Bimanual interference associated with the selection of movement targets

Three experiments were conducted to identify the locus of bimanual interference observed during the production of reaching movements. The movements had either the same or different amplitude and were directed towards identically or differently colored target circles. In Experiment 1, reaction times for movements of different amplitudes to targets of the same color were faster than for movements of the same amplitude to targets of different colors. This indicates that the cost to initiate responses of unequal amplitudes arises during selection of the target of the movement rather than during motor programming. Experiments 2 and 3 further specify the sources of interference found in target selection. Reaction time costs are found with unimanual responses when the target is presented among distractors associated with responses for the other hand. This indicates that the costs may arise through an assignment problem of response-rules to the respective hands. BIMANUAL INTERFERENCE IN TARGET SELECTION 3 Bimanual interference associated with the selection of movement targets We typically use our two hands in a coordinated fashion to achieve a common goal. When unscrewing a bottle or tying a knot, the actions of the two hands have to be finely tuned in relation to each other. Constraints on bimanual performance become apparent when we use our hands to achieve different goals at the same time. Consider for example the task of picking out good cherries among rotten ones from a tray in the supermarket, with either one or two hands. If our hands could work perfectly independently, we should be able to pick out twice as many cherries with both hands than with only one. This, however, is not the case. This limitation may be due to our inability to plan or execute independent movements with the two hands at the same time. Alternatively, we may have problems in selecting in parallel the goals of the two reaching movements based on a number of criteria as size, color, smoothness, etc. for the two hands. The focus of the present article is to determine the processing stage that gives rise to the performance limitations during bimanual movements. In the laboratory, these limitations of bimanual performance have been studied with tasks that require the simultaneous production of two movements with either symmetric or differing spatial characteristics. Compared to when the movements are the same for the two hands, producing different movements with the two hands prolongs initiation times (Franz, Eliassen, Ivry, & Gazzaniga, 1996; Spijkers, Heuer, Kleinsorge, & van der Loo, 1997) and significant distortions of the spatial trajectories can be observed (Franz, Zelaznik, & McCabe, 1991; Kelso, Putnam, & Goodman, 1983; Sherwood, 1990; Spijkers & Heuer, 1995). Spijkers and his associates have explored the question of when limitations arise in a series of papers (1997; Spijkers, Heuer, Steglich, & Kleinsorge, 2000). In one study, participants were required to execute fast lateral movements (outward and back) with their arms. The target amplitude of each movement was cued by the presentation of two bars, one on the left and one on the right of a computer monitor. Each bar could either be short or long. Reaction times were nearly 100 ms longer when the two movements were incongruent (i.e., one short and the other long) compared to when the two movements were congruent (i.e., both short or both long). The authors proposed that this increase BIMANUAL INTERFERENCE IN TARGET SELECTION 4 reflected interference at the stage of motor programming (see Rosenbaum, 1980; Rosenbaum & Kornblum, 1982). Programming was fast when the only unknown parameter, movement amplitude, was set to the same value for each hand. When different parameters needed to be specified, cross talk was hypothesized to occur between the programming process required for the right arm and that required for the left arm. This cross talk presumably led to the increase in RT on incongruent trials. A challenge to the programming interpretation comes from a recent study in which the target locations were specified directly (Diedrichsen, Hazeltine, Kennerley, & Ivry, in press). That is, visual signals were presented on the table surface and the participants were instructed to move as quickly as possible to the target locations. Under these conditions, no differences were found in RT between congruent and incongruent movements, regardless of whether incongruency was defined as a difference in movement amplitude or direction. Indeed, the RT on bimanual trials with direct cues was as fast as on unimanual trials. Assuming that motor programming and execution processes are similar for direct and symbolically cued movements, these results indicate that the interference observed in earlier studies is unlikely to be attributed to either of these stages. Rather, the interference visible in the RT data is likely due to a processing stage associated with identification of the stimulus and/or selection of the appropriate response. Although rarely discussed as such, bimanual reaching studies constitute a form of a dual-task paradigm. Limitations in dual task performance have been one of the central topics in cognitive psychology and have played a critical role in the development of analytic tools for specifying processing stages across a range of tasks (Pashler, 1998b). One example is the literature on the psychological refractory period (McCann & Johnston, 1992; Pashler, 1994; Telford, 1931). In these studies, participants have to respond to two stimuli in rapid succession. When the stimulus onset asynchrony (SOA) between the tasks is short, the RT for the second task is delayed. This limitation has been attributed to a response-selection bottleneck (but see Meyer & Kieras, 1997; Pashler, 1984, 1998a). By this view, it is assumed that stimulus identification and response execution for the two tasks can occur in parallel; the limitation is hypothesized to arise due to overlapping demands of the two tasks on a common response selection process. BIMANUAL INTERFERENCE IN TARGET SELECTION 5 Response selection for one task must be completed before this operation can be performed for the other task. With short SOAs, this delay will be manifest as an increase in RT for the second task. Spijkers et al. (2000) used a PRP paradigm with bimanual reversal movements of same or different amplitudes. They found that at short SOA the initiation of the second movement was considerably delayed, when the movement amplitudes were incongruent. They interpreted their results as interference of temporally overlapping motor programming processes. With their symbolic bar stimuli as cues, however, the cost on incongruent trials might be associated with the demands associated with processing nonidentical cues or because non-identical "abstract" movement codes had to be selected. The first hypothesis was ruled out by a control condition in which the first stimulus had to be identified, but no response was required. The participants showed similar RTs in this condition on congruent and incongruent trials. The second possibili ty was considered by the authors, but dismissed because “the distinction between response selection and amplitude specification is inappropriate for our experimental paradigm [...]. As argued by Rosenbaum (1983), selection of a movement is equivalent to specifying its parameters” (Spijkers et al., 2000, p. 1103 ). This argument may be valid in cases in which the movements are not directed to specific targets but are instead determined by symbolic stimuli that specify the movement in terms of kinematic parameters. Under these symbolic conditions, movements may be mentally represented in terms of these task-defined movement parameters. That is, when participants are told to make a long or short movement in response to the letter “L” or “S” , then they likely represent possible action as either a “long movement” or “short movement” . Selection of the response then operates on codes in terms of movement parameters. Bimanual interference arises when two actions with different representations have to be selected. However, when the movement is directed towards an object, the response code may encompass characteristics of the goal object rather than the movement parameters. For example, imagine you want retrieve a tool from a cluttered tool drawer. Deciding which object you want to pick out would constitute a response-selection process. The location of that object and its orientation would then dictate the motor programming requirements, a process that would follow selection. In this way, the BIMANUAL INTERFERENCE IN TARGET SELECTION 6 selection of the movement goal and the specification of the movement parameters to get to the goal can be separated. This distinction suggests two possible loci for bimanual interference. If cross talk arises during the selection of responses, then the pattern of interference should depend heavily on how the possible actions are represented. If cross talk arises during the specification of motor parameters, then the interference should be determined by the kinematic properties of the movements. In Experiment 1 participants were instructed to select an object for each hand and then directly reach for it. The reaches could require movements of the same or different amplitudes. The goal of this study was to examine whether the interference should be attributed to early stages (stimulus identification and response selection) as has been hypothesized in PRP studies and our earlier direct reaching study (Diedrichsen et al., in press) or late stages (motor programming and motor execution) as has been assumed in most bimanual reaching studies. Experiment 1 The experiment was conducted in an apparatus which allowed participants to reach directly for visually presented targets (Figure 1). Four target locations were defined on each trial by the presentation of four colored circles that forme