How Useful is Executive Control Training PDF

Developmental Science 12:6 (2009), pp 978–990 DOI: 10.1111/j.1467-7687.2009.00846.x PAPER How useful is executive control training? Age differences in near and far transfer of task-switching training Julia Karbach and Jutta Kray Department of Ps...

Developmental Science 12:6 (2009), pp 978–990 DOI: 10.1111/j.1467-7687.2009.00846.x PAPER How useful is executive control training? Age differences in near and far transfer of task-switching training Julia Karbach and Jutta Kray Department of Psychology, Saarland University, Germany Abstract Although executive functions can be improved by training, little is known about the extent to which these training-related benefits can be transferred to other tasks, or whether this transfer can be modulated by the type of training. This study investigated lifespan changes in near transfer of task-switching training to structurally similar tasks and its modulation by verbal self- instructions and variable training, as well as far transfer to structurally dissimilar ‘executive’ tasks and fluid intelligence. Three age groups (8–10; 18–26; 62–76 years of age) were examined in a pretest-training-posttest design. We found near transfer of task-switching training in all age groups, especially in children and older adults. Near transfer was enhanced in adults and impaired in children when training tasks were variable. We also found substantial far transfer to other executive tasks and fluid intelligence in all age groups, pointing to the transfer of relatively general executive control abilities after training. Introduction show smaller mixing costs than children (e.g. Crone, Ridderinkhof, Worm, Somsen & Van der Molen, 2004; Recently, much research has focused on executive Kray et al., 2008; Kray et al., 2004) and older adults (e.g. control, that is, on the ability to plan, guide, and Mayr, 2001; Meiran, Gotler & Perlman, 2001), while age monitor complex goal-directed actions, considered to be differences on the level of switching costs seem to be less a fundamental ability of human intelligent behavior. pronounced (e.g. Kray & Lindenberger, 2000; Mayr, Today, it is widely accepted that executive control 2001; Verhaeghen & Cerella, 2002). Consistently, consists of separate control components, such as lifespan studies showed U-shaped developmental switching, updating, and inhibition (e.g. Fisk & Sharp, functions for mixing costs, but not for switching costs 2004; Miyake, Friedman, Emerson, Witzki & Howerter, (cf. Kray et al., 2008; Kray et al., 2004; Reimers & 2000). For some of these components, substantial age- Maylor, 2005). related changes have been observed across the lifespan Based on these findings, developmental researchers have (e.g. Bedard, Nichols, Schachar, Schachar, Logan & investigated the potential range of cognitive plasticity in Tannock, 2002; Cepeda, Kramer & Gonzales De Sather, task-switching abilities. So far, a number of previous 2001; Kray, Eber & Karbach, 2008; Kray, Eber & studies have indicated that training can reduce age-related Lindenberger, 2004; Williams, Ponesse, Schachar, Logan differences in both types of switching costs (e.g. Cepeda & Tannock, 1999). Evidence for these differential age- et al., 2001; Kray et al., 2008; Kray & Lindenberger, 2000). related changes in executive control comes from a variety Cepeda and colleagues (2001), for instance, found that of experimental paradigms, among them the task- mixing costs were reduced after two sessions of task- switching paradigm. In task-switching studies, switching training, especially in children (10–12 years of participants are instructed to perform two simple tasks age) and in older adults. Similarly, Kray and Lindenberger A and B, either in single-task blocks (only A or B) or in (2000) showed a reduction of mixing as well as switching mixed-task blocks (switching between both tasks). This costs after six sessions of training in young and older design allows calculating two types of task-switching adults. The primary aim of our study was to examine costs: Mixing costs, defined as the difference in mean whether these training-related improvements in task- performance between mixed-task and single-task blocks, switching abilities can be transferred to new switching refer to the ability to maintain and select two tasks. tasks, and whether training strategies and the type of task- Switching costs, defined as the difference in mean switching training can modulate this transfer in different performance between switch and nonswitch trials age groups. within mixed-task blocks, measure the ability to Evidence for the transfer of executive control training flexibly switch between tasks. Young adults usually is rather scarce and comes from a wide variety of Address for correspondence: Julia Karbach, Department of Psychology, Saarland University, Campus A 1.3, D-66123 Saarbrcken, Germany; e-mail: [email protected]  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.  Transfer of task-switching training 979 experimental paradigms and training tasks. Regarding two aspects of executive control, namely task-set childhood, a number of developmental studies showed maintenance and selection as well as task-set switching. that different types of executive control training could be Similar to Minear et al. (2002), we compared transfer transferred to structurally similar (near transfer) and after task-switching training to transfer after training on dissimilar (far transfer) tasks after training. Kloo and the same two single tasks performed separately. To Perner (2003), for instance, trained 3- to 4-year-olds by examine whether specific training strategies or the type means of the Dimensional Change Card Sort (DCCS) of training can modulate transfer, this study included two and the false-belief task. They found that DCCS training additional training conditions. In one of these improved performance on the false-belief task and vice conditions, we investigated whether verbal self- versa (for transfer in autistic children, see Fisher & instruction strategies can be transferred to new, Happ, 2005). Also, Rueda, Rothbart, McCandliss, untrained tasks. Prior research indicated that verbal Saccomanno and Posner (2005) showed that training processes could support the retrieval of the phonological including a battery of executive control tasks generalized representation of currently relevant task goals (e.g. to similar new tasks as well as to aspects of intelligence Baddeley, Chinchotta & Adlam, 2001). This effect is quite remote from the training tasks (cf. Dowsett & particularly pronounced when external task cues are Livesey, 2000; Klingberg, Fernell, Olesen, Johnson, missing and the need for endogenous control is increased Gustafsson, Dahlstrçm, Gillberg, Forssberg & (e.g. Emerson & Miyake, 2003). Consistently, a recent Westerberg, 2005). Evidence for the near transfer of study showed that verbal self-instructions (i.e. naming executive control training in older age mostly comes from the next task goal during task preparation) facilitate the dual-task studies (cf. Kramer & Kray, 2006): Recently, maintenance and selection of task sets, especially in Bherer and colleagues (Bherer, Kramer, Peterson, childhood and older age. That is, they serve as effective Colcombe, Erickson & Becic, 2005) showed that dual- means to reduce age-related differences in task-switching task training benefits in young and older adults abilities (Kray et al., 2008). To examine whether these generalized to new tasks and stimuli (cf. Kramer, verbal self-instruction benefits can be transferred to a Larish, Weber & Bardell, 1999; Kramer, Larish & new task, we trained