1. Introduction
Incidental learning is broadly defined as the acquisition of information (sequences, patterns, motor programmes or other regularities) without conscious or explicit awareness of the process of learning or explicit knowledge of the information acquired [
1,
2]. Incidental learning (sometimes termed procedural or implicit) is a ubiquitous foundational cognitive ability thought to support diverse complex functions, including processing and production of language, musical and arithmetical ability, non-verbal social cue encoding and decoding, behavioural insight, acquisition of constructs, and myriad motor skills from juggling to hopscotch [
3,
4,
5,
6,
7]. Explicit learning by contrast is typically amenable to verbal description, accompanied by awareness of the process (the learning episode) and the “product” (the information learned).
While there is general agreement about the ubiquitous nature of incidental learning there is debate about the contribution of explicit processes [
8,
9,
10], specifically the role of attention and working memory, to incidental learning. Some researchers have proposed that working memory and attention resources are crucial to incidental learning [
11,
12,
13,
14], others have suggested minimal contribution [
15,
16,
17,
18], or that learning proceeds entirely automatically [
19].
Hsiao and Reber [
20] observed that most models of incidental sequence learning incorporate (usually tacitly), a “limited-capacity short-term memory system” (p. 341). The relative contribution of memory processes to incidental learning is typically investigated by either comparing working memory capacity with learning score, comparing learning performance of high and low memory span groups, or by comparing learning on an incidental learning task with no secondary load (single-task design) to learning with the presence of secondary memory/attention tasks (dual-task design). Methods derive from the assumption that working memory capacity constrains learning on tasks that load the same cognitive processes; working memory capacity should either correlate with incidental learning score, and/or secondary tasks should load working memory and diminish incidental learning score, if completion of both tasks depends upon the same cognitive processes. The standard approach utilises some variant of a standard Serial Reaction Time (SRT) task to measure incidental sequence learning and often includes a concurrent auditory tone-counting task as the secondary memory task/attentional load [
1,
11,
14,
21,
22,
23,
24,
25].
In the canonical version of the SRT task visual stimuli follow a repeating sequence and participants respond to the position of the target stimulus on screen by pressing a corresponding keypad as quickly as possible [
1,
11]. Reaction time (RT) responses are recorded and learning is shown by faster responses to sequence compared to non-sequenced stimulus learning trials. Early research suggested that the effect is robust across single tasks but may be diminished or abolished entirely with the introduction of a secondary concurrent tone counting task [
11,
23]. Effects of secondary tasks on incidental learning performance are variously thought to occur due to a processing bottleneck when separate motor responses are required for both tasks, interference effects in working memory, central resource (attention) capacity constraints or temporally asynchronous dual-task stimuli, depending upon theoretical position and experimental design [
11,
17,
21,
22,
23,
24,
25,
26]. Additionally, there is huge variability across dual-task design in different studies. In many studies secondary stimuli are embedded in some way within the primary task (either temporally or visually depending on the secondary task) and occur either before or after sequence trials (the convention with secondary memory tasks), or between Stimulus-Response of sequence trials (secondary auditory tone counting tasks), with few exceptions [
10]. However, in designs where secondary stimuli are not synchronous with primary stimuli it is possible for participants to switch attention between two sequential task stimuli rather than share memory and attentional resources across tasks. Synchronicity, on the basis of the current study and other work [
10], refers to dual-task designs whereby secondary task stimuli constitute both a primary sequence task trial and crucial secondary task target stimuli. This design is more likely to require allocation of shared resources across tasks if both tasks recruit the same cognitive processes.
What remains to be determined is whether key component functions of incidental learning are implicit/explicit, visual/spatial or a combination of visual and spatial and whether such functions load working memory resources since most dual-task designs have used auditory tone counting rather than working memory secondary task stimuli. This knowledge may lead to a better understanding of the underlying cognitive processes responsible for incidental learning and provide important insights into the organization of cognitive architecture governing complex behaviour.
Barker and Andrade [
27] proposed that incidental learning occurs due to an associative learning mechanism that automatically detects environmental co-variations resulting in processing fluency for sequential patterns and/or reliably co-varying stimuli. This hypothesis tacitly assumes a negligible role of working memory to incidental learning although studies have yielded mixed results. Frensch and Miner made similar assumptions but proposed a role of short-term memory for encoding sequence elements [
19]. However, Unsworth and Engle [
28] found that working memory span (high
versus low) was uncorrelated with incidental sequence learning and positively correlated with explicit sequence learning, suggesting that incidental learning does not depend upon working memory processes.
