Mental fatigue and temporal preparation in simple reaction-time performance

Abstract

Performance decrements attributed to mental fatigue have been found to be especially pronounced in tasks that involve the voluntary control of attention. Here we explored whether mental fatigue from prolonged time on task (TOT) also impairs temporal preparation for speeded action in a simple reaction-time task. Temporal preparation is enabled by a warning signal presented before the imperative stimulus and usually results in shorter reaction time (RT). When the delay between warning and imperative stimuli – the foreperiod (FP) – varies between trials, responses are faster with longer FPs. This pattern has been proposed to arise from either voluntary attentional processes (temporal orienting) or automatic trial-to-trial learning (trace conditioning). The former account suggests a selective RT increase on long-FP trials with fatigue; the latter account suggests no such change. Over a work period of 51 min, we found the typical increase in overall RT but no selective RT increase after long FPs. This additivity indicates that TOT-induced mental fatigue generally reduces cognitive efficiency but leaves temporal preparation under time uncertainty unaffected. We consider this result more consistent with the trace-conditioning account of temporal preparation.

Introduction

Fatigue from prolonged mental work has been found to impair performance in a variety of cognitive tasks (e.g., Bills, 1931, Helton and Warm, 2008, Kraepelin, 1902, Lorist et al., 2000, Sanders and Hoogenboom, 1970). The state of mental fatigue is characterized by the inability to allocate sufficient processing resources to the task at hand (Matthews and Desmond, 2002, Smit et al., 2004). Frequently, fatigued participants are still able to perform highly over-learned, automated tasks, whereas their performance significantly deteriorates when tasks require the voluntary allocation (i.e., top-down control) of attention (e.g., Boksem et al., 2005, Boksem et al., 2006, Lorist et al., 2005, Lorist et al., 2000). Because top-down control processes are transient (Smallwood and Schooler, 2006, Steinborn et al., 2008, Stuss et al., 2003, West, 2001), maintaining performance at optimal levels requires a mechanism that stabilizes control and ensures continuous task engagement. This stabilization is considered an effortful mechanism vulnerable to mental fatigue (e.g., Lorist et al., 2005, Sarter et al., 2006, Smit et al., 2004, Wright et al., 2008).

Our study examined effects of mental fatigue on speeded performance in a forewarned simple reaction-time (RT) task. Various studies have demonstrated that simple RT performance substantially deteriorates over time (e.g., Buck, 1966, Lisper and Ericsson, 1973, Lisper et al., 1973, Sanders et al., 1982, Van den Berg and Neely, 2006). It is not yet clear, however, whether this effect only results from an overall effect of fatigue on response speed or whether it also involves more specific fatigue-related changes in the timing behaviour under temporal uncertainty.

Preparation enhances performance, for example, by speeding up responses to an imperative signal in simple and choice RT tasks (Jennings & Van der Molen, 2005). Here, we only deal with purely temporal (i.e., nonspecific) preparation, which is based on the temporal contingencies between experimental events. In tasks involving nonspecific preparation, participants use temporal information to optimise performance by anticipating the imperative moment (i.e. the moment of target occurrence) (Coull, 2004). Typically, a warning signal (WS) precedes the imperative stimulus (IS), enabling nonspecific preparation for the impending IS. This usually improves RT substantially (e.g., Hackley and Valle-Inclan, 2003, Los and Schut, 2008).

The delay between WS and IS is called foreperiod (FP). When FP duration is variable within a block of trials, participants remain uncertain about the exact moment of IS occurrence on any given trial and thus cannot exactly synchronize their preparation with IS occurrence. In this variable-FP setting, responses are typically found to be relatively slow at early imperative moments but to become faster at later imperative moments during the FP interval (cf. Los et al., 2001, Niemi and Näätänen, 1981, Woodrow, 1914). This phenomenon, termed the variable-FP effect, has been traditionally explained by assuming that participants exploit the gradual increase in conditional probability of IS occurrence during the FP and transform them into a state of preparation (Niemi & Näätänen, 1981, p. 137).