one group of participants not only Strayer, 1995). Similarly, Minear, Shah and Park (2002) in task switching, but also in the use of verbal self- found that task-switching training transferred to similar instructions. tasks after training in young and older adults. However, In the other condition, we also tested the influence of this near transfer was found only for mixing costs, not the type of task-switching training on the amount of for switching costs. Thus, existing evidence for the transfer in different age groups. Previous studies have transferability of task-switching training seems to be provided considerable evidence indicating that restricted to near transfer in younger and older adults. conditions facilitating performance during training are In sum, there are a number of studies indicating that not always most effective in supporting the acquisition of children, young adults, and older adults can transfer a generalizable skill. In contrast, manipulations executive control training to untrained tasks. However, decreasing the speed of skill acquisition during these effects are highly variable and based on rather training, such as variable training tasks, can support its different types of training, such as working memory transfer to a new, untrained task (for reviews, see training (e.g. Klingberg et al., 2005), dual-task training Rosenbaum, Carlson & Gilmore, 2001; Schmidt & (e.g. Bherer et al., 2005; Kramer et al., 1999; Kramer Bjork, 1992). For instance, Sanders, Gonzalez, Murphy, et al., 1995), task-switching training (Minear et al., 2002) Pesta and Bucur (2002) showed that high variability or even a battery of several executive control tasks (e.g. training in mental calculation supported transfer to non- Rueda et al., 2005). Also, the range of transfer distance trained tasks in young adults (for similar results after (i.e. different types of near and far transfer) as well as the dual-task training, see Kramer et al., 1999). Thus, for age range of the participants was very diverse in these another group in the present study, training was variable, studies. Hence, the comparability of previous studies meaning that the stimuli and the type of tasks in each focusing on the transfer of training seems to be very task-switching training session were different. limited. Although most of these findings suggest that at In sum, the first goal of this study was to examine age least near transfer is possible in different age groups, differences in the transfer of task-switching training to a conditions supporting the occurrence of far transfer, similar switching task (near transfer) and its modulation differences between diverse types of training, and the by a training strategy and by the type of training. Since lifespan development of these effects are still not clear. prior evidence indicated that verbal self-instructions Therefore, the aim of the present study was to could support task switching (Kray et al., 2008), we systematically investigate age-related changes in the investigated whether these verbal self-instructions near and far transfer of training and the influence of performed during training influence the amount of different types of training within one study and across a transfer. Also, because variable training tasks can foster wide range of ages. transfer in adults (cf. Kramer et al., 1999; Sanders et al., In order to investigate transfer of cognitive training, 2002), we expected more near transfer after variable we applied the task-switching paradigm, tapping at least training, at least in adults.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 980 Julia Karbach and Jutta Kray Table 1 Outline of the training and transfer procedure Pretest Training Posttest Sessions 1 + 2 Sessions 3–6 Sessions 7 + 8 All groups: Group 1: Single-task training All groups: Single tasks (tasks C and D) Single tasks (tasks A and B) Group 2: Task-switching training (tasks A and B) Task switching (tasks C and D) Task switching (tasks A and B) Group 3: Task-switching (tasks C and D) (tasks A and B) Cognitive battery: + verbal self-instruction training Cognitive battery: Stroop task Group 4: Task-switching + verbal Stroop task Verbal working memory self-instruction training + training Verbal working memory Spatial working memory variability Spatial working memory Fluid intelligence (tasks C/D, E/F, G/H, I/J) Fluid intelligence Note: Subjects within each age group were matched to one of the four training groups based on their pretest performance in task switching (mixing costs), single-task reaction time, and Raven score to prevent differences in baseline performance between the training groups. Pretest 1/Posttest 1 included the measurement of verbal and spatial working memory as well as fluid intelligence abilities, and Pretest 2/Posttest 2 single tasks, task-switching, and the Stroop task. The second goal was to investigate the range of Transfer of training was assessed by means of a transfer. Therefore, we examined age-related changes in pretest-training-posttest design (see Table 1) and was the far transfer of task-switching training to other defined as performance improvement at posttest relative ‘executive control tasks’, that is, the Stroop test and to the baseline performance at pretest. The two pretest working memory tasks. Given that these tasks require sessions included baseline measurements of task executive control abilities that are also needed for task switching and single-task performance as well as a switching, such as the online maintenance of relevant battery of cognitive tasks. They were followed by four task goals and the inhibition of currently irrelevant training sessions. The two posttest sessions were information, far transfer to these tasks may be expected. identical to pretest sessions, each of them taking Finally, we also included measures of fluid intelligence to 60–70 minutes. For each participant, testing took investigate far transfer to another task domain. While we 6–8 weeks, that is, they performed approximately one expected near transfer of task-switching training in all session per week. age groups (cf. Bherer et al., 2005; Minear et al., 2002; Kramer et al., 1999; Kramer et al., 1995; Rueda et al., Pretest and posttest assessment 2005), age differences in the amount of far transfer and its modulation by the type of training were an open Task switching question. Since there is usually less transfer when the We used a modified version of the task-switching training and transfer tasks are less similar (for a review, paradigm, including performance in single-task (task A see Klauer, 2001), we expected more transfer of training or B only) and mixed-task blocks (switching between to structurally similar tasks than to structurally both tasks). In mixed-task blocks, subjects were dissimilar tasks. instructed to switch tasks on every second trial. Task A required participants to decide whether a picture showed a fruit or a vegetable (‘food’ task), and task B whether a Method picture was small or large (‘size’ task). The same two response keys were used for both task sets. Stimuli Participants consisted of 16 fruit and 16 vegetable pictures, each one presented in a large and a small version. Mixing and Fifty-six children (mean age = 9.2, SD = 0.6, switching costs were defined as two orthogonal contrasts range = 8.1–10.1 years, 43% female), 56 young adults for the factor trial type (single, nonswitch, switch trials): (mean age = 22.4, SD = 2.2, range = 18.0–26.3 years, Mixing costs were measured as the difference in mean 51% female), and 56 older adults (mean age = 68.7, performance between single-task and mixed-task blocks SD = 3.0, range = 62.3–76.8 years, 59% female) par- (contrast: -2 1 1), and switching costs as the difference ticipated in this study. They were recruited from the between nonswitch and switch trials within mixed-task subject pool at Saarland University, tested individually blocks (contrast: 0 1 -1). Participants performed two by one of the eight experimenters and were paid 60 Euros single-task practice blocks (17 trials) followed by 20 (95 USD) for participating in the eight sessions of the experimental blocks1 (eight single and 12 mixed blocks; study. 