Practice-related optimization and transfer of executive functions: a general review and a specific realization of their mechanisms in dual tasks

Abstract

Improvements in performing demanding and complex task situations are typically related to the optimization of executive functions and efficient behavioral control. The present study systematizes and reviews the optimization of different executive function types: Shifting, Inhibition, Updating, and Dual tasking. In particular, we focus on optimisations of these functions with training and on transfer effects of related training skills to non-trained situations. The aim of the study’s empirical part (see also Appendix) was to investigate the specific mechanisms of executive functions in the context of Dual tasking, leading to improved dual-task performance after practice. More specifically, we tested the Efficient Task Instantiation (ETI) model that includes specific assumptions regarding practice-related improvements of executive task coordination skills: Dual-task performance is improved with practice because of an efficient and conjoint instantiation of sets of relevant task information in working memory at the onset of a dual task. According to our knowledge, the ETI model is one of the first that allows illustrating the contribution of cognitive mechanisms underlying practice-related improvements in performing dual tasks and the impact of task coordination skills on this performance.

Appendices

Appendix

This section includes the Methods and Results that we applied to test predictions of the ETI model and one specific mechanism that contributed to improved dual-task coordination skills. The key assumption of this model is that dual-task performance is improved with dual-task practice because of an efficient and conjoint instantiation of relevant task information in working memory at the onset of a dual-task trial. The central prediction tested in this study is that this efficient instantiation is realized for information of two component tasks after dual-task practice and mixed single-task practice while there is no such instantiation for two tasks after single-task practice.

Methods

Participants

Twenty-eight participants were randomly assigned to one of the three experimental groups: the hybrid group, the single-task group, and the mixed single-task group. The hybrid group included 10 participants (5 female) with a mean age of M = 23.7 years (SD = 3.3 years, age range 19–29 years). Eight participants (4 female) were included in the single-task group with a mean age of M = 26.2 years (SD = 4.4 years, age range 19–32 years). The mixed single-task group included 10 participants (7 female) with a mean age of M = 23.8 years (SD = 4.7 years, age range 19–32 years). Participants were recruited at the Departments of Psychology at Humboldt University Berlin and at Ludwig-Maximilians-University Munich. They were right-handed, had normal or corrected to normal vision and were not informed of the purpose of the experiment. They were paid for participation (8 €/session plus performance-based bonuses).

Apparatus

Visual stimuli were presented on a 17-in. color monitor and auditory stimuli were presented via headphones which were connected to a Pentium I IBM-compatible PC. The RTs for manual responses were recorded with external response keys and the RTs of verbal responses were recorded via a voice key connected to the experimental computer. The experiment was controlled by the software package ERTS (Experimental Runtime System; Beringer, 2000).

Stimuli and component tasks

Participants conducted two speeded-choice RT tasks. In the auditory task, they responded to sine-wave tones presented at frequencies of 300, 950 or 1,650 Hz by saying “ONE”, “TWO”, or “THREE” (German: “EINS”, “ZWEI”, or “DREI”), respectively. An auditory single-task trial started with the presentation of three dashes on the computer screen. After an interval of 500 ms, the tones were presented for 40 ms. The trial was completed when the participant responded verbally or a 2,000-ms response interval had expired. To analyze the accuracy of each response, the experimenter recorded the verbal responses. In the visual task, participants responded manually by pressing a spatially compatible key with the index, middle, or ring finger of their right hand to white circles appearing at the left, central, or right position arranged horizontally on the computer screen, respectively. In visual single-task trials, three white dashes served as placeholders for the possible positions of the visual stimuli. They appeared as a warning signal 500 ms before the visual stimulus was presented. The stimulus remained visible until the participant responded or a 2,000-ms response interval had expired. After correct responses in the visual and in the auditory task, the RTs were presented for 1,500 ms on the screen. Following incorrect responses, the word “ERROR” (German: “FEHLER”) appeared. A blank interval of 700 ms preceded the beginning of the next trial in both component tasks.

Dual-task trials included the auditory and the visual task. These trials were identical to single-task trials with the exception that an auditory and a visual stimulus were presented simultaneously (SOA = 0 ms) and participants responded to both stimuli with equal emphasis.