To illustrate this point, consider an RT experiment in which the IS is presented with an equal a priori probability at three imperative moments, say 1000, 3000, and 5000 ms after the WS. At the first imperative moment (i.e., 1000 ms after the WS), the probability of IS occurrence is 33% (and 66% that the IS will be presented later). In trials where the first imperative moment is bypassed, the probability that the IS will occur at the next one (i.e., 3000 ms after the WS) increases to 50%. When this moment is also bypassed, participants will then have full certainty that the IS will be presented at the latest imperative moment (i.e., 5000 ms after the WS). In this situation, the typical finding is a decrease in RT with increasing FP length. That is, the fastest responses in this example occur with FPs of 5000 ms. The resulting downward-sloping FP–RT function has traditionally been taken to reflect a strategic process by which participants convert the objective increase in the conditional probability of IS occurrence into a subjective expectation. This strategic account assumes that participants actively track the flow of time after the WS and intentionally monitor the changing conditional probability to use this information for top-down regulation of their preparatory state (e.g., Näätänen & Merisalo, 1977). This top-down control of preparation is considered an effortful process since it arises from voluntarily orienting attention to specific moments in time after the WS (e.g., Correa et al., 2006, Coull et al., 2000, Lange et al., 2006, Nobre, 2001). The degree to which attention is directed to a specific time point during the FP is assumed to be directly related to the subjective probability (expectancy) of IS occurrence at this time point (e.g., Baumeister and Joubert, 1969, Karlin, 1966). Furthermore, studies have shown that a state of peak preparation can hardly be maintained for long, which underscores the importance of exact temporal predictions for being optimally prepared at the right time (e.g., Gottsdanker, 1975, Näätänen, 1972).

This strategic account, however, cannot explain sequential FP effects: Analyses that also considered FP length on the previous trial (FP_n−1) as a determinant of RT revealed that responses are relatively fast when the previous trial’s FP was short but are relatively slow when the previous trial’s FP was long. (e.g., Alegria and Delhaye-Rembaux, 1975, Karlin, 1959, Steinborn et al., 2008, Steinborn et al., 2009, Vallesi and Shallice, 2007, Van der Lubbe et al., 2004, Woodrow, 1914). These sequential FP effects are usually asymmetric, since RT is more strongly affected in trials with short FPs compared to trials with longer FPs, producing a typical FP_n−1 × FP_n interaction (see Fig. 1). To explain these asymmetric sequential FP effects within the traditional account (e.g., Alegria, 1975, Drazin, 1961, Klemmer, 1957), it has been argued that individuals expect a repetition of FP_n−1 on the current trial, so that optimal preparedness is reached at the same moment as on the preceding trial. If FP_n is shorter than FP_n−1, then optimal preparedness will not yet have been reached at IS occurrence, and RT will be relatively slow. If instead FP_n is longer than FP_n−1 and the repetition-expectancy-based moment of optimal preparedness is bypassed without IS occurrence, then it is assumed that individuals extend the period of optimal preparedness or cyclically re-prepare at later moments. Thus they achieve relatively fast responses in long-FP_n trials even after short-FP_n−1 trials (i.e., in non-repetition trials), which accounts for the asymmetry in the sequential effects (cf. Vallesi and Shallice, 2007, Van der Lubbe et al., 2004). This asymmetry, in turn, has the potential to also explain the FP_n effect, although it is as yet unclear to what extent. The assumption of two different processes for explaining the FP_n effect and the asymmetric sequential FP effect is, however, a general disadvantage of the strategic account.

Los and co-workers (e.g., Los and Heslenfeld, 2005, Los and Van den Heuvel, 2001, Los et al., 2001) have recently challenged the strategic view by proposing a unified and parsimonious account for both effects. They argued that response-related temporal preparation is driven by trace conditioning, a nonstrategic process of trial-to-trial associative learning that determines preparatory behaviour across subsequent trials (see also Gallistel and Gibbon, 2000, Machado, 1997). The conditioning account of temporal preparation maintains that the FP_n and the asymmetric FP_n−1 effects are two outcomes of one process. Specifically, it is assumed that the asymmetry of the sequential effect drives the FP_n main effect. That is, the FP_n−1 × FP_n interaction is considered responsible for the negatively accelerating slope of the FP_n–RT function.

