Distinct transfer effects of training different facets of working memory capacity
Highlights
► Each of three groups trained one functional category of working memory capacity. ► Training groups were compared to an active control group. ► Linear mixed-effects modeling revealed distinct transfer profiles for the groups. ► Both storage-processing and supervision training led to transfer to reasoning. ► Effects were still observed after 6 months without training.
Introduction
Distinct Transfer Effects of Training Different Facets of Working Memory Capacity Working memory is a cognitive system providing temporary access to representations needed for complex cognition. The purpose of the present work is to investigate whether working memory capacity (WMC) can be improved by training. Our study builds on a model of the factorial structure of working-memory capacity, the facet model (Oberauer et al., 2000, Oberauer et al., 2003, Süß et al., 2002).
According to the facet model, WMC can be classified into three functional categories: simultaneous storage and processing, relational integration, and supervision. Storage and processing comprises the simultaneous maintenance and manipulation of information. Relational integration involves the coordination of information elements into new structures. The third functional category, supervision, refers to the selective activation of relevant information and inhibition of irrelevant information. In studies of the facet model of working memory (Oberauer et al., 2000, Oberauer et al., 2003), the concept of supervision has so far been measured through the efficiency of task switching, which is largely equivalent to the concept of shifting in Miyake et al.’s model of executive functions (2000). Miyake et al. (2000) distinguish shifting, inhibition, and working-memory updating as three lower-level factors of executive functions. Factor-analytic studies have shown that the constructs storage and processing and relational integration are closely correlated, and both are more weakly related to supervision (Buehner et al., 2005, Oberauer et al., 2003). Therefore, here we regard WMC (constituted by storage and processing and relational integration) and supervision (or shifting in the context of executive functions) as two related but distinct high-level constructs.
WMC is assumed to be a largely stable trait, which predicts other cognitive abilities such as fluid intelligence and reasoning (Conway et al., 2003, Engle et al., 1999, Kyllonen and Christal, 1990, Oberauer et al., 2008, Süß et al., 2002). There is also a considerable number of published studies linking impairments of WMC to a wide range of neurological disorders, such as attention-deficit hyperactivity disorder (Martinussen, Hayden, Hogg-Johnson, & Tannock, 2005). Likewise, there is evidence that individual differences in supervision abilities (i.e., executive functions) are highly stable, and impact everyday behavior (Friedman et al., 2007; Mischel et al., 2011). Hence, it is an important question whether WMC and supervision can be improved by training, and whether this improvement affects related cognitive abilities.
Previous research findings concerning training and transfer effects of WMC are inconsistent if not contradictory. Several recent studies indicate that adequate training can lead to an increase in WMC test performance and also to transfer to the performance in non-trained cognitive tasks (for a review see Klingberg, 2010). For example, Chein and Morrison (2010) examined the effectiveness of storage and processing training, operationalized by a complex span task. They found a generalization of the training effect to reading comprehension and inhibition in a subgroup of successfully trained participants. There is even evidence that fluid intelligence can be increased by computer-based training of WMC (for a review see Buschkuehl & Jaeggi, 2010; see also Borella et al., 2010, Jaeggi et al., 2008, Jaeggi et al., 2010, Karbach and Kray, 2009, Klingberg et al., 2005, Schmiedek et al., 2010). The work of Colom et al. (2010) suggests that such increases in intelligence are not only due to retest effects, since they found that retest gains in WMC tasks were not related to increases in intelligence test scores.
However, other studies do not reveal such convincing training and transfer effects. For example, Holmes, Gathercole, and Dunning (2009) found no transfer of working-memory training to fluid intelligence, although these authors used the same training paradigm and a sample comparable to Klingberg et al. (2005). Moreover, Owen et al. (2010) recently showed that an online cognitive training that led to large practice effects on the trained tasks induced no measurable generalization to other cognitive tasks. Contradictory results are also present in training studies focusing explicitly on executive processes. On the one hand, there are several studies supporting plasticity of executive processes (Dahlin et al., 2008, Karbach and Kray, 2009, Li et al., 2008). On the other hand, one study found no transfer of inhibition training for preschool children to cognitive control (Thorell, Lindqvist, Bergman, Bohlin, & Klingberg, 2008).
There are multiple possible reasons for the inconsistency in previous findings (cf. Conway & Getz, 2010; Moody, 2009; Shipstead, Redick, & Engle, 2012). First, it is problematic that many training regimens lack underlying theoretical models. Hence, it often remains unclear which cognitive processes, broadly defined as “inhibition” or “working memory”, were actually trained.