17 trials). Trials started with a fixation-cross (1400 ms), followed by the target until the subject responded. After Materials and procedure We used IBM-compatible computers for data collection. 1 Block sequence: 2 single – 2 mixed – 2 single – 2 mixed – Stimuli were presented on a 17-inch CRT color monitor single – 2 mixed – single – 2 mixed – single – 2 mixed – single – and an external keypad registered manual responses. 2 mixed.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.  Transfer of task-switching training 981 25 ms, the next fixation-cross appeared. Subjects were Set sizes for all WM tasks ranged from two to five instructed to respond as fast and as accurately as items, with a total of eight sets (i.e. two items per set possible, and they were offered a short break after half size). The test score refers to the number of sets of the blocks had been completed. correctly recalled. 4. Fluid intelligence: Figural reasoning/letter series (cf. Lindenberger, Mayr & Kliegl, 1993) and Raven’s Cognitive test battery Standard Progressive Matrices (Raven, 1988). In the To examine whether task-switching training also figural reasoning task, items followed the format, transfers to structurally dissimilar ‘executive’ tasks and ‘A is to B as C is to ?’ In the letter series task, fluid intelligence, the cognitive battery included tests for subjects saw items consisting of five letters followed four constructs (each measured with two or three by a questions mark (e.g. a c e g i ?), and named indicators to increase the reliability of the measurement): the letter that would logically fill the position of the question mark. In both tasks, five response 1. Inhibitory control: Color-Stroop/Number-Stroop (cf. alternatives were presented along with the items. Salthouse & Meinz, 1995). In the Color-Stroop task, The experimenter terminated the task when subjects subjects saw words (e.g. ‘red’, ‘tree’) presented in red, committed three consecutive errors or after they blue, green, or yellow letters. Participants indicated answered all 16 items (for details, see Lindenberger the letter color as quickly as possible by pressing one et al., 1993). In the Raven’s task, subjects of four response buttons. In the Number-Stroop task, completed 30 trials in which they selected one of participants saw characters (e.g. 2, HHH) presented eight figures that best completed a pattern (for one-, two-, three-, or fourfold and decided how many details, see Raven, 1988). Test scores refer to the stimuli were presented. Stroop interference was number of correctly solved items. defined as the difference in performance between ‘neutral’ (e.g. ‘tree’ in red ink, ‘HH’) and incongruent (e.g. ‘blue’ in red ink, ‘44’) trials. Participants Training sessions performed two practice blocks (12 trials) and four For the four training sessions, participants within each experimental blocks (24 trials) for each of the tasks. age group were assigned to one of the following four Stimuli were presented for 2000 ms or until the training groups (see Table 1): During single-task subject responded, followed by a response–stimulus training (group 1), subjects practiced only the two interval of 700 ms. single tasks, so that executive control demands during 2. Verbal WM: Reading span/counting span (see Kane, training were relatively low (control condition). During Hambrick, Tuholski, Wilhelm, Payne & Engle, 2004). task-switching training they practiced only mixed-task In the reading span task, participants recalled letters blocks, so that executive control demands during against a background reading task; in the counting training were high (group 2). In the task-switching + span task, they recalled digits against a background verbal self-instruction training group, participants counting task (for details, see Kane et al., 2004). also trained on mixed-task blocks. In addition, they 3. Spatial WM: Symmetry span/navigation span verbalized the upcoming task goal (e.g. ‘transportation’ (adapted from Kane et al., 2004). In the symmetry or ‘number’, see below) to the onset of the span task, subjects recalled sequences of locations in a fixation-cross in each trial (group 3). Finally, the task 4 · 4 matrix against a background symmetry- switching + verbal self-instruction + training variability judgment task. In the navigation span task, they group received the same training as the third group, but recalled the paths of moving balls across the screen the tasks and stimuli were different in each training against a background rotation task (for details, see session (group 4). Kane et al., 2004).2 In the single-task training group, participants performed alternating blocks including tasks A or B. 2 The adaptation of the tasks from Kane et al. (2004) included The task-switching procedure during training was the following details: In the original version of the symmetry structurally similar to the one applied at pretest and span task, the symmetry judgment required participants to decide whether two complex geometric matrices were posttest except that subjects performed different tasks. symmetrical along a vertical axis. Since pilot testing indicated In task C (‘transportation’ task), subjects had to decide that this task was too difficult for children, subjects were shown whether the pictures showed planes or cars, and in task two letters instead of the complex matrices and they were D (‘number’ task) whether one or two planes/cars were instructed to decide whether these letters were symmetrical presented. The design of the additional tasks applied to along a vertical axis. Similarly, the navigation span task from group 4 (tasks E–J) was similar to tasks C and D, but Kane et al. (2004) included a distraction task requiring included different stimuli and response categories. That participants to count the corners of bold uppercase letters from a certain starting point in a designated direction. Given is, while participants in this fourth group also that children in particular had problems performing this task, performed tasks C and D in the first training session, we substituted the letters with polygons (including the same they were instructed to classifying pictures according to number of to-be-counted corners). task E (‘hobby’ task: Sport [e.g. a football] or music  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 982 Julia Karbach and Jutta Kray [e.g. a piano]?) and F (‘stoplight’ task: Red or green?) in for instance, indicates that the mean difference between the second training session, task G (‘animal’ task: Fish pretest and posttest corresponds to one standard or bird?) and H (‘direction’ task: Normal or rotated?) in deviation. the third training session, and task I (‘plant’ task: Tree or flower?) and J (‘color’ task: Black-and-white or colored?) in the fourth training session. Training Results sessions for all groups took about 30–40 minutes. Training data They started with two practice blocks followed by 24 experimental blocks (17 trials), so that all groups To investigate training-related benefits (i.e. a reduction of performed 1768 training trials. switching costs from the beginning to the