Journal of the International Neuropsychological Society (2014), 20, 333-341. Copyright © INS. Published by Camhridge University Press, 2014. doi;10.I017/S13556177I3001446
Simple and Complex Rule Induction Performance in Young and Older Adults: Contribution of Episodic Memory and Working Memory
Joukje M. Oosterman,' Merle S. Boeschoten,^ Paul A.T. Eling,' Roy P.C. Kessels,''-' AND Joseph H.R. Maes' 'Radhoud University Nijmegen, Donders Institute for Brain, Cognition and Behaviour, Nijmegen, The Netherlands -Trudy Bron Institute, Bilthoven, The Netherlands 'Radhoud University Medical Center, Department of Medical Psychology, Nijmegen, The Netherlands (RECEIVED February 13, 2013; FINAL REVISION December 11, 2013; ACCEPTED December 11, 2013)
Abstract This study tested the hypothesis that part of the age-related decline in perfonnance on executive function tasks is due to a decline in episodic memory. For this, we developed a rule induction task in which we manipulated the involvement of episodic memory and executive control processes; age effects and neuropsychological predictors of task performance were investigated. Twenty-six younger (mean age, 24.0; range, 19-35 years) and 27 community-dwelling older adults (mean age, 67.5; range, 50-91 years) participated. The neuropsychological predictors consisted of the performance on tests of episodic memory, working memory, switching, inhibition and flexibility. Performance of the older adults was worse for the leaming and memorization of simple rules, as well as for the more demanding executive control condition requiring the manipulation of informational content. Episodic memory was the only predictor of performance on the simple leaming and memorization task condition whereas an increase in rule induction complexity additionally engaged working memory processes. Together, these findings indicate that part of the age-related decline on rule induction tests may be the result of a decline in episodic memory. Further studies are needed that examine the role of episodic memory in other executive function tasks in aging. {JINS, 2014, 20, 333-341) Keywords: Executive function. Aging, Flexibility, Prefrontal cortex. Hippocampus, Dementia
in the initial leaming and memorization of rules. Episodic memory has similarly been implicated in a computational model of set switching, and may be crucial for memorizing task codes representing the rule (Altmann & Gray, 2008). A handful of studies have additionally shown that episodic memory predicts performance on executive functioning tasks in older people. For example, a correlational study revealed that episodic memory in older people with vascular risk factors uniquely predicts the performance on the Trail Making Task (Oosterman et al., 2010). Similarly, a strong association between episodic memory impairment and impaired performance on executive function tasks was reported in Alzheimer patients (Baudic et al, 2006). Recent studies also indicate that the medial temporal lobes, brain regions that are crucially involved in episodic memory performance, also are important for executive function performance. Such results have been found in healthy young participants (Takahashi et al., 2007, 2008), and in normal (Oosterman et al., 2008, 2010) and pathological (Bastos-Leite et al., 2007; Nagata et al., 2011; Oosterman, Oosterveld, Olde-Rikkert, Claassen, & Kessels, 2012) aging.
INTRODUCTION Aging is commonly found to be associated with diminished performance on tasks purportedly measuring executive functioning or cognitive control (e.g.. Keys & White, 2000). Little is known, however, about tbe underlying cognitive processes that mediate this diminished perfonnance. Next to executive processes, such as the capacity to display behavioral flexibility and working memory, many executive function tests, such as card-sorting tasks [e.g., the Wisconsin Card Sorting Test (WCST) and its variants] tap episodic tnemory as they also require leaming and memorization of rules. Hence, a deficit in memory may negatively influence the performance on executive tests. In agreement with this suggestion, Giovagnoli (2001) observed a decline in WCST performance in epilepsy patients with left hippocampal sclerosis, and suggested that this may be the result of deficits Correspondence and reprint requests to: Joukje M. Oosterman, Radboud University Nijmegen, Donders Institute for Brain, Cognition and Behaviour, Montessorilaan 3, 6500 HE Nijmegen, The Netherlands. E-maü; j.oosterman® donders.m.nl 333
334
J.M. Oosterman et al.
These studies all suggest that episodic memory is involved in tasks that were specifically designed to tap executive functioning. However, these previous studies are limited in being fully based on correlational results and between-task comparisons. The present study therefore further investigated the hypothesis that age-related decrements in associative leaming and memory predict a substantial pait of the age-related decline in executive function performance. To disentangle the contribution of episodic memoiy and executive control processes to executive function task performance, a rule induction task was developed in which we established different conditions varying in episodic memory and executive function load. It was expected that older adults would perform worse than young adults on this task, not only at a high level of complexity that presumably requires both episodic and executive control processes, but also at an easier level of rule induction petformance, which solely relies on episodic memory. This would then suggest that the age-related decline in rule induction performance is indeed partly due to episodic memory difficulties. In addition, neuropsychological tests of memory and executive function were administered to substantiate the involvement of these functions in the different conditions of the rule induction task. Here, we distinguished between verbal and visual episodic memory, working memory, inhibition, flexibility, and switching, with the latter four functions representing different aspects of executive functioning.
