Introduction
Even though we are able to perform with ease most everyday tasks that require asynchronous bimanual coordination, such as driving our car or ironing our shirts, experimental studies have repeatedly and consistently demonstrated that there are temporal as well as spatial coordination constraints between the two hands (for review see, Swinnen & Wenderoth,
2004). For example, Kelso, Southard, and Goodman (
1979) showed that even when the two hands have to perform movements of varying difficulty and to different positions in space, participants show a strong tendency to initiate and terminate both movements at the same time (but see also, Fowler, Duck, Mosher, & Mathieson,
1991; Marteniuk, MacKenzie, & Baba,
1984). In addition to these temporal constraints, limitations to produce independent bimanual hand movements can also be observed in the spatial domain (e.g., Franz,
1997; Franz, Zelaznik, & McCabe,
1991; Kelso, Putnam, & Goodman,
1983; Spijkers & Heuer,
1995; Spijkers, Heuer, Kleinsorge, & van der Loo,
1997). For instance, Franz et al. (
1991) observed spatial assimilation effects when the hands had to produce simultaneous asymmetric movements such as drawing a line with the left hand and a circle with the right hand (see also, Albert & Ivry,
2009). Furthermore, when movements are spatially incongruent (different amplitudes and/or different directions), reaction times (RTs) are usually prolonged indicating that processing times increase when the movements become more complex (Heuer,
1986; Spijkers et al.,
1997). The observation that there are general RT costs for planning hand movements with asymmetric amplitudes (or directions) is often referred to as the spatial interference effect. However, there is considerable debate about the underlying mechanism(s) of this phenomenon (e.g., Blinch et al.,
2014; Diedrichsen, Hazeltine, Kennerley, & Ivry,
2001; Hazeltine, Diedrichsen, Kennerley, & Ivry,
2003; Heuer & Klein,
2006; Spijkers et al.,
1997).
One suggestion has been that the interference effects in bimanual coordination tasks are likely to be caused by some kind of neural cross-talk (Marteniuk et al.,
1984; Spijkers & Heuer,
1995; Swinnen & Walter,
1991). More specifically, Heuer and colleagues argued that longer reaction times observed when different movement amplitudes have to be specified for the two hands (as compared to identical movement amplitudes) can be attributed to transient coupling during the movement programming phase (e.g., Heuer,
1993; Heuer, Spijkers, Kleinsorge, van der Loo, & Steglich,
1998; Spijkers et al.,
1997). According to this transient cross-talk hypothesis, mutual inhibition occurs during the movement programming phase when distinct movement parameters have to be specified for the two hands simultaneously (Spijkers et al.,
1997; Spijkers, Heuer, Steglich, & Kleinsorge,
2000). If participants have the opportunity to prepare their movements in advance such that movement parameter specification no longer needs to occur during the RT interval, then RTs no longer differ between symmetric and asymmetric movements (Spijkers et al.,
1997). To put it simply, according to this hypothesis prolonged RTs for asymmetric movements are caused by increased processing demands during response programming. Once movement programming for both hands is finished, no further cross-talk is assumed to happen (see also Schmidt,
1975, generalised motor programing theory).
However, the hypothesis that the RT costs for asymmetric bimanual movements occur at the level of motor programming was later challenged by Diedrichsen et al. (
2001). They argued that increased RTs for asymmetric movements only occur when symbolic cues are used to specify the movement targets but not when the movement targets are defined directly (spatially). In other words, in most of the initial studies on the bimanual spatial interference effect, targets were either defined by words (e.g., “short” or “long”), letters (e.g., “S” or “L”) or bars indicating the length of the movement amplitude (Heuer & Klein,
2006; Spijkers et al.,
1997,
2000). Hence, to initiate the correct movements, these cues have to be identified and then translated into the required actions. In contrast, if the movement targets are presented directly such that there are only two target locations present in the workspace, no cue–response translation process is required. By comparing RTs in conditions employing either direct spatial cues or symbolic (letter) cues, Diedrichsen and colleagues could show that RT costs for asymmetric movements are limited to conditions in which the movements are cued symbolically. Based on these findings, they suggested that asymmetry costs for bimanual movements are related to response selection processes and not to increased processing demands during motor programming as initially suggested. Thus, increased RTs for asymmetric movement amplitudes are likely to be linked to the fact that two different stimulus response mapping rules have to be retrieved and applied in the incongruent (different amplitude) condition while the same mapping can be used in the congruent (same amplitude) condition (see also Albert, Weigelt, Hazeltine, & Ivry,
2007).