Design and procedure

Hybrid group

The hybrid practice condition included single-task blocks and dual-task blocks. In the single-task blocks, participants performed either 45 single-task trials of the auditory task or they performed 45 trials of the visual task. During dual-task blocks, participants performed a random mixture of 18 dual-task trials and 30 single-task trials, 15 of the auditory task and 15 of the visual task.

In a first familiarization session, participants performed six auditory and six visual single-task blocks that were presented in an alternating order. As in all sessions, half of the participants started with an auditory single-task block and the other half with a visual single-task block. The first practice session, i.e. Session 1, included six single-task blocks (three auditory and three visual task blocks) and eight dual-task blocks. After two initial single-task blocks (one auditory and one visual single-task block), sequences of two dual-task blocks and one single-task block followed; the type of single-task blocks were alternated. The design in Sessions 2–7 was identical to that in Session 2, but these sessions included two additional dual-task blocks at the end. Different sessions were conducted on successive days.

Single-task group

The single-task group practiced single-task blocks (Sessions 1–6) with the exception of a few dual-task blocks for manipulation check purposes (see below). To keep the number of stimulus contacts constant across dual-task and single-task practice groups, one dual-task trial in the hybrid practice group was replaced by one single-task trial of each task in the single-task group. Consequently, there were single-task blocks with 45 trials (short blocks) and single-task blocks with 66 trials (long blocks). Session 1 included 12 single-task blocks (six auditory and six visual single-task blocks) and two dual-task blocks; these dual-task blocks were included to analyze initial dual-task performance in the single-task group at the beginning of practice and to match this performance between practice groups. In Session 1, these two initial dual-task blocks were introduced after two short single-task blocks. Then, sequences of one short and two long single-task blocks followed. In Sessions 2–6, we presented 16 single-task blocks (eight auditory and eight visual single-task blocks). After two initial short single-task blocks, sequences of two long single-task blocks and one short single-task block followed. In Sessions 1–6, blocks with the auditory and visual task were alternated and the first type of block (either auditory or visual task) was counterbalanced between subjects. Session 7 was identical to the procedure in the hybrid group.

Mixed single-task group

The procedure of the mixed single-task group was similar to the procedure in the single-task group except that long single-task blocks were replaced by mixed single-task blocks. These mixed single-task blocks included 66 single-task trials, 33 auditory task and 33 visual task trials, presented randomly and in unpredictable order.

Results

The Results section is structured as follows: first, the practice findings in the hybrid group and a comparison of single-task performances in the three practice groups (hybrid, single task, mixed single task) are briefly summarized. Second, we analyzed the dual-task performance at the end of single-task, mixed single-task, and hybrid practice. In all analyses, we present the RT and error data separately for both the auditory and the visual tasks. The indicator of dual-task performance was the dual-task cost (i.e., the mean performance on dual-task trials compared with the mean performance on single-task trials in single-task blocks) that is considered a strong and reliable criterion for dual-task performance (e.g., Hazeltine et al., 2002; Liepelt et al., 2011; Strobach et al., 2012d; Tombu & Jolicoeur, 2004). The familiarization session served to help participants getting acquainted with the tasks and the material and was thus not included in the analyses.

Practice effects

The dual-task practice results from Session 1 to 7 were similar to those of Experiment 1 of Schumacher et al. (2001, see also Hazeltine et al., 2002, Tombu and Jolicoeur, 2004) and demonstrated an optimization of single- and dual-task performance as illustrated in Table 1. Importantly, this improvement was larger in visual and auditory dual tasks compared to single tasks and a practice-related reduction of dual-task RT costs (as the difference in mean performance between dual-task and single-task trials in single-task blocks) from 169 ms, t(9) = 8.207, p < 0.001, to 41 ms, t(9) = 6.993, p < 0.001, in the auditory task and from 120 ms, t(9) = 4.913, p < 0.001, to 27 ms, t(9) = 4.158, p < 0.01, in the visual task.

