Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Apr 1.
Published in final edited form as: J Exp Psychol Hum Percept Perform. 2020 Apr;46(4):434–441. doi: 10.1037/xhp0000725

Item-specific priming of voluntary task switches

Yu-Chin Chiu 1, Kerstin Fröber 2, Tobias Egner 3
PMCID: PMC7147980  NIHMSID: NIHMS1574385  PMID: 32223294

Abstract

The ability to switch efficiently between different tasks underpins cognitive flexibility, and is impaired in various psychiatric disorders. Recent research suggests that the control processes mediating switching can be subject to learning, as “switch readiness” can become associated with, and primed by, specific stimuli. In cued task switching, items that are frequently associated with the need to switch incur a smaller behavioral switch cost than items associated with a low probability of switching, known as the item-specific switch probability (ISSP) effect (Chiu & Egner, 2017). However, it remains unknown whether ISSP associations modulate the efficiency only of cued switching or also impact people’s voluntary choice to switch tasks. Here, we addressed this question by combining an ISSP manipulation with a protocol that mixed 75% standard cued task trials with 25% free choice trials, allowing us to measure the effect of ISSP on voluntary switch rate (VSR). We observed robust ISSP effects on cued trials, replicating previous findings. Crucially, we also found a greater VSR for items associated with a high than with a low switch likelihood. We thus demonstrate that associating specific stimuli with frequent switch requirements not only reduces switch costs but also enhances participants’ tendency to switch voluntarily.

Keywords: cognitive control, learning and memory, voluntary task switching, flexibility

Introduction

Cognitive control refers to a collection of cognitive processes that allow us to employ internal goals and current context to guide our actions (Egner, 2017; Miller & Cohen, 2001). One key aspect of cognitive control is the ability to configure, maintain, and flexibly update “task sets”, including a set of currently relevant rules that link stimuli to responses (Monsell, 2003). The control processes associated with task-set coordination have been extensively studied using cued task-switching paradigms. In a typical protocol, participants are cued on each trial which one out of two possible tasks to perform on a given stimulus. For instance, in the task employed in the present study, participants are asked to judge whether a pictured object is a living or nonliving thing (animacy task), or whether it is smaller or larger than a shoebox (size task). The general finding is that response times are longer (and error rates higher) on trials where the current task differs from the one performed on the previous trial (task switch trials) than when it is the same (task repeat trials), known as “switch cost” (reviewed in Kiesel et al., 2010; Monsell, 2003; Vandierendonck, Liefooghe, & Verbruggen, 2010).

The interpretation of switch costs is contentious, with some researchers focusing on the contribution of active control processes putatively required to configure the new set (Meiran, 1996; Rogers & Monsell, 1995), and others on the contribution of possible interference from lingering activation (Allport, Styles, & Hsieh, 1994) or cued associative retrieval (Waszak, Hommel, & Allport, 2003, 2004, 2005) of the previous task set. A key development in the task switching literature in this regard has been the introduction of the voluntary task switching (VTS) procedure. Here, subjects choose the task (e.g., animacy vs. size task) to be performed on a given trial themselves; nevertheless, robust switch costs are observed (Arrington & Logan, 2004). This procedure is important because, according to some researchers, it is the only way to guarantee that subjects have to perform an active, volitional act of control when switching tasks, as opposed to relying on externally cued retrieval of learned stimulus-response associations to perform the instructed tasks (Arrington & Logan, 2004; 2005).

Regardless of the exact interpretation of switch costs, a smaller cost is usually taken to indicate a more efficient implementation of task-set updating processes, reflecting greater cognitive flexibility or switch-readiness (e.g., Goschke, 2000). Moreover, one’s degree of switch-readiness can be shaped by the statistical properties of the task sequence, suggesting that the control or cued retrieval processes underpinning task-set updating are subject to associative learning. For instance, switch costs are smaller in blocks of trials where switches are frequent than when they are rare (Dreisbach & Haider, 2006; Monsell & Mizon, 2006). Others have shown switch costs are modulated when individual stimuli are associated with specific tasks (Koch & Allport, 2006; Waszak et al., 2003, 2004, 2005), specific location (Crump & Logan, 2010; Leboe, Wong, Crump, & Stobbe, 2008) and the likelihood of switching (Chiu & Egner, 2017; Leboe et al., 2008). In particular, Chiu and Egner (2017) employed a typical cued task-switching paradigm but, without mentioning explicitly to the participants, some of the stimuli were associated with a high switch probability (i.e., these items occurred more often on switch than on repeat trials) whereas other stimuli were associated with a low switch probability. The overall switch rate was 50%, and the stimuli were neither predictive of a specific task nor a particular response. Nevertheless, we found that stimuli that predicted a high switch probability were associated with a smaller switch cost than stimuli that predicted a low switch probability. This finding, referred to as the item-specific switch probability (ISSP) effect, documents that specific stimuli can be trained to serve as bottom-up cues for cognitive flexibility, presumably by priming (i.e., rapidly retrieving) control or associative processes implicated in task-set updating (see also Crump & Logan, 2010; Leboe, Wong, Crump, & Stobbe, 2008). In other words, initially associative learning enables the linking between a specific stimulus and a particular control state and subsequently, the encounter of the same stimulus leads to activation of the associated control state (Abrahamse, Braem, Notebaert, & Verguts, 2016; Bugg & Crump, 2012; Chiu, 2019; Chiu & Egner, 2019; Egner, 2014).