Song, Marks, Howard and Howard [
29] tested healthy older adults on a cued variant of a probabilistic SRT task to distinguish effects of explicit learning on incidental learning from effects on motor performance. The authors argued that reported variability in incidental learning ability in groups with reduced memory capacity (stroke patients and healthy elderly [
30,
31]) reflected effects of explicit learning, rather than diminished incidental learning, on motor performance. Findings showed that incidental sequence learning was not influenced by concurrent explicit learning in older adults, and was not correlated with working memory span, although explicit learning did negatively impact motor performance. Similarly, Feldman, Kerr and Streissguth found that incidental learning on an SRT task was not significantly correlated with reasoning, processing speed and working memory tasks, whilst explicit learning measured by a generation task correlated reliably with most measures [
32]. Finally, Remillard [
18] reported that the incidental sequence learning mechanism operates over a range of at least seven sequence elements exceeding the purported capacity of explicit spatial working memory. Remillard concluded that incidental sequence learning is not bound by the capacity limits of working memory [
18].
In contrast, Bo, Jennet and Seidler [
12] found that visual and verbal working memory capacity correlated with rate of reaction time change on a SRT task suggesting a relationship between working memory and incidental sequence learning. They also found that visuospatial working memory ability explained a significant portion of the variance in rate of SRT performance change across individual participants. The authors concluded that working memory processes make a key contribution to incidental learning on SRT tasks. Similarly, Barker
et al. found that some patients with frontal brain injury (the region thought to subserve working memory) and visual memory deficits were impaired on an incidental SRT task compared to matched controls [
2]. However findings were not clear-cut as patients with intact visual memory (but other cognitive deficits) also showed diminished incidental learning compared to controls hinting that functions separate from visual memory also contribute to incidental learning performance.
Hsiao and Reber [
20] manipulated RSOA (response-secondary stimulus onset asynchrony) in a series of dual-task experiments. The authors posited that any account of the varying impact of RSOA on sequence learning must include a working memory component that sequentially codes stimulus representations in a time-locked manner, so that incoming stimulus items will override previous representations if they overlap and so compromise sequence learning. Frensch, Lin and Buchner [
24] similarly proposed that dual-task interference effects occur because secondary tasks disrupt the availability of successive sequence information through “interference mechanisms” in short-term memory based on findings that learning is diminished when secondary task inter-stimulus intervals increase. There is also some evidence of an interaction between age, incidental learning and memory/attentional resources with a secondary task load impeding incidental learning in an elderly group compared to a younger group thought to occur due to age-related attenuation of working memory capacity [
33].
Other researchers investigating the putative contribution of working memory processes to SRT learning have adapted the standard dual-task approach (secondary task = an auditory tone counting task), by embedding visual and verbal memory task stimuli before, after or between primary SRT task stimuli. Stadler [
16] presented participants with a series of letters (5, 7 or 9) for a duration of 5 s preceding a SRT task and found that memory load (letters accurately recalled after RT trials) did not impede sequence learning. Heuer and Schmidtke [
25] presented Brooks visuo-spatial or Brooks verbal task stimuli [
34] before SRT task stimuli; items were recalled 90 s later after the SRT task block was completed. They found no interference effects of secondary memory tasks on learning or the expression of learning, although a secondary auditory go/no go task did impede sequence learning. The authors concluded that secondary task interference is not capacity-based but occurred for the go/no go condition via disruption of contiguous elements of the sequence. These studies devised dual-task designs whereby secondary memory stimuli were not synchronous with primary SRT task stimuli, although they did require representations to be maintained in memory during performance of the sequence task. In addition, both studies reported almost 100% accuracy scores on the secondary tasks suggesting possible ceiling effects.
On the basis of earlier findings of secondary task effects (auditory tone counting, “go-no go”), and no effect of secondary (embedded memory) tasks on SRT learning, Schmidtke and Heuer [
35] predicted that synchronous primary and secondary task stimuli should result in unimpeded learning on both tasks. They concluded that task integration and task-relevance of secondary stimuli was a critical component of whether secondary stimuli became integrated with the primary task to produce unimpeded learning. The authors found that when the secondary task consisted of a random series of tones sequence learning was impeded whilst a repeated sequence of tones that contained the same number of elements as the visual task resulted in higher learning scores than for the single-task alone. They concluded that task integration accounted for these data and that the task-relevance of secondary stimuli was a critical component of whether secondary stimuli became integrated. When tones were presented without instruction the secondary sequence was only minimally learned suggesting that mechanisms of implicit learning are not wholly non-attentional.
