Readiness for Perception and Action: Towards a More Mechanistic Understanding of Phasic Alertness

Where is the locus of alerting effects in the cognitive processing chain?

Decades of research have been devoted to the quest of finding the locus of alerting (or warning signal effects) within the chain of cognitive processing. Some accounts posited that alerting reduced reaction times in simple decision tasks by speeding up motor processing and response execution (Posner, 1978; Sanders, 1980). Other accounts suggested perceptual processing (Jepma et al., 2009; Rolke, 2008) or later perceptual and response selection processes as cognitive loci of the alerting effect on reaction time (Fischer et al., 2010; Hackley, 2009; Hackley & Valle-Inclán, 1998; 1999; 2003). It was long thought that this issue had been settled with studies that partitioned reaction times according to components in the electroencephalogram (the lateralized readiness potential). These studies showed that processes before motor processes of response execution were shortened by alerting (Hackley & Valle-Inclán, 1998). Based on these findings, it had been assumed that phasic alertness elicits its effects by speeding up (“late”) perceptual processes and/or response selection (Hackley, 2009). More recent work demonstrated that alerting also affected early perceptual encoding of visual information. That is, alerting increased the sensitivity of visual perception (Kusnir et al., 2011), speeded up visual processing for encoding into visual short-term memory, and led to an earlier start of visual processing (Haupt et al., 2019; 2021; Matthias et al., 2010; A. Petersen et al., 2017; Wiegand et al., 2017). As such, these findings provide unequivocal evidence that alerting also affects perceptual processing that is assumed to happen before response selection. We should note, however, that this does not mean that alerting solely affects perceptual encoding. As explained above, alerting affects a wide range of cognitive functions, and these functions could be provided by mechanisms at different processing phases or stages. Thus, phasic alertness seems to propagate through the entire cognitive system (e.g., by means of modulation of cortical processing by norepinephrine (NE), Posner & Petersen, 1990, see below). The sizes of alerting effects at different phases or stages of processing may differ, and could combine so that the largest effect emerged at the response selection stage (not later, as alerting effects seem to happen before the motor state, see above Hackley & Valle-Inclán, 1998).

The described “processing-stage accounts” attribute alerting effects to an influence on one or (simultaneously) more of several stages of processing (for instance, stages such as early visual processing, figure-ground segregation, attention-dependent encoding into visual working memory, response selection; see, Bundesen et al., 2005; Pashler, 1994; Treisman & Gelade, 1980, for some examples of stage-based processing models). For example, alerting could selectively shorten the time required for processing at the response selection stage (see above). In contrast to processing-stage accounts, one may also suppose that alerting influenced all cognitive processing for a certain time-window after the alert, irrespective of what process was currently active and irrespective of the processing stage at which the process was located (Jankovic et al., 2022). This idea seems in line with the proposal that the phasic noradrenaline release that could underlie phasic alertness implemented a temporal filter that amplified processing in a time-window (Aston-Jones & Cohen, 2005). The idea also fits to the conception of phasic alertness as non-specific with respect to specific cognitive processes, which matches the diversity of cognitive functions that have been found modulated by alerting (see above, section Phasic alertness and behavior). That is because the cognitive functions should differ in the processing stages of their underlying mechanisms, so that it would seem unlikely that a single stage accounted for the diversity of effects. What is problematic, however, is that the time-based account seems difficult to test empirically. For instance, a recent study investigated how alerting affected performance in a simple visual search task supposed to mainly require target detection as compared with a more complex compound visual search task supposed to require target detection and discrimination (Jankovic et al., 2022). This study found alerting effects only for the simple and not for the complex task. This was interpreted in favor of the time-based account. For simple tasks, phasic alertness sped up processing for detection. In contrast, for the complex task, detection was sped up by phasic alertness as well. However, the modulation by phasic alertness was short-lasting, and so that a supposed additional and subsequent process of task-set reconfiguration that was necessary for the discrimination was not sped up. Now, it was assumed that this process was sped up when there was no alert, because the onset of the visual search stimuli themselves induced phasic alertness. Therefore, the speeding up of the detection and the speeding up of task-set reconfiguration for discrimination canceled out, resulting in the absence of alerting effects. What argues against these hypotheses, are opposing findings showing that alerting in fact also improved visual search performance in highly difficult compound visual search tasks (Dietze & Poth, 2024). In addition, we should note that the described processes of detection and discrimination as sequential operations are not the only way to conceptualize attention in visual search, and may also be realized within the same process in parallel (Bundesen, 1990).