end of training) Subjects were matched to these training groups in the three task-switching training groups, data4 were based on their pretest performance in task switching subjected to a four-way ANOVA with the between- (RT mixing costs), single-task RT, and Raven score to subjects factors Age (children/young adults/older adults) prevent baseline differences between the training and Training (group 2/3/4), and the within-subjects groups. In order to test whether this matching factors Session (training 1/training 4) and Trial Type procedure was successful, pretest data for the three (nonswitch/switch). matching criteria were subjected to a two-way analysis We found a quadratic age effect, indicating that young of variance (ANOVA) with the between-subjects adults responded faster than children and older adults, factors Age (children/young adults/older adults) and age2: F(1, 116) = 144.33, p <.0001, g2 =.46, and a Training (group 1/2/3/4). Neither the main effect for main effect for session, showing a speeding of RT from training nor its interaction with age reached the first to the last training session, F(1, 116) = 205.01, significance for any of the matching criteria (all p <.0001, g2 =.63, that was more pronounced for ps >.31), indicating that there were no baseline children than for adults and also larger for the groups differences between the training groups (see Table A1 performing verbal self-instructions (groups 3 and 4) than in Appendix). for group 2 (both ps =.01). We found significant switching costs, that were larger for children and older adults than for young adults, F(1, 116) = 311.63, Data analysis p <.0001, g2 =.67, and age2: F(1, 116) = 22.80, Analyses for task switching and the Stroop task were p <.0001, g2 =.05. Switching costs were smaller in the restricted to mean RT for correct responses.3 Practice groups performing verbal self-instructions (groups 3 and blocks and the first trial in each block were not analyzed. 4) than in the group without verbalizations (group 2), To control for age differences in baseline performance, we F(1, 116) = 7.48, p <.01, g2 =.01. Switching costs were ran ANOVAs based on log-transformed RT (cf. Kray & reduced from the first to the last training session, Lindenberger, 2000). Unless reported otherwise, these F(1, 116) = 113.48, p <.0001, g2 =.46, but this results were consistent with those based on mean RT. reduction was less pronounced in the variability group We also analyzed error rates, but there were no significant (group 4) than in the remaining groups, interactions with the factor Training on the level of F(1, 116) = 11.33, p =.001, g2 =.04. This interaction accuracy; therefore, the presentation of results focuses on was not modulated by age (p =.78). RT. Data were corrected for multiple comparisons using a Bonferroni correction at p <.05. For the remaining tasks Near transfer, verbal processes and training variability (WM, fluid intelligence), the analyses were based on accuracy (% correct) relative to baseline performance at Next, we examined near transfer of task-switching pretest. training to a structurally similar switching task (i.e. a To examine the range of transfer effects across training reduction of mixing and switching costs from pretest to conditions of near and far transfer tasks, we also posttest) and its modulation by verbal processes and calculated Cohen’s (1977) d, or the standardized mean training variability. Data were subjected to a four-way difference in performance between pretest and posttest ANOVA with the between-subjects factors Age (children/ (cf. Verhaeghen, Marcoen & Goossens, 1992). That is, young adults/older adults) and Training (group 1/2/3/4), the pretest–posttest difference (for each training and age and the within-subjects factors Session (pretest/posttest) group) was divided by the pooled standard deviation for and Trial Type (single/nonswitch/switch). Young adults both test occasions. We then corrected all d-values for responded faster than children and older adults, age2: small sample bias using the Hedges and Olkin (1985) F(1, 156) = 195.39, p <.0001, g2 =.53, and a main correction factor (d¢). A pretest-posttest effect size d¢ = 1, effect for session pointed to faster RTs at posttest, F(1, 156) = 363.40, p <.0001, g2 =.67. There were 3 For task switching, latencies > 4000 ms were excluded from 4 the analyses (Training: children: 1.43%; young adults: 0.01%; Data for one child in the variability group were lost, so the older adults: 0.15%. Pretest and posttest: children: 2.35%; analysis of training data was restricted to 125 instead of 126 young adults: 0.09%; older adults: 0.81%). subjects.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.  Transfer of task-switching training 983 reliable mixing and switching costs, F(1, 156) = 666.65, Far transfer to other executive tasks and other task p <.0001, g2 =.78, and F(1, 156) = 658.66, p <.0001, domains g2 =.80. Mixing costs were generally larger for children Second, we investigated far transfer to a structurally and older adults than for young adults, age2: dissimilar ‘executive’ task, namely the Stroop task.7 Since F(1, 156) = 22.52, p <.0001, g2 =.03, but there were the modulation of far transfer by verbal self-instructions no age differences for switching costs5 (p =.19). Both and training variability was an open question, we first types of costs were reduced from pretest to posttest (both examined whether the task-switching training groups ps <.0001). However, the outcome of greatest interest in (2–4) showed different amounts of transfer (see this study was whether training modulated these Table A2). We found no interactions of training with interactions. Indeed, we found interactions between age, session, or trial type (all ps >.10), so we collapsed session and training, F(3, 156) = 6.11, p <.001, data across groups 2–4 to increase the statistical power. g2 =.03, session, trial type, and training, F(6, Data were then subjected to a four-way ANOVA with the 312) = 10.61, p <.0001, g2 =.08, as well as between factors Age (children/young adults/older adults), session, trial type, training, and age, F(12, 312) = 3.07, Training (single-task/task-switching), Session (pretest/ p <.001, g2 =.04. To disentangle these interactions, we posttest), and Trial Type (neutral/incongruent). We specified three contrasts for the factor Training: found a main effect for session, indicating that Comparing groups 1 and 2 showed that the reduction participants responded faster at posttest than at of mixing and switching costs from pretest to posttest pretest, F(1, 162) = 65.15, p <.0001, g2 =.25, as well was larger after task-switching training (group 2) than as a reliable interference effect,8 F(1, 162) = 254.20, after single-task training (group 1), F(1, 156) = 32.05, p <.0001, g2 =.60, showing longer latencies on p <.0001, g2 =.07, and F(1, 156) = 14.10, p <.001, incongruent than on neutral trials, but the main effect g2 =.04, respectively. For mixing costs, this transfer of training was not significant (p =.35). Again, of effect was more pronounced in children and older adults greatest interest in this study were interactions with the than in young adults, age2: F(1, 39) = 4.57, p <.05, factors Session and Training. Indeed, there was a session g2 =.03 (see Figure 1). A comparison of groups 2 and 3 · training · trial type interaction, F(1, 162) = 9.25, indicated that near transfer was not modulated by verbal p <.01, g2 =.05, indicating that interference was self-instructions performed during task-switching reduced from pretest to posttest after task-switching training (p =.64). Finally, a comparison between training, F(1, 123) = 13.11, p <.001, g2 =.09 (but not groups 2/36 and 4 revealed that transfer on the level of after single-task training, p =.14), pointing to far mixing costs was reduced in children and increased transfer of task-switching training to interference in adults when training tasks were variable, control in the Stroop task (see Table 2). However, this F(1, 52) = 11.89, p =.001, g2 =.10, and F(1, 104) = far transfer was not modulated by age (p =.18). 