Schumacher and Schwarb [
28] investigated the effects of secondary task load on incidental sequence learning in an experiment where the secondary tone identification task was presented concurrently (occurred simultaneously with the visual SRT task stimulus), or occurred after a brief stimulus onset asynchrony (SOA) interval. Surprisingly, they found impaired dual-task learning in the first condition but not the second and concluded that simultaneous dual-task processing that requires response selection disrupts sequence learning as opposed to attentional switching between two tasks. Of note, the authors proposed that secondary task stimuli were “synchronous” with primary task stimuli in these experiments because they co-occurred with visual sequence trials. Secondary task stimuli were not synchronous in the sense that they simultaneously represented a sequence trial maintaining the temporal integrity of the sequence, and also represented secondary task target stimuli. Similarly, secondary task-stimuli were presented aurally not visually as with primary visual SRT task stimuli, and presentation rate varied to correlate or not with the primary SRT task. So it is possible that secondary task effects on sequence learning might occur due to perceptual integration errors between incoming visual and auditory stimuli rather than as a result of competing response selection or shared limited capacity resources across tasks. Importantly, Schumacher and Schwarb [
28] reported that 15 of 21 dual-task SRT studies utilising auditory tone counting as the secondary task showed dual-task learning with only six studies showing dual-task interference effects. Thus, the weight of evidence suggests that SRT task learning is not impeded by a secondary auditory tone counting task, or secondary memory tasks that are embedded within, but not synchronous with primary SRT task stimuli. There is variability however across studies between presentation rates, mode of instruction, inter-trial interval, and modality of secondary task that might somewhat account for contrasting findings.
Other researchers have developed dual-task designs whereby secondary task stimuli appear as targets for reaction time responses in the primary incidental task, and simultaneously constitute target stimuli for the secondary learning or memory task [
10]. Jiménez and Méndez [
10] assessed the differential effects of incidental and explicit “shape” learning on a probabilistic (grammar learning) SRT task using synchronised secondary stimuli. The authors relied upon the same four stimuli to specify the location of the sequence trials, to predict the next position of the sequence (on a shape learning task) and to comprise the secondary counting task. Findings showed that incidental learning was acquired and expressed regardless of the influence of explicit knowledge. However, secondary tasks were not designed to access specific verbal and visual working memory resources although explicit shape learning did require visual processing concurrent with incidental SRT task learning. In other work, Jiménez and Vázquez [
17] reported that interference effects of secondary tasks diminished when a secondary auditory tone task co-varied with the location of primary task stimuli. These findings suggest that when secondary tasks are synchronous with [
10], or reliably co-vary with the primary task, learning on incidental and explicit tasks may proceed undiminished in line with Barker and Andrade’s [
27] and Schmidtke and Heuer’s assumptions [
35].
Finally, there are conflicting data on the role of attentional resources separate from memory processes to multiple contingency learning in incidental learning paradigms. Rowland and Shanks [
13] found that learning of a secondary sequence in a dual-task SRT design was abolished under conditions of high perceptual load suggesting attentional limitations on learning of multiple contingencies, though there was evidence of learning across both tasks when perceptual load was low. However, in this study the secondary task constituted a statistically independent secondary sequence task, appearing at one of four locations above the visual array (four locations) for the primary sequence. The authors concluded that it remains to be established whether learning of multiple contingencies is incidental or explicit and whether findings reflect effect of secondary load on incidental or explicit learning, or a combination. Emberson, Conway and Christiansen [
14] also proposed that directed attention is a prerequisite for incidental learning because they found no evidence of learning in any unattended information stream regardless of stimulus modality or presentation timing. However, it is generally accepted in incidental learning paradigms that a stimulus array must be attended too, even in cases where allocation of attentional resources is covert, for learning to occur [
27]. In contrast, other work has shown that dual-motor sequences can be acquired without significant interference effects between tasks [
36].
To conclude, incidental learning is a robust phenomenon when measured by single-task SRT tasks. Dual-task designs have been employed to investigate the putative contribution of attention and working memory to incidental and secondary task learning with the typical convention of adopting an auditory tone-counting task as the secondary task. One problem is that whilst a secondary auditory task will load attention and to some extent verbal working memory, it is unlikely to load visuospatial processes. Studies investigating the contribution of visual and verbal memory processes to incidental learning typically present secondary stimuli before, within (inter-trial stimulus) or after the primary SRT task stimuli [
11,
16,
19,
23,
24,
25], a convention that is also prevalent when a secondary auditory task is used, rather than present secondary stimuli synchronous with primary task stimuli. Findings generally show no effect of secondary tasks on memory although there is evidence from correlational studies that working memory processes may contribute to incidental learning. Other studies have developed novel dual-task designs that embedded and synchronised visual stimuli [
10], or presented auditory stimuli that co-varied with sequence stimuli [
17,
28], and showed conflicting findings that may reflect perceptual modality of secondary task used or some other aspect of experimental design. Importantly, in these studies working memory tasks did not comprise the secondary load. In light of these mixed findings the contribution of explicit visual/verbal and visuospatial working memory processes to incidental SRT task learning, with secondary task stimuli embedded within and synchronised with the primary task, remains to be determined.