Table 1 Mean RTs (ms) and error rates (percent; in brackets) for the dual-task, mixed single-task, and single-task groups in Sessions 1–7

To test whether hybrid, mixed single-task, and single-task practice yielded similar effects on single-task performance, we analyzed single-task block performance in all three groups from Session 1 to 7. In the auditory and visual task, RTs decreased similarly in the single tasks across the three experimental groups from Session 1 to 7, Fs(6, 150) > 63.059, ps < 0.001; note that there was no modulation of this effect by the different practice types, Fs(6, 150) < 1.363, ps > 0.27 (see Table 1). A comparison of the error rates yielded equivalent results. This data demonstrates an equivalent level of automaticity of stimulus–response mappings across practice conditions and contrasts assumptions of Kramer et al. (1995).

Dual-task performance after hybrid, mixed single-task, and single-task practice

Second, we compared the dual-task performance at the beginning of practice (i.e., pre-test) and at the end of practice (i.e., post-test) in the hybrid, mixed single-task, and the single-task groups. For the pre-test comparison, we analyzed the dual-task performance at the beginning of practice by comparing the RTs and error rates of the first two single-task blocks with that of the dual-task trials of the two following dual-task blocks in Session 1. The data of Session 7 (in which the single-task, the mixed single-task, and the hybrid groups performed single- and dual-task trials) served as the measure of single- and dual-task performance at the end of practice. We performed mixed-measures ANOVAs on the RTs and error rates of the auditory and visual task with the within-subject factors Testphase (pre-test vs. post-test) and Trial type (single-task trials vs. dual-task trials), and the between-subject factor Group (dual-task vs. mixed single-task vs. single-task groups). Since our primary focus is on effects resulting from the different types of practice, in the following analyses, we exclusively report main effects of and interactions with Group.

For the auditory task RTs, Testphase interacted with Group, F(2, 25) = 3.494, p < 0.05, as well as with Trial type, F(1, 25) = 46.775, p < 0.001. Most important, these interactions were qualified by a three-way interaction of Testphase, Trial type, and Group, F(2, 25) = 3.641, p < 0.05, suggesting that single-task and dual-task RTs changed differently from pre- to post-test in the three groups of participants. As illustrated in Fig. 2a, dual-task performance at pre-test was similar across all groups, as the main effect and the interactions with the factor Group did not reach significance [Fs(2, 25) < 1.073, ps > 0.36]. Therefore, between-group differences in dual-task performance at post-test cannot be explained by baseline differences in dual-task performance at pre-test. In fact, the post-test analysis indicated an interaction between Trial type and Group, F(2, 25) = 8.094, p < 0.01, providing evidence for differences in dual-task costs after hybrid, mixed single-task, and single-task practice. RTs in dual-task trials were similar in hybrid and mixed single-task groups, t(18) < 1, and the dual-task RTs of these two groups were both decreased relative to the RTs in the single-task group, ts(16) > 2.230, ps < 0.05. Importantly, single-task RTs were similar in the three groups of participants, ts(16) < 1; thus, differences in dual-task performance cannot be attributed to different levels of component-task processing skill after practice. The corresponding analysis of the auditory error rates showed no main effect and/or interaction with Group (Table 2A). In sum, the present findings of the auditory task are consistent with the prediction of the ETI model.

Fig. 2

Single-task and dual-task RT data at pre-test (first four blocks in Session 1) and post-test (Session 7) in the dual-task, mixed single-task, and single-task groups. a Data in auditory task, b data in visual task

Table 2 Single-task and dual-task error data in percent at pre-test (first four blocks in session 1) and post-test (session 7) in the dual-task, mixed single-task, and single-task groups

In the visual task (see Fig. 2b), RTs were also slower in the single-task group compared with the mixed single-task and hybrid groups, F(2, 25) = 130.620, p < 0.05. Across pre- and post-test, there were similar single-task RTs in the three experimental groups; RTs in dual tasks, however, were generally increased across pre- and post-test in the single-task group when compared with the mixed single-task and hybrid practice groups, F(2, 25) = 3.581, p < 0.05. However, the main effect of and/or any other interaction with Group was not significant. The corresponding analysis based on error rates also revealed no such main effect or interactions (Table 1B).