However, while it has thus been shown that externally cued switch performance (as measured in RT switch costs) can be improved via the associative linking of specific stimuli to switch-likelihood, an important question remains unanswered: would presenting a stimulus that had previously been predictive of frequent cued switching actually enhance a participant’s tendency to switch tasks if they were given a choice? In other words, it is presently not known whether ISSP associations only modulate the efficiency of cued switching (which may not necessarily be an active act of control) or whether they also impact people’s voluntary choice to switch task, which is thought to necessarily involve active control (Arrington & Logan, 2004; 2005).

Two recent studies suggest that voluntary switching can in fact be affected by learning. In one study, participants who were rewarded more on switch trials in a cued task-switching phase exhibited a higher voluntary switch rate in a subsequent phase where they could freely select which task to perform from trial to trial (Braem, 2017). In the other study, participants performed a “hybrid” task, where they were cued which task to perform on some trials, and free to choose which task to perform on others. It was found that when participants were forced to switch on a high proportion of the cued trials, they also produced a higher rate of voluntary switches than when forced switches were less frequent (Fröber & Dreisbach, 2017).

While there is thus some evidence that voluntary switching can be promoted by task or reward statistics, no prior study has tested whether ISSP, previously shown to reduce switch costs in cued task switching, may also influence voluntary switch rates. This question has important theoretical and practical implications, as a positive finding would indicate that item-specific switch learning is not cue dependent and – importantly - can modulate the volitional choice to engage in a cognitive control process. The latter would considerably raise the potential value of developing control learning regimes for clinical application. To examine whether ISSP can promote voluntary switching, we combined an ISSP design with the hybrid approach of Fröber and Dreisbach (2017), by mixing 75% standard cued task trials with 25% voluntary task selection trials, where participants were free to choose which task to perform. We predicted that participants would deliberately switch more often when encountering items associated with a high switch probability (high switch items) as compared to items associated with a low switch probability (low switch items).

Method

Participants.

Two cohorts of participants provided informed consent: 65 Amazon Mechanical Turk (AMT) workers (Experiment 1) and 60 undergraduate students at Purdue University (Experiment 2). The two samples were collected under the protocol approved by the Duke University and Purdue University Institutional Review Board, respectively. The online Experiment 1 was run first, and Experiment 2 was conducted subsequently as an in-person replication experiment. AMT workers were compensated with $4 and Purdue undergrads were given course credits. The sample size was determined based on the ISSP effect in Chiu & Egner (2017). With a medium effect size, a desired power of .8 and a type I error of 0.05, the estimated sample size was 60. We included 5 more participants in anticipation of data loss due to technical errors on AMT. Experiment 1 included 58 participants (Mage = 31, SD = 8, 42 males) in the final analysis after excluding 7 participants (5 due to making no response on more than 20% of trials, 2 due to mean accuracy outside of the group mean ± 2.5 SD. Experiment 2 included 54 participants (Mage = 19.5, SD = 1.5, 32 females) in the final analysis after excluding 6 participants (4 due to aborting the experiment early, 2 due to mean accuracy outside group mean ± 2.5 SD).

Stimuli.

We used the same stimulus set as in Experiment 2 of Chiu & Egner (2017), consisting of 40 color photographs of objects (~8° in width and 5.7° in height) from Moreno-Martínez & Montoro (2012), with 10 objects belonging to each of four categories: living objects smaller than a shoebox, living objects larger than a shoebox, nonliving objects smaller than a shoebox and nonliving objects larger than a shoebox. For each participant, a total of 8 images were used in the experiment (two from each stimulus category, randomly selected). In each switch probability condition, there were 4 unique stimuli (8 total). We chose this small number of stimuli in order to maximize learning while minimizing the total number of trials required to likely obtain stimulus-switch associations (and hence the length of the experiment). A different set of 8 images was used for practice.

Design and Procedure.

Participants went through two task-switching phases sequentially. First was a 100% cued task phase where they were cued which task to perform on each trial. This phase served to ensure that participants had the opportunity to form strong item-specific switch probability associations. There were 128 trials per block for two blocks in this experiment phase. On these cued trials, participants categorized stimuli either as living versus nonliving or as larger versus smaller than a shoebox, based on the color of a frame (red or blue, counterbalanced across participants; thickness: ~0.3°) placed around the stimuli, which was presented simultaneously with the stimuli. The incidence of the two tasks and trial types (switch/repeat) were both equated. Without explicitly mentioning to the participants, four of the stimuli were associated with a 75% and four with a 25% chance of switching, with an overall switch rate fixed at 50%. Note that even though we did not tell participants about the switch probability manipulation, we also did not assess participants’ awareness of this manipulation at the end of the experiment (see Discussion). Four keys on a standard QWERTY keyboard (F and V for one task, M and K for the other task) were used to indicate the category of the stimulus. Participants placed their left middle and index finger on F and V, respectively, and placed their right index and middle finger on M and K, respectively. Stimulus category-to-response key mappings were counterbalanced across participants. To ensure that the switch probability manipulation was not confounded by biased stimulus-task associations, trial sequences were generated with a randomization procedure with the constraints that the incidence of the two tasks were equated and the difference in trial count for each stimulus appearing in one task versus the other task was less than two trials. Participants were instructed to respond as fast as possible without sacrificing accuracy.