METHODS Participants Twenty-six younger (aged 19-35 years) and 27 comtnunitydwelling older participants (aged 50-91 years) were included in Correct response
Left button
the present study. Participants were recruited via the network of the researcher (i.e., acquaintances, famñy members). Exclusion criteria for all participants were neurodegenerative disorders (e.g., dementia, Parkinson's disease), history of severe psychiatric diseases (e.g., current severe depression, schizophrenia), and neurological disorders (e.g., stroke, multiple sclerosis) (self-report questionnaire). In addition, all older participants completed the Mini Mental State Examination (Folstein, Folstein, & McHugh, 1975) to exclude potential severe cognitive decline (all obtained a score of 26 or higher). The study was conducted in accordance with the Helsinki Declaration. Educational level was measured with a 7-point ordinal rating scale in accordance with the Dutch educational system, ranging from less than primary education (level 1) to university degree (level 7). These ordinal scores equal approximately the following number of years of education as used in the United States (Bouma, Mulder, Lindeboom & Schmand, 2012, p. 19): Level 1, incomplete primary education: 1-5 years; Level 2, primary education: 6 years; Level 3, incomplete lower secondary education: 7-8 years; Level 4, lower general secondary education: 7-9 years; Level 5, vocational education: 7-10 years; Level 6, higher general secondary/higher vocational/pre-university education: 7-16 years; Level 7, academic degree: 17-20 years.
Rule Induction Task In the employed rule induction task, participants had to induce the rules enabling them to respond correctly to different stimuli, based on accuracy feedback. Three conditions were included (see Figure 1). Condition 1 consisted of a simple two-rule leaming task, in which a single feature was indicative of the response. For example, participants had to Right button
Condition 1
Condition 2
Condition 3
Fig. 1. An illustration of tbe simple two-rule condition (first row), tbe four-rule condition in wbicb responses were based on a single feature (e.g., color, second row) and tbe more complex condition requiring bi-conditional learning (third row). Performance is expressed as tbe number of trials needed to complete a condition. Eight consecutive correct answers were needed; if a participant failed to complete a condition, the maximum number of trials was recorded (50 for Condition 1 and 100 for Conditions 2 and 3).
Rule induction in aging leam that a white balloon indicated pressing the left button, whereas a puiple balloon indicated pressing the right button. In this baseline condition, only two exemplars of the "color" feature had to be acquired and memorized and coupled to a specific response. Therefore, the task complexity was low, with petformanee reflecting mainly basic associative learning and (episodic) memory processes and requiring the memorization of only two rules. In Condition 2, the number of rules was increased in that four simple rules had to be acquired and memorized (e.g., a single shape presented in four different colors: respond left to yellow and orange peppers, respond right to red and green peppers). These rules can still be assumed to have a low complexity, since they are based on the coupling of exemplars of a single feature to a response. The number of exemplars that have to be memorized (four) is increased compared to the baseline condition (two), thereby increasing episodic memory load (see also Maes & Eling, 2007). Finally, Condition 3 increased complexity by requiring biconditional learning. A specific combination of exemplars of different features had to be learned to respond cotTectly, rather than learning single stimulus-response associations. Such biconditional tasks are known to engage preñ"ontal cortex processes (Haddon & Killcross, 2006). For example, when a small flower was presented in blue, the left button should be pressed, whereas the right button was correct when a large flower was presented in blue. The opposite keys corresponded to these flowers being presented in pink; this time, the right button was correct when the small flower was presented and the left button should be pressed in response to the large flower. Therefore, neither exemplar of the features "shape" nor "color" in itself is sufficient to indicate the correct response, but it is the unique conjunction of exemplars from the two features that defines the correct response. Compared to Condition 1, episodic memory load was increased in Condition 2, since more exemplars and rules had to be memorized. A disproportionate age-related performance decline was therefore expected in Condition 2 compared to Condition 1. Also, strong correlations between episodic memory performance and performance on Conditions 1 and 2 were expected. The crucial difference between Conditions 2 and 3 is their complexity. In both conditions, the participant had to leam and memorize four stimulus-response associations (either four colors in Condition 2, or four combinations of two colors and two shapes in Condition 3), resulting in four rules in both conditions. The complexity of the rule was, however, increased, since it was based on a unique combination of features. Condition 3 should theoretically recruit both episodic memory and executive function processes, a claim whieh was empirically evaluated by correlational analysis using the performance on standard neuropsychological tests. Because of the proposed additional involvement of executive function in Condition 3, we expected the largest difference between the young and older participants to become evident in this condition. Finally, hierarchical regression analyses (see below) were used to address our primary research question, whether episodic
335 memory performance indeed significantly contributes to the expected age-related difference in Condition 3 performance.
