Introduction

In daily life, adapted motor behaviors are goal-directed. To achieve these goals, we must constantly process and integrate multiple competitive sources of information in order to readjust behaviors. Even for a casual daily activity, such as driving to the grocery store, the cognitive functions involved to achieve our goal are complex. They include keeping track of the grocery list and updating it in memory, retrieving and remembering the route to the destination, but above all, focusing on the road and monitoring drivers’ and pedestrians’ behaviors to adjust our own actions. The optimal coordination of all the involved cognitive processes is even more necessary in complex and unpredictable environments such as crowded streets. In such a situation, to ensure that our goal to stay safe until arriving to the grocery store is achieved, two control strategies can be adopted. The first is to be constantly alert of traffic information and to slow down when approaching the crowded area to more easily avoid potential accidents. The second, which is the opposite strategy, consists in not modifying our behavior when approaching the crowded area, but in relying on our ability to react when an accident is likely to occur (e.g., by braking if a pedestrian suddenly crosses the road). Although both strategies lead to the same goal, namely to avoid the accident, they differ in the way this goal is achieved. Investigating inter-individual differences in the implementation of these cognitive control strategies may be the key to gain a better understanding of adaptive behavior [1].

Cognitive control is a set of basic executive functions that guide and adjust goal-directed actions according to internal and external demands [2, 3]. According to the dual mechanisms of control framework [1], on the one hand, proactive control aims to anticipate future difficulties by facilitating the resolution of the competition between the multiple sources of information. To do so, it relies on the active maintenance of the goal in memory and the anticipatory selection of goal-relevant information to achieve it. Therefore, proactive control is costly as it implies working memory resources and sustained attention [1, 4]. A driver who turns down the radio to pay more attention to the road-relevant information engages proactive control resources. On the other hand, reactive control aims to resolve the competition at the time it occurs and consists in the retrieval of goal-relevant information when a difficulty is detected, leading to a novel decisional process and the late correction of erroneous behaviors [1, 4]. Reactive control is thus less costly than proactive control since it is recruited on an as-needed basis. The driver who brakes at the last minute if a pedestrian decides to cross illustrates reactive control. Therefore, proactive and reactive mechanisms rely on different temporal dynamics, but both use goal-relevant information in order to adjust goal-directed behaviors to external and internal demands.

Proactive and reactive control are two complementary but independent cognitive control mechanisms that co-exist in the cognitive control system [1]. They are not mutually exclusive as they are often simultaneously engaged during a cognitive control task to optimize behavioral outcomes [5, 6]. However, their recruitment to perform a task is weighted according to their respective cognitive costs, i.e., load of the cognitive resources needed for an optimal efficiency. Proactive control mechanisms rely on sustained attention to goal-relevant information and thus, highly consumes cognitive resources while enabling to resist attentional capture generated by distractors. In contrast, reactive control mechanisms rely on flexibility and inhibitory capacities and thus, enable fast adjustments to changes in the environment. Their drawbacks are that they are more sensitive to disturbances and thus, less robust [7]. Also, since they can take effect too late, i.e., only after the need for adjustment is detected, they are the last opportunity to successfully adjust actions. The weighted recruitment of the two control mechanisms depends, at least to a large extent, on these specific cognitive costs and benefits that are known to be modulated by both contextual and inter-individual characteristics [1, 4, 7]. It has been found that proactive control is favored by individuals with a high working memory capacity, as the cost induced by sustained attention is lighter for them compared to that experienced by those with a low working memory capacity [8, 9]. Older adults showing impairments in goal maintenance also favor the use of reactive control [10]. However, when these deficits have little or no impact on performance, for example in a low cognitive load condition, ageing was associated with a tendency toward proactive control [11, 12]. Indeed, the environmental context can also modulate the favored strategy by shifting it towards proactive or reactive control. High task-context load generally decreases the use of proactive control, as the goal-relevant information is too heavy to process continuously [11, 13]. On the contrary, a high motivational context favors the use of proactive control to ensure optimal efficiency of control [14, 15]. Therefore, the costs of the deployment of the control strategies vary as a function of contextual and inter-individual (e.g., personality, age, cognitive capacities) characteristics [e.g., 8, 9, 13, 16].