10.95, p =.001, g2 =.07, but there was no such effect for Although the reduction of interference effects from switching costs (p =.90). pretest to posttest after task-switching training was These results were supported by the pretest-posttest consistent with the initial expectations, there also was effect sizes (ES). For both types of costs, ES were larger an unexpected deterioration of performance (i.e. after task-switching (d¢ =.88–2.12) than after single- increased interference effects) in adults after the single- task training (d¢ =.11–.60), particularly for children task training (see Table 2). Control analyses for younger (see Figure 2). ES for mixing costs in adults increased and older adults indicted that the increased again when the switching training was combined with interference effects were due to a larger pretest–posttest verbalizations (d¢ = 1.44–1.46) and variability (d¢ = improvement in the baseline condition (i.e. neutral 1.28–1.66), while we found the reverse effect for trials), F(1, 26) = 16.01, p <.001, g2 =.59, and not children: The verbalizations (d¢ = 1.55) performed to impairments in high-interference conditions (i.e. during training, and even more the variable training incongruent trials), F(1, 26) = 7.52, p <.05, g2 =.29, (d¢ =.65) resulted in substantially smaller ES (see resulting in larger interference effects for the single-task Figure 2). Results for switching costs were similar, training group at posttest. with maximized ES in young adults in the variability Finally, we investigated two additional executive group (d¢ = 1.59) and decreased ES in children domains (verbal and visuospatial WM) and another (d¢ =.68). 5 7 Based on mean RT, switching costs were larger for children Correlations between task versions were high (neutral trials: and older adults than for young adults, age2: F(1, r =.82*** incongruent trials: r =.76***) and the pattern of 156) = 25.67, p <.0001, g2 =.04. results was similar, so the data were collapsed across the color 6 and the number version. Since we found no difference between groups 2 and 3, data 8 were collapsed across both groups to increase statistical power. Analyses based on mean RT showed larger interference in We found the same pattern when we compared groups 2 and 3 children and older adults than in young adults, age2: F(1, separately to group 4 (all ps <.05). 162) = 6.08, p =.01, g2 =.02.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 984 Julia Karbach and Jutta Kray Mixing Costs Switching Costs Children 500 500 Pretest Switching Costs (ms) Posttest Mixing Costs (ms) 400 400 300 300 200 200 100 100 0 0 Single Switch Verbalization Variability Single Switch Verbalization Variability Young adults 500 500 Switching Costs (ms) Mixing Costs (ms) 400 400 300 300 200 200 100 100 0 0 Single Switch Verbalization Variability Single Switch Verbalization Variability Older adults 500 500 Switching Costs (ms) 400 Mixing Costs (ms) 400 300 300 200 200 100 100 0 0 Single Switch Verbalization Variability Single Switch Verbalization Variability Figure 1 Mixing costs (left panel) and switching costs (right panel) as a function of session (pretest, posttest), training (single = single-task training; switch = task-switching training; verbalization = task-switching + verbal self-instruction training; variability = task-switching + verbal self-instructions + training variability), and age (children, young adults, older adults). Error bars refer to standard errors of the mean. 2.5 2.5 Children Pretest-Posttest Effect Size d' Young Adults 2.0 2.0 Older Adults 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 Single-Task Task-Switching + Verbalization + Variability Single-Task Task-Switching + Verbalization + Variability Training Training Training Training Mixing Costs Switching Costs Figure 2 Effect size d ¢ for near transfer of task-switching training based on mixing costs (left panel) and switching costs (right panel) as a function of training (single-task training, task-switching training, task-switching + verbal self-instruction training, task-switching + verbal self-instructions + training variability) and age (children, young adults, older adults).  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.  Transfer of task-switching training 985 Table 2 Mean performance (SD) for far transfer tasks (Stroop, verbal WM, spatial WM, and fluid Intelligence) as a function of session (pretest/posttest), training (single-task training/task-switching training), and age (children/young adults/older adults) Children Young adults Older adults Pretest Posttest Pretest Posttest Pretest Posttest M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) Stroop interference (ms) STT 70 (42) 72 (49) 30 (31) 48 (41) 57 (57) 72 (55) TST 48 (61) 24 (53) 57 (41) 27 (34) 77 (81) 56 (46) Verbal working memory (% correct) STT 47.3 (14.4) 47.8 (24.0) 66.1 (12.7) 68.3 (16.3) 56.3 (16.8) 58.0 (15.8) TST 45.8 (13.4) 56.0 (17.2) 71.6 (16.1) 81.3 (15.3) 55.1 (15.9) 62.4 (18.6) Spatial working memory (% correct) STT 20.1 (10.7) 23.7 (12.8) 44.2 (14.0) 46.4 (17.1) 25.9 (18.0) 26.3 (21.4) TST 17.7 (11.6) 27.2 (16.4) 46.0 (17.6) 56.1 (17.4) 21.7 (16.6) 26.3 (16.8) Fluid intelligence (% correct) STT 75.2 (12.0) 76.8 (16.0) 92.7 (5.0) 93.7 (4.1) 78.6 (10.2) 79.5 (17.2) TST 73.3 (10.5) 79.8 (9.8) 89.8 (6.3) 93.4 (6.2) 75.1 (10.9) 79.7 (12.5) Note: STT = Single-task training, TST = Task-switching training. 2.4 2.4 2.4 2.0 2.0 2.0 Single-Task Training 1.6 1.6 1.6 Task-Switching Training Effect Size (d') 1.2 1.2 1.2 0.8 0.8 0.8 0.4 0.4 0.4 0.0 0.0 0.0 –0.4 –0.4 –0.4 –0.8 –0.8 –0.8 –1.2 –1.2 –1.2 MC SC IC VWM SWM FI MC SC IC VWM SWM FI MC SC IC VWM SWM FI Children Younger adults Older adults Figure 3 Effect size d ¢ for near and far transfer of training as a function of training (single-task training, task-switching training), transfer measure (MC = mixing costs; SC = switching costs; IC = interference control; VWM = verbal working memory; SWM = spatial working memory; FI = fluid intelligence), and age (children, young adults, older adults). task domain (fluid intelligence). For each domain,9 we 162) = 4.60, p <.05, g2 =.02, and also for fluid again tested whether the task-switching training groups intelligence, F(1, 162) = 5.37, p <.05, g2 =.03 (see (2–4) showed different amounts of transfer (see Table 2). Thus, the performance improvements from Table A3). Since we found no effect of training for pretest to posttest were larger after task-switching any domain (all ps >.13), data were collapsed across training than after single-task training, and were not groups 2–4 and subjected to a two-way ANOVA with modulated by age (all ps >.61). the between-subjects factors Age (children/young In order to investigate the range of far transfer, we adults/older adults) and Training (single-task/task- calculated the pretest-posttest ES (see Figure 3). switching). The results indicated that the task- Consistent with previous findings (cf. Klauer, 2001), switching training groups showed more far transfer ES were smaller for far transfer to other executive tasks than the single-task groups for both verbal and spatial and fluid intelligence than for near transfer. However, WM, F(1, 162) = 4.94, p <.05, g2 =.02, and F(1, ES after task-switching training were still relatively large even for far transfer, with most values >.70 for 9 children, >.60 for adults, and >.40 for older adults, Correlations between the tasks were high (Verbal WM: and were quite consistent across the far transfer tasks. counting span–reading span, r=.50**; spatial WM: symmetry In contrast, ES for the single-task training were span–navigation span, r =.56**; fluid intelligence: letter series–figural reasoning, r =.45**; letter series–Raven, generally small or even negative, and substantially r =.59**; figural reasoning–Raven, r =.52**) and the smaller than for task-switching training under all results were similar across tasks for the respective constructs. experimental conditions.