Who is talking in backward crosstalk? Disentangling response- from goal-conflict in dual-task performance
Introduction
Humans have considerable difficulty performing multiple tasks at once, so that multi-tasking typically results in performance decrements in at least one of the tasks. Such performance costs have been attributed to serial processing in one or more of the involved processing stages. Serial models often identify the selection of appropriate responses as such a serial processing stage, while perceptual or motor execution processes are assumed to be carried out in parallel to other processes (e.g., Pashler, 1994, Pashler and Johnston, 1989; Welford, 1952; see Fig. 1a). However, a number of empirical observations have casted some doubts on the validity of serial models. The present study aimed at revisiting the serial-stage assumption by better characterizing the supposedly serial processes, if any. We did so by making use of the backward crosstalk effect in dual-task situations.
In addition to general dual-task performance costs, specific task demands and characteristics determine how well two or more tasks go together. For example, mental rotation is facilitated if preceded by, or performed simultaneously with a manual rotation in the same direction (Wexler, Kosslyn, & Berthoz, 1998; Wohlschläger & Wohlschläger, 1998). It is even more intriguing that such inter-task facilitation works in the reverse direction as well: Responding in the first-performed task is facilitated if the features of the corresponding response overlap with features of the response in the subsequently performed second task (e.g., Hommel, 1998). This is particularly interesting because it suggests that features of the second response are activated before or while the first response is being selected, which challenges the idea that response selection operates strictly serially. Apparently, the second response “works back” on the first response, which is why effects of Task 2 on Task 1 performance have been coined “backward crosstalk effects” (BCEs).
The first demonstration of a BCE was reported by Hommel (1998): A colored letter stimulus was presented in each trial and participants carried out a manual left/right response (R1) to the color of this letter and a vocal response (R2; the German words for “left” or “right”) to the letter identity. Response times (RTs) in the first task were shorter when R1 and R2 were compatible (e.g., when pressing the left key was followed by saying “left”) as compared to incompatible relations. Later studies showed BCEs with various kinds of feature overlap (e.g., Ellenbogen and Meiran, 2008, Ellenbogen and Meiran, 2011, Hommel and Eglau, 2002, Logan and Delheimer, 2001, Logan and Gordon, 2001, Logan and Schulkind, 2000, Miller, 2006), ranging from physical features, such as spatial correspondence (e.g., Lien & Proctor, 2000), to rather abstract features, such as if responding in the first task is systematically delayed when the second task requires withholding a response (Miller, 2006) or influences of Task 2 response force requirements on Task 1 responses (Miller & Alderton, 2006).
Accommodating the phenomenon of BCEs requires the assumption of some degree of parallel processing even at stages typically assumed to operate serially.1 Hommel (1998) subdivided central stages into a parallel response activation sub-stage that is followed by a (perhaps strictly serially operating) response selection sub-stage (see also Lien & Proctor, 2002; see Fig. 1b). According to that model, Task 2 characteristics can affect Task 1 performance during the response activation sub-stage, thus yielding BCEs. Indeed, there is converging evidence that some of the processes mapping a Task 2 stimulus to its appropriate response are already active, while the same processes are still ongoing for Task 1 (Schubert, Fischer, & Stelzel, 2008).
In most previous demonstrations of BCEs, crosstalk arises between responses that indicate certain classifications of experimental stimuli. Accordingly, most BCEs might best be seen as an instance of response–response conflict (see also Navon & Miller, 1987). In line with this reasoning, Logan and Gordon (2001) suggested that many BCEs are “most likely a response repetition effect” (p. 412). However, this idea immediately poses another question: If BCEs really reflect interactions between features of the two responses, which features are actually critical for producing BCEs?
In the present study we identified the response characteristics/features that eventually give rise to BCEs, i.e., which aspects of the second response “cross-talk” with processing of Task 1. Answering this question was thought to provide deeper insight into which response characteristics can be processed in parallel and how dual-task performance might be improved. To do so, one first needs to dissect responses as used in typical studies on BCEs into different features. In fact, even simple key press responses can comprise various features as we will discuss in the next section.