The observation of similar dual-task performance after mixed single-task and hybrid practice in the auditory-task RTs is inconsistent with general assumptions of Meyer and Kieras (1999) as well as Kramer et al. (1995, Bherer et al., 2005, 2008). Meyer and Kieras assumed that only dual-task practice provides conditions for optimal dual-task performance. We showed that a specific type of single-task practice (i.e., mixed single-task practice) also enabled such optimized performance. Similarly, Kramer and colleagues assumed that the acquisition of task coordination skills requires simultaneous dual-task practice and there is no such acquisition under single-task conditions. The present finding of similar levels of optimizing dual-task performance after hybrid practice (including dual-task trials) as well as mixed single-task practice is inconsistent with this assumption. These similar levels should be achieved with the contribution of similar practice mechanisms such as the acquisition of improved task coordination skills. Since we provided evidence for these similar optimization levels, there should be skill acquisition not only under conditions of hybrid practice, but also under mixed single-task practice conditions without dual-task practice trials. This latter assumption indicates that improved task coordination skills, in contrast to the assumption of Kramer et al., can be acquired without the simultaneous presentation and execution of two tasks.

Task overlap after hybrid, mixed single-task, and single-task practice

The next section tested a second prediction of the ETI model: mixed single-task practice and hybrid practice should result in a different overlap of component task processing than single-task practice (see also Strobach, Schubert, Pashler, & Rickard, 2014, for the case of two memory-retrieval tasks). Participants in the former two practice conditions were expected to perform an efficient conjoint instantiation of two component tasks at the onset of a dual-task trial. This conjoint instantiation should facilitate a fast initiation of task processes in a long task after the completion of a short task. As a consequence, this fast initiation should result in an increased overlap in simultaneous task processing and shorter intervals between responses to the shorter and the longer task (i.e., inter-response interval, IRI). Thus, we tested the assumption of shorter IRIs after mixed single-task and hybrid practice in contrast to single-task practice.

This IRI distribution analysis was conducted on the data of the three different practice groups at pre- and post-test by calculating the IRIs as the difference of auditory dual-task RTs minus visual dual-task RTs. Thus, positive IRIs reflects larger auditory than visual-task RTs while negative IRIs indicate larger visual than auditory-task RTs. In order to calculate IRI distributions, we divided them into bins of 20 ms, computed the frequency of trials for each IRI bin (from small to large IRI bins), and plotted the corresponding IRI distributions as a function of practice group and test phase in Fig. 3. We reasoned that large IRIs indicated a small temporal overlap between both tasks and vice versa. Also, different IRI slopes across practice conditions may point to a different degree of temporal overlap and differential practice effects on dual-task performance (in the context of the ETI model), which may be associated with the acquisition of task coordination skills (i.e., optimized instantiation of task information of the longer auditory task). In particular, the model assumed a comparable reduction of IRIs as a result of hybrid and mixed single-task practice in contrast to single-task practice.

Fig. 3

Cumulated frequency distribution (percent) of inter-response intervals (IRIs) in dual-task trials of 20 ms bins at pre- and post-test in the dual-task, mixed single-task, and single-task groups

As illustrated in Fig. 3, IRIs were similarly distributed at pre-test in all practice groups; thus, temporal overlap between tasks was similar at the beginning of practice. However, it increased after hybrid and mixed single-task practice at the end of practice (i.e., post-test) relative to single-task practice. Analyses of the median IRI supported this observation: an ANOVA including the within-subjects factor Testphase (pre-test vs. post-test) and the between-subjects factor Group (hybrid vs. mixed single-task vs. single-task) provided a significant main effect of Testphase and interaction of Testphase and Group, F(1, 25) = 3.338, p < 0.05, and, F(2, 25) = 5.467, p < 0.05, respectively. While there was no GROUP effect at pre-test, F(1, 25) < 1, median IRIs were reduced in the mixed single-task and the hybrid group in contrast to the single-task group, ts(17) > 5.623, ps < 0.05. Thus, the effects of practice on the temporal relation of dual-task processing are consistent with the prediction of the ETI model, suggesting an increased task processing overlap after initiating conjoint task information at the onset of dual-task trials as a consequence of practice.