To our knowledge, no study has investigated whether potential limitations in reactive control capacities may be associated with a higher propensity to use proactive control. Indeed, if reactive control is insufficiently efficient, the successful online resolution of a competition between responses, i.e., before a response is executed, is uncertain. Therefore, to ensure the efficiency of the cognitive control system, proactive control should be strongly engaged to prevent the emergence of such competition. The weighted recruitment of proactive control mechanisms according to reactive control capacities would thus respect the principle of parsimony of cognitive systems, optimizing the efficiency of cognitive control while reducing its costs. Consistent with these statements, our general hypothesis was that individuals with effective reactive capacities would generally less rely on proactive control processes ensuring adapted behaviors without substantial cognitive cost. Conversely, individuals with low reactive capacities should be those who generally rely more on proactive control strategy ensuring adapted behaviors by maintaining the effectiveness of the cognitive control system despite the generated cognitive cost.

To test this hypothesis, two behavioral indices were chosen to reflect reactive and proactive control mechanisms. As reactive control rely on inhibitory capacities to correct actions in face of changes in the environment, we used the Stop Signal task to measure reactive control capacities through the Stop Signal Reaction Time (SSRT) [17, 18]. Indeed, the SSRT is a precise and reliable task evaluating the capacity to inhibit an ongoing but no longer appropriate response. Thanks to the adjustment of task difficulty to one’s performance, the SSRT is also more sensitive to inter-individual variability in a healthy sample than other simpler inhibition indices, such as the error rates in the Go/NoGo task [18]. Secondly, even if the literature does not suggest a large panel of behavioral measures of proactive control recruitment, the post-error slowing (PES) was considered in the current study as a satisfactory and reliable candidate to reflect the spontaneous use of proactive control [19]. Despite the ongoing debate on the involvement of control mechanisms in the PES [20, 21], the performance slowing after an error measured by the PES was showed to be associated with an increase in response caution [22]. More specifically, the slowing of RTs after an error might reflect the use of the error as a cue to proactively refocus attention on goal-relevant information. Thus, in the present study, SSRT and PES were used as indices for to reactive and proactive control mechanisms, respectively.

In the current study, we tested the general hypothesis postulating that individuals with low reactive capacities should rely more on proactive control strategy than individuals with effective reactive capacities (H1). Moreover, as adaptive behaviors rely on this flexible weighting of proactive and reactive control mechanisms [1], one could hypothesize that an inflexible weighting of the two control mechanisms would result in maladaptive behaviors (i.e., insufficiently controlled behaviors). Risk-taking is largely associated with many maladaptive behaviors such as substance abuse [e.g., 23, 24], eating disorders [e.g., 25–27], and aggression [28] and can be objectively measured with the Balloon Analog Risk Task [29]. Among the various questionnaires and experimental paradigms that can be used to assess risk-taking [30–32], the BART particularly benefits from a large literature on its relationship with real-life maladaptive behaviors such as drug use [33–35] and delinquency [36]. Our second hypothesis was that the effect of limited reactive capacities on the increased propensity to use proactive control would be less observed in individuals with high risk-taking propensity (H2). In other words, we hypothesized that higher risk-taking propensity would be associated with less proactive counterbalance as a function of reactive inhibition capacities.

Method

Results

Analysis of the confounded variables

The effects of two potential confounded variables (i.e., age and sex) were tested on the included sample (N = 176). The effect of the discrete variable on the three variables of interest was analyzed using Student’s t-test (cf. Table 1). There were no significant effects of sex on PES, SSRT or on RTI, all p-values > .050.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Means (M) and standard deviations (SD) of the post-error slowing (ms), the Stop Signal reaction time (ms) and the press duration (s) in the BART.

https://doi.org/10.1371/journal.pone.0277246.t001

The effect of the continuous variable on the three variables of interest was analyzed using simple linear regression. Results showed a significant effect of age both on the PES and on the SSRT, β = 0.47, t(174) = 2.43, SE = 0.19, p = .016, adjusted R² = 0.03 and β = 1.54, t(174) = 6.71, SE = 0.23, p < .001, adjusted R² = 0.20, respectively (see Fig 1A and 1B). However, there were no significant effects of age on the RTI, β = 0.00, t(174) = -0.70, SE < 0.01, p = .483, adjusted R² < 0.01 (see Fig 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Effects of age on the three variables of interest.