After completing the 100% cued task phase, participants progressed to a “hybrid phase”, consisting of 75% cued trials and 25% “choice” trials, where the absence of a colored frame indicated that they could freely choose which of the two tasks to perform. The same four keys were used in the choice condition, thus allowing us to infer their choice. There were 128 trials per block for six blocks in this experiment phase. Each trial started with a blank for 500 ms, followed by an image surrounded by a colored frame (or no frame) for 1500 ms appearing at the center of the screen, during which participants could make a response, and after which they received written feedback (correct or incorrect) for 500 ms.

Before the main task, participants familiarized themselves with the assigned stimulus-response mapping in one or more practice blocks without the ISSP manipulation. After the participants reached 80% accuracy on these cued trials, they were given a hybrid block with 75% cued trials and 25% choice trials and were instructed that they were free to choose a task on choice trials. This was to familiarize participants with the idea of choosing a task freely on choice trials in the hybrid blocks. The trial timing was the same as in the main task.

Analysis.

RTs that were too fast (<300 ms) were excluded. We first analyzed the correct trial response times (RTs) and error rate (ERs) on cued trials using a 2 (phase: 100% cued vs. hybrid) x 2 (switch probability: 25% vs. 75%) x 2 (trial type: switch vs. repeat) repeated measure analysis of variance (ANOVA). This analysis allows us to assess whether participants acquired item-specific switch readiness (i.e., a significant ISSP effect) during the 100% cued phase and maintained the effect during the hybrid phase. Similarly, we also analyzed the RTs and ERs on choice trials using a 2 (switch probability: 25%, 75%) x 2 (task switch: switch, repeat) repeated-measure ANOVA to evaluate the ISSP effect on these trials. Importantly, to address our main question, we analyzed the voluntary switch rate (VSR) on choice trials during the hybrid phase using a one-tailed paired sample t-test as we expect the VSR to be larger for high than for low switch probability items. VSR is the percentage of switches made on choice trials (regardless of the trial accuracy). To figure out the VSR, we first inferred the task that participants had performed based on the response hand, and a change of task from the previous trial (which could be a cued or a choice trial) was marked as a switch. Choice trials with no response and choice trials following no response trials (cued or choice) were excluded. Incorrect choice trials were also excluded as well as those following an error trial (which could be a cued or choice trial). In addition, we also assessed the task bias on choice trials as a function of switch probability using a paired t test. The task bias was quantified by the percentage of choice trials that participants had chosen the animacy task.

Results

Cued trials.

RTs.

Complete descriptive statistics are shown in Table 1. Both experiments showed almost exactly the same pattern (Figure 1). Participants took longer to respond on switch trials than on repeat trials in both experiments, Experiment 1: F(1, 57) = 100, p < .01, η2p = .64; Experiment 2: F(1, 53) = 347, p < .01, η2p = .87. Mean RT on cued trials was overall faster in the hybrid phase than in the 100% cued phase, Experiment 1: F(1, 57) = 14, p <.01, η2p = .19; Experiment 2: F(1, 53) = 21, p <.01, η2p = .28. Switch costs were larger in the 100% cued phase than in the hybrid phase, as indicated by a significant phase x trial type interaction effect, Experiment 1: F(1, 57) = 6, p < .05, η2p = .10; Experiment 2: F(1, 53) = 6, p < .05, η2p = .11. These two effects are likely due to a practice effect that facilitated responding during later stages of the experiment.

Table 1.

Means (standard deviations) for Response Time (in ms) on cued trials

Switch Probability
 Experiment 1 Experiment 2
Cued Hybrid Cued Hybrid
Repeat Switch Repeat Switch Repeat Switch Repeat Switch
Low 805 (100) 890 (131) 788 (94.7) 858 (120) 827 (69.1) 951 (89.5) 800 (77.3) 917 (108)
High 808 (98.2) 876 (134) 795 (101) 849 (119) 836 (76.3) 952 (87.2) 813 (81.2) 909 (103)
Open in a new tab
Figure 1.

Figure 1.

Open in a new tab

Switch cost in the cued condition, i.e., (a) response time (in ms) and (b) error rate (in %) on switch trials vs. repeat trials. Note that error bars represent “within-subject” standard error of means (SEM), which is estimated from SEM of all pairwise mean differences (Franz & Loftus, 2012) in the full ANOVA.

Importantly, the switch probability x trial type interaction was highly significant in both experiments, Experiment 1: F(1, 57) = 13, p < .01, η2p = .18; Experiment 2: F(1, 53) = 10, p < .001, η2p = .15. Unpacking this interaction, the switch costs for high switch items (Experiment 1: M = 59.34, SD = 53.68; Experiment 2: M = 101.81, SD = 48.37) were significantly smaller than those for low switch items (Experiment 1: M =73.24, SD = 52.88; Experiment 2: M = 118.00, SD = 53.19), and this effect did not interact with phase, F’s < 1. These results (Figure 1a) replicate Chiu & Egner (2017) and demonstrate a robust ISSP effect that remained stable in the hybrid phase, where cued trials were intermixed with free choice trials. No other main or interaction effect was significant.

Error.

See Table 2 for complete descriptive statistics. Similar to the RT results, participants responded less accurately on switch trials than on repeat trials (Figure 1b), Experiment 1: F(1, 57) = 42, p < .01, η2p = .43; Experiment 2: F(1, 53) = 51, p < .01, η2p = .49. The switch probability x trial type interaction (i.e., an ISSP effect) was significant in Experiment 1, F(1, 57) = 4, p < .05, η2p = .07, but not in Experiment 2, F < 1. Main effect of phase was not significant in Experiment 1, F < 1, but was in Experiment 2, F(1, 53) = 11, p < .01, η2p = .17. No other main effect or interaction was significant.

