摘要：Human behaviour is characterized both by action and reaction; the action mode can be conceived of as being embedded in anticipatory control, whereas the reactive mode requires the instantaneous processing of stimuli. In the reaction mode, tasks have to be completed (most of the time) as fast as possible; in the action mode, the task is to act “at the right time” anticipating the consequences of one's actions. A control machinery for the action mode has been suggested some time ago by the “reafference principle” (von Holst & Mittelstaedt, 1950), the theoretical concept already being formulated by Ernst Mach and Hermann von Helmholtz in the 19thcentury. The basic idea is that an action program not only initiates a motor program, but an efference copy is drawn from this program, which is compared after the completion of the movement with the re-afference. When efference copy and reafference match, the movement program is cancelled, i.e., the movement has come to an end. There are plenty of examples to support the basic notion of the re-afference principle, but there remains one essential shortcoming, i.e., the temporal domain. The re-afference principle is conceived of as being temporally neutral which is both theoretically and practically not satisfying. How can one test temporal constraints that might shed some more light on behavioural control?Quite often, observations from everyday behaviour may provide an incentive for experimental paradigms. Everybody may have been confronted with the feeling that a planned action with an anticipated movement pattern is experienced to happen at the wrong time; sometimes one acts too early and sometimes too late, the reason often being that the attentional control was disrupted. Examples can, for instance, be observed in driving an automobile (Tanida & Pöppel, 2006) or in some sport activities like football, tennis or squash. Practical experience shows that it is often necessary to delay an action for better temporal control. We decided on the basis of theoretical considerations and practical experiences to look into this temporal control problem in more detail by examining delayed actions using a modified experimental paradigm of a previous experiment (Szelag et al., 2001).Participants were asked to respond to visual targets presented continuously with pre-defined delays between 400 and 5000 milliseconds; ten such different delays were chosen (400, 600, 800, 1000, 1500, 2000, 25000, 3000, 4000, 5000ms). For each delay time, 40 trials were sequentially employed, and after each trial the participants was provided a feedback of the response time. The goal of the participants was to delay the action as accurately as possible. A critical variable was also the standard deviation (SD) of the 40 trials for each pre-defined interval being normalized; i.e., each SD was divided for comparison reasons by the average response time to the target delay time. As a control the fastest reaction times for each subject were also measured. The results are quite clear and surprising: Participants were very fast to adjust their response time to the pre-defined delay time; the best response time was actually observed for the longest delay time with an average of 4992ms for 5000ms delay time. More interesting, however, were the results for the normalized SDs: The largest value with some 60ms was observed for the shortest delay time of 400ms. These normalized SDs became smaller for the longer delay times being characterized an exponential decay and reaching a plateau of some 20ms at approximately 3000ms delay time. As 40 trials were employed, it was possible to compute the normalized SDs for the first and second 20 trials; it turned out that there was a clear learning effect, i.e. the second 20 trials showed much smaller SDs for all pre-defined delay times. The fastest reaction with on average 220ms was much shorter than the selected delay times, thus, lying outside the operating range of the pre-defined delay times.The higher variance for shorter delay times comes as a surprise and might even be considered to be a paradoxical phenomenon. One might expect for shorter delay times smaller variances compared to longer delay times, but the contrary is the case. Better temporal control is reached if the “waiting time” is extended beyond the temporal limit of approximately three seconds. If attention is focused on the passage of time, an optimal action mode is switched on only after a few seconds; our experiments do not indicate how long such an anticipatory “temporal window” may last. But what the experiment indicates is that up to such an optimal delay time responses are characterized by temporal instability. It appears as if between fastest reactions and optimal actions our behaviour is characterized by a “temporal twilight” zone. If this is a case as the data indicate, it certainly makes sense to delay actions for some time, in case a fast reaction is not required.These observations also shed new light on basic features of the re-afference principle. If actions cannot be precisely pre- programmed because their execution shows too high temporal variance, the “efference-copy” is not clearly defined. This temporal instability makes it next to impossible to match the efference-copy with the re-afference; this then results in behavioural instability. Only if sufficient time is allowed to pass may such a match between anticipation and satisfaction become operative.
关键词：Delayed responses;reafference principle;reaction time;anticipation;movement control