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 986 Julia Karbach and Jutta Kray Discussion adapting to new task demands in each training session supported the acquisition of a generalizable switching skill The primary aim of this study was to investigate the in adults, but hindered it in children. Regarding adults, this usefulness of task-switching training. To answer this finding is consistent with the literature, suggesting that question, we examined the amount of near and far variable training can promote transfer (cf. Schmidt & transfer of task-switching training in children, young Bjork, 1992), although the training-related benefits (i.e. adults, and older adults under different training the reduction of switching costs from training session 1 conditions. Our results identified several important new to 4) were smaller than in the remaining two task-switching findings. First, we found evidence for substantial transfer training groups. However, regarding children, it seems that of task-switching training to a structurally similar new the increased cognitive load associated with variable switching task after training. Consistent with a prior training tasks did not leave enough processing capacity study (Minear et al., 2002), the reduction of mixing costs to implement the abilities improved during training and from pretest to posttest was much larger after task- to develop cognitive representations of the task structure switching training (mean d¢ = 1.44) than after single-task (cf. van Merrinboer, Kester & Paas, 2006). This training (mean d¢ =.26). From a theoretical point of interpretation is consistent with theoretical accounts view, this finding is particularly important because it (e.g. Sweller, 1999) emphasizing that complex tasks and shows that the trainability and transferability of stimuli result in higher working memory demands while executive control processes is not merely mediated by performing a given task. Since working memory capacity is automatization of single-task components (cf. Kramer more limited in children than in adults (for a review, see et al., 1999). In contrast to the Minear et al. study, we Hitch, 2006), the increased cognitive load associated with also found near transfer of task-switching training on the the variable training would be more likely to affect level of switching costs (mean d¢ = 1.17). children’s performance. Hence, the implementation of Second, and particularly interesting from a the trained abilities and the representation of the task developmental perspective, the near transfer on the structure are impaired, especially on the level of mixing level of mixing costs was most pronounced in children costs, which include a substantial working memory and older adults. Thus, in particular the age groups component (i.e. the ability to maintain two task sets). usually characterized by marked deficits in task-set Future studies may test this hypothesis by including a selection and maintenance were able to transfer cognitively less demanding variable training condition. training-related benefits to a new task. This finding has A potentially critical point for the interpretation of this important implications for the application of training finding is the fact that the variable training was programs to individuals with executive deficits in the combined with verbal self-instruction training. clinical and educational contexts. Although a comparison of training groups 2 and 3 A third result was that the type of training modulated indicated that verbal self-instructions did not influence the amount of near transfer. On the one hand, verbal the amount of transfer, it may be argued that the self-instructions did not promote transfer of task- decreased transfer after variable training found in switching training. There are at least two possible children is the result of an interaction between the explanations for this finding. First, the groups trained variable training and the verbalizations performed in task switching without verbal self-instructions used an during training, which makes the training even more internal verbal strategy similar to the overt self- complex. However, in order to ultimately disprove this instructions anyway so that we found no difference in point, a variable training condition without verbal self- the amount of transfer between these groups. Second, if instructions would have been necessary. training and transfer tasks were more similar for the The fourth and most striking result concerns far verbal self-instruction group, that is, if participants were transfer of task-switching training. Our data clearly allowed to verbalize at posttest (and not only during show that in contrast to single-task training, task- training), then transfer may occur. This idea is in line switching training resulted in improved performance in with results from Healy, Wohldmann, Parker and Bourne an interference control task, in verbal and spatial WM (2005), showing that participants performing a tasks, and even in fluid intelligence tasks. Although there secondary verbal task during training in a prospective is some evidence for far transfer of executive control paradigm performed worse during transfer when the training in children (Fisher & Happ, 2005; Klingberg secondary task was not required during transfer. The et al., 2005; Kloo & Perner, 2003; Rueda et al., 2005), authors suggested that the training task and the verbal most training programs in previous studies focusing on task are integrated into a single, more complex task adults resulted in large improvements on the training during practice, and that transfer only occurs when the task itself while transfer to other tasks was very limited, cognitive operations acquired during training can be suggesting that transfer was quite domain and process applied at transfer. specific (e.g. Ball, Berch, Helmers, Jobe, Leveck, On the other hand, training variability resulted in Marsiske, Morris, Rebok, Smith, Tennstedt, Unverzagt differential age effects. Specifically, the requirement for & Willis, 2002; Jennings, Webster, Kleykamp &  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.  