Table 2.

Means (standard deviations) for Error Rate (in %) on cued trials

Switch Probability
 Experiment 1 Experiment 2
Cued Hybrid Cued Hybrid
Repeat Switch Repeat Switch Repeat Switch Repeat Switch
Low 9.8 (8.6) 13.4 (12.2) 8.8 (6.1) 13.2 (9.3) 8.9 (5.5) 13.4 (9.3) 7.9 (4.1) 11.9 (7.0)
High 8.8 (6.9) 11.7 (8.9) 9.5 (7.9) 12.3 (9.1) 10.4 (6.4) 14.6 (6.3) 8.2 (5.1) 12.0 (5.9)
Open in a new tab

Choice trials.

Voluntary Switch Rate.

See Table 3 for complete descriptive statistics. In both experiments, the mean VSR was significantly greater for high switch items than for low switch items, Experiment 1: t(57) = 1.83, p < .05 (one-tailed), Cohen’s d = 0.24; Experiment 2: t(53) = 2.45 , p < .01 (one-tailed), Cohen’s d = 0.33. This finding shows that participants chose to switch tasks more often when encountering items that were associated with a high switch probability on cued trials as compared to when encountering low switch probability items (Figure 2).

Table 3.

Means (standard deviations) for Voluntary Switch rate (VSR, in %)) and Task Choice Probability (Prob., in %) on choice trials

Switch Probability
 Experiment 1 Experiment 2
VSR Prob. VSR Prob.
Low 31.6 (9.6) 51.1 (26.6) 23.8 (9.7) 54.9 (17.8)
High 33.1 (9.1) 51.3 (27.6) 26.3 (9.9) 51.1 (19.0)
Open in a new tab
Figure 2.

Figure 2.

Open in a new tab

Voluntary switch rate (VSR, in %) in the choice condition. The item-specific switch probability (low vs. high) refers to the switch probability associated with the same items in the 100% cued phase. Note that error bars represent “within-subject” standard error of means (SEM), which is estimated from SEM of all pairwise differences (Franz & Loftus, 2012) in the full ANOVA.

To probe whether VSR faded with time, we analyzed mean VSR as a function of time by dividing the VSR phase into 3 blocks. This did not reveal a significant main effect of time nor an interaction of switch probability and time (Experiment 1 & 2: F’s <2, p’s >.05), suggesting that item-specific switch probability had a sustained effect on VSR.

Task Bias.

See Table 3 for complete descriptive statistics. On choice trials, participants did not seem to have a task bias in Experiment 1, as the task probability was around 50% and did not differ between items with low versus high switch probability, p > .05 (low switch: M = 51, SD = 27; high switch: M = 51, SD = 28). However, participants in Experiment 2 displayed a slight task bias as they chose the animacy task over the size task more often when encountering low switch items (M = 55, SD = 18) as compared to high switch items (M = 51, SD =19), t(53)= 2.45, p < .05.

RTs.

See Table 4 for complete descriptive statistics. On choice trials, participants also exhibited switch costs, as they took longer to respond on switch trials than on repeat trials, Experiment 1: F(1, 57) = 63, p < .01, η2p = .52; Experiment 2: F(1, 53) = 39, p < .01, η2p = .42, thus replicating prior voluntary task switching studies (e.g., Arrington & Logan, 2004). Mean RT was not different between high and low switch items, F’s < 1, and the RT ISSP effect on choice trials was not significant, F’s < 1.

Table 4.

Means (standard deviations) for Response Time (in ms) on choice trials

Switch Probability
 Experiment 1 Experiment 2
Repeat Switch Repeat Switch
Low 744 (84.2) 793 (111) 800 (82.5) 862 (124)
High 752 (92.1) 792 (112) 804 (77.3) 861 (104)
Open in a new tab

Error.

See Table 5 for complete descriptive statistics. Mean error rate on choice trials was low in both experiments. No main effects or interactions effects on error rate were observed on choice trials, F’s < 1, except that for Experiment 2, error rate was significantly higher on switch than on repeat trials, F(1, 53)= 14, p <.01, η2p= .21.

Table 5.

Means (standard deviations) for Error rate (in %) on choice trials

Switch Probability
 Experiment 1 Experiment 2
Repeat Switch Repeat Switch
Low 3.9 (3.8) 4.5 (5.6) 4.1 (3.7) 5.9 (6.5)
High 4.4 (4.1) 5.3 (7.6) 3.9 (2.7) 7.2 (6.8)
Open in a new tab

Control analysis.

Our favored interpretation is that the increased VSR in high switch probability items is due to incrementally acquired item-specific switch associations. Alternatively, it could be due to, or confounded with, more short-lived priming effects based on previous trial characteristics, including stimulus features (i.e., exact stimulus repetition from the previous trial), whether the previous trial was a low or high switch probability item, and whether it was a choice or cued trial. We performed additional control analyses, as prior studies have documented that these kinds of factors can modulate voluntary task switching (e.g., Arrington & Logan, 2005; Arrington, Weaver, & Pauker, 2010; Fröber & Dreisbach, 2017).

