Journal of Experimental : Copyright 1997 by the American Psychological Association, Inc. , , and Cognition 0278-7393/97/$3.00 1997, Vol. 23, No. 1,239-245 The Relationships Between Psychometric and Learning in an Explicit and an Implicit Task

Peter McGeorge, J. R. Crawford, and S. W. Kelly University of Aberdeen

An experiment is reported examining the relation of implicit grammar learning and series completion tasks to a standard measure of psychometric intelligence, the Wechsler Adult Intelligence Scale---Revised (WAIS-R; D. Wechsler, 1981). The results replicate and extend an earlier study by A. S. Reber, F. F. Walkenfeld, and R. Hernstadt (1991) and provide the following support for the differences between explicit and implicit tasks: (a) The implicit task was less strongly related to Full Scale IQ, and (b) the implicit task appeared to be independent of age. The implicit and explicit tasks exhibited a quite different pattern of relations to the factors known to underlie WAIS-R performance. Although both tasks showed significant links with a Perceptual Organization factor, only the series completion task showed a significant link with the Attention factor.

The distinction between explicit and implicit processes or principles derived from evolutionary biology, Reber argued systems represents one of the most active areas of research that these earlier emerging functions will be the foundations in (for reviews, see Reber, 1989, 1993; on which later developing conscious functions will have Roediger & McDermott, 1993). One aspect of this research been built. In addition, on the basis of these operating is the distinction, first articulated by Reber (1967), between principles, Reber derived five important predictions concern- explicit and implicit learning. Implicit learning has been ing the nature of implicit processes; these are as follows: characterized as the process through which Implicit processes should be more robust in the face of concerning complex rule-governed environments is ac- psychiatric and neurological insults; implicit processes should quired independent of of both the process and the display tighter distributions in the population with fewer products of acquisition (Reber, 1989). Implicit learning has individual differences; implicit functions should show little been reported in a wide range of experimental settings, for relationship to performance on standard measures of cogni- example, the control of dynamic systems (e.g., Berry & tive capability, such as psychometric tests of intelligence; Broadbent, 1984; McGeorge & Burton, 1989; Stanley, implicit processes should show a commonality across spe- Mathews, Buss, & Kotler-Cope, 1989), the acquisition of cies; and implicit processes should be little affected by age covariances between personality and facial characteristics and developmental level. (e.g., Lewicki, 1986), the acquisition of complex procedural To test the assumptions that implicit learning is indepen- information (e.g., Lewicki, Czyzewska, & Hoffman, 1987), dent of IQ and has a lower variability in the population, the learning of invariances (McGeorge & Burton, 1990), and Reber, Walkenfeld, and Hernstadt (1991) examined the the learning of artificial grammars (e.g., Mathews et al., relationships between performance on a standard intelli- 1989; Reber, 1967, 1969). Explicit learning is seen as the gence test (a four-subtest short form of the Wechsler Adult converse of this with awareness of both the processes and Intelligence Scale--Revised, WAIS-R; Wechsler, 1981), the products being its central aspects. performance on a task thought to favor an explicit mode of Recently, Reber (1993) argued that the study of implicit learning, and performance on a task thought to favor an processes should be seen from an evolutionary perspective. implicit mode of learning. The explicit learning task was In evolutionary terms, is a relatively recent based on series solution problems. In this task, participants phenomenon and one predated by many sophisticated percep- are presented with ordered sequences of letters. Each tual and cognitive functions. Invoking so-called operating sequence represents a pattern, and the task of the participant is to discover what the pattern is in order to predict the subsequent letter or letters of the sequence, For example, Peter McGeorge, J. R. Crawford, and S. W. Kelly, Department of participants might be given the sequence ~a3ACAa~A and Psychology, University of Aberdeen, Old Aberdeen, Scotland. asked to predict the next letter of the sequence (in this case S. W. Kelly is now at the Department of Psychology, University of E). In Reber et al.'s (1991) study, the selection of the next Glasgow, Glasgow, Scotland. letter in the sequence was based on a two-alternative forced Preparation of this article was supported by Grant R000234247 from the Economic and Social Research Council. We thank Pat choice procedure. The implicit learning task was based on a Chalmers for her help in collecting the data. finite state grammar. Permissible (grammatical) combina- Correspondence concerning this article should be addressed to tions of letters were specified by pathways within a state Peter McGeorge, Department of Psychology, University of Aber- transition diagram. Combinations of letters that violated deen, Old Aberdeen, AB9 2UB Scotland. Electronic mall may be these pathways were referred to as nongrammatical. Follow-" sent via Internet to psy 144 @abdn.uk.ac. ing an incidental learning phase, often disguised as a simple