The present study addressed the possible contribution of explicit working memory processes to incidental learning in a novel design comparing learning on a single- compared to dual-task SRT with verbal and visual memory tasks as secondary task stimuli [
1,
11]. In the present study a conventional SRT was replicated from Seger’s [
1] study for the single-task learning condition and modified for dual-task conditions. Previous data show robust learning in healthy controls on the selected SRT task indicating that it reliably captures incidental learning [
1,
2,
3,
4,
5]. Results of the present study should be comparable to other dual-task studies and may help resolve some of the conflicting findings.
The present experiment had seven conditions; a single-SRT task, Verbal dual-task, Verbal-Spatial Above, Verbal-Spatial Below, Visual dual-task, Visual-Spatial Above and Visual-Spatial Below. In Visual and Verbal dual-task conditions secondary visual or verbal memory stimuli appeared at screen locations in the primary SRT in place of the stimulus circle for approximately half of stimulus trials. Thus, secondary memory task stimuli were embedded and synchronous with the primary SRT task stimuli simultaneously constituting a SRT trial requiring a RT response and target stimuli for the secondary memory task similar to the Jiménez and Méndez study [
10]. In the Verbal-Spatial and Visual-Spatial Above conditions secondary memory task stimuli appeared at sequence locations in place of the SRT stimulus circle but were spatially decoupled from the sequence by appearing above the locus of the sequence array where the circle stimuli appeared (that were not replaced by icons/digits). Thus, secondary task stimuli were embedded and synchronous with the primary SRT (appearing in place of a circle and requiring a RT response and thus following the sequence) but were
spatially displaced to evaluate the putative contribution of spatial processing to incidental learning separate from visual and verbal memory processes. Verbal-Spatial Below and Visual-Spatial Below conditions again presented secondary stimuli that were temporally embedded and synchronous with the primary task, but were spatially decoupled, appearing below the screen location where the stimulus circle would normally appear. Visual-Spatial and Verbal-Spatial Below conditions were thought likely to snare spatial attention more effectively than “Above” conditions because scanning below a horizontal visual array is more demanding than diverting attention above a horizontal visual array [
37]. It was anticipated that secondary visual memory tasks should disrupt learning more than verbal tasks even though the verbal task was presented visually. Concomitantly, Visual-Spatial and Verbal-Spatial Above and Below conditions were expected to make increased demands on attention and memory processes compared to the Single- and Visual and Verbal Dual-tasks, because the participant had to divert attention above or below the central visual array where target circles appeared to process secondary icons/digits. Thus incidental and explicit learning score were expected to incrementally decrease from Single-, through Verbal-Dual, Verbal-Spatial A, Verbal-Spatial B, Visual-dual, Visual-Spatial A and Visual-Spatial B conditions if secondary task load made increased demands on memory/attentional processes across conditions and recruited the same resources required for incidental and explicit sequence learning. In the present study incidental sequence learning, explicit sequence learning and acquisition of the secondary learning/memory task proceed concurrently with primary and secondary tasks simultaneously recruiting attentional and memory processes representing a significant departure from conventional dual-task designs.
3. Results
Mean incidental learning scores, explicit learning scores and percent accuracy scores were calculated for each condition (see
Table 2). Percentage accuracy scores, rather than overall accuracy score, were used in subsequent analyses because this value was considered a more accurate and fine-grained representation of participants’ performance than total accuracy score. Mean percent accuracy scores were calculated for each participant for each digit value response in Verbal/Verbal-Spatial A and B conditions and for mean overall accuracy in the Visual/Visual-Spatial A and B conditions as follows: Verbal experiment—where “a” signifies participant’s response for digits counted across block (23 in this example) and “b” represents the correct answer (26), the participant’s response is 88% accurate:
In circumstances where there were false positives (the correct answer was 26 and the participant reported 27), the false positive (one value) was deducted from the correct total counted (26 in this example) to produce an adjusted response of 25. Percentages were combined for the seven acquisition blocks and averaged to produce a mean percentage accuracy score for each participant across blocks. The same procedure was followed for the Visual experiment where “a” represented the number of correct yes/no responses and “b” represented total possible number of correct responses (n = 7).
Table 2.
Descriptive data for incidental learning, explicit learning and percentage accuracy scores × condition.