At first glance, a simple act such as pressing a key seems to comprise just a few features. For instance, it seems intuitively plausible that pressing the left of two alternative response keys with the left hand, say, is compatible with pressing the left of two other response keys with the right-hand—implying that an entire response can be exhaustively characterized as “left”. But a closer look reveals that even a simple key press has more features than that and can thus be cognitively represented in multiple ways. For one, there is the overt behavior resulting from the motor act of key pressing that can be observed from the actor’s but also from another person’s perspective. Traditionally, this feature is referred to as “response feature” and we will therefore denote the task-relevant feature of observable, overt behavior as “R”. In addition, R produces kinesthetic and tactile re-afferent sensory input from pressing the key that is only perceived from the actor’s perspective. Further, often visual and perhaps auditory input from the action and its immediate environmental consequences follows. All these kinds of input can be considered effects of R and will therefore be denoted as (features of) “E”. Note that in many instances, all these R and E features suggest coding a left key press as “left” and thus are confounded.
However, such effects (Es) are likely to play an important role in action selection. This has been proposed by the ideomotor approach to action control, which dates back to philosophical ideas of the 19th century (see Pfister and Janczyk, 2012, Stock and Stock, 2004). Basically, this approach and its more modern versions assume that actions are cognitively represented by codes of their perceivable effects (Greenwald, 1970; Hommel, Müsseler, Aschersleben, & Prinz, 2001). Selecting a response thus means to retrieve and anticipate its sensory effects, which then spread activation to the associated motor pattern. Among other evidence, this assumption has received empirical support by studies on response-effect compatibility (REC; Kunde, 2001). For instance, Kunde had his participants respond to the color of a stimulus by pressing one of four keys. In the compatible REC condition, the key press switched on a spatially corresponding light, while in the incompatible REC condition a non-corresponding light was switched on. Although these action effects occurred only after the key press (i.e., not during response selection), RTs were faster in the compatible REC condition (see also Kunde et al., 2002, Pfister et al., 2010). Similar REC effects were demonstrated with other and even more abstract relations of responses and effects, such as vocal color responses producing semantically compatible or incompatible color words (Koch & Kunde, 2002) or vocal number responses resulting in visually displayed compatible or incompatible numbers (Badets, Koch, & Toussaint, 2013). Such results are well in line with the ideomotor assumption that action selection operates on codes representing the anticipated sensory features of a given action (cf. Hommel, 2009). This general principle of effect-based action control raises the question of which of the multiple features of actions and their effects are involved in producing BCEs.
Most authors seem to assume that actions are in some way represented by the relative location of the effectors being involved. For instance, Lien and Proctor (2000) had participants press one set of two response keys with the middle and index finger of the left hand and another set with the index and middle finger of the right hand, assuming that pressing the “left” key of the left-hand set and the “left” key of the right-hand set would be compatible. Indeed, BCEs were obtained under these conditions and the direction of the effect was in line with this definition of compatibility. As the two left-hand keys were operated by different fingers, this means that the key locations to which R was made were more important than the anatomical status of the operating finger – a common observation in compatibility studies (e.g., Wallace, 1971). Now, remember that any given R leads to various Es, such as the proprioceptive feedback from moving the particular finger, the feeling of pressing down a key, and, possibly, ensuing changes on the computer screen. Thus, it might be any of these E-features that were anticipated for response selection (instead or in addition to activating R-features). In other words, the available evidence does not tell us whether BCEs are driven by R-features or E-features. Sketched in a schematic originally developed by Hommel (1998), stimuli can automatically – via direct links (DL) or transient links (TL) – activate either R or E (see Fig. 2 for an illustration). To carefully distinguish these features and to identify the features that are crucial for BCEs was at the core of the present research.
Unfortunately, effects related to kinesthetic or proprioceptive feedback are coupled so tightly to R that it is almost impossible to distinguish between R and these Es experimentally (see, e.g., Stenneken, Aschersleben, Cole, & Prinz, 2002, for work with a deafferented patient). A less invasive approach is to introduce effects in the environment that are coupled to R by an experimental manipulation of R–E relations.2 Indeed there is increasing evidence that responses can be represented by novel, experimentally induced effects (Elsner & Hommel, 2001), especially if they are defining the actual action goal (Hommel, 1993). For instance, the otherwise robust advantage of responding simultaneously with homologous rather than non-homologous finger combinations (e.g., Cohen, 1971, Heuer, 1993) can be reversed if actions with homologous fingers lead to different visual effects while actions with non-homologous fingers produce similar effects (Janczyk, Skirde, Weigelt, & Kunde, 2009; for converging evidence see also Kunde et al., 2009, Kunde and Weigelt, 2005, Mechsner et al., 2001).