Interestingly, studies further investigating this suggestion came to mixed conclusions with some confirming the absence of asymmetric RT costs in direct cueing conditions (e.g., Albert et al.,
2007; Diedrichsen, Ivry, Hazeltine, Kennerley, & Cohen,
2003; Hazeltine et al.,
2003) and others showing that there are small but still significant costs even when movements are cued directly (e.g., Blinch et al.,
2014; Blinch, Cameron, Franks, Carpenter, & Chua,
2015; Heuer & Klein,
2006). Based on this inconsistency, it was proposed that the two suggested forms of interference processes are not mutually exclusive but can occur concurrently (Diedrichsen, Grafton, Albert, Hazeltine, & Ivry,
2006; Heuer & Klein,
2006): Firstly, there are (relatively small) costs due to an increased complexity of motor programming (constraint on motor level) and secondly, there are larger costs related to increased demands of cue translation and response selection (constraint on perceptual and cognitive level; for review see Wenderoth & Weigelt,
2009). The notion that interferences during bimanual movements do not exclusively arise on a motor outflow level but are strongly mediated by cognitive factors is further supported by studies showing that RT costs for asymmetric movements are attenuated in situations in which movements are performed to identical target symbols (e.g., two circles out of circles and crosses) suggesting that selecting target positions with similar features enhances bimanual performance and eliminates RT costs for incongruent movements (Diedrichsen et al.,
2003; Weigelt, Rieger, Mechsner, & Prinz,
2007; Wenderoth & Weigelt,
2009).
The phenomenon that choice RTs depend on the stimulus–response (S–R) compatibility has been studied extensively using different paradigms (Hazeltine et al.,
2003; Hommel,
1997; Kornblum, Hasbroucq, & Osman,
1990; Neumann,
1990; Prinz,
1990). In short, it has been shown that response specification is generally facilitated when the similarity between stimulus and response is increased. In other words, a high compatibility between the stimulus and the required response permits a more direct parameter specification resulting in faster RTs (or reduced RT costs for incompatible movements). Hence, response selection and associated RT costs can vary strongly with the properties of the presented cues. Following this line of argument, Hazeltine et al. (
2003) suggested that while the cue–response mapping in symbolic cueing conditions is highly abstract requiring a (cognitively demanding) translation of the cue into the appropriate response, direct cueing conditions place only minimal demands on the response selection process (excluded stage hypothesis).
In our study, we wanted to further investigate the claim that interference effects disappear for directly cued movements as central processes required for cue translation and response selection are bypassed. Specifically, we hypothesised that the occurrence of interference effects may be more generally linked to the task difficulty and thus the cognitive resources available for response selection and movement preparation. In the symbolic cueing conditions employed in previous studies, the cues needed to be selected, identified and subsequently translated into a motor response (applying mapping rules that needed to be retrieved from working memory). The translation of movement cues into actions, therefore, requires cognitive resources and hence may leave reduced capacity for response selection and motor programming when asymmetric movements are required. In contrast, in the direct cueing conditions, stimulus–response translation requirements—and thus cognitive demands—are negligible. If, as we propose, RT costs for asymmetric movements are linked to a limitation in central cognitive resources, they should also occur in dual-task situations in which the secondary task is completely unrelated to the movement task.
To test this idea, we asked participants to perform symmetric and asymmetric bimanual movements in two conditions in which the movements were cued directly; in one block of trials participants had to perform an additional movement-irrelevant but highly demanding attentional task shortly before or during movement preparation, while in another block no such task was required (Experiment 1). We also implemented two symbolic cueing conditions with varying cue–response compatibility mappings. In the mapping condition with high cue–response compatibility, centrally presented arrows pointed directly toward the relevant movement targets. In contrast, in the mapping condition with low cue–response compatibility, the arrows indicating the relevant movement targets were not clearly associated with the movement targets. We predicted that RT costs for asymmetric movements should arise (or at least significantly increase) in (1) the condition in which targets were directly cued and a cognitively demanding secondary task had to be performed; and (2) in conditions in which symbolic cues with low stimulus–response compatibility were implemented (requiring a demanding cue translation process). Finally, to test the generality of our cognitive resource limitation hypothesis, we conducted a second experiment testing a different secondary task. Specifically, we asked participants to execute directly cued bimanual movements while simultaneously performing a (movement unrelated) working memory task with either no-, low- or high-working memory load conditions. Generally, our findings seem to support the view that the occurrence of bimanual interference effects depends on the overall task demands.