Procedure Participants were instmcted to respond to a stimulus presented on a computer screen by pressing either a left or a right button on the keyboard. They were instmcted that certain rules detennined which key should be pressed in response to each stimulus, but that they had to induce these rules based on feedback they received following each response. Each condition terminated after eight consecutive correct trials or if the maximum number of trials was reached (50 for Condition 1 and 100 for Conditions 2 and 3). The total number of trials needed to complete a condition was used as outcome variable; the maximum number of trials (50 for Condition 1 and 100 for Conditions 2 and 3) was recorded in case a condidon was not successfully completed. Since the first two conditions consisted of a single stimulus feature only (e.g., color) and the third condition of two features (color and shape), two task versions were created. One version used color as the feature in Conditions 1 and 2 and the other version used shape as the feature. In both tasks, the third condition consisted of the same color-shape combinations. Participants were pseudo-randomly assigned to the different task versions, with task version being counterbalanced across the two age groups.
Neuropsychological Tests Next to the rule induction task, participants completed the immediate and delayed recall measures of the Rey Auditory Verbal Learning Test (RAVLT; Van der Eist, Van Boxtel, Van Breukelen, & Jolies 2005) and the Visual Paired Associates (VPA) test of the Wechsler Memory Scale - Revised (WMS-R; Wechsler, 1987) to measure verbal and visual episodic memory respectively. Working memory was assessed with the Letter-Number Sequencing (LNS) task of the Wechsler Adult Intelligence Scale m (WAIS-IH; Wechsler, 2000), switching with the Modified Card Sorting Test (MCST, categories and total etTors: Nelson, 1976), flexibility with the TMT-B (Bowie & Harvey, 2006: using the ratio score TMT-B/ TMT-A), and inhibition using the Stroop Color/Word (C/W) card (Van der Eist, Van Boxtel, Van Breukelen, & Jolies 2006: using the interference score Stroop C/W divided by Stroop Color card). With regard to the TMT-B and Stroop C/W tasks, proportion scores were used since these provide more pure measures of executive function processes (Oosterman et al., 2010; Stuss, Bisschop, et al., 2001; Stuss, Roden, Alexander, Levine, Katz, 2001). All tasks were administered in a fixed order: MMSE, RAVLT-immediate recall, MCST, RAVLTdelayed recall, VPA-immediate recall, TMT, Stroop, LNS, VPA-delayed recall, and the rule induction task.
Statistical Analysis To examine whether an effect of task version was present, we performed a repeated-measures analysis of variance
336 (ANOVA) with the number of trials to complete the task condition as dependent variable. Condition (1-3) as withinsubject factor and Task Version as between-subject factor. Because of distribution differences between the age groups, nonparametric Mann-Whitney V tests were used to test for potential age differences in the three conditions of the mle induction test. To examine whether an increase in memory load (i.e., from Condition 1 to Condition 2) and an increase in executive function load (i.e., from Condition 2 to Condition 3) indeed induced a disproportionate age-related decline in task performance, difference scores (Condition 2 minus 1, Condition 3 minus 2) were calculated and age effects were examined using Mann-Whitney U tests. To examine the relationship between the mle induction task and the neuropsychological tests, the following steps were taken. Eirst, cognitive domain scores were calculated for some neuropsychological tests to reduce the number of outcome variables. To accomplish this, scores were transformed to standardized Z-scores based on the average performance of the younger age group. Next, these standardized scores were unified into domain scores, which consisted of a verbal episodic memory domain (composed of the immediate and delayed recall measures of the RAVLT test), a visual episodic memory domain (composed of the immediate and delayed recall measures of the VPA test), and a switching score (composed of MCST categories and errors). Eor working memory, flexibility, and inhibition, the single test results were used for the analyses. Spearman correlations between the performance on these tasks and the mle inducfion conditions were calculated. Identified significant correlations were subsequently subjected to stepwise regression analysis to determine the unique contribution of the neuropsychological test performances to rule induction performance. Since immediate and delayed memory indices were included in the same domain (e.g., immediate