Transfer of task-switching training 987 Dagenbach, 2005). Also, in previous studies reporting we used in this study required a number of different far transfer, the transfer distance and the type of training executive control processes. First, demands on goal were not systematically varied. In contrast, the present maintenance were high because subjects received no study shows broad transfer that was stable even for external task cues. Second, stimuli were highly different measures of far transfer and to domains quite ambiguous; that is, they always represented features remote from the training tasks. It also provides the first relevant to both tasks, and the currently irrelevant evidence that the near and far transfer of executive feature had to be suppressed. Consequently, interference control training can indeed be achieved across a wide control was constantly required. Finally, because subjects range of ages. Still, it may seem surprising that far had to perform two rather than only one task during task- transfer was neither modulated by age nor by the type of switching training, task-set selection demands were high. task-switching training. Given that there is no prior Thus, assuming that all these executive processes were evidence with respect to these aspects, our expectations trained, it seems less surprising that our task-switching were relatively unspecific. Based on the present results, training showed broad transfer to other executive and one may assume that the different types of task- cognitive task domains. Nevertheless, since the transfer switching training were equally efficient, and that the distance is an important aspect for evaluating training training was equally beneficial for all age groups. programs, it seems that this type of task-switching training However, this conclusion should be drawn cautiously. is suitable for promoting not only one, but several In line with previous results (cf. Klauer, 2001; Salomon executive control abilities; therefore, it is probably useful & Perkins, 1989), effects sizes were generally smaller for for a number of clinical and educational applications. It far transfer than for near transfer in this study. The should also be noted that compared with other studies smaller effects are, the harder they are to verify in small investigating the transfer of training (cf. Klauer, 2001), the samples (for a meta-analysis, see Lipsey & Wilson, ES were relatively large for near transfer, particularly for 1993), indicating why it may have been hard to find a children, and consistently remained on a high level even modulation of far transfer by age group or training type. across far transfer tasks. In order for the age group or training type differences to Determining the relative training potential regarding reach statistical significance, the sample would have to different executive control components and the long- be relatively large, which is usually hard to realize in term effects of task-switching training is a matter for training studies. future research. Also, it may be important to consider Considering the findings of the present study, the individual differences regarding transfer benefits to obvious question is what kinds of processes were actually clarify how the training improves performance, so that transferred after task-switching training? Our data suggest it can be optimized when applied to those who need it that subjects transferred more than the mere ability to most (cf. Bissig & Lustig, 2007; van Merrinboer et al., switch between tasks. However, the task-switching version 2006). Appendix Table A1 Means (M) and standard deviations (SD) for the training group matching criteria (single-task RT, mixing costs, Raven score) as a function of age (children, young adults, older adults) and training group (single-task training, task-switching training, task- switching + verbal self-instruction training, task-switching + verbal self-instruction + variability training) at pretest Matching variables Single-task RT Mixing costs Raven score Training group M SD M SD M SD Children Single-task training 1000 184 363 227 23.6 3.5 Task-switching training 1040 194 363 148 23.6 2.1 + verbalization 973 148 400 182 23.1 2.3 + + variability 996 297 357 179 22.6 2.3 Young adults Single-task training 570 118 170 184 27.8 1.6 Task-switching training 545 67 149 91 26.8 1.9 + verbalization 525 105 174 88 26.6 1.9 + + variability 604 91 186 118 27.8 1.8 Older adults Single-task training 758 175 345 243 23.9 3.9 Task-switching training 818 262 363 217 22.1 2.7 + verbalization 705 121 364 184 23.2 3.3 + + variability 765 215 394 243 23.9 1.9  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 988 Julia Karbach and Jutta Kray Table A2 Stroop task mean RT (M) and standard deviation (SD) as a function of age (children, young adults, older adults), training (single-task training, task-switching training, task-switching + verbal self-instruction training, task-switching + verbal self-instruction + variability training), session (pretest, posttest), and trial type (neutral, incongruent) Pretest Posttest Neutral Incongruent Neutral Incongruent Training group M SD M SD M SD M SD Children Single-task training 904 91 974 98 927 170 999 192 Task-switching training 992 156 1017 146 933 123 954 154 + verbalization 960 88 1029 103 889 103 929 111 + + variability 952 187 1004 182 876 145 889 144 Young adults Single-task training 631 92 661 99 573 71 621 93 Task-switching training 582 76 634 95 542 62 572 85 + verbalization 628 115 680 129 554 92 586 119 + + variability 569 95 636 116 509 65 528 79 Older adults Single-task training 821 172 878 206 754 149 832 209 Task-switching training 863 113 954 161 828 128 897 149 + verbalization 791 107 853 114 733 104 787 97 + + variability 736 127 802 159 678 112 725 124 Table A3 Mean performance (M) and standard deviation (SD) for verbal WM, visuospatial WM, and fluid intelligence (% correct) as a function of age (children, young adults, older adults), training (single-task training, task-switching training, task-switching + verbal self-instruction training, task-switching + verbal self-instruction + variability training), and session (pretest, posttest) Verbal WM Visuospatial WM Fluid intelligence Pretest Posttest Pretest Posttest Pretest Posttest Training group M SD M SD M SD M SD M SD M SD Children Single 47.3 14.4 47.8 24.0 20.1 10.7 23.7 12.8 75.2 12.0 76.8 16.0 Switch 43.3 15.6 56.3 23.8 16.1 13.4 26.3 17.9 74.3 11.4 81.4 7.3 + verbalization 47.8 14.8 58.0 16.3 17.9 8.4 28.1 16.0 73.2 8.9 78.9 11.8 + + variability 46.4 9.4 53.6 11.7 19.2 13.1 27.2 16.4 72.3 11.6 78.9 10.3 Young adults Single 66.1 12.7 68.3 16.3 44.2 14.0 46.4 17.1 92.6 5.0 93.6 4.1 Switch 73.7 11.5 81.3 12.3 43.8 20.9 55.8 17.9 88.1 7.1 93.1 7.2 + verbalization 69.2 20.4 79.0 17.9 51.3 16.7 55.8 21.9 90.4 4.4 92.9 6.4 + + variability 71.9 15.3 83.5 16.0 42.9 14.7 56.7 12.6 90.9 7.2 94.3 5.1 Older adults Single 56.3 16.8 58.0 15.8 25.9 18.0 26.3 21.4 78.6 10.2 79.5 17.2 Switch 52.7 16.4 59.4 17.1 14.3 13.7 21.4 15.1 68.9 10.0 74.3 14.1 + verbalization 55.8 18.4 64.7 21.1 25.4 20.1 29.9 17.7 75.4 12.7 81.1 13.3 + + variability 56.7 13.3 62.9 18.3 25.4 13.8 27.7 17.4 81.0 6.1 83.7 8.3 Acknowledgements Ball, K., Berch, D.B., Helmers, K.F., Jobe, J.B., Leveck, M.D., Marsiske, M., Morris, J.N., Rebok, G.W., Smith, D.M., This research was funded by the Deutsche Tennstedt, S.L., Unverzagt, F.W., & Willis, S.L. (2002). Effects of cognitive training interventions with older adults: a Forschungsgemeinschaft (grant Kr 1884/3-3). Thanks randomized controlled trial. Journal of the American Medical to Katharina Engelke, Claudia Kersken, Anna Orth, and Association, 288, 2271–2281. Daniel Straß for their help running the experiments. Bedard, A.-C., Nichols, S., Schachar, J.A., Schachar, R., Logan, G.D., & Tannock, R. (2002). The development of selective inhibitory control across the life span. Develop- References mental Neuropsychology, 21, 93–111. Bherer, L., Kramer, A.F., Peterson, M.S., Colcombe, S., Baddeley, A., Chincotta, D., & Adlam, A. (2001). Working Erickson, K., & Becic, E. (2005). Training effects on dual-task memoryand the control of action: evidence from task switching. performance: are there age-related differences in plasticity of Journal of Experimental Psychology: General, 130, 641–657. attentional control? Psychology and Aging, 20, 695–709.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.  