General discussion
In two studies we investigated a possible explanation for why RTs for asymmetric (or incongruent) bimanual movements are usually longer when the movements are cued symbolically but not (or to a much smaller extent) when they are cued directly. Previous studies have suggested two different, but not mutually exclusive, mechanisms that may be responsible for increased RTs for incongruent movements. Initially, interference was suggested to occur at the motor programming level as the generation of two distinct motor commands may cause mutual inhibition due to neural cross-talk during amplitude specification (Heuer,
1986,
1993; Heuer et al.,
1998; Spijkers et al.,
1997). However, a few years later, it was proposed that interference mostly arises at a cognitive level. According to this view, the observed RT costs for incongruent movements are attributed to the resource-demanding cue–response translation process necessary in symbolic cueing conditions (Albert et al.,
2007; Diedrichsen et al.,
2001,
2003; Hazeltine et al.,
2003). To date, it is considered that in fact both processes may play a role in creating the bimanual congruency effect (Diedrichsen et al.,
2006; Heuer & Klein,
2006). In other words, increased RTs for incongruent movements can be attributed to a small cost arising at the motor level (i.e., preference of the motor system to plan and execute symmetric movements) which occurs for both direct and symbolically cued movements and a larger cost arising at a cognitive level when cues have to be identified and translated into movement goals in the symbolic cueing conditions (for review see Wenderoth & Weigelt,
2009). Here we suggest that the overall size of the interference effect does not necessarily depend on whether or not cue identification and translation are required but depends more directly on the overall task demands.
We based our study on the view that the symbolic cueing conditions create a dual-task situation. In other words, in addition to movement programming and execution, participants have to identify the cues, retrieve and select the correct stimulus–response mapping rules (keeping the associated mapping rules in working memory) and subsequently select the appropriate motor responses. In other words, symbolic cueing requires participants to develop internal codes for each movement and associate these with the presented symbolic cues (such as letters, bars or words). Hence, the process of cue translation requires cognitive resources and may therefore leave less capacity for the relevant processes related to response selection and motor preparation (see also, Albert et al.,
2007; Hazeltine et al.,
2003). In contrast, in the direct cueing conditions, no resource-demanding cue–response translation process is required as there is a direct mapping between the stimulus and the required response. Hence, we hypothesised that RT costs for asymmetric movements vary with the difficulty of the secondary task and may be relatively independent of whether or not this task is movement related. We tested this prediction in Experiment 1 in two ways: firstly, we introduced two different symbolic cueing conditions that varied the compatibility between the presented cue and the required response. In line with our prediction, we found asymmetry costs for movements performed in the symbolic condition with low cue–response compatibility (i.e., high translational load) but not in conditions with high cue–response compatibility (i.e., low translational load). Secondly, we tested whether bimanual interference occurs in direct cueing conditions when participants perform a secondary cognitively demanding, but movement unrelated, task. Interestingly, we found RT costs for asymmetric movements in the dual-task condition suggesting that any kind of dual-tasking coinciding with response selection and action preparation may be sufficient to evoke a movement congruency effect. This finding makes it unlikely that interference effects observed in previous studies are a direct consequence of cue translation and corresponding response selection processes but can instead, more generally, be attributed to increased cognitive demands in symbolic cueing tasks. In other words, interference effects in bimanual actions may only become apparent in more complex (difficult) movement tasks.
To further confirm this notion, we conducted a second experiment in which we introduced a different secondary task that varied the amount of working memory load during movement preparation and execution. In line with our hypothesis that RT costs for asymmetric movements vary with the cognitive task demands, we found longer RTs for incongruent movements when the working memory load was high. However, even though the RT costs occurred reliably, they were overall smaller for the working memory task than for the dual-task condition in Experiment 1 (about 40 ms in Experiment 1 vs. 20 ms in Experiment 2). There are a couple of possible reasons for this discrepancy. On the one hand, the reduced RT costs in Experiment 2 may reflect that a mere working memory task requires less resources than a task comprising a combination of visual attention and working memory components as used in Experiment 1 (note that participants had to keep the identified number in working memory until the end of the trial). On the other hand, the secondary task in Experiment 2 may just have been simpler than the task used in Experiment 1. Tentative support for this suggestion comes from the finding that the amount of correctly reported target numbers was much higher in Experiment 2 (Exp. 2: 93.5 % vs. Exp.1: 69.8 %).