and delayed RAVLT performance for the verbal memory domain), and it is known that immediate and delayed memory constitute partially separable processes dependent on different neural correlates (Neuner et al., 2007), additional analyses were performed in which the mle induction conditions were correlated with immediate memory and delayed memory perfonnance of each task separately. Einally, (hierarchical) regression analyses were used to determine the extent to which the different cognitive functions mediate the age-related decline in mle induction performance. Eirst, the proportion of variance explained in mle induction performance was analyzed with regression analyses in which age group was the only predictor. These analyses were next compared to the outcomes of hierarchical regression analyses in which those neuropsychological scores that significantly predicted perfonnance on the mle induction task were entered before entering age group as a predictor variable. With these analyses, one can directly estimate the proportion of variance accounted for by age that is due to the decline in episodic memory and executive function performance. This was accomplished with the following formula: (/?age-A/?age)//?age, ¡n which RI¡,^ represents the proportion of variance accounted for by age, and
J.M. Oosterman et al. represents the addition of age after controlling for the respective neuropsychological score(s). Eor these analyses, if necessary, data were normalized using the minimum amount of transformation (Osborne, 2002) that provided a good fit of the data [square root, logarithmic, or rank-based inverse normal transformation (Blom transformation)]. Analyses with regard to age effects and neuropsychological predictors were performed one-tailed, alpha was set at 0.05 for all analyses.
RESULTS Significant group differences with regard to the neuropsychological test scores were present for all test variables except the VPA-delayed recall, the MCST-errors and the TMT-B ratio score (see Table 1). Two older participants failed to complete even the most simple two-mle condition of the task; these participants were therefore excluded from the subsequent analyses. Inspection revealed that these participants had a high to very high age (79 and 91, respectively), average levels of education (scores 4 and 5 on a scale of 1-7, reflecting completed general and pre-vocational secondary education, respectively) and normal levels of general cognitive functioning (MMSE of 28 for both participants). One older participant failed to complete Condition 2, and 4 older participants failed to complete Condition 3; in these instances, the maximum number of trials (100) was recorded as score. Characteristics of the participants are presented in Table 1. No significant group differences in education (C/=246.0; z = - 1 . 5 9 ; p = .U) or sex distribution (X"(l) = 0.17;/? = .68) were present. ANOVA did not reveal a main (F(l,49) = 0.21; /? = .65, T^^ = 0.00) or interaction (Greenhouse-Geisser adjustment: F(1.47,72.21)=L98;/) = .16;
Table 1. eharacteristics of the younger and older participants
Age Sex (M/F) Education RAVLT-IR RAVLT-DR VPA-IR VPA-DR LNS MeST-categories MCST-errors TMT-B ratio Stroop interference
Younger group
Older group
24.0 (4.0) 15/11 6 (5-7) 48.9 (7.5) 10.4 (2.5) 15.4 (3.0) 5.8 (0.5) 12.1 (2.3) 5.9 (0.3) 5.5 (3.4) 2.2 (0.6) 1.5 (0.2)
66.1 (11.8) 13/12 5 (3-7) 35.2 (10.3)* 6.5 (2.9)* 12.3 (4.5)* 5.2(1.3) 9.0 (2.3)* 5.2 (1.0)* 9.0 (7.2) 2.1 (0.7) 1.8(0.5)*
Note. *Level of significance was set at p < .01 to correct for multiple tests. Means (standai'd deviations) ai'e reported for age and the neuropsychological test scores of the different groups. Frequencies are reported for sex distribution, education represents median score (range). The /-tests were performed to compare neuropsychological test performance between the different groups. DR = delayed recall; IR = immediate recall; LNS = Letter-Number Sequencing; MCST = Modified Card Sorting Test; RAVLT = Rey Auditory Verbal Learning Test; TMT-B ratio = Trail Making Test B/A; VPA = Visual Paired Associates.
Rule induction in aging
337
70
60
Table 2. Correlations between the rule induction conditions and the neuropsychological scores
^•Younger 6SS3 Older
Condition 1 Condition 2 Condition 3
50
Verbal episodic memory Visual episodic memory Working memory Switching Flexibility Inhibition
r 40 —
30
g 20 10
Condition 1
Condition 2
Condition 3
-0.26* -0.06 -0.17 -0.08 0.12 0.09
-0.14 -0.29* -0.06 -0.14 0.21 0.13
-0.36** -0.18 -0.36** -0.35** 0.04 -0.03
Note. A higher score represents better performance with the exception of the three conditions of the rule induction test, the flexibility and the inhibition score. *p < .05.