239 240 OBSERVATIONS memory task, participants were tested on their ability to Concentration factor (e.g., Crawford, Allan, Stephen, Parker, distinguish between well-formed or grammatical strings and & Besson, 1989; Kaufman, 1990; Le,ckliter, Matarazzo, & nongrammatical strings (well-formedness test). Silverstein, 1986). The Attention--Concentration factor is of The results of Reber et al.'s (1991) study indicated that particular interest as low scores on the subtests that serve as although IQ was significantly correlated with the series indicators of this factor (Digit Span and Arithmetic) are solution task, accounting for nearly 50% of the variance in often associated with individuals from a range of special performance, the correlation between IQ and performance populations with problems in attention, including individu- on the well-formedness test was not significant and ac- als with reading problems and learning disabilities, alcohol- counted only for just over 6% of the variance in perfor- ism, brain injuries, and Huntington's disease (Kaufman, mance. In addition and as predicted, the distribution of 1990). Reber (1993) has argued that these populations show scores from the explicit task showed much higher variance levels of implicit learning equivalent to the normal popula- than the scores obtained for the implicit task. Together these tion. This suggests that performance on the implicit task findings provide support for the contentions, derived from shows little relationship to the Attention-Concentration the phyletic primacy of implicit processes, that traditional factor. psychometric tests do not measure the same cognitive So, although the correlation between an implicit function components as are required by implicit tests (such as the and IQ may always be significantly lower than the correla- well-formedness test) and that implicit tasks show fewer tion between tQ and the corresponding explicit function-- individual differences than equivalent explicit tasks. and indeed it may not show a significant correlation with As Reber (1993) pointed out, Rozin (1976) has argued Full Scale IQ--an examination of the relationships between that the development of intelligence should also be seen these learning functions and the various factors thought to from an evolutionary perspective. Rozin argued that evolu- underlie the WAIS-R may provide significant insights into tion will give rise to a series of adaptive specializations. the cognitive functions important in the two learning tasks. These are highly modular processes, which are initially Reber et al. (1991) used a short form of the WAIS-R, inaccessible to other processes and are often hardwired. comprising 4 of the 11 subtests, and so were unable to Together this collection of hardwired, limited-access machin- explore the relationships between the learning tasks and the ery forms the cognitive unconscious (Rozin, 1976). Part of factors underlying the WAIS-R. Using a full-length WAIS-R the evolution of more intelligent organisms would be to establish Full Scale IQ would allow these relationships to increased communication between these unconscious modu- be tested. lar processes. This would result in the development of the Finally, there are a number of methodological issues conscious executive controls that mark intelligent behavior. raised by Reber et al.'s (1991) study that complicate the Within this framework, intelligence can be conceived as a interpretation of their results. The foremost of these issues hierarchy in which these encapsulated modular processes relates to the size and nature of the sample on which the become available to more and more systems (Rozin, 1976). analysis was based. Reber et al. used a relatively small Under such hierarchical conditions, traditional psychometric sample of participants (N = 20), all of whom were students. tests such as the WAIS-R will, in measuring this conscious The mean IQ for the group was 110.0 with a standard executive, be measuring the involvement of the more deviation of 21.2. Reber et al. (1991) pointed out that this primitive modules from which it is derived. Because, almost mean is in line with what would be expected in a student by definition, it is the primitive encapsulated modules that sample. However, given that the WAIS-R is standardized to are responsible for implicit functions, psychometric tests have a standard deviation of 15 (with a mean of 100), the will measure both the implicit and the explicit functions. large standard deviation and the large range (73-150) Cattell (1963) has argued that intelligence, as measured suggest that the sample may have been unusual, particu- by standard psychometric tests such as the WAIS-R, can be larly as a student sample would be expected to have a divided into two types of abilities: crystallized and fluid. restricted range of ability. Crystallized intelligence refers to those abilities that are We report a subset of data from a study concerned with the culturally determined and verbally encoded. Fluid intelli- relationships between explicit tasks, implicit tasks, and gence refers to basic cognitive capacities that are mainly intelligence. The data reported here address the issues biologically determined (Valsiner & Leung, 1994). Fluid discussed above concerning the relationships between learn- intelligence is often considered an indicator of reasoning ing an artificial grammar, solving series completion prob- ability, the ability to educe correlatesMa basic relation- lems, and the underlying factor structure of the WAIS-R. perceiving ability (Kline, 1991), whereas crystallized intelli- gence is considered an indicator of what has been learned Method (Cattell, 1971). On the basis of this, it might be predicted that both explicit and implicit tests should show a relation- Series Completion Task ship to measures of fluid intelligence. Further, factor analytic Series completion problems were constructed with the rules studies based on the full 11 subtests of the WAIS-R have described by Brown and Ferrara (1985). In each sequence, letters consistently indicated that it can be characterized by a could be repeated (identity or I), followed by the next letter in the Global factor, a Verbal Ability-Comprehension factor (linked alphabet (next or N), or followed by the preceding letter of the to crystallized intelligence), a Perceptual Organization fac- alphabet (backward or B). The number of letters between each tor (linked to fluid intelligence), and an Attention- regularity, referred to as the period, could also be varied. For OBSERVATIONS 241 example, in the sequence ~d3^CAD , each regularity is For the series completion task, participants were presented with separated by one other letter (period = 2) and beginning with the a sheet on which were printed 28 series completion problems. They first letter, the regularities are identity and next. The correct were given 20 rain in which to complete as many of the problems as completion is ~. We constructed 28 sequences by using they could. They were encouraged not to spend too long on any one different combinations of identity, next, and backward regularities problem and to attempt all 28 problems. Each problem required and by varying the period. These 28 sequences were of varying four letters for completion. Problems were scored as correct if three degrees of complexity. of the four letters were correctly reported. Following a further unrelated learning task, participants were given a questionnaire to assess their knowledge and strategy for each of the learning tasks. Artificial Grammar The second session consisted of the WAIS-R. Participants were administered the 11 subtests of the WAIS-R, following the The materials for the grammar were taken from the set reported procedures described in the test manual and measures of Verbal, by Dulaney, Carlson, and Dewey (1984). Performance, and Full Scale IQ derived from the scaled scores.