Based on the above considerations, we can arrive at the opinion that both studies remain insufficient for testing the time-based account of phasic alertness. This also points us at a general problem of assuming sequential processing stages on the one hand and time-based effects on the other hand. That is, when we assume processing stages, for example, target detection, task-set reconfiguration, and target discrimination, in visual search, we typically leave their durations unspecified. Thus, as in the example of the compound visual search task, alerting by the alerting cue may speed up one process, whereas the onset of the search stimuli may speed up another process, so that both improvements cancel out. Therefore, it seems necessary to independently manipulate or measure the durations of the different processing stages, in order to be able to quantify whether or not such a cancelling-out had taken place. Only this would allow to interpret null effects of alerting as evidence for time-based accounts. In other words, finding alerting effects in different tasks that rely on different processing stages cannot alone decide between processing-stage or time-based accounts: Even if the time-based account was true, alerting effects (rather than null effects) should still emerge even in more complex tasks that involve more processing stages. If one assumes that the onset of a stimulus such as the search display constituted a kind of sensory transient in the input stream changing the level of alertness, and if one assumes that the timing of this effect differed from that of the alerting effect induced by the alerting cue, then the two effects may not cancel out each other in the comparison of no-cue and alerting-cue condition. One way to approach this problem may be to vary alerting-cue-target onset asynchronies (CTOAs) and alerting cue intensities randomly, and then reverse-correlate their effects with task performance (cf. Okazawa et al., 2018). However, this endeavor may be non-trivial, because, as argued below (see section Phasic alertness and top-down temporal preparation: Distinct processes?), the probability distributions chosen for this random variation may greatly affect the results by inducing their own effects on perceptual and cognitive processing (e.g., by inducing temporal expectations, Vangkilde et al., 2012; 2013; Weinbach & Henik, 2012a) or threshold adjustment to the mean intensities across trials (Levison & Restle, 1968).

To conclude, given the diversity of cognitive functions that are affected by alerting and the hierarchical organization of perceptual and cognitive processing, it seems unlikely that there was a single locus in the processing chain at which alerting effects arise. Rather, it seems more likely that alerting effects arise at multiple levels of processing and propagate forward through the levels of processing.

Does phasic alertness affect perception and cognition online or offline?