Second, previous studies vary widely regarding general training conditions such as intensity and number of training sessions. For example, in Owen et al.’s (2010) paradigm, training sessions were very short (only 10 min a day) and the number of completed sessions varied between participants from 2 to 188, whereas others who found transfer effects controlled very carefully how much training participants went through, showing increased transfer with larger amounts of training (Jaeggi et al., 2008). Moreover, it is important to control the quality of training and the commitment of participants. A plausible reason for the absence of transfer effects in the study by Owen et al. (2010) is the uncontrolled and anonymous setting of the training regimen. For instance, Smith et al. (2009) showed that in a more controlled setting, with face-to-face contact at pre and post training assessment, self-administered training interventions at home resulted in transfer effects.
Third, only few studies include an active control group that completes an alternative intervention. Implementing an active control group differentiates training and transfer effects not only from repetition effects (as a passive or waiting control group would do), but also from intervention effects (e.g., effects of sticking to a regular training schedule), and expectancy effects influencing cognitive performance (Oken et al., 2008). For the latter purpose it is important that the alternative training is perceived by participants as a potentially effective cognitive training. The importance of including an active rather than a passive control group was recently demonstrated by Redick et al.’s (in press) failure to replicate the findings of Jaeggi et al., 2008, Jaeggi et al., 2010 when evaluating transfer effects in comparison to an active control group.
With the present study, we wanted to answer the following questions: (1) can WMC (with its two aspects, storage and processing, and relational integration) and supervision be improved by extensive training, (2) do training effects transfer to non-trained tasks measuring the same construct, and (3) does transfer to related cognitive abilities – such as inhibition and reasoning – occur?
In our study, we aimed to avoid the possible weaknesses occasionally observed in previous research. Therefore, our study was designed to meet the following four requirements. First, the choice of training tasks should be based on a theoretical model. Our training was based on the facet model of WMC (Oberauer et al., 2003), thus, we chose training tasks which are assumed to measure a particular functional category in that model (i.e., storage and processing, relational integration, and supervision). To distinguish training effects between the functional categories, each of three groups trained only one function.
Second, transfer should be measured by a broad test battery in order to reveal a fine-grained picture of transfer effects (Li et al., 2008, Shipstead et al., 2010). It should consist of tests covering a broad range of transfer distances. Our test battery comprised several tasks measuring the same constructs as the trained tasks (i.e., the WMC-constructs storage and processing and relational integration, and the supervision construct of shifting), together with tasks measuring related constructs that were not trained but are assumed to correlate with the trained constructs (i.e., inhibition and reasoning). Additionally, we included a face matching test which is assumed to demand only a minimum of WMC and supervision, so we expected no effect of training on this measurement.
Third, the training should be extensive and include a certain degree of variability to facilitate transfer effects (Schmidt & Bjork, 1992). Additionally, difficulty of the training tasks should be adapted stepwise to trainees’ performance in order to induce plastic processes (Klingberg et al., 2005, Lövden et al., 2010). Participants of the present study completed 20 training sessions (each approximately 30–40 min) in 4 weeks. Three tasks were used to train each functional category, and the level of difficulty was adapted to individual performance.
Fourth, training effectiveness should be evaluated by comparing an experimental with an active control group. Participants should be assigned randomly to experimental and control groups to ensure internal validity (Shipstead et al., 2010). Moreover, the general training conditions should be as similar as possible for all included groups in order to control for motivational and psychological effects such as the Hawthorne effect, which refers to improvements in performance simply due to increased attention to the participants’ behavior (Green and Bavelier, 2008, McCarney et al., 2007, Shipstead et al., 2010). Therefore, we implemented a control training where participants practiced visual discrimination with the same training conditions as the experimental groups.
Section snippets
Method
Participants completed 4 weeks of extensive cognitive training. They were randomly assigned to one of four groups (matched by gender): Storage-Processing training, Relational Integration training, Supervision training, or to the Active Control group that completed a visual discrimination training. Assignment of participants was double blind, so that neither the experimenter who had contact with the participants, nor the participants themselves, knew which group they were assigned to.