Measures of Psychometric Intelligence Results The intelligence measure consisted of a full-length WAIS-R. The mean Full Scale IQ was 107.0 (SD = 13.2) with a range of 78.0 to 136.0. Mean Verbal and Performance IQs Participants were 107.0 (SD = 13.7) and 105.4 (SD = 12.6), respec- tively. The mean proportion correct for the well-formedness There were 123 participants who took part in the study (60 male and the series completion tests was .60 (SD = .06) and .29 and 63 female). These were recruited from a wide range of sources (SD = .23), respectively. Performance on the well-formed- and were free from neurological, psychiatric, or sensory disorders. ness test was significantly above chance, t(122) = 18.4,p < All participants received a small honorarium for their participation. .01. Measures of variability in task performance (coefficient Mean age was 45.1 years (SD = 17.1) with a range of 18 to 77. of variability; Howell, 1992) were calculated for the well- Goodness-of-fit chi-square tests revealed that there were no formedness and series completion tasks. The values of these significant differences between the distributions in the sample and the distributions in the United Kingdom census data with respect to were 0.11 and 0.80, respectively. socialclass, X2 (4,N = 123) = 0.31,p > .1;age, x2 (3,N = 123) = Table 1 shows the mean proportion of correct items for 0.75,p > .1; and sex, X2 (1,N = 123) = 0.04,p > .1. each of the four age groups for each of the two learning tasks. For the well-formedness test, an analysis of variance Procedure (ANOVA) indicated no significant difference between age Testing took place over two sessions, each of which lasted bands, F(3, 119) = 1.71, p > .1. For the series completion approximately 1.5 hr. Each participant was tested individually. The tests, an ANOVA indicated a significant difference, first session consisted of the learning tasks. The grammar and series F(3, 119) = 5.72, p < .01. An analysis of linear trend completion tasks were administered as part of a more general (coefficients = -3, -1, 1, 3) indicated a significant effect, battery of learning tasks. The grammar task was always adminis- F(1,119) = 15.1,p < .001. tered first, and following two fitrther tasks, participants were There was a significant correlation between performance presented with the series completion problems. Previously, Reber on the series completion task and Full Scale IQ (r = .58, et al. (1991) reported no order effects for the well-formedness and p < .05). The correlations between the well-formedness test series completion tests, and so the order of testing was held and Full Scale IQ, and between the two learning tasks, were constant for all participants. The grammar task comprised two phases. In the first phase, the not significant, (r = .04 and r = .12, respectively). With study phase, participants were seated in front of a personal Williams's test (Williams, 1959; see Steiger, 1980), there computer. They were informed that they would be shown a series of was a highly significant difference between the correlations letter strings and should try to remember each of these as they of the two learning tasks and Full Scale IQ, t(120) = 5.23, would be tested on them later. Following these instructions, the p < .0001. experimenter initiated the presentation of 20 grammatical strings. Using IQ summary statistics in samples where partici- Each string appeared individually and remained on the screen for 6 pants may vary widely in age can result in a considerable s before being replaced by the next string. These 20 strings were loss of information. It could be argued that in examining the presented three limes in random orders, Following this learning phase, the participants were then presented with an unexpected test phase in which they were informed that the strings that they had viewed previously all conformed to a complex rule structure. They Table 1 were informed that they would be shown another series of letter Mean Proportion of Correct Responses for the strings, some of which conformed to the rule structure and some of Well-Formedness and Series Completion Testfor which did not. Each letter string was presented on a card, and for Each Age Group each string they were asked to indicate whether they thought it did or did not conform to the rule structure (well-formedness test). Age group Twenty new grammatical strings and 20 nongrammatical strings Measure 16--29 30 A.A 45-59 60+ were presented to the participant in a random order. These 40 test items were presented twice, and the participant's response, gram- Series completion .43 .28 .25 .20 Well-formedness .60 .62 .60 .58 matical or nongrammatical, was recorded. 