Thus, the asymmetric FP_n−1 × FP_n interaction is at the core of the conditioning model, and it is explained as follows: In cases where FP is repeated, fast responses occur, because responding was just previously reinforced at the same imperative moment. In cases where FP alters from long on the preceding trial to short on the current one, slow responses occur, because the imperative moment was just previously bypassed. This bypassing without response is thought to extinguish previous moment–response associations or at least to reduce the strength of their association. Finally, in cases where FP alters from short to long, again fast responses occur, because later imperative moments were not just previously bypassed, and, thus, their response associations were not extinguished or loosened. As a result, responses in trials with the longest FP_ns are predicted to be consistently fast and not subject to sequential effects. In sum, the conditioning account predicts asymmetric sequential FP effects, since a long FP_n−1 slows RT on a short-FP_n trial but not on a long-FP_n trial. A short FP_n−1, however, should not produce any slowing, neither on short- nor on long-FP_n trials. According to this model, the strength of the association between any given imperative moment and a response should increase with increasingly long FP_ns and should be maximal at the latest imperative moment. The strength of this moment–response association is assumed to be directly related to the preparatory state at this moment, which, in turn, is thought to facilitate responding (Los and Heslenfeld, 2005, Van der Lubbe et al., 2004).

As mentioned above, mental fatigue from prolonged time on task (TOT) has been shown to impair simple RT performance. Previous research, however, mainly focussed on the effects of TOT on overall RT performance. To our knowledge, only one study has investigated TOT modulations of FP effects so far: Björklund (1992) reported that RT after long FP_ns increased more over 80 min than did RT after short FP_ns. Unfortunately, he did not analyse sequential FP effects. Here we explored whether TOT-induced mental fatigue affects the complex sequential dependencies within RT patterns, which are typical of temporal preparation in variable-FP designs. Based on the premises (1) that TOT-induced mental fatigue mainly impairs tasks involving top-down attentional control, (2) that the processes underlying temporal preparation according to the strategic view do involve top-down attentional control, and (3) that the processes underlying temporal preparation according to the conditioning view do not involve top-down attentional control but rather bottom-up learning, we derived the following hypothesis: If the typical RT pattern in variable-FP experiments were mainly based on strategic, top-down attentional processes, it should suffer from mental fatigue, whereas if it were mainly based on trace conditioning, it should remain rather unaffected by fatigue.

Specifically, from a strategic view of temporal preparation (e.g., Näätänen & Merisalo, 1977), one would predict that mental fatigue reduces or even eliminates the FP_n effect and subtly changes the asymmetric sequential FP effect: regarding the FP_n effect, a strategic view would predict that TOT impairs the ability to increase preparation with an increase in the conditional probability of IS occurrence in long-FP_n trials, since this requires an effortful process of conditional probability monitoring during the FP interval (cf. Vallesi & Shallice, 2007, for a recent discussion). As a result, RT in long-FP_n trials would increase, and the typical downward slope of the FP_n–RT function would dwindle or even vanish. This prediction is consistent with the results of Björklund (1992). Analogously, regarding the asymmetry of the sequential FP effect, a strategic view would predict that TOT impairs the maintenance or restoration of a prepared state when imperative moments occur later than expected (in long-FP_n trials following a shorter FP_n−1). Further, it can be reasonably assumed that the FP-repetition expectancy does not change with TOT, since expecting a repetition appears to be the rather effortless default option. In sum, TOT-induced fatigue should bereave later imperative moments of their benefit from increased preparation after FP repetitions or maintained/restored preparation after shorter FP_n−1s, whereas it should spare the benefits of a short-FP repetition as well as the costs of an earlier-than-expected imperative moment (see Fig. 1, Panel C, for an idealized visualization of the predicted outcome pattern). In contrast, from the perspective of the trace-conditioning model (Los and Van den Heuvel, 2001, Los et al., 2001), no significant changes in the RT pattern with TOT would be predicted, since the mechanisms assumed to underlie temporal preparation in this model do not involve top-down control processes. Instead, the model is solely based on the associative learning of temporal contingencies between warning signals and imperative stimuli. Accordingly, both the FP_n effect and the asymmetry of the sequential FP effect result from effortless, automatic processes and should not be affected by fatigue (see Fig. 1, Panel B, for an idealized visualization of the predicted outcome pattern).

To summarize, we investigated whether or not TOT-induced mental fatigue influences temporal preparation under time uncertainty. We derived two competing predictions based on two different explanations for the RT pattern typically found in variable-FP tasks: the traditional, strategic account, assuming top-down guidance of preparation, would predict a pronounced TOT-related RT increase at late imperative moments (in long-FP_n trials); the conditioning account, assuming bottom-up trial-to-trial learning, would predict no such interaction with TOT. To test these predictions, we conducted an experiment in which a warned simple RT task with variable FPs was performed over a time period of about 50 min.