To measure
Treatment of missing data
Due to technical difficulties at pre- and posttest assessment, we lost data of three participants on their performance in three tasks. Data loss concerned figural task switching and the Flanker task at the pretest, and memory updating at the posttest. For the analyses of variances including these tasks, we excluded the three participants. To maximize power for the structural equation and the mixed-effects model, we maintained these participants and imputed their missing values by using multiple
Discussion
The goal of the present work was to investigate training and transfer effects of working memory training with a theory-based and well controlled experimental protocol. Our training focused on the three functional categories storage and processing, relational integration, and supervision, which are based on a factor-analytic investigation of the structure of WMC and related constructs (Oberauer et al., 2003). Our structural equation model of the pretest data replicated the three-factor
Conclusion
The present study contributes evidence suggesting that a theory-based, carefully controlled and extensive training has an impact on cognitive abilities such as working memory and reasoning that were believed to be constant traits. However, distinct transfer profiles for the three functional categories storage and processing, relational integration, and supervision illustrated considerable variations regarding the magnitude of transfer to certain constructs – providing a possible explanation for
Acknowledgments
This research was supported by a grant to the first author from the Forschungskredit of the University of Zurich and is part of the first author’s doctoral thesis. We thank André Locher and Michael Ruflin for programming the training tasks in Tatool, and Veronica Heusser, Sandra Scherer, Marc Züst, Eszter Montvai and Semra Atsiz for assistance with collecting the data. We also thank Simone Eberhart and Nicole Cruz for their help in piloting the training tasks, and Julia Karbach and Jutta Kray
References (70)
- et al.
Reasoning = working memory ≠ attention
Intelligence
(2005) - et al.
Improvement in working memory is not related to increased intelligence scores
Intelligence
(2010) - et al.
Cognitive ability: Does working memory training enhance intelligence?
Current Biology
(2010) - et al.
Working memory capacity and its relation to general intelligence
Trends in Cognitive Sciences
(2003) - et al.
The relationship between n-back performance and matrix reasoning – Implications for training and transfer
Intelligence
(2010) Training and plasticity of working memory
Trends in Cognitive Sciences
(2010)- et al.
Computerized training of working memory in children with ADHD – A randomized, controlled trial
Journal of the American Academy of Child & Adolescent Psychiatry
(2005) - et al.
Reasoning ability (is little more than) working-memory capacity?!
Intelligence
(1990) - et al.
A meta-analysis of working memory impairment in children with attention-deficit/hyperactivity disorder
Journal of the American Academy of Child & Adolescent Psychiatry
(2005) - et al.
The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis
Cognitive Psychology
(2000)
Task switching
Trends in Cognitive Sciences
Working memory capacity – Facets of a cognitive ability construct
Personality and Individual Differences
The multiple faces of working memory: Storage, processing, supervision, and coordination
Intelligence
Which working memory functions predict intelligence?
Intelligence
Working-memory capacity explains reasoning ability – And a little bit more
Intelligence
Mixed-effects modeling with crossed random effects for subjects and items
Journal of Memory and Language
Working memory training in older adults: Evidence of transfer and maintenance effects
Psychology and Aging
The psychophysics toolbox
Spatial Vision
Some tests of the decay theory of immediate memory
Quarterly Journal of Experimental Psychology
Improving intelligence: A literature review
Swiss Medical Weekly
Human cognitive abilities: A survey of factor-analytic studies
Expanding the mind’s workspace: Training and transfer effects with a complex working memory span task
Psychonomic Bulletin & Review
Working memory span tasks: A methodological review and user’s guide
Psychonomic Bulletin & Review
Confidence intervals in within-subjects designs: A simpler solution to Loftus and Masson
Tutorial in Quantitative Methods for Psychology
Plasticity of executive functioning in young and older adults: Immediate training gains, transfer, and long-term maintenance
Psychology and Aging
Individual differences in working memory capacity and what they tell us about controlled attention, general fluid intelligence, and functions of the prefrontal cortex
Effects of noise letters upon the identification of a target letter in a nonsearch task
Perception & Psychophysics
Greater attention problems during childhood predict poorer executive functioning in late adolescence
Psychological Science
The relations among inhibition and interference control functions: A latent-variable analysis
Journal of Experimental Psychology: General
Exercising your brain: A review of human brain plasticity and training-induced learning
Psychology and Aging
Multilevel models for the experimental psychologist: Foundations and illustrative examples
Behavior Research Methods
Adaptive training leads to sustained enhancement of poor working memory in children
Developmental Science
Improving fluid intelligence with training on working memory
Proceedings of the National Academy of Sciences of the United States of America
Cited by (118)
Cognitive training based on human-computer interaction and susceptibility to visual illusions. Reduction of the Ponzo effect through working memory training
2024, International Journal of Human Computer StudiesMutual interference in working memory updating: A hierarchical Bayesian model
2022, Journal of Mathematical PsychologyThe domain-specific approach of working memory training
2022, Developmental ReviewSelf-Regulated Learning and Working Memory Determine Problem-Solving Accuracy in Math
2022, Spanish Journal of PsychologyCan we enhance working memory? Bias and effectiveness in cognitive training studies
2024, Psychonomic Bulletin and Review