242 OBSERVATIONS relationship between the learning measures and the WAIS-R WAIS-R factors and the learning tasks was the same, a Walt abilities, either all the measures should be uncorrected for test (Bentler, 1989) was conducted for each model. The age or all the measures should be corrected for age. The Wald test is used to examine whether any of the parameters learning measures are uncorrected for age. Measures of of the model that currently have nonzero values could be set WAIS-R performance uncorrected for age were derived by to zero without any statistically significant loss in the summing the appropriate sealed subtest scores. For the sums model's overall fit to the data. For the model incorporating of the scaled scores of the WAIS-R subtests, all the the series completion task, the results indicated that the path correlations with the series completion test were significant. between series completion and the Verbal Ability factor For the well-formedness test, only the correlation with the could be dropped from the model. For the model incorporat- sum of the scaled scores of the performance subtests was ing the well-formedness task, the results indicated that the significant (see Table 2). paths between well-formedness and the Verbal Ability and Confirmatory factor analyses were conducted on the Attention-Concentration factors could be dropped from the learning task and scaled WAIS-R subtest scores with the model. EQS program (Bentler, 1989). Factor analytic studies of the Bentler (r989) has suggested that the fit indices of a 11 subtests of the WAIS-R have consistently indicated that it well-fitting model should be greater than 0.9. Neither of the can be characterized by a global factor, a Verbal Ability models evaluated in the present study achieved values in this factor, a Perceptual Organization factor, and an Attention- range. To test which parameters would need to be altered to Concentration factor (Kaufman, 1990). With these factors, significantly improve the model's fit a Lagrange multiplier models were constructed for each learning task and for the (LM) test was also conducted (Bentler, 1989). The LM test is WAIS-R subtests. These models can be seen as representing used to examine whether there are parameters in the model the null hypothesis that both learning tasks show the same currently fixed with zero values that if allowed to be relations to the factors underlying WAIS-R performance. nonzero, would lead to a significant improvement in the All the performance subtests were assumed to have nonzero model's fit. The results of the LM test indicated that there loadings on the Perceptual Organization factor and zero were a number of paths associated with the WAIS-R loadings on the other factors. All verbal subtests except Digit subtests that if allowed nonzero parameters, would lead to an Span and Arithmetic were assumed to have nonzero loadings improvement in the overall fit of the model. Improvements on the Verbal Ability factor and zero loadings on the other in the fit between the model and the data could be achieved factors. Digit Span and Arithmetic were assumed to have with nonzero path coefficients for the Similarities subtest on nonzero loadings on the Attention-Concentration factor and the Perceptual Organization factor, the Picture Completion zero loading on all other factors. The learning measures subtest on the Verbal Ability factor, and the Object Assembly were assumed to have a nonzero loading on all three factors. test on the Attention--Concentration factor. In addition, all factors were assumed to be correlated and so to implicitly contain a second-order factor, General Intel- ligence. The models, with the standardized parameter Discussion estimates, are represented in Figure 1. For the model In line with previous studies of artificial grammar learn- incorporating the series completion task, the chi-square ing, participants were significantly better than chance at goodness-of-fit value was ×2(49, N = 123) = 130.3, p < discriminating between grammatical and nongranunatical .001, and the Comparative Fit index was .888. There were strings, and the overall level of performance was of a similar significant loadings on both the Perceptual Organization and magnitude to that reported previously (e.g., Reber, 1989). Attention-Concentration factors (.49 and .52, respectively). The results of the correlations replicate the findings For the model incorporating the well-formedness task, the reported by Reber et al. (1991). Performance on the explicit chi-square goodness-of-fit value was ×2(49, N = 123) = series completion test shows a strong positive relationship 134.8, p < .001, and the Comparative Fit index was .866. with IQ measures, whereas performance on the well- There was a significant positive loading on the Perceptual formedness test and performance on the IQ measures are not Organization factor (.32), and the loading on the Attention- correlated. Indeed, the correlation, which is based on a larger Concentration factor was negative but nonsignificant (- .29). sample, is much smaller than that found in Reber et al.'s To test whether the pattern of relations between the study. However, the factor analyses reported here indicated that the IQ-test correlations represent a simplification of the relationships between performance on the psychometric test Table 2 and performance on the well-formedness and series comple- Correlations Between the Sum of the Scaled Subtest Scores tion tests. As predicted, the results indicate that both tests for Full, Verbal, and Performance IQ Tests and load on the Perceptual Organization factor, the factor most the Learning Tasks closely linked to fluid intelligence. However, only the series Test completion task shows a significant loading on the Attention- Sum of scaled Concentration factor. The implication is that only series subtests Well-formedness Series completion completion is dependent on the deployment of conscious All subtests .12 .67* attention. Verbal subtests .04 .53" Several studies have examined the role of attention in Performance subtests .19" .65* implicit learning, particularly in relation to sequence learn- *p < .05. ing. In the standard sequence learning task, a visual stimulus OBSERVATIONS 243