Accordingly, we performed several ANOVAs on VSR as a function of switch probability (25% vs. 75%) and previous trial characteristics in order to (a) probe whether the main effect of switch probability on VSR remained significant, and (b) whether there were any significant interactions between switch probability and previous trial characteristics, as these interactions would indicate that VSR is modulated not only by switch probability but also by priming effects based on previous trial characteristics. To have sufficient statistical power, we combined the two cohorts (N=112) and included experiment (1 vs. 2) as a between-subjects factor in these ANOVAs. Note that based on the numerical differences in the mean VSR in the two experiments (Experiment 1: M = 32.35; Experiment 2: M = 25.05), it was expected that the main effect of Experiment would be significant.

The first ANOVA examined the effect of stimulus repetition. However, due to the low trial count for conditions involving stimulus repetition (M = 8.5, SD = 1.9), we did not perform a full 3-factor ANOVA here. Instead, we excluded these rare trials and performed an ANOVA involving only experiment and switch probability to establish whether our findings are in any way dependent on the presence of direct stimulus repetitions. As expected, the main effects of experiment and switch probability remained significant (experiment, F(1, 110) = 16, p <.01, η2p= .13; switch probability, F(1, 110) = 5, p <.05, η2p= .04) and the experiment x switch probability interaction was not significant (experiment x switch probability, F(1,110) = 2, p >.05).

The second ANOVA involved switch probability and previous trial type (cued vs. choice), revealing significant main effects of experiment, previous trial type, switch probability, and a significant experiment x previous trial type interaction (experiment, F(1, 110) = 10, p < .01, η2p = .08; previous trial type, F(1, 110) = 89, p < .01, η2p = .45; switch probability, F(1, 110) = 6, p < .05, η2p = .05; experiment x previous trial type, F(1, 110) = 8, p < .05, η2p = .07). While the main effect of switch probability was significant (as expected), none of the interactions involving switch probability were significant (experiment x switch probability, F(1, 110) = 1, p > .05; previous trial type x switch probability, F(1, 110) = 1, p >.05, experiment x trial type x switch probability, F(1, 110) < 1, p > 05).

The third ANOVA involved switch probability and previous item switch probability (25% vs. 75%), and it revealed significant main effects of experiment, previous item type and switch probability (experiment, F(1, 110) = 19, p < .01, η2p = .15; previous item type, F(1, 110) = 16, p < .01, η2p = .13; switch probability, F(1, 110) = 11, p < .01, η2p = .10). However, none of the interactions were significant (Experiment x previous item type, F(1, 110) < 1, p > .05; experiment x switch probability, F(1, 110) < 1, p > .05; previous item type x switch probability, F(1, 110) = 3, p >.05; experiment x previous item type x switch probability, F(1, 110) = 3, p >.05).

In sum, these control analyses did not yield evidence for the effect of switch probability being confounded by previous trial characteristics, thus supporting the interpretation that the effect of switch probability on VSR was driven by longer-term ISSP associations as opposed to short-lived priming effects.

Discussion

We tested whether ISSP would generalize from influencing the efficiency of performing a cued switch to the choice domain, that is, whether items associated with frequent cued switching would promote voluntary task switching. We found clear support for this notion: The ISSP manipulation resulted both in reduced switch costs in items predicting a high probability of switching when participants were cued to perform a specific task, as well as in an increased likelihood that participants would choose to switch a task deliberately. While the VSR effect was relatively small, it proved robust across two experiments, one conducted online and the other in the lab.

Our findings dovetail well with the idea that control operation can be activated by bottom-up priming via learned associations between a stimulus and appropriate control requirements (Abrahamse et al., 2016; Bugg & Crump, 2012; Bugg, Jacoby, & Chanani, 2011; Chiu & Egner, 2019; Chiu, Jiang, & Egner, 2017; Egner, 2014; Verguts & Notebaert, 2008), and – importantly - they expand the scope of previously documented effects, with important implication for theories of how learning and cognitive control processes interact. In particular, the VSR effect observed here shows that the learned item-switch associations not only enhance switch readiness on cued task switching trials but also on uncued trials, where participants must actively control the choice of the task to be performed. In other words, the present data document, for the first time, that item-based switch associations can not only affect how efficiently participants perform a task switch, but also whether they switch tasks at all. Moreover, the effect of ISSP on VSR show that item-specific switch associations can modulate “pure” volitional control, independent of any processes primed by task cue retrieval (Arrington & Logan, 2004, 2005; Reiman, Weaver, & Arrington, 2014).

We attribute the VSR effect to long-term learning based on accumulation of item-specific experiences. What was learned could be a greater reliance on episodic retrieval processes for low switch items than for high switch items (Waszak et al., 2003, 2004, 2005), or a faster task reconfiguration process (Meiran, 1996; Rogers & Monsell, 1995) for high switch items than for low switch items. This account naturally predicts that the longer the training, the larger the VSR effect, which should be tested in future studies. In our current experiments, each unique item was only encountered 32 times during the 100% cued phase. This amount seems to be sufficient to produce a robust ISSP effect and a small but reliable VSR effect. Thus, to evaluate the scope for maximizing the VSR effect for practical impact, future studies should administer item-specific associations with various lengths.

As a contrast to the long-term learning account, we also explored other short-term priming effects that might contribute to our findings. Our control analyses showed that the observed VSR effect was not driven by short-term priming effects that are based on previous trial characteristics, such as whether the previous trial was a cued vs. a choice trial or whether the previous item was associated with a high vs. low switch probability. Although these previous trial characteristics contribute to performance, in line with previous studies (Arrington & Logan, 2005; Fröber & Dreisbach, 2017), they did not interact with the factor of switch probability. Therefore they did not seem to drive the observed VSR effects. Nonetheless, as we created these conditions post hoc, systematic manipulations of these factors along with the ISSP manipulation is required in the future to tease apart the contribution of short-term priming and long-term control learning in modulating switch readiness.