Table 2.
Descriptive data for incidental learning, explicit learning and percentage accuracy scores × condition.
Condition | Incidental learning scoreMean (SD) | Explicit learning scoreMean (SD) | Digits counted: % accuracyMean (SD) | Icons identified: % accuracyMean (SD) |
---|
Single-task | 83.4 (49.6) | 13.8 (4.4) | | |
Verbal dual-task | 64.9 (33.8) | 7.5 (3.8) | 94.7 (4.9) | |
Verbal-Spatial A | 77.3 (53.1) | 4.5 (2.3) | 93.7 (7.7) | |
Verbal-Spatial B | 62.1 (40.8) | 5.6 (3.1) | 89.7 (11.8) | |
Visual dual-task | 73.5 (36.6) | 8.3 (3.3) | | 85.6 (11.4) |
Visual-Spatial A | 71.9 (39.7) | 3.8 (2.3) | | 84.9 (15.7) |
Visual-Spatial B | 64.8 (26.1) | 3.9 (1.8) | | 82. 8 (15.8) |
Total group | 71.1 (39.9) | 6.7 (3.0) | 92.7 (8.1) | 84.4 (13.5) |
Descriptive data presented in
Table 2 show that incidental learning scores were similar across conditions and did not appear to incrementally diminish from Single- to Visual-Spatial B conditions as predicted on the basis of increasing demands on shared processes across primary SRT and secondary tasks. The Single-task condition mean incidental learning score was greater than incidental learning mean scores across conditions but standard deviations were also large across groups. Verbal-Spatial A condition had a relatively greater mean incidental learning score than other dual-task conditions, but the standard deviation value was also greater for this compared to other dual-task conditions arguably indicating greater variability across scores for Verbal-Spatial A compared to other conditions. A One-Way ANOVA conducted with learning score as the dependent variable and condition the between-groups factor showed that incidental learning score did not differ as an effect of condition
F(6,133) = 0.71,
p = 0.64 two-tailed. There was no significant linear trend for incidental learning score
t(1,133) = 0.82,
p = 0.36 two-tailed, indicating that learning score did not incrementally diminish from the Single-task condition to Verbal dual-, Verbal-Spatial A, Verbal-Spatial B, Visual dual-, Visual-Spatial A and Visual-Spatial B conditions as expected if primary SRT and secondary memory task performance depended on shared cognitive resources.
Explicit learning/awareness scores presented in
Table 2 indicated a general trend of incremental diminution as an effect of condition as predicted. Results of a One-Way ANOVA showed that explicit scores differed significantly as an effect of condition
F(6,133) = 25.7,
p ≤ 0.01 two-tailed. Results of Tukey’s
post-hoc comparisons (corrected for multiple analyses) showed that explicit scores in the Single-Task (
M = 13.8, 95% CI [11.7, 15.9]) condition differed significantly from Verbal dual condition scores (
M = 7.5, 95% CI [5.7, 9.3])
p ≤ 0.01, Visual dual condition scores (
M = 8.3, 95% CI [6.7, 9.8])
p ≤ 0.01, Verbal-Spatial A condition scores (
M = 4.5, 95% CI [3.3, 5.5])
p = 0.00, Verbal-Spatial B condition scores (
M = 5.6, 95% CI [4.1, 7.1])
p ≤ 0.01, Visual-Spatial A condition scores (
M = 3.8, 95% CI [2.7, 4.9])
p ≤ 0.01, and Visual-Spatial B condition scores (
M = 4.1, 95% CI [3.3, 4.9])
p ≤ 0.01. In sum, Single-task explicit scores differed significantly from the explicit mean scores of each other experimental condition. There were also differences between explicit learning/awareness scores for dual-task conditions. For the Verbal dual-task condition, mean explicit score differed significantly from Verbal-Spatial A
p = 0.03, Visual-Spatial A,
p ≤ 0.01, and Visual-Spatial B conditions,
p = 0.01, but not for the Visual dual-task,
p = 0.99, or Verbal-Spatial B conditions,
p = 0.44. Similarly, the Visual dual-task explicit scores differed significantly from Verbal-Spatial A,
p ≤ 0.01, Visual-Spatial A,
p ≤ 0.01 and Visual-Spatial B conditions,
p = 0.01, but not Verbal-Spatial B,
p = 0.10. There were no other differences. These findings indicate that spatially asynchronous conditions (with the exception of Verbal-Spatial B) made greater demands on resources required for expression of explicit sequence learning than Visual and Verbal dual-task conditions comprised of spatially embedded and temporally synchronous secondary stimuli. The lack of difference between Visual/Verbal dual-task and Verbal-Spatial B conditions might be due to participants’ attending less to secondary task stimuli; descriptive percentage accuracy scores indicated that participants were least accurate in the Verbal-Spatial B condition of all Verbal conditions. Hence, there may have been some trade-off between performance accuracy, incidental learning and explicit learning in this condition. Descriptive data also indicated that Visual dual-task conditions were generally more difficult than Verbal conditions based on mean accuracy scores, which is not surprising since the SRT task is primarily a visuospatial task.
