly by the cognitive marker and age is, by design, cohort-free. And finally, some words of caution. The conclusions of this study are based on comparison of linear models. Selecting an adequate set of variables for a linear model is of crucial importance for that model’s validity. Admittedly, a three-variable model proposed here presents a simplified and limited view of cognitive ageing. Only one task in one sensory modality was sampled as a representative of basic information processing; only one test per factor represented fluid and crystallized intelligence. Therefore, the findings reported here should be treated as a prologue to a truly multivariate exploration of the relationship between information processing and intelligence in the context of ageing. Acknowledgements-We thank Lee Willerman and Eric Turkheimer for helpful comments on the earlier versions of this manuscript. This research was supported in part by the Biomedical Research Support Grant S07-RRO-5366-23 to N. Raz through The Chicago Medical School.
REFERENCES Bergman, M., Blumenfeld, V. G., Cascardo, D., Dash, B., Levitt, H. & Margulies, M. K. (1976). Age-related decrement in hearing for speech: Sampling and longitudinal studies. Journal of Gerontology, 31, 533-538. Birren, J. E. (1965). Age changes in speed of behavior: Its central nature and physiological correlates. In Welford, A. T. & Birren, J. E. (Eds), Behavior, aging, and the nervous sysfem. Springfield, Ill.: Thomas. Brand, C. R. & Deary, I. J. (1982). Intelligence and ‘inspection time’. In Eysenck, H. J. (Ed.), A modelfor intelligence. New York: Springer. Buss, A. R. (1973). An extension of the developmental models that separate ontogenetic changes and cohort differences. Psychology Bulletin, 80, 466-479.
Cattell, R. B. & Cattell, A. K. S. (1973). Handbook for the individual or group Culture-Fair Inrelligence Test: Scale 2, Champaign, Ill.: IPAT. Cerella, J. (1985). Information processing rates in elderly. Psychology Bulletin, 98, 67-83. Corso, J. F. (1981). Aging, sensory systems und perception. New York: Praeger. Ekstrom, R. B., French, J. W., Harman, H. H. & Derman, D. (1976). Manual for kit offictor-referenced cognitive tests. Princeton, N.J.: Educational Testing Services. Folstein, M. F.. Folstein, S. E. & McHugh, P. R. (1975). Mini-mental state: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189-198. Hertxog, C. (1989). Influences of cognitive slowing on age differences in intelligence. Development Psychology, 25.636-651. Horn, J. L. (1985). Remodeling old models of intelligence. In Wolman, B. B. (Ed.), Hundbook of intelligence: Theories, measurements, and applications. New York: Wiley. Hunt, E. (1978). Mechanics of verbal ability. Psychological Review, 85, 109130. Jensen, A. (1982). Reaction time and intelligence. In Eysenck, H. J. (Ed.), A model for infelligence. New York: Springer. Johnson, D. M.. Watson, C. S. & Jensen, J. K. (1987). Individual differences in auditory capabilities. Journal of the Acoustical Society of America, 81, 427-438.
Kausler, D. H. (1982). Experimenrul psychology and human uging. New York: Wiley. Levitt, H. (1971). Transformed up-down method in psychoacoustics. Journal of the Acoustical Society of America, 49, 467-477.
Nettelbeck, T. & Lally, M. (1976). Inspection time and measurement of intelligence. British Journal of Psychology, 67, 17-22. Pedhazur. E. J. (1982). Mulriple regression in behavioral reseurch (2nd edn). New York: Holt. Raz. N. (1985). Auditory information processing and intelligence: Beyond mental speed. Unpublished doctoral dissertation. University of Texas at Austin. Rar, N., Millman, D. & Moberg, P. J. (1989). Auditory memory and age-related differences in two-tone frequency discrimination: Trace decay and interference. Experimenfal Aging Research, 15, 43-49. Raz, N., Willerman, L. & Yama, M. (1987). On sense and senses: Auditory information processing and intelligence. Personality and Individual Dlyerences, 7, 20 l-2 IO. Raz, N., Willerman, L., Ingmundson, P. & Hanln, M. (1983). Aptitude-related differences in auditory recognition masking. Intelligence, 7, 7 l-90.
1152
NAF~ALI RAZ et ul.
Salthouse, T. A. (1985). Speed of behavior and its implications for cognition. In Buren, J. E. & Schaic, K. W. (Eds), Handbook of the psychology of aging (Chap. 15, pp. 400-426). New York: Reinhold. Specht, D. A. (1975). On the evaluation of causal models. Social Science Research, 4, 113-133. Stankov. L. (1988). Single tests, competing tasks and their relationship to the broad factors of intelligence. Personulity and Individual D#erences,
9, 25-33.
Stankov, L. & Horn, J. L. (1980). Human abilities revealed through auditory tests. Journal of Educurionul Psychology.
72,
21-44.
Steiger, J. H. (1980). Tests for comparing elements of a correlation matrix. Psychological RuNetin, 87, 245-251. Watson, C. S. (1980). Time course of auditory perceptual learning. Annuls of Otofogy, Rhinology and Luryngology, Supplemenr,
74, 96-102.
Wechsler, D. (1981). A ~Uunuul for Wechsler Adull Intelligence Sale-Revised. New York: Psychological Corporation. White, L. R., Cartwright, W. S., Comoni-Huntley, J. & Brock, D. B. (1986). Geriatric epidemiology. Annuul Reuiew of Gerontology Geriutrics, 6, 215-3 11.