(A) the post-error slowing (ms), an index of proactive behavioral adjustments, (B) the Stop Signal reaction time (ms), an index of reactive inhibition capacity and on (C) the mean button press duration (s) the risk-taking index in the BART task.

https://doi.org/10.1371/journal.pone.0277246.g001

As age had an effect on both PES and SSRT (cf. Fig 1), age was set in the model as a random effect to control for its confounded influence. As sex had no effect on the variables of interest (cf. Table 1), sex was not added to the model.

Modulation of proactive adjustments as a function of reactive inhibition and risk-taking propensity

In the current study, we aimed at exploring the predictive value of the SSRT, and its interaction with risk-taking propensity (RTI) on the PES, while controlling for the effect of age. Longer SSRT significantly predicted greater PES, β = 1.36, t(77.94) = 2.16, SE = 0.63, p = .034. The RTI alone did not significantly predict the PES, β = 37.35, t(96.37) = 1.58, SE = 23.62, p = .117. However, results showed a significant interaction effect between SSRT and RTI on the PES, β = -0.17, t(82.17) = -2.08, SE = 0.08, p = .040. The greater the RTI, the less SSRT predicted PES (cf. Fig 2). This effect appeared to be robust. Indeed, when including participants whose performances on at least one of the two tasks did not meet the inclusion criteria (N = 571), we observed a similar significant interaction effect, β = -0.06, t = -1.98, SE = 0.03, p = .048.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Proactive behavioral adjustments as a function of reactive inhibition capacity and risk-taking propensity.

The SSRT (Stop Signal Reaction Time) was used as an index of reactive inhibition capacities. The PES (Post-Error Slowing) was used as an index of proactive behavioral adjustments. The risk-taking groups were created for visualization purpose only. We used the median of the RTI (Risk-Taking Index, the mean duration of the button-press at the BART task) distribution to categorize participants as high (N = 87) or low risk-takers (N = 86).

https://doi.org/10.1371/journal.pone.0277246.g002

Discussion

The aim of our study was to determine whether reactive inhibition capacities could constrain a general tendency toward proactive control and whether the weakening of this effect would be associated with risk-taking propensity. We first hypothesized that an individual with inefficient reactive control capacities would have a higher tendency to use proactive mechanisms to guarantee adapted behaviors. Results showed that longer Stop Signal Reaction Time (SSRT) significantly predicted greater post-error slowing (PES), thus confirming H1. Secondly, we hypothesized that the propensity to take risks would alter the increased weighting of proactive mechanisms in response to limited reactive capacities. We observed that the higher the risk-taking index (RTI), the less the SSRT predicted PES, thus confirming H2. These findings suggest that higher risk-taking is associated with a smaller weighting of proactive control to counterbalance poorer reactive inhibition capacities. Imbalanced weighting between reactive and proactive control could be a fine-grained index for the detection of maladaptive behavior.

Cognitive control relies on processes that can be defined as part of either reactive or proactive mechanisms. These control mechanisms are independently engaged to handle conflicting situations [1] and are associated with different cognitive costs and benefits [7]. The balanced weighting of proactive and reactive control is a function of inter-individual and contextual characteristics and favors one control mechanism over the other to limit cognitive costs and optimize behavioral outcomes. Interestingly, some studies have reported the effects of inter-individual factors (e.g., cognitive capacities, age, personality traits) on the preferential use of one mechanism over the other [e.g., 8–10, 16]. Two previous studies have demonstrated a tendency towards greater use of reactive control when proactive control was limited by poor working memory capacities, on which active maintenance of goal-relevant information is based [8, 9]. To the best of our knowledge, the literature had not yet addressed the opposite effect, i.e., the effect of limited reactive capacities on the tendency to use proactive control. In the current study, we thus hypothesized that the involvement of proactive behavioral adjustments would be increased in individuals with inefficient reactive inhibition capacities. Our results were in favor of this hypothesis. Indeed, the SSRT predicted the PES: larger SSRT (i.e., poorer reactive inhibition capacities) were associated with larger PES (i.e., stronger proactive behavioral adjustments after an error). Our current findings were in line with previous research that reported larger PES in individuals with lower cognitive resources, suggesting proactive control as a compensatory mechanism [47, 48]. Nonetheless, other studies have reported opposite results: impaired older adults showed less reliance on proactive control [10], which could seem inconsistent with our present findings reporting an increase in the PES with age and replicating the literature [47]. However, it is important to note that the PES reflect one proactive process of a more broad cognitive mechanism. Other proactive-related cognitive capacities, such as proactive inhibition, working memory or sustained attention, might be impaired in older adults, leading to the global less reliance on proactive mechanism reported in Paxton et al. [10]. Moreover, in our study, the task was performed in a low cognitive load context, that might have been not sufficient to reveal age-related cognitive deficits, explaining why elderly individuals could still rely on proactive control processes for this particular task [11, 12]. To complement these findings, future researches should investigate the balance between reactive capacities and the recruitment of proactive processes through age in a more cognitive challenging context.