Linear contrast analyses were conducted to investigate whether the suggested trend across conditions in explicit awareness descriptive data was a significant effect of secondary task performance on explicit learning. Contrast data were input in the same order as for other analyses to establish whether a significant linear trend, not seen in incidental learning data, was present in explicit learning data supporting a proportionate effect of condition on cognitive resources. Analyses revealed a significant trend for explicit scores to incrementally decrease from the Single- to Verbal dual-, Verbal-Spatial A, Verbal-Spatial B, Visual dual-, Visual-Spatial A and Visual-Spatial B conditions, t(1,133) = 6.77, p ≤ 0.01 two-tailed. However, despite the significant trend descriptive data showed similar explicit mean learning scores for Visual- and Verbal-dual tasks, suggesting that these conditions were more similar in terms of secondary task effects on explicit learning than Verbal- and Visual dual-tasks were to Verbal- and Visual-Spatial conditions.
Overall group means presented in
Table 2 show that participants were 84.4% accurate overall for Visual conditions and 92.7% accurate overall for Verbal conditions. Results of a One-way ANOVA showed no effect of condition on percentage accuracy of secondary task performance for Verbal conditions,
F(2,57) = 1.9,
p = 0.15 two-tailed. Similarly, there was no effect of condition on percentage accuracy of secondary task performance,
F(2,57) = 0.44,
p = 0.64 two-tailed for Visual conditions. However, descriptive data indicated that performance accuracy for the secondary task was greatest for the Verbal dual-task condition compared to other conditions and that performance accuracy for the Visual-Spatial B condition was least accurate, although incidental learning scores were similar for both conditions. These data support the assumption that Visual-Spatial B condition should make most demands on resources required for explicit learning and secondary memory task performance during concurrent SRT learning compared to other conditions. Consequently, accuracy data for Visual and Verbal dual-task conditions were compared in a One-Way ANOVA with linear contrasts to establish whether accuracy data followed a significant trend corresponding to explicit data potentially revealing an explicit learning/accuracy trade-off. Analyses revealed that condition had a significant effect on accuracy
F(5,114) = 3.37,
p ≤ 0.01 two-tailed.
Post hoc analyses corrected for multiple comparisons revealed that Verbal dual-task condition accuracy score (
M = 94.7, 95% CI [92.4, 97.0]) was significantly greater than Visual-Spatial B accuracy score (
M = 82.8, 95% CI [75.4, 90.2]),
p = 0.02 two-tailed, and Verbal-Spatial A condition accuracy scores (
M = 93.7, 95% CI [90.1, 97.93]) condition accuracy scores were also marginally significantly greater than Visual-Spatial B condition scores,
p = 0.05 two-tailed. Notably, percentage accuracy scores significantly and proportionately decreased from Verbal dual-, to Verbal-Spatial A, Verbal-Spatial B, Visual dual-, Visual-Spatial A and Visual-Spatial B conditions
t(1,114) = 16.2,
p ≤ 0.01 two tailed corresponding to results of analyses for explicit data across conditions.
Overall results of analyses showed no significant effect of secondary task load on expression of incidental learning, but significant effects of condition on expression of explicit learning and percentage accuracy scores for secondary working memory tasks that followed a similarly significant linear trend across conditions.
4. Discussion
In the present experiment there was one Single- and six dual-task conditions designed to increase in difficulty from Single-task through Verbal dual-, Verbal-Spatial A, Verbal-Spatial B, Visual dual-, Visual-Spatial A and Visual-Spatial B condition order based on assumed attentional demands of secondary tasks. For Verbal and Visual conditions secondary visual icons and verbal digits were synchronous with the primary SRT task by constituting a sequence trial in place of a stimulus circle for approximately half of trials across acquisition blocks and also comprising target stimuli for visual/verbal secondary working memory tasks. However, in Visual-Spatial and Verbal-Spatial conditions icons and digits were temporally synchronous with the primary SRT task but were spatially asynchronous (decoupled) from the central sequence circle stimulus array. It was anticipated that Visual dual-task conditions would recruit resources required for expression of SRT and explicit sequence task learning more than Verbal dual-tasks because the SRT task is primarily a visuospatial motor task. It was also anticipated that secondary spatial conditions would have a greater effect on sequence learning than Visual and Verbal conditions, because spatial displacement of secondary task stimuli necessitated gaze shift away from the central SRT stimulus circle array, with Visual-Spatial B making the greatest demands on memory/attentional processes compared to other conditions during concurrent SRT task learning.