A appears in one of four different spatial locations. The task of the participant is to press the corresponding response key as E1 0.4g --* r~ v..~,,. quickly as possible. Typically, participants exposed to a 0.87" repeating (hence predictable) sequence become much faster E2 -- 0.42 ~J'~0.91"~ than participants exposed to arandom sequence of locations (e.g., Nissen & BuUemer, 1987). This speedup occurs even F.3- 0.61 [~] 0.63 though the participants are not told of the repeating sequence prior to the test and profess no knowledge of it after testing. Nissen and Bullemer (1987) found that the addition of a . \\ secondary task interfered with participants' ability to learn 0.,45" \ :o.14 the repeating sequence, suggesting that attention was impor- tant if learning was to occur. However, both Cohen, Ivry, and Keele (1990) and Frensch, Buchner, and Lin (1994) demon- strated that learning of a sequence could occur even with the ._ o. addition of a secondary task and that the effect of the 0.Sg-E12 secondary task depended on the nature of the sequence. Stadler (1995) has argued that the interference effect ob- 7- served for the secondary task in these °'/I// studies arises not because it requires participants to divert attentional resources from the sequence learning task to the I secondary task but more as a side effect of the secondary task interfering with the production of consistent sets of memory traces. The role of attention has also been examined in learning an artificial grammar. Dienes, Broadbent, and Berry (1991) reported that the addition of a secondary task, .o_ o.. random number generation, had a detrimental effect on the acquisition of information necessary to make correct well- Ell-- 0.61 --~ I'~ formedness judgements. These studies have in general addressed the question of whether attention is necessary for implicit learning to occur, but they have not directly addressed the issue of how demanding of attention the application of implicit knowl- B edge is. Curran and Keele (1993) examined the effect of the addition of a secondary task to sequence learning following st - 0.,~ -~[';'1-~ I,.ml ~ a period of learning with the sequence task alone. Their 0.87* results indicated that although, overall, response times were E2 -- 0.43 --~] x-~. 0.91.~-- much longer under secondary task conditions, participants ~VA~ still responded faster to the repeating sequence than to a E3 0.61 -~. r~ x---- o.8o*~ random sequence. The results from the present study could o.83" I\ \ be interpreted as suggesting that the addition of a secondary task during the test phase of the well-formedness test would ,, \\ have tittle detrimental effect. The addition of the secondary 0.45- \ 0.. task to the series completion test would be expected to have a highly detrimental effect. Whatever process underlies the application of the information acquired during artificial grammar learning, it is far less demanding of conscious - 0- attention than the application of the knowledge necessary to 0.96-E12 complete the series completion problems. The series completion task used in this study differed in " - 7- • o.~/ / F.8 0.61 --~ ['~ ~"/'0.56" / "0.29 Figure 1. Confirmatory factor analytic models of Wechsler Adult 0.55" Intelligence Scale--Revised subtests and learning tasks. A: three- factor model incorporating series completion. B: three-factor ., / model incorporating well-formedness test. El-E12 = error terms associated with each of the paths; I = Information; V = Vocabu- lary; C = Comprehension; S = Similarities; PC = Picture Completion; PA = Picture Arrangement; B = Blocks Design; ElO-- 0.74 --~ ~ <--0.67"~ OA = Object Assembly; DS = Digit Symbol; DSP = Digit Span; A = Arithmetic; VA = Verbal Ability; PO = Perceptual Organiza- 0.80" Ell-- 0.59 ---~J & J tion; A-C = Attention-Concentration; SC = Series Completion; i--i GR = Grammar Well-Formedness. 244 OBSERVATIONS both the nature of the rules used to construct the materials Cohen, A., Ivry, R. I., & Keele, S. W. (1990). Attention and and the nature of the response. Reber et al. (1991) used a structure in sequence learning. Journal of Experimental Psychol- two-alternative forced-choice procedure in which partici- ogy: Learning, Memory, and Cognition, 19, 17-30. pants selected which of two letters would complete the Crawford, J. R., Allan, K. M., Stephen, D. W., Parker, D. M., & sequence. Although this format differs from the response Besson, J. A. O. (1989). The Wechsler Adult Intelligence Scale--Revised (WAIS-R): Factor structure in a UK sample. format in this study where the participants were required to Personality and lndividual Differences, 10, 1209-1212. generate their own completions, it is unlikely that this would Curran, T., & Keele, S. W. (1993). Attentional and nonattentional have had a major impact on the attentional requirements of forms of sequence learning. Journal of Experimental Psychol- the tasks. Both tasks require participants to parse the test ogy: Learning, Memory, and Cognition, 19, 189-202. strings to identify the rule structure and then generate Dienes, Z., Broadbent, D., & Berry, D. (1991). Implicit and explicit appropriate responses. knowledge bases in artificial grammar learning. Journal of The results also provide some support for two other of the : Learning, Memory, and Cognition, hypothesized differences between implicit and