**p
Simple and Complex Rule Induction Performance in Young and Older Adults: Contribution of Episodic Memory and Working Memory
Joukje M. Oosterman,' Merle S. Boeschoten,^ Paul A.T. Eling,' Roy P.C. Kessels,''-' AND Joseph H.R. Maes' 'Radhoud University Nijmegen, Donders Institute for Brain, Cognition and Behaviour, Nijmegen, The Netherlands -Trudy Bron Institute, Bilthoven, The Netherlands 'Radhoud University Medical Center, Department of Medical Psychology, Nijmegen, The Netherlands (RECEIVED February 13, 2013; FINAL REVISION December 11, 2013; ACCEPTED December 11, 2013)
Abstract This study tested the hypothesis that part of the age-related decline in perfonnance on executive function tasks is due to a decline in episodic memory. For this, we developed a rule induction task in which we manipulated the involvement of episodic memory and executive control processes; age effects and neuropsychological predictors of task performance were investigated. Twenty-six younger (mean age, 24.0; range, 19-35 years) and 27 community-dwelling older adults (mean age, 67.5; range, 50-91 years) participated. The neuropsychological predictors consisted of the performance on tests of episodic memory, working memory, switching, inhibition and flexibility. Performance of the older adults was worse for the leaming and memorization of simple rules, as well as for the more demanding executive control condition requiring the manipulation of informational content. Episodic memory was the only predictor of performance on the simple leaming and memorization task condition whereas an increase in rule induction complexity additionally engaged working memory processes. Together, these findings indicate that part of the age-related decline on rule induction tests may be the result of a decline in episodic memory. Further studies are needed that examine the role of episodic memory in other executive function tasks in aging. {JINS, 2014, 20, 333-341) Keywords: Executive function. Aging, Flexibility, Prefrontal cortex. Hippocampus, Dementia
in the initial leaming and memorization of rules. Episodic memory has similarly been implicated in a computational model of set switching, and may be crucial for memorizing task codes representing the rule (Altmann & Gray, 2008). A handful of studies have additionally shown that episodic memory predicts performance on executive functioning tasks in older people. For example, a correlational study revealed that episodic memory in older people with vascular risk factors uniquely predicts the performance on the Trail Making Task (Oosterman et al., 2010). Similarly, a strong association between episodic memory impairment and impaired performance on executive function tasks was reported in Alzheimer patients (Baudic et al, 2006). Recent studies also indicate that the medial temporal lobes, brain regions that are crucially involved in episodic memory performance, also are important for executive function performance. Such results have been found in healthy young participants (Takahashi et al., 2007, 2008), and in normal (Oosterman et al., 2008, 2010) and pathological (Bastos-Leite et al., 2007; Nagata et al., 2011; Oosterman, Oosterveld, Olde-Rikkert, Claassen, & Kessels, 2012) aging.
INTRODUCTION Aging is commonly found to be associated with diminished performance on tasks purportedly measuring executive functioning or cognitive control (e.g.. Keys & White, 2000). Little is known, however, about tbe underlying cognitive processes that mediate this diminished perfonnance. Next to executive processes, such as the capacity to display behavioral flexibility and working memory, many executive function tests, such as card-sorting tasks [e.g., the Wisconsin Card Sorting Test (WCST) and its variants] tap episodic tnemory as they also require leaming and memorization of rules. Hence, a deficit in memory may negatively influence the performance on executive tests. In agreement with this suggestion, Giovagnoli (2001) observed a decline in WCST performance in epilepsy patients with left hippocampal sclerosis, and suggested that this may be the result of deficits Correspondence and reprint requests to: Joukje M. Oosterman, Radboud University Nijmegen, Donders Institute for Brain, Cognition and Behaviour, Montessorilaan 3, 6500 HE Nijmegen, The Netherlands. E-maü; j.oosterman® donders.m.nl 333
334
J.M. Oosterman et al.
These studies all suggest that episodic memory is involved in tasks that were specifically designed to tap executive functioning. However, these previous studies are limited in being fully based on correlational results and between-task comparisons. The present study therefore further investigated the hypothesis that age-related decrements in associative leaming and memory predict a substantial pait of the age-related decline in executive function performance. To disentangle the contribution of episodic memoiy and executive control processes to executive function task performance, a rule induction task was developed in which we established different conditions varying in episodic memory and executive function load. It was expected that older adults would perform worse than young adults on this task, not only at a high level of complexity that presumably requires both episodic and executive control processes, but also at an easier level of rule induction petformance, which solely relies on episodic memory. This would then suggest that the age-related decline in rule induction performance is indeed partly due to episodic memory difficulties. In addition, neuropsychological tests of memory and executive function were administered to substantiate the involvement of these functions in the different conditions of the rule induction task. Here, we distinguished between verbal and visual episodic memory, working memory, inhibition, flexibility, and switching, with the latter four functions representing different aspects of executive functioning.