In both experiments, subjects were not told about the switch probability manipulation, and we also did not assess their explicit awareness of the item-specific switch probability at the end of the experiment. However, several previous context-control learning studies showed that participants were unable to report the task structure (Bejjani, Zhang, & Egner, 2018; Crump & Logan, 2010; King, Korb, & Egner, 2012), suggesting that unless participants are explicitly informed about them, context-control associations are likely to be learned and applied implicitly. For example, in Crump & Logan’s study (2010) where switch probability was manipulated across specific locations, participants performed at chance when estimating the probability of task switch (and repeat) for both high and low switch probability locations. In Bejjani et al.’s study (2018) where context-control learning was item-specific and involved only 4 unique pairs of pictures, including/exluding participants with explicit awareness did not alter the result patterns. We therefore suspect (but cannot be certain) that participants were unware of the item-specific switch probability (at least in the 100% cued phase) and the ISSP effect is unlikely to be modulated by awareness. However, it remains intriguing to determine whether awareness plays a role in modulating the VSR in the hybrid phase. As voluntary task switches may involve processes not involved in the cued task switches (Arrington & Logan, 2005), the free choice procedure might promotes explicit awareness of task structures in some individuals, resulting in a different VSR pattern in those individuals than those that remain unaware of the structure.

We did not find a significant ISSP effect on RTs in choice trials. However, this might be due to the fact that, after averaging across both cued and choice trials in the hybrid phase, the actual switch probability associated with high vs. low items were smaller than the intended 25% and 75% in the 100% cued phase, respectively. For example, the actual switch probability ended up to be about 54% for Experiment 1 and 51% for Experiment 2 for high switch items in the hybrid phase. In other words, these new stimulus-switch associations likely worked against the prior established stimulus-switch association that promotes switch readiness in high switch items. Nonetheless, low VSRs are fairly common in voluntary task switching studies, often falling in the range of 15% to 40% (e.g., Arrington & Logan, 2004, 2005; Fröber & Dreisbach, 2017), and it has been shown that the shorter the interval between trials, the lower the VSR (Arrington & Logan, 2004, 2005). The trial interval in our experiments was 500 ms, which likely contributed to the overall low VSR.

Alternatively, it could be due to an RT scaling effect, as the RTs on choice trials were shorter than those on cued trials. This speeding could be due to our specific design, where choice trials were presented without a colored cue frame. Subjects might be able to process the objects immediately as opposed to processing the frame first (as on cued trials), and then the objects. Future studies might consider equating this processing difference between the cued and the choice trials by also presenting a differently colored frame on choice trials. Relatedly, our design employed one cue per task (same as Leboe et al., 2008), which did not allow us to separately assess the modulation of cue-repetition benefit and task-switch cost by item-specific control learning (Logan & Bundesen, 2003; Mayr & Kliegl, 2003). This very issue was specifically addressed in Crump & Logan (2010) in the context of location specific switch probability manipulation, however. Their results showed that the contextual modulation affected primarily task switching as opposed to cue-encoding processes. Nonetheless, this is also an intriguing issue that remains to be addressed in the context of ISSP manipulation.

Also note that in our paradigm, the zero cue to stimulus interval (CSI) likely encourages participants to use a bottom-up strategy, which works in favor of any stimulus-based priming effect (Koch & Allport, 2006), including priming of switch readiness intended here. If the cue were presented much earlier than the stimulus, participants would be able to engage in preparation for a task repetition or switch before the switch-associated stimulus shows up, such that target stimulus driven priming of switch readiness would likely have less impact. This intuition is borne out by previous papers on the list-wide switch proportion effects. For instance, Monsell & Mizon (2006) found that block-wise switch-likelihood only modulates switch costs when CSIs are short. Likewise, we predict that context-specific modulation of switch costs would decrease with increasing cue-stimulus intervals.

The finding that ISSP affects not only cued switching but also the tendency to switch sets voluntarily enhances the potential significance of this effect in terms of translational impact. Many psychiatric conditions, including Autism Spectrum Disorder (ASD), Eating Disorders, Schizophrenia, and Attention Deficit Hyperactivity Disorder, are marked by deficits in cognitive flexibility (Goldberg & Weinberger, 1988; Hill, 2004; Morice, 1990). For example, individuals with ASD often performed worse than matched controls on tasks requiring task-set switching (D’Cruz et al., 2013; Roberts, Tchanturia, Stahl, Southgate, & Treasure, 2007). The present findings suggest that it might be possible to employ bottom-up cueing of switch-readiness to not only facilitate the efficiency of switching but to actually promote behavioral (choice) tendencies in clinical populations with deficits in cognitive flexibility.

Public Significance Statements.

A hallmark of human cognition is our cognitive flexibility, reflected in the ability to efficiently switch between different tasks, which is impaired in many psychiatric disorders. A recent study demonstrated that the efficiency with which people switch tasks, or “switch readiness”, can be improved through learning: people become better at switching tasks for stimuli that are frequently associated with the need to switch compared to stimuli that are rarely associated with switching. In the present study, we show that frequent-switch stimuli also increase people’s tendency to switch tasks when they are allowed to choose which task to perform. This finding suggests that it might be possible to employ stimulus-specific cuing of switch readiness to enhance both the ability and the choice to behave more flexibly in clinical populations with deficits in cognitive flexibility.