explicit pro- 17, 875-887. cesses. First, the coefficients of variation (Howell, 1992) Dulaney, D. E., Carlson, R. A., & Dewey, G. I. (1984). A case of support the contention that individual variability in task syntactical learning and judgment: How conscious and how performance is greater for explicit processes than it is for abstract? Journal of Experimental Psychology: General, 113, 541-555. implicit processes. Second, the analysis of the four arbitrary Frensch, P. A., Buchner, A., & Lin, J. (1994). Implicit learning of age groups provides some support for the independence of unique and ambiguous serial transitions in the presence and implicit processes, but not explicit processes, from age and absence of a distractor task. Journal of Experimental Psychol- developmental level. Performance on the well-formedness ogy: Learning, Memory, and Cognition, 20, 567-584. test remains relatively stable with age, whereas performance Howell, D. C. (1992). Statistical methods for psychology (3rd ed.). on the series completion test shows a significant negative Belmont, CA: Duxbury Press. effect of age. The relative independence of performance on Kaufman, A. D. (1990). Assessing adolescent and adult intelli- the well-formedness test and age parallels the results re- gence. London: Allyn & Bacon. ported for (e.g., Light, Singh, & Capps, Kline, P. (1991). Intelligence: The psychometric view. London: 1986). However, as with all cross-sectional designs that Routledge. examine changes in performance with age, the decline in Leckliter, I. N., Matarazzo, J. D., & Silverstein, A. B. (1986). A literature review of factor analytic studies of the WAIS-R. performance observed in the series completion test may Journal of , 42, 332-342. represent nothing more than a cohort effect. Lewicki, P. (1986). Processing information about covariations that In summary, the results lend support to the hypotheses cannot be articulated. Journal of Experimental Psychology: derived from the phylogenetic primacy of implicit processes Learning, Memory, and Cognition, 12, 135-146. (Reber, 1993). The weU-formedness test results in smaller Lewicki, P., Czyzewska, M., & Hoffman, H. (1987). Unconscious individual differences in performance and is unaffected by acquisition of complex procedural knowledge. Journal of Experi- age. Performance on the series completion task results in mental Psychology: Learning, Memory, and Cognition, 13, greater individual differences and shows a negative effect of 355-362. increasing age. Only the series completion task shows a Light, L. L., Singh, A., & Capps, J. L. (1986). Dissociation of significant correlation with IQ as measured by the WAIS-R. memory and awareness in young and older adults. Journal of Clinical and Experimental , 8, 62-74. The correlation between the well-formedness test and IQ is Mathews, R. C., Buss, R. R., Stanley, W. B., Blanchard-Fields, E, negligible. The results of the confirmatory factor analyses Cho, J.-R., & Druhan, B. (1989). The role of implicit and explicit indicate that the principal difference between the two processes in learning from examples: A synergistic effect. learning tasks is that only performance on the series Journal of Experimental Psychology: Learning, Memory, and completion test is dependent on conscious attention. Whether Cognition, 15, 1083-1100. this is specific to these tasks or generalizes to other perhaps McGeorge, E, & Burton, A. M. (1989). The effects of concurrent more closely matched tasks remains to be determined. verbalisation on performance in a dynamic systems task. British Journal of Psychology, 80, 455--465. References McGeorge, P., & Burton, A. M. (1990). Semantic processing in an incidental learning task. Quarterly Journal of Experimental Bentler, P. (1989). EQS structural equations program manual. Los Psychology: Human Experimental Psychology, 42(A), 597-610. Angeles: BMDP Statistical Software. Nissen, M. J., & Bullemer, P. (1987). Attentional requirements of Berry, D. C., & Broadbent, D. E. (1984). On the relationship learning: Evidence from performance measures. Cognitive Psy- between task performance and associated verbalizable knowl- chology, 19, 1-32. edge. Quarterly Journal of Experimental Psychology: Human Reber, A. S. (1967). Implicit learning in artificial grammars. Experimental Psychology, 36(A), 209-231. Journal of Verbal Learning and Verbal Behavior, 6, 855-866. Brown, A. L., & Ferrara, R. A. (1985). Diagnosing zones of Reber, A. S. (1969). Transfer of syntactic structure in synthetic proximal development. In J. V. Wertsch (Ed.), Culture, communi- languages. Journal of Experimental Psychology, 81, 115-119. cation, and cognition: Vygotskian perspectives (pp. 273-305). Reber, A. S. (1989). Implicit learning and tacit knowledge. Journal Cambridge, England: Cambridge University Press. of Experimental Psychology: General, 118, 219-235. Cattell, R. B. (1963). Theory of fluid and crystallized intelligence: Reber, A. S. (1993). Implicit learning and tacit knowledge: An A critical experiment. Journal of , 54, essay on the cognitive unconscious. New York: Oxford Univer- 1-22. sity Press. Cattell, R. B. (1971). Abilities: Their structure growth and action. Reber, A. S., Walkenfeld, F. F., & Hemstadt, R. (1991). Implicit and New York: Houghton Mifflin. explicit learning: Individual differences and IQ. Journal of OBSERVATIONS 245