METHODS Participants Twenty-six younger (aged 19-35 years) and 27 comtnunitydwelling older participants (aged 50-91 years) were included in Correct response
Left button
the present study. Participants were recruited via the network of the researcher (i.e., acquaintances, famñy members). Exclusion criteria for all participants were neurodegenerative disorders (e.g., dementia, Parkinson's disease), history of severe psychiatric diseases (e.g., current severe depression, schizophrenia), and neurological disorders (e.g., stroke, multiple sclerosis) (self-report questionnaire). In addition, all older participants completed the Mini Mental State Examination (Folstein, Folstein, & McHugh, 1975) to exclude potential severe cognitive decline (all obtained a score of 26 or higher). The study was conducted in accordance with the Helsinki Declaration. Educational level was measured with a 7-point ordinal rating scale in accordance with the Dutch educational system, ranging from less than primary education (level 1) to university degree (level 7). These ordinal scores equal approximately the following number of years of education as used in the United States (Bouma, Mulder, Lindeboom & Schmand, 2012, p. 19): Level 1, incomplete primary education: 1-5 years; Level 2, primary education: 6 years; Level 3, incomplete lower secondary education: 7-8 years; Level 4, lower general secondary education: 7-9 years; Level 5, vocational education: 7-10 years; Level 6, higher general secondary/higher vocational/pre-university education: 7-16 years; Level 7, academic degree: 17-20 years.
Rule Induction Task In the employed rule induction task, participants had to induce the rules enabling them to respond correctly to different stimuli, based on accuracy feedback. Three conditions were included (see Figure 1). Condition 1 consisted of a simple two-rule leaming task, in which a single feature was indicative of the response. For example, participants had to Right button
Condition 1
Condition 2
Condition 3
Fig. 1. An illustration of tbe simple two-rule condition (first row), tbe four-rule condition in wbicb responses were based on a single feature (e.g., color, second row) and tbe more complex condition requiring bi-conditional learning (third row). Performance is expressed as tbe number of trials needed to complete a condition. Eight consecutive correct answers were needed; if a participant failed to complete a condition, the maximum number of trials was recorded (50 for Condition 1 and 100 for Conditions 2 and 3).
Rule induction in aging leam that a white balloon indicated pressing the left button, whereas a puiple balloon indicated pressing the right button. In this baseline condition, only two exemplars of the "color" feature had to be acquired and memorized and coupled to a specific response. Therefore, the task complexity was low, with petformanee reflecting mainly basic associative learning and (episodic) memory processes and requiring the memorization of only two rules. In Condition 2, the number of rules was increased in that four simple rules had to be acquired and memorized (e.g., a single shape presented in four different colors: respond left to yellow and orange peppers, respond right to red and green peppers). These rules can still be assumed to have a low complexity, since they are based on the coupling of exemplars of a single feature to a response. The number of exemplars that have to be memorized (four) is increased compared to the baseline condition (two), thereby increasing episodic memory load (see also Maes & Eling, 2007). Finally, Condition 3 increased complexity by requiring biconditional learning. A specific combination of exemplars of different features had to be learned to respond cotTectly, rather than learning single stimulus-response associations. Such biconditional tasks are known to engage preñ"ontal cortex processes (Haddon & Killcross, 2006). For example, when a small flower was presented in blue, the left button should be pressed, whereas the right button was correct when a large flower was presented in blue. The opposite keys corresponded to these flowers being presented in pink; this time, the right button was correct when the small flower was presented and the left button should be pressed in response to the large flower. Therefore, neither exemplar of the features "shape" nor "color" in itself is sufficient to indicate the correct response, but it is the unique conjunction of exemplars from the two features that defines the correct response. Compared to Condition 1, episodic memory load was increased in Condition 2, since more exemplars and rules had to be memorized. A disproportionate age-related performance decline was therefore expected in Condition 2 compared to Condition 1. Also, strong correlations between episodic memory performance and performance on Conditions 1 and 2 were expected. The crucial difference between Conditions 2 and 3 is their complexity. In both conditions, the participant had to leam and memorize four stimulus-response associations (either four colors in Condition 2, or four combinations of two colors and two shapes in Condition 3), resulting in four rules in both conditions. The complexity of the rule was, however, increased, since it was based on a unique combination of features. Condition 3 should theoretically recruit both episodic memory and executive function processes, a claim whieh was empirically evaluated by correlational analysis using the performance on standard neuropsychological tests. Because of the proposed additional involvement of executive function in Condition 3, we expected the largest difference between the young and older participants to become evident in this condition. Finally, hierarchical regression analyses (see below) were used to address our primary research question, whether episodic
335 memory performance indeed significantly contributes to the expected age-related difference in Condition 3 performance.