Transfer of task-switching training 989 Bissig, D., & Lustig, C. (2007). Who benefits from memory younger and old adults. Journal of Experimental Psychology: training? Psychological Science, 18, 720–726. Applied, 1, 50–76. Cepeda, N.J., Kramer, A.F., & Gonzales De Sather, J.C.M. Kramer, A.F., Larish, J.E., Weber, T., & Bardell, L. (1999). (2001). Changes in executive control across the life span: Training for executive control: task coordination strategies examination of task-switching performance. Developmental and aging. In D. Gopher, & A. Koriat (Ed.), Attention and Psychology, 37, 715–730. performance XVII: Cognitive regulation of performance: Cohen, J. (1977). Statistical power analysis for the behavioral Interaction of theory and application (pp. 617–652). sciences (rev. edn.). New York: Academic Press. Cambridge, MA: The MIT Press. Crone, E.A., Ridderinkhof, K.R., Worm, M., Somsen, Kray, J., Eber, J., & Karbach, J. (2008). Verbal self-instructions R.J.M., & Van der Molen, M.W. (2004). Switching be- in task switching: a compensatory tool for action-control tween spatial stimulus–response mappings: a developmental deficits in childhood and old age? Developmental Science, 11, study of cognitive flexibility. Developmental Science, 7, 443– 223–236. 455. Kray, J., Eber, J., & Lindenberger, U. (2004). Age differences Dowsett, S.M., & Livesey, D.J. (2000). The development of in executive functioning across lifespan: the role of verbali- inhibitory control in preschool children: effects of ‘execu- zation in task preparation. Acta Psychologica, 115, 43–165. tive skills’ training. Developmental Psychobiology, 36, Kray, J., & Lindenberger, U. (2000). Adult age differences in 161–174. task switching. Psychology and Aging, 15, 126–147. Emerson, M.J., & Miyake, A. (2003). The role of inner speech Lindenberger, U., Mayr, U., & Kliegl, R. (1993). Speed and and task switching: a dual-task investigation. Journal of intelligence in old age. Psychology and Aging, 8, 207–220. Memory and Language, 48, 148–168. Lipsey, M.W., & Wilson, D.B. (1993). The efficacy of psy- Fisher, N., & Happ, F. (2005). A training study of theory of chological educational, and behavioral treatment: confirma- mind and executive function in children with autistic spec- tion from meta-analysis. The American Psychologist, 48, trum disorders. Journal of Autism and Developmental Disor- 1181–1209. ders, 35, 757–771. Mayr, U. (2001). Age differences in the selection of mental sets: Fisk, J.E., & Sharp, C.A. (2004). Age-related impairment in the role of inhibition, stimulus ambiguity, and response set executive functioning: updating, inhibition shifting, and ac- overlap. Psychology and Aging, 16, 96–109. cess. Journal of Clinical and Experimental Neuropsychology, Meiran, N., Gotler, A., & Perlman, A. (2001). Old age is 26, 874–890. associated with a pattern of relatively intact and relatively Healy, A.F., Wohldmann, E.L., Parker, J.T., & Bourne, L.E., Jr impaired task-set switching abilities. Journal of Gerontology: (2005). Skill training, retention, and transfer: the effects of a Psychological Sciences, 56, 88–102. concurrent secondary task. Memory and Cognition, 33, Minear, M.E., Shah, P., & Park, D. (2002). Age, task switching, 1457–1471. and transfer of training. Poster presented at the 9th Cognitive Hedges, L.V., & Olkin, I. (1985). Statistical methods for meta- Aging Conference, Atlanta, 2002. analysis. Orlando, FL: Academic Press. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., & Hitch, G.J. (2006). Working memory in children: a cognitive Howerter, A. (2000). The unity and diversity of executive func- approach. In E. Bialystok, & F.I.M. Craik (Eds.), Lifespan tions and their contributions to complex ‘frontal lobe’ tasks: a cognition (pp. 112–127). Oxford: Oxford University Press. latent variable analysis. Cognitive Psychology, 41, 49–100. Jennings, J., Webster, L., Kleykamp, B., & Dagenbach, D. Raven, J.C. (1988). Standard progressive matrices. Weinheim: (2005). Recollection training and transfer effects in older Beltz. adults: successful use of a repetition-lag procedure. Aging, Reimers, S., & Maylor, E.A. (2005). Task switching across the Neuropsychology, and Cognition, 12, 278–298. lifespan: effects of age on general and specific switch costs. Kane, M.J., Hambrick, D.Z., Tuholski, S.W., Wilhelm, O., Developmental Psychology, 41, 661–671. Payne, T.W., & Engle, R.W. (2004). The generality of Rosenbaum, D.A., Carlson, R.A., & Gilmore, R.O. (2001). working memory capacity: a latent-variable approach to Acquisition of intellectual and perceptual-motor skills. An- verbal and visuospatial memory span and reasoning. Journal nual Review of Psychology, 52, 453–470. of Experimental Psychology: General, 133, 189–217. Rueda, M.R., Rothbart, M.K., McCandliss, M.D., Saccom- Klauer, K.J. (2001). Handbuch Kognitives Training [Handbook anno, L., & Posner, M.L. (2005). Training, maturation, and of cognitive training]. Gçttingen: Hogrefe. genetic influences on the development of executive attention. Klingberg, T., Fernell, E., Olesen, P., Johnson, M., Gustafsson, Proceedings of the National Academy of Sciences, 102, P., Dahlstrçm, K., Gillberg, C.G., Forssberg, H., & Wester- 14931–14936. berg, H. (2005). Computerized training of working memory Salomon, G., & Perkins, D.N. (1989). Rocky roads to transfer: in children with ADHD – a randomized, controlled trial. rethinking mechanisms of a neglected phenomenon. Educa- Journal of the American Academy of Child and Adolescent tional Psychology, 24, 113–142. Psychiatry, 44, 177–186. Salthouse, T.A., & Meinz, E.J. (1995). Aging, inhibition, Kloo, D., & Perner, J. (2003). Training transfer between card working memory, and speed. Journals of Gerontology Series sorting and false belief understanding: helping children apply B: Psychological Sciences and Social Sciences, 50, 297–306. conflicting descriptions. Child Development, 74, 1823–1839. Sanders, R.E., Gonzalez, D.J., Murphy, M.D., Pesta, B.J., & Kramer, A.F., & Kray, J. (2006). Aging and divided attention. Bucur, B. (2002). Training content variability and the effec- In E. Bialystok, & F.I.M. Craik (Eds.), Lifespan cognition: tiveness of learning: an adult age assessment. Aging, Neuro- Mechanisms of change (pp. 57–69). Oxford: Oxford Univer- psychology, and Cognition, 9, 157–174. sity Press. Schmidt, R.A., & Bjork, R.A. (1992). New conceptualizations Kramer, A.F., Larish, J.F., & Strayer, D.L. (1995). Training for of practice: common principles in three paradigms suggest attentional control in dual task settings: a comparison of new concepts for training. Psychological Science, 3, 207–217.  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd. 990 Julia Karbach and Jutta Kray Sweller, J. (1999). Instructional design in technical areas. Verhaeghen, P., Marcoen, A., & Goossens, L. (1992). Camberwell, Victoria, Australia: Australian Council for Improving memory performance in the aged through mne- Educational research. monic training: a meta-analytic study. Psychology and Aging, van Merrinboer, J.G.J., Kester, L., & Paas, F. (2006). 7, 242–251. Teaching complex rather than simple tasks: balancing Williams, B.R., Ponesse, J.S., Schachar, J.A., Logan, G.D., & intrinsic and germane load to enhance transfer of learning. Tannock, R. (1999). Development of inhibitory control Applied Cognitive Psychology, 20, 343–352. across the lifespan. Developmental Psychology, 35, 205–213. Verhaeghen, P., & Cerella, J. (2002). Aging, executive control, and attention: a review of meta-analyses. Neuroscience and Received: 6 November 2007 Biobehavioral Reviews, 26, 849–857. Accepted: 9 September 2008  2009 The Authors. Journal compilation  2009 Blackwell Publishing Ltd.