Experimental Psychology: Learning, Memory, and Cognition, Steiger, J. H. (1980). Tests for comparing elements of a correlation 17, 888-896. matrix. Psychological Bulletin, 87, 245-251. Roediger, H. L., HI, & McDermott, K. B. (1993). Implicit memory Valsiner, J., & Leung, M. (1994). From intelligence to knowledge in normal subjects. In E Boiler & J. Cn'afman (Eds.), Handbook construction: A sociogenetic approach. In R. J. Steinberg & R. K. ofneuropsychology (Vol. 8, pp. 63-131). Wagner (Eds.), Mind in comext: Interactionist perspectives on Rozin, P. (1976). The evolution of intelligence and access to the human intelligence (pp. 202-217) Cambridge, England: Cam- cognitive unconscious. Progress in Psychobiology and Physi- bridge University Press. ological Psychology, 6, 245-280. Wechsler, D. (1981). Manual for the Wecheler Adult Intelligence Stadler, M. A. (1995). The role of attention in implicit learning. Scale--Revised. New York: Psychological Corporation. Journal of Experimental Psychology: Learning, Memory, and Williams, E. J. (1959). The comparison of regression variables. Cognition, 21, 674--685. Journal of the Royal Statistical Society, 21(B), 396-399. Stanley, W. B., Mathews, R. C., Buss, R. R., & Kotler-Cope, S. (1989). Insight without awareness: On the interaction of verbal- ization, instruction and practice in a simulated process control Received October 10, 1995 task. Quarterly Journal of Experimental Psychology: Human Revision received April 22, 1996 Experimental Psychology, 41(A ), 553-578. Accepted May 1, 1996 •