The current notion of phasic alertness is that it is a temporary state of readiness for perception and action (S. E. Petersen & Posner, 2012; Posner & Petersen, 1990; Sturm & Willmes, 2001). This means that being phasically alert prepared the cognitive system for perception and action. The state of phasic alertness was reached first, and exerted its effects prospectively by modulating the cognitive processing that followed (Brown et al., 2015; Hackley & Valle-Inclán, 2003; Jepma et al., 2009; Tona et al., 2016). Therefore, the current state of alertness should influence ongoing cognition in an online fashion. Phasic alertness as an online-influence on cognition seems intuitive because the sequence of cognitive processing would match the sequence of events in experimental paradigms: I) The alert induced phasic alertness, II) phasic alertness modulated current cognition (online), III) impacting on the processing of target stimuli and responding within the task. Critically, however, we should also consider that the timing and sequence of events in the experimental paradigm need not be the same as the timing and sequence of cognitive events happening in the brain. Thus, in addition to prospective influences on cognition that happened in an online-fashion, phasic alerting may also have offline effects on perception and cognition. Specifically, alerts could in some way be encoded with the information about the target stimuli of the task and then affect their processing later on, offline, once they are decoded for selecting a response, for example. For instance, alerts (simultaneous accessory stimuli) can reduce the temporal uncertainty of target stimuli that are to be reported later on (Lippert et al., 2007). We could imagine that the target stimuli of a task are encoded as a signal in a temporal stream of information and noise (see Figure 1, upper panel). Now, an alert could be encoded in some way within this stream as well. For example, the alert could modulate the input stream by increasing its signal or noise amplitude (see Figure 1, lower panel) or by adding noise. The perceptual analysis of the input stream happens later on, after initial encoding, and after the modulation of the input stream by the alert. Now, before the input stream is analyzed perceptually to extract the signal from the stream, it is processed by a temporal filter (cf. Aston-Jones & Cohen, 2005, as discussed below). The filter could be set in a way that lets through only those parts of the input stream that deviated from the rest of the input stream according to their temporal and intensity characteristics. So, when the alert increased the amplitude of the input stream (Figure 1, lower panel, red graph), it enabled the later filter to restrict the later perceptual analysis to only a part of the input stream, rather than requiring analysis of the full stream. Such an influence of alerting on cognitive processing would be “offline”, in the sense that the filter was applied to the input stream after the alert, and then affected the efficiency of the subsequent perceptual analysis stage, but did not affect the computational processes at this stage directly (only passively, by modulating the input stream beforehand, Figure 1). Thus, the alert would be similar to a temporal landmark allowing to temporally filter the signal (cf. Aston-Jones & Cohen, 2005), exclude parts of the stream that are unnecessary, and focus computationally demanding perceptual processing only on the interval after the alert. This would explain findings that two successive alerts (i.e. an alert and a so-called accessory stimulus appearing simultaneously with the target) offered smaller benefits for choice reaction than a single alert (Poth, 2020). The two alerts could force the analysis stage to consider two parts of the input stream, increasing the amount of input that needs to be analyzed and induce some temporal uncertainty regarding where in the input stream the target signal could be found. This would then decrease the benefits from the alerting cues as compared to a situation with only one alerting cue. In addition, if the two parts of the input stream were represented separately at the perceptual analysis stage, there could also arise a competition between them for processing, calling for attention mechanisms allocating available processing resources to the parts (Bundesen, 1990; Bundesen et al., 2005). Prioritizing each part equally would lead to a situation of divided attention (Bundesen, 1990; Bundesen et al., 2005), compared with a situation of only one alert, this would mean that the additional alert would be “distracting”, in the sense that it consumed processing capacity that would otherwise be available for processing the other part. Thus, even though it seems reasonable to assume that alerting worked prospectively in an online-fashion (S. E. Petersen & Posner, 2012; Posner & Petersen, 1990; Sturm & Willmes, 2001), it seems fruitful to speculate about potential offline-effects on subsequent perception and cognition, as this forces us to consider the role of the processing architecture (i.e. which temporal filter is applied at what stage of processing). This idea is in line with recent calls for entertaining both, prospective and retrospective thoughts about cognition more generally, rather than restricting models to one of these alternatives (Herzog et al., 2020; Hogendoorn, 2022).

Figure 1

Extrinsic stimulus-driven versus intrinsic phasic alertness

The effects of phasic alerting on other cognitive processes should depend on the baseline level of tonic alertness, and this relationship should be nonlinear, for instance, inverse U-shaped. Alerting effects should be large if tonic alertness had been intermediate, but should be small if tonic alertness had been low or high (Figure 2).