The present study also aimed at investigating whether maladaptive behaviors could be associated with a maladapted weighting in the recruitment of control mechanisms according to the environmental context, and in particular in the current study, to one’s own capacities. The second aim was thus to investigate the influence of risk-taking propensity on the observed weighted recruitment of proactive and reactive mechanisms. Our results showed that higher risk-taking propensity was associated with a smaller increase in post-error slowing in response to limited reactive inhibition capacities. In other words, larger SSRT predicted to a lesser degree the PES with an increase in risk-taking propensity. The imbalanced weighting between the two control mechanisms with high risk-taking propensity reduced the global cost of control, since less proactive and costly behavioral adjustments were engaged. However, this strategy may also reduce control efficiency. Therefore, these findings have direct applicability in the general population, for which the understanding of the emergence of maladaptive behaviors is crucial. Currently, maladaptive behaviors are mostly viewed as resulting from the lack of inhibition, but numerous studies are more recently showing an absence of association between disinhibition and various maladaptive behaviors [49–52]. The current findings suggest that the assessment of the adapted weighting of reactive and proactive control mechanisms could be useful to create a finer-grained index of an individual’s tendency to adopt maladaptive behaviors. At the behavioral level, the efficient recruitment of proactive control mechanisms could hide and compensate for limited reactive inhibition capacities. An individual knowing that he/she is unable to stop the car in time to avoid an accident should slow down to efficiently drive. If such strategy is adopted then, this individual should not be considered as an at-risk individual. However, the same individual would be at-risk in another context requiring the use of reactive control. On the other hand, an excessive recruitment of proactive control mechanisms in individuals whose reactive capacities are efficient would generate a cognitive overload for the control system. Such individuals would be characterized by high cognitive fatigue, resulting in decreased performance [53].

In the current study, three specific behavioral measures were chosen to assess inter-individual variability in the spontaneous tendency to use proactive control, reactive control capacities and risk-taking propensity. However, as already mentioned, the literature offers other indices that could have also been used to differently evaluate reactive capacities and the general tendency to use proactive control, as other ways to assess risk-taking. Therefore, the performed analysis of their relationships reported in the current paper is one way to test our hypotheses. In future studies, alternative indices could be used to investigate different facets of reactive and proactive control and hence, consolidate the reported findings. Firstly, it would be interesting to use an index of reactive inhibition that is internally generated, such as reflected in the partial-errors and the correction ratio [54, 55]. Indeed, the SSRT reflects an externally triggered reactive inhibition, whereas the correction of partial-errors operates without an external cue. Secondly, the PES is as a short-term proactive adjustment as it decays with inter-trial interval [56]. Hence, the PES is not an index of long-term strategic adjustments that are reflected in the global slowing of reaction times as a function of the Stop likelihood in the Stop Signal task [57]. Also, the proactive behavioral index calculated on the performances at the AX-CPT task would also be an interesting way to measure the reliance on proactive control strategy as a function of reactive capacities. The analysis of the relationship between partial-error correction ratio and the global slowing of reaction times or the proactive behavioral index could be a further step in the understanding of the weighted recruitment of reactive and proactive control mechanisms in healthy individuals. Finally, other maladaptive behaviors than risk-taking could be considered. The relationship between the imbalanced weighting of control mechanisms and disadvantageous decision-making, evaluated in the Delay Discounting task [58], could support our findings.

Conclusions

The present study demonstrated that reactive inhibition capacities predict the general tendency towards proactive control, but this effect is reduced with higher risk-taking propensity. To our knowledge, these are the first empirical arguments that confirm a stronger reliance on proactive control in individuals with low reactive control capacities. Also, poor inhibition capacities may not be sufficient enough to index the tendency to adopt maladaptive behaviors, but the absence of a proactive counterbalancing could be a finer-grained index to identify at-risk individuals.

Acknowledgments

We thank the company ECCA Conduite and their team of psychologists for their help in data collection.