AMERICAN PSYCHOLOGICAL ASSOCIATION

SUBSCRIFHON CLAIMS INFORMATION Today's l)ate:

We provide this form to assist members, institutions, and nonmember individuals with any subscziption problems. With the appropriate information we can begin a resolution. If you use the servioes of an ageat, please do NOT duplicate claims through them and directly to us. PLEASE PRINT CLEARLY AND IN INK IF POSSIBLF~

PRINT FULL NAMB Og KEY NAME OF ~¢ ~I~I~UTION M]~d]~RR OR CUSTOI,4E~ 0~AY BE FOUND ON A.NYPAST L~J'JB LAB~)

ADDRESS DAIE YOUR ORD]~. WAS ~n' ~ (OR PHONIC)

PPd~AID CHECK CHAX~ CHBCK/C..ARD CLEARID DA'I~ crrY STAS~/COUN'I'~Y 23P (If pmsibl~ send a copy, from s/~d back, d yore csnc~l ~ Io I~p m. in our rm~..h ot your dslm.) YOUR NAME AND PHONE HUMB]~ ~: MlSS~O ~ DAMAG6D

TITLE VOLUME OR YEAR NUMBI~ OR MONTH

Thank you. On~ a daim is recd~d and r*#ol~#~ delivery of replac~nmt b~ rom~e~ ta~ 4-6 ~d~.

(TO BE FILLED OUT BY APA STAFF) , DATE RECEIVED: DATE OF ACTION: ACTION TAKEN: INV. NO. & DATE: STAFF NAME: LABEL NO. & DATE:

Send this form to APA SubKripl/on Cl*Im%750 First Street, K W,d, ln~ton, IE 200~..4242

PLEASE DO NOT REMOVE. A PHOTOCOPY MAY BE USED.