Acknowledgments

This work was supported in part by National Institute of Mental Health Award R01 MH087610 (T. E.).

We thank Zhimei Niu and Georgia Gross for their assitance with data collection.

The raw data are available at https://osf.io/d2c9v/.

References

  1. Abrahamse E, Braem S, Notebaert W, & Verguts T (2016). Grounding cognitive control in associative learning. Psychological Bulletin, 142(1), 693–728. 10.1037/bul0000047 [DOI] [PubMed] [Google Scholar]
  2. Allport A, Styles E. a, & Hsieh S (1994). Shifting Intentional Set: Exploring the Dynamic Control of Tasks. Attention and Performance XV, (June), 421–452. [Google Scholar]
  3. Arrington CM, & Logan GD (2004). The cost of a voluntary task switch. Psychological Science, 75(9), 610–615. 10.1111/j.0956-7976.2004.00728.x [DOI] [PubMed] [Google Scholar]
  4. Arrington CM, & Logan GD (2005). Voluntary task switching: Chasing the elusive homunculus. Journal of Experimental Psychology: Learning Memory and Cognition, 31(4), 683–702. 10.1037/0278-7393.3l4.683 [DOI] [PubMed] [Google Scholar]
  5. Arrington CM, Weaver SM, & Pauker RL (2010). Stimulus-Based Priming of Task Choice During Voluntary Task Switching. Journal of Experimental Psychology: Learning Memory and Cognition, 36(4), 1060–1067. 10.1037/a0019646 [DOI] [PubMed] [Google Scholar]
  6. Braem S (2017). Conditioning task switching behavior. Cognition, 166, 272–276. 10.1016/j.cognition.2017.05.037 [DOI] [PubMed] [Google Scholar]
  7. Bugg JM, & Crump MJC (2012). In Support of a Distinction between Voluntary and Stimulus-Driven Control: A Review of the Literature on Proportion Congruent Effects. Frontiers in Psychology, 3(September), 1–16. 10.3389/fpsyg.2012.00367 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bugg JM, Jacoby LL, & Chanani S (2011). Why it is too early to lose control in accounts of item-specific proportion congruency effects. Journal of Experimental Psychology. Human Perception and Performance, 37(3), 844–859. 10.1037/a0019957 [DOI] [PubMed] [Google Scholar]
  9. Chiu Y-C (2019). Automating adaptive control with item-specific learning Psychology of Learning and Motivation - Advances in Research and Theory (1 st ed., Vol. 71). Elsevier Inc; 10.1016/bs.plm.2019.05.002 [DOI] [Google Scholar]
  10. Chiu Y-C, & Egner T (2017). Cueing Cognitive Flexibility: Item-Specific Learning of Switch Readiness. Journal of Experimental Psychology. Human Perception and Performance, 43(12), 1950–1960. 10.1037/xhp0000420 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chiu Y-C, & Egner T (2019). Neuroscience and Biobehavioral Reviews Cortical and subcortical contributions to context-control learning. Neuroscience and Biobehavioral Reviews, 99, 33–41. 10.1016/j.neubiorev.2019.01.019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Chiu Y-C, Jiang J, & Egner T (2017). The Caudate Nucleus Mediates Learning of Stimulus-Control State Associations. Journal of Neuroscience, 37(4), 1028–1038. 10.1523/JNEUROSCI.0778-16.2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Crump MJC, & Logan GD (2010). Contextual control over task-set retrieval. Attention, Perception & Psychophysics, 72(8), 2047–2053. 10.3758/APP.72.8.2047 [DOI] [PubMed] [Google Scholar]
  14. D’Cruz AM, Ragozzino ME, Mosconi MW, Shrestha S, Cook EH, & Sweeney JA (2013). Reduced behavioral flexibility in autism spectrum disorders. Neuropsychology, 27(2), 152–160. 10.1037/a0031721 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Dreisbach G, & Haider H (2006). Preparatory adjustment of cognitive control in the task switching paradigm. Psychonomic Bulletin & Review, 13(2), 334–338. 10.3758/BF03193853 [DOI] [PubMed] [Google Scholar]
  16. Egner T (2014). Creatures of habit (and control): a multi-level learning perspective on the modulation of congruency effects. Frontiers in Psychology, 5(November), 1–11. 10.3389/fpsyg.2014.01247 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Egner T (2017). The Wiley Handbook of Cognitive Control. (Egner T, Ed.). Chichester, UK: John Wiley & Sons, Ltd. 10.1002/9781118920497 [DOI] [Google Scholar]
  18. Franz VH, & Loftus GR (2012). Standard errors and confidence intervals in within-subjects designs: generalizing Loftus and Masson (1994) and avoiding the biases of alternative accounts. Psychonomic Bulletin & Review, 19(3), 395–404. 10.3758/s13423-012-0230-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Fröber K, & Dreisbach G (2017). Keep flexible – Keep switching! The influence of forced task switching on voluntary task switching. Cognition, 162, 48–53. 10.1016/j.cognition.2017.01.024 [DOI] [PubMed] [Google Scholar]