Figure 2

This is in line with commonly held views of the neurophysiological underpinnings of phasic alertness. Alertness is assumed to rely on cortical arousal that is modulated by the the locus-coeruleus (LC) norepinephrine system (Aston-Jones & Cohen, 2005; S. E. Petersen & Posner, 2012; Posner & Petersen, 1990). The LC has widespread connections to cortical neurons implementing diverse cognitive functions, which makes the LC-NE system a candidate for underlying the various behavioral effects of alerting (McBurney-Lin et al., 2019; Poe et al., 2020; Sara, 2009). Alerting is thought to result in a brief release of norepinephrine by the LC in the brain stem (e.g., Brown et al., 2015; Coull et al., 2001; Tona et al., 2016). The LC-NE system seems to operate in two different modes, a tonic and a phasic mode (Aston-Jones & Cohen, 2005). The tonic mode of LC-NE activity sets the baseline norepinephrine release (Aston-Jones & Cohen, 2005), and at the level of the cognitive system, this baseline may correspond to tonic alertness. Tonic LC-NE activity has been found to relate to cognitive performance in a nonlinear, inverted-U-shaped function, with poor performance for low and high tonic activity, and highest performance at intermediate activity levels (Aston-Jones et al., 1999). In this way, the relationship follows the classic Yerkes-Dodson-Law (Aston-Jones & Cohen, 2005; Yerkes & Dodson, 1908), as an instance of the more general principle of hormesis biology and pharmacology (Calabrese, 2008). In the phasic mode, short bursts of NE release happen time-locked to the decision processes involved in the current task, which supports task performance (Aston-Jones & Cohen, 2005). Specifically, the bursts of NE release could implement a temporal attention filter that is general and non-specific to given processes and works by enhancing just the current processing for the task (Aston-Jones & Cohen, 2005). The pattern of phasic and tonic LC-NE activity resembles the relationship between phasic and tonic alertness described above: At low and at high tonic activity, phasic bursts of NE release cannot be observed, thus preventing any phasic modulation of NE at floor or ceiling levels. In contrast, at intermediate levels of tonic activity, phasic burst of NE become visible and could affect perception and action. Thus, taken together, and according to this assumption of a common mechanistic basis of phasic and tonic alertness, the effects of phasic alertness at the behavioral and the neurophysiological level should directly depend on the current level of tonic alertness (see above). Currently, however, this problem is mostly ignored by research on alerting. Studies on phasic alerting use experimental paradigms that neither assess nor explicitly manipulate participants’ current level of tonic alertness and instead only measure the effect of the alerting manipulation (cf. Figure 2, see the studies in section Phasic alertness and behavior) (but see Asanowicz & Marzecová, 2017; Martella et al., 2014; even though experimental paradigms exist that combine phasic alerting and measures of tonic alertness in the sense of vigilance, Luna et al., 2018). Thus, most studies are unable to detect where participants are on the inverted-U function that relates alerting effects to the level of tonic alertness (Figure 2). It is important to consider when this would be a problem for studying phasic alertness and when it would be of little concern. It would be problematic, when the source of low or high tonic alertness was located in the experiment. At high levels of tonic alertness, the effects of alerting compared with situation without alerting could be concealed. This could be the case in experiments involving challenging tasks or in which a challenging task is required in addition. Likewise, at low levels of tonic alertness, such as in situations with a low overall stimulation and low frequency of events (e.g., Hancock, 2017) alerting effects may likewise be concealed, either because processing happens at such a low level that it cannot be scaled by alerting (cf. Bundesen et al., 2015 for a similar idea for attention) or because the alert would come as a surprisingly strong stimulus that may inhibit responding and thereby hinder the manifestation of alerting effects in behavior (Dietze et al., 2024, see below). The potential effects of tonic alertness on phasic alerting effects would be of smaller concern, if the source of tonic alertness resided in individual participants (e.g. persons with naturally higher or lower tonic alertness as a personal trait), as long as phasic alerting effects are assessed within subjects or groups for between-subjects designs are large enough and randomly sampled from the population. As mentioned above, phasic alertness not only means that alertness is increased phasically, that is, in a short-lived fashion. Rather, a sometimes neglected feature of of phasic alertness is that it is defined to occur in response to an eliciting external stimulus (Sturm & Willmes, 2001). For this reason, phasic alertness is also called extrinsic alertness, as opposed to intrinsic alertness that is used synonymously with tonic alertness (Sturm & Willmes, 2001). These distinctions are problematic for several reasons. First, the tonic intrinsic alertness and phasic extrinsic alertness confound the time-scale of variation (longer vs. shorter) with the source of alertness (internal vs. external). Thus, by using this definition one commits to exclude intrinsic and short-lasting variations of alertness and extrinsic longer term variations of alertness, even though both seem perfectly possible. Second, for phasic alertness in fact, we can assume that it can be elicited not only by external stimuli but also by internal processes. For example, urgency (time-pressure) has been found to phasically boost stimulus-driven action, even in violation of current behavioral goals (Krause & Poth, 2023; Poth, 2021; Salinas et al., 2019). This observation resembles effects of alerting in cognitive control paradigms (Callejas et al., 2005; Fan et al., 2002; Fischer et al., 2010; MacLeod et al., 2010; McConnell & Shore, 2011; D. W. Schneider, 2018; 2019a; 2019b; Weinbach & Henik, 2012b) and in fact has been speculated to be due to phasic alertness (Poth, 2021). This is in line with neurophysiological findings showing that the LC-NE boosts supposed to underlie phasic alerting also occur in response to internal events, such as the outcomes of decision processes (Nieuwenhuis et al., 2005). Thus, in sum, we should assume an extrinsic form of phasic alertness that is elicited by external stimuli, but should also be open to intrinsic forms of phasic alertness triggered by internal events (estimated elapsed time in urgent tasks, Krause & Poth, 2023; Poth, 2021; Salinas et al., 2019).