Results showed that the expression of incidental learning was not different as an effect of condition. Single-condition mean incidental learning score was greater than for dual-task conditions, but standard deviations were also relatively large across conditions. Analyses revealed that incidental learning scores did not follow an anticipated significant linear trend across conditions, and there was no evidence of a proportionate diminution in learning score through Single- to Visual-Spatial B as anticipated. Although there were only twenty participants in each condition, this task has been shown to be sensitive to differences in amount of learning in small groups across studies (2–5), and the overall cohort was relatively large making it unlikely that null effects represent small sample size, particularly since there were significant differences for other variables as an effect of condition. In contrast, explicit sequence learning/awareness scores were significantly greater for the single-task compared to dual-task conditions with explicit learning means following a significant linear trend from the Single- through to Visual-Spatial B condition, as anticipated if secondary load made increasing demands on the expression of explicit learning as an effect of condition. Interestingly, secondary task accuracy scores also followed a similar significant linear trend across conditions from Single- through to Visual-Spatial B conditions corresponding to the pattern of explicit data. These results indicate an accuracy/learning trade-off for the expression of explicit sequence learning in the present experiment that was not present for incidental sequence learning data. Single-task explicit scores differed significantly from the explicit mean scores of each other experimental condition. There were also differences between explicit learning/awareness scores for dual-task conditions. Results indicated that spatially asynchronous conditions (with the exception of Verbal-Spatial B) made greater demands on expression of explicit sequence learning than Visual and Verbal dual-task conditions comprised of spatially embedded and temporally synchronous secondary stimuli.
Present findings concur with other studies showing minimal effect of working memory secondary tasks on incidental learning, although there is broad variability across study design [
18,
28,
29,
36]. Gobel, Sanchez and Reber recently found that explicit scores dissociated from overall sequence learning ability in two experiments, similar to present findings, and concluded that incidental sequence learning proceeded implicitly [
22]. Other findings have shown that two concurrent implicit tasks can be acquired without competition for cognitive resources [
17].
Previous studies have typically presented secondary memory or tone counting tasks before, after or between incidental sequence stimulus trials [
11,
14,
15,
16,
21,
22,
23], with only a few reported exceptions [
11,
17]. In this way stimuli can be considered “embedded” within the sequence but asynchronous with the sequence. That is, separate responses are required for primary and secondary task trials, and/or secondary stimuli do not simultaneously comprise a component of the primary sequence task. Several findings have revealed that pairing a SRT task with a secondary tone-counting task results in diminished incidental sequence learning [
11,
13,
41]. However, reasons for this remain unclear though it has been hypothesised that secondary task trials might impact on incidental learning by simultaneously drawing on a limited capacity attentional resource [
12,
13,
19,
26], interfering with the automatic maintenance of sequential elements in working memory [
24], producing a processing bottleneck because separate motor responses must be generated for each task stimulus [
21], or by disrupting integrated elements of the sequence [
39]. Present findings show that secondary visual, verbal, visual spatial and verbal spatial conditions did not disrupt the expression of incidental learning inconsistent with
attentional and
working memory interference hypotheses, although these theories provide plausible explanations for dual-task effects on explicit learning and secondary task performance accuracy.
The reported significant incremental decrease in explicit learning and performance accuracy scores across conditions in the present study shown by corresponding linear contrast data results for explicit learning and accuracy data, suggests that processes required for deliberative encoding (participants were instructed to memorize or count secondary stimuli), maintenance, recollection and recall of stimulus information are crucial to the expression of explicit sequence learning but less important for incidental learning. Jiménez and Vázquez similarly found that a SRT and contextual cueing task could be acquired concurrently without diminution of learning on either task so long as learning on both tasks remained incidental [
17]. Inter-task interference only occurred when sequence information became explicit resulting in diminution of the expression of contextual cue learning not SRT task learning. Present findings also indicate an explicit learning/secondary memory task trade-off supporting the assumption that explicit learning depends upon finite resources and that incidental learning may proceed independently and relatively automatically.