Before discussing the implications of our study we need to address one important methodological difference to many previous studies employing a direct cueing paradigm (e.g., Albert et al.,
2007; Diedrichsen et al.,
2001; Hazeltine et al.,
2003). In these studies, direct cues were presented as a sudden onset within the visual field (i.e., the two targets to which participants have to move their hands appeared). In contrast, in our study, we presented all four possible target locations during the preview period (similar to the symbolic cueing conditions) and defined the targets by a visual offset of the non-target locations (for a similar procedure see also, Blinch et al.,
2014). We chose this procedure as it was pointed out by Hazeltine and colleagues (
2003) that many studies investigating differences between symbolic and direct cueing conditions (e.g., Diedrichsen et al.,
2001; Hazeltine et al.,
2003) displayed all relevant movement targets before the movement was required in the symbolic cueing conditions but not in the direct cueing conditions. Hence, partial movement pre-programming may have taken place in the symbolic cueing conditions before cue presentation. Adjusting these pre-planned movement programs after cue presentation may in turn have induced the observed cross-talk in these conditions. By always displaying all possible movement targets in both the direct and the symbolic cueing conditions during the preview period, this potential confound is avoided. Finally, we think that it is unlikely that this procedure substantially changes our findings, compared to studies using target onsets, as it has been shown that (when attention is unfocused as in the current study) visual onsets and offsets are equally effective in attracting attention to different locations in space (Theeuwes,
1991).
Overall, this is the first study that indicates that the occurrence of the bimanual interference effect does not merely depend on the type of cues used (symbolic vs. direct) but rather seems to be related to the general cognitive demands the task poses. In other words, even when a secondary task that is completely unrelated to the movement task is performed, interference effects can be observed. Notably, these findings may partly resolve the debate of why interference effects have consistently been found in symbolic cueing conditions but rarely (and to a much smaller extent) in spatial cueing conditions. It is important to point out that the link between bimanual movement studies and dual-task performance was originally suggested by Hazeltine and colleagues (
2003). However, our findings that bimanual interference can (1) be abolished in symbolic cueing tasks by minimising the response requirements and (2) be created in direct cueing conditions by maximising processing demands provides the first convincing empirical evidence for the notion that bimanual interference effects are primarily related to dual-task demands and overall task difficulty.
Regarding the question of how our account relates to the previous notion that interference occurs at two stages, i.e., during motor programming and cue translation, we think that it has the advantage that it can explain previously observed effects without assuming two different and independent underlying processes. Specifically, our results may help to understand why some, but not all, studies found bimanual interference effects in direct cueing tasks. For instance, Blinch et al. (
2014) found RT costs when participants performed directly cued asymmetric movements without visibility of their hands using a handheld stylus. It stands to reason that it is a much more demanding task to perform movements with a tool and without visual feedback than it is to point directly with both fingers while having both hands fully visible. Consequently, the task is likely to require more attentional resources yielding the observed congruency effect. Similarly, movement congruency effects were found to be larger in direct cueing conditions when reversal movements rather than discrete pointing movements were investigated (e.g., Heuer & Klein,
2006) suggesting again that movement complexity may affect the size of the observed interference effect. However, even though our suggestions seem to fit nicely with some of the findings from previous studies there are also instances in which a simple explanation in terms of task demands is not instantly obvious. For example, in a recent paper, Blinch et al. (
2015) reported small but significant interference effects (12 ms) in a relatively simple direct cueing task employing discrete pointing movements. One methodological difference to previous studies was, however, that target buttons were used as movement goals requiring participants to perform relatively accurate movements which are potentially more resource demanding (Hesse & Deubel,
2010). In other words, we suggest that factors that relate to movement difficulty (such as endpoint accuracy, target visibility and movement speed) may determine the amount of bimanual interference measured in different paradigms.
However, we also need to point out that our findings cannot provide a definite answer on the question at which exact processing stages the interference effect arises. Clearly the current findings can be reconciled with the proposition that interference occurs at a motor level as increasing the task demands in a direct cueing task may leave less capacity for movement programming thereby enabling transient coupling to occur. On the other hand, engaging in a movement-related (e.g., cue translation) or movement-unrelated (e.g., attentional) secondary task also leaves less resources for stimulus identification and response selection thereby allowing interference effects to emerge.
Finally, our finding that RTs are prolonged when a movement-unrelated cognitively demanding task has to be performed indicates that movement planning relies on the same central resources as needed for the execution of conscious perceptual tasks. This is in line with previous studies on unimanual reaching and grasping movements showing that movement preparation is less efficient (as indicated by longer RTs) when resources have to be shared between concurrent perceptual and visuomotor tasks (Hesse & Deubel,
2011; Hesse, Schenk, & Deubel,
2012; Kunde, Landgraf, Paelecke, & Kiesel,
2007; Similä & McIntosh,
2015). Therefore, our study provides further evidence against the view that perception and action processes may be controlled by separate attentional mechanisms allowing for efficient task sharing between visuomotor and perceptual processes without dual-task costs (Enns & Liu,
2009; Norman,
2002).
In conclusion, we showed that RT costs for incongruent bimanual movements do not depend on whether the movements are cued symbolically or directly, but on the overall processing demands of the task at hand. The harder the task, the more likely it is that dual-task interferences become apparent, suggesting that perceptual/cognitive and visuomotor tasks compete for the same limited resources.