Procedure Participants were instmcted to respond to a stimulus presented on a computer screen by pressing either a left or a right button on the keyboard. They were instmcted that certain rules detennined which key should be pressed in response to each stimulus, but that they had to induce these rules based on feedback they received following each response. Each condition terminated after eight consecutive correct trials or if the maximum number of trials was reached (50 for Condition 1 and 100 for Conditions 2 and 3). The total number of trials needed to complete a condition was used as outcome variable; the maximum number of trials (50 for Condition 1 and 100 for Conditions 2 and 3) was recorded in case a condidon was not successfully completed. Since the first two conditions consisted of a single stimulus feature only (e.g., color) and the third condition of two features (color and shape), two task versions were created. One version used color as the feature in Conditions 1 and 2 and the other version used shape as the feature. In both tasks, the third condition consisted of the same color-shape combinations. Participants were pseudo-randomly assigned to the different task versions, with task version being counterbalanced across the two age groups.
Neuropsychological Tests Next to the rule induction task, participants completed the immediate and delayed recall measures of the Rey Auditory Verbal Learning Test (RAVLT; Van der Eist, Van Boxtel, Van Breukelen, & Jolies 2005) and the Visual Paired Associates (VPA) test of the Wechsler Memory Scale - Revised (WMS-R; Wechsler, 1987) to measure verbal and visual episodic memory respectively. Working memory was assessed with the Letter-Number Sequencing (LNS) task of the Wechsler Adult Intelligence Scale m (WAIS-IH; Wechsler, 2000), switching with the Modified Card Sorting Test (MCST, categories and total etTors: Nelson, 1976), flexibility with the TMT-B (Bowie & Harvey, 2006: using the ratio score TMT-B/ TMT-A), and inhibition using the Stroop Color/Word (C/W) card (Van der Eist, Van Boxtel, Van Breukelen, & Jolies 2006: using the interference score Stroop C/W divided by Stroop Color card). With regard to the TMT-B and Stroop C/W tasks, proportion scores were used since these provide more pure measures of executive function processes (Oosterman et al., 2010; Stuss, Bisschop, et al., 2001; Stuss, Roden, Alexander, Levine, Katz, 2001). All tasks were administered in a fixed order: MMSE, RAVLT-immediate recall, MCST, RAVLTdelayed recall, VPA-immediate recall, TMT, Stroop, LNS, VPA-delayed recall, and the rule induction task.
Statistical Analysis To examine whether an effect of task version was present, we performed a repeated-measures analysis of variance
336 (ANOVA) with the number of trials to complete the task condition as dependent variable. Condition (1-3) as withinsubject factor and Task Version as between-subject factor. Because of distribution differences between the age groups, nonparametric Mann-Whitney V tests were used to test for potential age differences in the three conditions of the mle induction test. To examine whether an increase in memory load (i.e., from Condition 1 to Condition 2) and an increase in executive function load (i.e., from Condition 2 to Condition 3) indeed induced a disproportionate age-related decline in task performance, difference scores (Condition 2 minus 1, Condition 3 minus 2) were calculated and age effects were examined using Mann-Whitney U tests. To examine the relationship between the mle induction task and the neuropsychological tests, the following steps were taken. Eirst, cognitive domain scores were calculated for some neuropsychological tests to reduce the number of outcome variables. To accomplish this, scores were transformed to standardized Z-scores based on the average performance of the younger age group. Next, these standardized scores were unified into domain scores, which consisted of a verbal episodic memory domain (composed of the immediate and delayed recall measures of the RAVLT test), a visual episodic memory domain (composed of the immediate and delayed recall measures of the VPA test), and a switching score (composed of MCST categories and errors). Eor working memory, flexibility, and inhibition, the single test results were used for the analyses. Spearman correlations between the performance on these tasks and the mle inducfion conditions were calculated. Identified significant correlations were subsequently subjected to stepwise regression analysis to determine the unique contribution of the neuropsychological test performances to rule induction performance. Since immediate and delayed memory indices were included in the same domain (e.g., immediate