Results of the current study also concur with data showing that when a single motor response is required for primary and secondary stimuli (because secondary stimuli are embedded within and synchronous with the primary task), incidental learning is unaffected by the secondary task [
10,
17]. Considered together these findings provide indirect support for the processing
bottleneck theory of dual-task effects. Schumacher and Schwarb demonstrated successful learning under various dual-task conditions but found that sequence learning was diminished when parallel responses were required for both tasks [
21]. In contrast, sequence learning proceeded normally under serial response selection conditions. The authors argued that findings conflict with attentional [
12,
13,
19,
26], automatic [
24], dual-learning mechanism [
23], and integrative [
35,
39] hypotheses, proposing instead that findings occurred due to overlap of central processes involved in successful dual-task performance. However the processing bottleneck theory could arguably be accused of suffering from similar limitations to other attentional and/or limited capacity resource explanations, because the theory is conceptually similar to these approaches. Cheyne, Ferrari and Cheyne [
42] have suggested an alternative explanation for effects of parallel response selection on expression of incidental learning. They used MEG (magnetoencephalography) techniques to track the time course of neural activity in frontal and motor regions thought to reflect conscious and controlled inhibition of pre-potent responses and selection of alternate responses to an infrequent switch cue. The authors concluded that automatic (implicit/incidental) and controlled processes engage in parallel during rapid motor response tasks, and that strength and timing of these processes potentially underlies optimal task performance. Gobel, Sanchez and Reber similarly found that when precise timing is necessary for task performance, sequence learning depends on an integrated representation of sequential action and interaction timing information, indicating the importance of both timing (temporal information) and stimulus integration [
22]. Other studies adopting similar embedded/synchronous dual-task designs to the one presented here have shown that when tones co-vary with sequence locations the disruptive effects of secondary auditory tone counting on expression of incidental learning decreases [
17,
35], again suggesting that stimulus integration abolished dual-task effects on incidental learning. However, in support of the “bottleneck” hypothesis, McBride
et al. reviewed evidence showing significant overlap between brain regions active during consciously and unconsciously triggered action control [
43]. These findings provide a potential neural analogue to Schumacher and Schwarb’s notion of central cognitive resource overlap during parallel motor response selection in dual-SRT tasks [
21].
Importantly, spatially asynchronous stimuli similarly did not impede the expression of incidental learning in the present study. This may have been because secondary stimuli were embedded within and
temporally synchronous with the primary task across conditions regardless of spatial synchronicity or asynchronicity, simultaneously constituting a sequence element
and secondary task target. However, contrasting with this explanation Heuer and Schmidtke found that embedded but
temporally asynchronous visual and verbal secondary tasks also did not disrupt incidental learning, although sequential presentation of primary and secondary task stimuli may have enabled participants to switch attention between stimulus items in their study [
25]. Interestingly, a go/no go task requiring a foot pedal response interfered with primary sequence learning, again supporting the notion that conflicting parallel motor responses diminish SRT task learning [
21]. Present findings indicate that expression of incidental sequence learning is robust to spatial displacement of sequence trials, even when those sequence elements require visuospatial processing to meet secondary task requirements. Thus, spatially displaced secondary targets do not disrupt the expression of incidental sequence learning so long as primary and secondary stimuli are temporally synchronous requiring only a single motor response to both stimulus targets.
On the basis of current and previous findings, degree of synchronicity of primary and secondary task stimuli seems an important determinant of whether dual-task effects are seen when there are no competing motor responses for task completion. Results reported here suggest that timing/temporal ordering of learning trials may be a crucial, and likely automatic process contributing to incidental motor learning above the contribution of memory and attentional processes. Future work might use embedded and synchronous stimuli but specify a different or multiple competing response(s) to some element of secondary stimuli to explore this possibility further. Additionally, it is likely that several other boundary conditions govern how attention is allocated to particular task-stimuli and the capacity restraints of these resources. Task-relevance and intentional stance may be important factors, although Jiang and Leung showed that even task-irrelevant stimuli were “attended to” evidenced by facilitation effects on visual search when previously irrelevant cues became relevant after a switch [
44]. Their findings suggest that attentional filtering occurs late in the learning process. In contrast, Schmidtke and Heuer found that when secondary tones were presented without instruction the secondary sequence was only minimally learned indicating that mechanisms of incidental learning are not wholly non-selective [
35]. Similarly Knee, Thomasen, Ashe and Willingham [
45] investigated the possibility that explicit sequence learning depends upon acquisition of stimulus locations, and incidental learning depends upon acquisition of motor sequences. This hypothesis goes some way to explain current findings of linear contrasts that visuospatially asynchronous conditions were more difficult than synchronous conditions, and had greater impact on explicit sequence learning. Knee
et al. [
45] have proposed that explicit learning in the SRT task is visuospatial in nature, whereas incidental learning is not. One possible way to provide further support for this explanation might be to modify the current experiment to include motor sequence and visuospatial sequence transfer blocks.