Even more interesting, experimentally induced action effects have been shown to be involved in interference in dual-task situations, as with more or less simultaneous mental and manual rotations. As already mentioned, mental rotations are typically facilitated by preceding or simultaneous manual rotations into the same direction (e.g., Wexler et al., 1998, Wohlschläger and Wohlschläger, 1998). In a recent study, Janczyk, Pfister, Crognale, and Kunde (2012) disentangled the contributions of effector and goal-effect features by combining a manual rotation with an effect rotating in either the same or the opposite direction. Participants performed a mental rotation task followed by the manual rotation, and once attention was directed to the visual effects, facilitation depended on the directional overlap of the mental and the effect rotation, not the manual rotation itself.
Given the evidence for a contribution of action effects to both single- and dual-task performance it is interesting to ask whether they may represent the critical response feature that becomes activated automatically and creates BCEs. The present study pursued this question in three experiments in which the overt motor response to Task 1 (R1) was a simple left/right key press with the left hand. The response in Task 2 (R2) was carried out with the right hand and was either a left/right key press (Exp. 1 and 3) or a continuous left/right movement (Exp. 2).
In Experiments 1 and 2, the actual goal in Task 2 was to produce a visual effect (E2): A response-contingent visual event that in one group of participants (Exp. 1) or in one condition (Exp. 2) appeared on the same side as R2 (R–E-compatible condition) and in another group or condition on the opposite side (R–E-incompatible condition). In the R–E-compatible condition, all spatial features of the entire action event, including both R2 and E2, referred to the same relative location. Accordingly, the entire action event would be coded “left” or “right” and affect compatible or incompatible responses in Task 1 accordingly, irrespective of whether the resulting BCE would reflect an impact of R codes, E codes, or both. We thus expected a standard R2–R1 BCE to emerge in this situation with faster Task 1 responses for spatially compatible R2–R1 relations.
The crucial question was what would happen in the R–E-incompatible situation, where a right response would contingently produce an intended action effect on the left side and a left response would contingently produce an intended action effect on the right side. Note that, due to this R–E relationship, a compatible R1–R2 relation (e.g., left key press in Task 1 and left key press in Task 2) implied an incompatible relation between R1 and E2 (left key press in Task 1 and right light flash in Task 2), while incompatible R1–R2 relations implied compatible R1–E2 relations. A comparison of the size and sign of the BCE in R–E-compatible and R–E-incompatible conditions is thus diagnostic for the relative contributions of response and effect features to the BCE. Consider, for instance, that only the E feature would matter: This would mean that the BCE should completely reverse from R–E-compatible to R–E-incompatible conditions.3 If, in turn, only R features would matter, the compatibility between R2 and E2 should have no impact on the BCE. A probably more realistic outcome is a pattern laying somewhere in between, however. As shown by Hommel (1993) with single-task conditions, people tend to consider multiple spatial relationships when coding their responses. That study was designed to disentangle the degree to which people code their actions in terms of the anatomical status of the active effectors, the effectors’ (and keys’) relative locations, and the location of the intended action effect. The findings suggest a dominant contribution from the latter source (i.e., action goal), but measurable contributions from the other two variables as well. This indicates that people can weigh the feature codes referring to a given action according to their relative importance (Memelink & Hommel, 2013). If so, we would expect that making R2 and E2 incompatible reduces and perhaps even tends to reverse the BCE in sign without necessarily producing a significantly reversed BCE of the same size as with a compatible R2–E2 relation. In this case, however, the interaction of R1–R2 relation and R2–E2 relation should still be significant. Experiment 3 complements these experiments by coupling both responses with intended visual action effects.
Section snippets
Experiment 1
Participants were presented with single letters of varying identity and color as stimuli. Both tasks required a left/right manual key press with the left (Task 1) or the right hand (Task 2). Both responses were thus compatible when both hands performed a key press on the same side but incompatible if the key locations did not match (see Hommel, 1998, and Lien & Proctor, 2000, Exp. 2 and 3, for comparable setups).