Phasic alertness as a bottom-up modulation of perception and action

The term alerting has been used with considerable heterogeneity throughout the literature, sometimes referring to the process leading to phasic alertness (A. Petersen et al., 2017; Poth, 2020), but sometimes also referring to cuing attention to prioritize processing at a specific time (Coull et al., 2001), or to increasing the temporal expectation for stimulus occurrence (as discussed by Weinbach & Henik, 2012a). Therefore, it is often not clear which underlying mechanisms should be implicated for interpreting empirical results (Hackley, 2009; Weinbach & Henik, 2012a).

Phasic alertness is assumed to come from a bottom-up process driven by the stimulus of the alerting cue (for discussion, see Hackley, 2009; Poth, 2020). As such, it should depend on the physical characteristics of the stimulus, such as intensity, timing and duration, and its stimulus modality (e.g., Dietze & Poth, 2023). Congruent with this assumption, alerting effects on reaction times in choice tasks increase (up to a point) with increasing intensity of the alerting cue (Adams & Behar, 1966; Petersen et al., 2017; and increasing CTOAs, Dietze & Poth, 2023). Bottom-up phasic alertness should be relatively short-lived. It should be long-lasting enough to improve fast perception and action after a warning stimulus. But it should be short-lasting enough to regain susceptibility to subsequent warning stimuli and to prevent that alerting-induced impairments in resolving cognitive conflicts (or interference between cognitive processes) last long and lead to large detriments of behavior. Recent experiments show that alerting effects last at least for cue-target onset asynchronies of up to 1500 ms (Dietze & Poth, 2023). This may seem already quite long for striking the balance between supporting perception and action and impairing cognitive conflict resolution. For situations without cognitive conflicts, the duration of phasic alertness seems to be optimal for supporting the next action (Dietze, Recker, & Poth, 2023). Specifically, the next manual action in sequence of actions is speed up after an alert, whereas the next action after that falls already in the normal range of reaction times (Dietze et al., 2023).

In terms of stimulus modality, alerting traditionally has been assumed to be strongest for auditory stimuli (Bertelson & Tisseyr, 1969; Davis & Green, 1969; Harvey, 1980), in line with the idea that audition was the primary “warning sense” for humans (Dykstra et al., 2017). However, recent experiments reveal that audition and vision lead to similar alerting effects once their stimulus intensities are equated for processing speed based on the reaction times for a simple detection (Dietze & Poth, 2023). Thus, audition is not a primary warning sense with direct impact on phasic alertness. Rather, phasic alertness seems to be triggered qualitatively similar by the different sensory modalities. The differences in alerting effects seem to reflect low-level characteristics of the senses, such as their basic processing speed which is for example higher for audition than for vision (Recanzone, 2009; Torre et al., 1995).

Phasic alertness and top-down temporal preparation: Distinct processes?

Top-down temporal orienting of attention occurs when a cue indicates the time when the target stimulus for the current task was about to appear (Correa et al., 2010; Coull et al., 2001; Coull & Nobre, 1998). Similar to phasic alertness, temporal attention supports behavioral performance by reducing reaction times and accuracy in detection and discrimination tasks (Correa et al., 2004; 2005; 2010; Coull & Nobre, 1998). For temporal attention, participants must learn (by explicit instruction or implicit experience), that the target occurs at a specific time relative to the cue, so that processing at this time can be enhanced by attentional prioritization. What distinguishes temporal attention from phasic alertness is that temporal attention is flexible: A given cue could indicate that the target was going to appear shortly or after a longer interval (Correa et al., 2004; Coull & Nobre, 1998), and in both cases the cues can be used to support behavior. In contrast, phasic alertness is inflexible. It supports processing for a more or less ballistic time-window, so that processing after this time-window cannot be enhanced over the level within the time-window (Hackley, 2009; Hackley et al., 2009). After a manual action, the modulation of perception and action could be shut down, so that alerting effects are limited to only the next action (Dietze et al., 2023).