  20. Goldberg TE, & Weinberger DR (1988). Probing prefrontal function in schizophrenia with neuropsychological paradigms. Schizophrenia Bulletin, 14(2), 179–183. 10.1093/schbul/14.2.179 [DOI] [PubMed] [Google Scholar]
  21. Goschke T (2000). Intentional reconfiguration and involuntary persistence in task-set switching. In Control of cognitive processes: Attention and performance XVIII (pp. 331–355). [Google Scholar]
  22. Hill EL (2004). Executive dysfunction in autism. Trends in Cognitive Sciences, 8(1), 26–32. 10.1016/j.tics.2003.11.003 [DOI] [PubMed] [Google Scholar]
  23. Kiesel A, Steinhauser M, Wendt M, Falkenstein M, Jost K, Philipp AM, & Koch I (2010). Control and interference in task switching--a review. Psychological Bulletin, 136(5), 849–874. 10.1037/a0019842 [DOI] [PubMed] [Google Scholar]
  24. Koch I, & Allport A (2006). Cue-based preparation and stimulus-based priming of tasks in task switching. Memory & Cognition, 34(2), 433–444. 10.3758/BF03193420 [DOI] [PubMed] [Google Scholar]
  25. Leboe JP, Wong J, Crump M, & Stobbe K (2008). Probe-specific proportion task repetition effects on switching costs. Perception & Psychophysics, 70(6), 935–945. 10.3758/PP [DOI] [PubMed] [Google Scholar]
  26. Logan GD, & Bundesen C (2003). Clever Homunculus: Is There an Endogenous Act of Control in the Explicit Task-cuing Procedure? Journal of Experimental Psychology: Human Perception and Performance, 29(3), 575–599. 10.1037/0096-1523.29.3.575 [DOI] [PubMed] [Google Scholar]
  27. Mayr U, & Kliegl R (2003). Differential effects of cue changes and task changes on task-set selection costs. Journal of Experimental Psychology. Learning, Memory, and Cognition, 29(3), 362–372. 10.1037/0278-7393.29.3.362 [DOI] [PubMed] [Google Scholar]
  28. Meiran N (1996). Reconfiguration of processing mode prior to task performance. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22(6), 1423–1442. 10.1037/0278-7393.22.6.1423 [DOI] [Google Scholar]
  29. Miller EK, & Cohen JD (2001). An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24, 167–202. 10.1146/annurev.neuro.24.1167 [DOI] [PubMed] [Google Scholar]
  30. Monsell S (2003). Task switching. Trends in Cognitive Sciences, 7(3), 134–140. 10.1016/S1364-6613(03)00028-7 [DOI] [PubMed] [Google Scholar]
  31. Monsell S, & Mizon GA (2006). Can the task-cuing paradigm measure an endogenous task-set reconfiguration process? Journal of Experimental Psychology: Human Perception and Performance, 32(3), 493–516. 10.1037/0096-1523.32.3.493 [DOI] [PubMed] [Google Scholar]
  32. Moreno-Martínez FJ, & Montoro PR (2012). An ecological alternative to Snodgrass & Vanderwart: 360 high quality colour images with norms for seven psycholinguistic variables. PLoS ONE, 7(5), 34–42. 10.1371/journal.pone.0037527 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Morice R (1990). Cognitive Inflexibility and Pre-frontal Dysfunction in Schizophrenia and Mania. British Journal of Psychiatry, 157(1), 50–54. 10.1192/bjp.157.1.50 [DOI] [PubMed] [Google Scholar]
  34. Reiman KM, Weaver SM, & Arrington CM (2014). Encoding and choice in the task span paradigm. Psychological Research, 79(2), 267–281. 10.1007/s00426-014-0556-6 [DOI] [PubMed] [Google Scholar]
  35. Roberts ME, Tchanturia K, Stahl D, Southgate L, & Treasure J (2007). A systematic review and meta-analysis of set-shifting ability in eating disorders. Psychological Medicine, 37(8), 1075–1084. 10.1017/S0033291707009877 [DOI] [PubMed] [Google Scholar]
  36. Rogers RD, & Monsell S (1995). Costs of a predictible switch between simple cognitive tasks. Journal of Experimental Psychology: General, 124, 207–231. 10.1037/0096-3445.124.2.207 [DOI] [Google Scholar]
  37. Vandierendonck A, Liefooghe B, & Verbruggen F (2010). Task switching: interplay of reconfiguration and interference control. Psychological Bulletin, 136(4), 601–626. 10.1037/a0019791 [DOI] [PubMed] [Google Scholar]
  38. Verguts T, & Notebaert W (2008). Hebbian learning of cognitive control: dealing with specific and nonspecific adaptation. Psychological Review, 115(2), 518–525. 10.1037/0033-295X.115.2.518 [DOI] [PubMed] [Google Scholar]
  39. Waszak F, Hommel B, & Allport A (2003). Task-switching and long-term priming: Role of episodic stimulus-task bindings in task-shift costs. Cognitive Psychology, 46(4), 361–413. 10.1016/S0010-0285(02)00520-0 [DOI] [PubMed] [Google Scholar]
  40. Waszak F, Hommel B, & Allport A (2004). Semantic generalization of stimulus-task bindings. Psychonomic Bulletin & Review, 11(6), 1027–1033. 10.3758/BF03196732 [DOI] [PubMed] [Google Scholar]
  41. Waszak F, Hommel B, & Allport A (2005). Interaction of task readiness and automatic retrieval in task switching: Negative priming and competitor priming. Memory and Cognition, 33(4), 595–610. 10.3758/BF03195327 [DOI] [PubMed] [Google Scholar]

RESOURCES