and delayed RAVLT performance for the verbal memory domain), and it is known that immediate and delayed memory constitute partially separable processes dependent on different neural correlates (Neuner et al., 2007), additional analyses were performed in which the mle induction conditions were correlated with immediate memory and delayed memory perfonnance of each task separately. Einally, (hierarchical) regression analyses were used to determine the extent to which the different cognitive functions mediate the age-related decline in mle induction performance. Eirst, the proportion of variance explained in mle induction performance was analyzed with regression analyses in which age group was the only predictor. These analyses were next compared to the outcomes of hierarchical regression analyses in which those neuropsychological scores that significantly predicted perfonnance on the mle induction task were entered before entering age group as a predictor variable. With these analyses, one can directly estimate the proportion of variance accounted for by age that is due to the decline in episodic memory and executive function performance. This was accomplished with the following formula: (/?age-A/?age)//?age, ¡n which RI¡,^ represents the proportion of variance accounted for by age, and
J.M. Oosterman et al. represents the addition of age after controlling for the respective neuropsychological score(s). Eor these analyses, if necessary, data were normalized using the minimum amount of transformation (Osborne, 2002) that provided a good fit of the data [square root, logarithmic, or rank-based inverse normal transformation (Blom transformation)]. Analyses with regard to age effects and neuropsychological predictors were performed one-tailed, alpha was set at 0.05 for all analyses.
RESULTS Significant group differences with regard to the neuropsychological test scores were present for all test variables except the VPA-delayed recall, the MCST-errors and the TMT-B ratio score (see Table 1). Two older participants failed to complete even the most simple two-mle condition of the task; these participants were therefore excluded from the subsequent analyses. Inspection revealed that these participants had a high to very high age (79 and 91, respectively), average levels of education (scores 4 and 5 on a scale of 1-7, reflecting completed general and pre-vocational secondary education, respectively) and normal levels of general cognitive functioning (MMSE of 28 for both participants). One older participant failed to complete Condition 2, and 4 older participants failed to complete Condition 3; in these instances, the maximum number of trials (100) was recorded as score. Characteristics of the participants are presented in Table 1. No significant group differences in education (C/=246.0; z = - 1 . 5 9 ; p = .U) or sex distribution (X"(l) = 0.17;/? = .68) were present. ANOVA did not reveal a main (F(l,49) = 0.21; /? = .65, T^^ = 0.00) or interaction (Greenhouse-Geisser adjustment: F(1.47,72.21)=L98;/) = .16;
Table 1. eharacteristics of the younger and older participants
Age Sex (M/F) Education RAVLT-IR RAVLT-DR VPA-IR VPA-DR LNS MeST-categories MCST-errors TMT-B ratio Stroop interference
Younger group
Older group
24.0 (4.0) 15/11 6 (5-7) 48.9 (7.5) 10.4 (2.5) 15.4 (3.0) 5.8 (0.5) 12.1 (2.3) 5.9 (0.3) 5.5 (3.4) 2.2 (0.6) 1.5 (0.2)
66.1 (11.8) 13/12 5 (3-7) 35.2 (10.3)* 6.5 (2.9)* 12.3 (4.5)* 5.2(1.3) 9.0 (2.3)* 5.2 (1.0)* 9.0 (7.2) 2.1 (0.7) 1.8(0.5)*
Note. *Level of significance was set at p < .01 to correct for multiple tests. Means (standai'd deviations) ai'e reported for age and the neuropsychological test scores of the different groups. Frequencies are reported for sex distribution, education represents median score (range). The /-tests were performed to compare neuropsychological test performance between the different groups. DR = delayed recall; IR = immediate recall; LNS = Letter-Number Sequencing; MCST = Modified Card Sorting Test; RAVLT = Rey Auditory Verbal Learning Test; TMT-B ratio = Trail Making Test B/A; VPA = Visual Paired Associates.
Rule induction in aging
337
70
60
Table 2. Correlations between the rule induction conditions and the neuropsychological scores
^•Younger 6SS3 Older
Condition 1 Condition 2 Condition 3
50
Verbal episodic memory Visual episodic memory Working memory Switching Flexibility Inhibition
r 40 —
30
g 20 10
Condition 1
Condition 2
Condition 3
-0.26* -0.06 -0.17 -0.08 0.12 0.09
-0.14 -0.29* -0.06 -0.14 0.21 0.13
-0.36** -0.18 -0.36** -0.35** 0.04 -0.03
Note. A higher score represents better performance with the exception of the three conditions of the rule induction test, the flexibility and the inhibition score. *p < .05.
**p