Temporal expectation supports behavior again akin to phasic alertness. It reduces reaction times in choice tasks (for a review, see Nobre et al., 2007), increases visual perceptual accuracy, and increases the speed at which visual information is encoded into visual short-term memory (Vangkilde et al., 2012; 2013). However, in contrast to phasic alertness, temporal expectation builds up over an experimental trial, according to the hazard rate at which targets occur after cues across the experiment (Vangkilde et al., 2012; 2013; for a discussion, see Weinbach & Henik, 2012a). The hazard rate specifies the probability that the target occurs now, given that it has not appeared up until now (Luce, 1991). For example, consider an experiment in which cue-target intervals range from 0.5 to 3 s, and in which all intervals within this range are uniformly distributed (i.e. equally probable). Across the experiment, participants thus learn that the more time has already elapsed during a trial, the higher was the probability that the target appeared now, eventually reaching certainty at an interval of 3 s. Probability distributions in which the hazard rate increases over time are said to be aging. In contrast, non-aging probability distributions of cue-target intervals have a constant hazard rate. This means that the probability that the target occurs at an instance, given it has not appeared before, is constant over trials (Luce, 1991; Näätänen, 1971; Näätänen et al., 1974). These distributions contain many more short intervals and fewer and fewer longer intervals. Examples for such distributions include the exponential and the geometric probability distribution, both of which have been used to control for the build-up of temporal expectation across trials (A. Petersen et al., 2017; Poth, 2020; Poth et al., 2014; Vangkilde et al., 2012; 2013; Weinbach & Henik, 2013). When cue-target intervals come from non-aging distributions, participants experience more often a short interval, and then less and less often a longer interval. Thus, they develop a strong expectation for short intervals, and with increasing interval duration, their expectation decreases. Now, evidence dissociating phasic alertness from temporal expectation comes from experiments using these non-aging probability distributions. Even though the cue did not predict the onset of the target stimulus, substantial alerting effects on choice reaction times (Lu et al., 2014; Poth, 2020; Weinbach & Henik, 2013), temporal perception thresholds, and the speed of visual processing (A. Petersen et al., 2017) have been found. These findings thus argue that temporal expectation and phasic alerting effects stem from different processing sources. However, at least one recent study found that if waiting times to a target in a condition without alert followed a non-aging probability distribution, choice reaction times still decreased with increasing waiting time (Dietze & Poth, 2023). These findings suggest that some form of temporal preparation still occurred, arguing that non-aging waiting times offer no guarantee that temporal expectation was controlled.

It is a crucial question whether one can dissociate phasic alertness from temporal expectation or temporal preparation more generally. It has been argued that the crucial difference between phasic alertness and temporal expectation is that the latter is based on temporal contingencies between cue and target stimulus (Weinbach & Henik, 2012a). However, we could speculate that it still reflected a rudimentary form of temporal preparation. The phasic alerting effect seems to develop most strongly at CTOAs below 500 ms (Fan et al., 2002; 2005; Hackley, 2009; Hackley et al., 2009) but seems to last at least up until 1500 ms (Dietze & Poth, 2023). These intervals could match the statistics of environmental cues that predict events relevant for action or action sequences (Dietze et al., 2023) in everyday life. Thus, through life-long learning, temporal preparation mechanisms could be established that then give rise to a phasic alertness that would be fast, occurred only within the learned time-window, and would otherwise seem bottom-up (note, however, that learning effects can be hard to categorize as top-down or bottom-up, Awh et al., 2012). This would argue that it is impossible to clearly dissociate phasic alertness from temporal preparation.

Another point against a distinction of phasic alertness and temporal preparation comes from the hierarchy assumed to be present in cognitive processing (e.g., Norman & Shallice, 1986; Posner, 1978). Specifically, it is assumed that cognitive processes happen in successive stages or phases, where one stage or phase received the processing result of the previous, which may be organized in processing episodes (Poth et al., 2015; Poth & Schneider, 2016; 2018; W. X. Schneider, 2013). Now, if the successive processing stages or phases operated at different processing speeds (e.g. due to temporal summation in neuronal processing, Zhou et al., 2018), alerts that have constant hazard for target appearance at the early stage could still deliver temporal information about target appearance at the later stage. For example, imagine that a non-aging distribution of cue-target intervals between 0 and 200 ms was encoded into the earlier stage, and from that stage, the later stage sampled information at 5 Hz (5 samples/s of 200 ms each). Then, all the possible waiting times for the target after the cue at the early stage would fall into one sampling cycle of the later stage. As a result, at the later stage, target appearance could be predicted and used to modify subsequent processing mechanisms. Thus, taken together, it may be impossible conceptually to dissociate phasic alertness from temporal preparation mechanisms completely (even in the absence of temporal contingencies, see e.g. Lawrence & Klein, 2013). We see that this calls for a change of strategy in phasic alertness research: Rather than asking whether or not it seems to offer more insight to ask to what degree and how phasic alerting effects were due to temporal preparation or expectation processes (experimental manipulations).