Importantly, R2 switched on a light that was located on the left or right side of
Experiment 2
The results of Experiment 1 suggest that the intended action effects were critical for BCEs to a considerable degree, although other features (e.g., the actual motor response) likely also played a role. To further corroborate this finding, we conducted Experiment 2, in which the major change relates to the responses and intended action goals for Task 2. The discrete key press response was changed into a continuous right-hand movement that produced a spatially compatible or spatially
Experiment 3
Experiments 1 and 2 have demonstrated the importance of intended action effects in Task 2 and their compatibility with R1 for BCEs. However, in both experiments it was only Task 2 where responses were less important than their visual effects. This might have created asymmetries between the two tasks and perhaps made Task 2 more difficult or attention-demanding. Our aim was therefore to create a more balanced relationship between Task 1 and Task 2 by equipping both R1 and R2 with task-relevant
General discussion
To understand the limits of dual-task performance it is important to know which aspects of concurrent actions can be processed in parallel and which cannot. A tool to study this question is the backward crosstalk effect (BCE), the observation that the response in an upcoming Task 2 can influence performance in the temporally preceding Task 1 in a dual-task situation. Arguably, those features of Task 2 responses that are able to engage in crosstalk with those of Task 1 responses, are not subject
Conclusions
The present study aimed at investigating what features of responses can become activated automatically in dual-task situations. To address this question, we investigated the role of intended action effects, i.e., action goals, for backward crosstalk effects (BCEs). Our findings attribute to them an important role in determining the direction of BCEs. It is thus not so much the features of the overt response that are activated in parallel with other processes, but rather the anticipated and
Acknowledgments
We thank Eric Ruthruff and Iring Koch for valueable comments and suggestions on a previous version of this manuscript. Parts of this research were supported by the Deutsche Forschungsgemeinschaft (DFG; German Research Council), project KU 1964/2-2.
References (67)
- et al.
Visual and tactile action effects determine bimanual coordination performance
Human Movement Science
(2009) - et al.
The impact of anticipated action effects on action planning
Acta Psychologica
(2002) - et al.
Adaptive control of ideomotor effect anticipations
Acta Psychologica
(2010) A critical path generalization of the additive factor method: Analysis of a Stroop task
Journal of Mathematical Psychology
(1978)- et al.
Motor processes in mental rotation
Cognition
(1998) - et al.
Role of an ideomotor mechanism in number processing
Experimental Psychology
(2013) Synchronous bimanual movements performed by homologous and non-homologous muscles
Perceptual and Motor Skills
(1971)- et al.
The representation of instructions in working memory leads to autonomous response activation: Evidence from the first trials in the flanker paradigm
The Quarterly Journal of Experimental Psychology
(2007) - et al.
The representation of instructions operates like a prepared reflex: Flanker compatibility effects found in first trial following S-R instructions
Experimental Psychology
(2009) - et al.
Working memory involvement in dual-task performance: Evidence from the backward compatibility effect
Memory & Cognition
(2008)
Objects and events as determinants of parallel processing in dual tasks: Evidence from the backward compatibility effect
Journal of Experimental Psychology: Human Perception and Performance
Effect anticipation and action control
Journal of Experimental Psychology: Human Perception and Performance
Assessing bimodality to detect the presence of a dual cognitive process
Behavior Research Methods
Sensory feedback mechanisms in performance control: With special reference to the ideo-motor mechanism
Psychological Review
The dip test of unimodality
The Annals of Statistics
Statistical methods for meta-analysis
Two modes of sensorimotor integration in intention-based and stimulus-based actions
The Quarterly Journal of Experimental Psychology
Action-effect bindings and ideomotor learning in intention- and stimulus-based actions
Frontiers in Psychology
Structural constraints on bimanual movements
Psychological Research
Inverting the Simon effect by intention
Psychological Research
The cognitive representation of action: Automatic integration of perceived action effects
Psychological Research
Automatic stimulus–response translation in dual-task performance
Journal of Experimental Psychology: Human Perception and Performance
Action control according to TEC (theory of event coding)
Psychological Research
Control of stimulus–response translation in dual-task performance
Psychological Research
The theory of event coding: A framework for perception and action
Behavioral and Brain Sciences
The principles of psychology (orig. 1890)
Level-2 perspective taking entails two processes: Evidence from PRP experiments
Journal of Experimental Psychology. Learning, Memory, and Cognition
The benefit of no choice: Goal-directed plans enhance perceptual processing
Psychological Research
Instant attraction: Immediate action-effect bindings occur for both, stimulus- and goal-driven actions
Frontiers in Psychology
Effective rotations: Action effects determine the interplay of mental and manual rotations
Journal of Experimental Psychology: General
On the persistence of tool-based compatibility effects
Journal of Psychology
Exceptions to the PRP Effect? A comparison of prepared and unconditioned reflexes
Journal of Experimental Psychology. Learning, Memory, and Cognition
Verbal response-effect compatibility
Memory & Cognition
Cited by (82)
Given the option, people avoid incongruent responses in a dual-tasking situation
2022, Acta PsychologicaInvestigation of reward effects in overlapping dual-task situations
2022, Acta Psychologica