Elsevier

Cognition

Volume 205, December 2020, 104426
Cognition

Decision making in slow and rapid reaching: Sacrificing success to minimize effort

https://doi.org/10.1016/j.cognition.2020.104426Get rights and content

Highlights

  • Results reveal motor efficiency and cognitive limitations in the same task.

  • Humans are unable to flexibly adapt their movement strategies to task demands.

  • Consistent bias to select targets that required the least biomechanical effort.

  • Selected movements fail to maximize chance of success.

Abstract

Current studies on visuomotor decision making come to inconsistent conclusions regarding the optimality with which these decisions are made. When executing rapid reaching movements under uncertainty, humans tend to automatically select optimal movement paths that take into account the position of all potential targets (spatial averaging). In contrast, humans rarely employ optimal strategies when making decisions on whether to pursue two action goals simultaneously or prioritise one goal over another. Here, we manipulated whether spatial averaging or pre-selection of a single target would provide the optimal strategy by varying the spatial separation between two potential movement targets as well as the time available for movement execution. In Experiment 1, we aimed to determine the time needed to reach for targets with small and large separation between them and to measure baseline strategies under low time pressure. Given generous time limits, participants did not employ a pure averaging approach but instead tended to pre-select the target that was easiest to reach and corrected their movement path in-flight if required. In Experiment 2, a strict time limit was set such that the optimal strategy to reach the correct target depended on the separation between the potential targets: for small separations, there was enough time to employ averaging strategies, but higher success for larger separations required pre-selecting the final target instead. While participants varied in the strategies they preferred, none of them flexibly adjusted their movement strategies depending on the spatial separation of the targets. In Experiment 3, we confirm the bias toward targets that are easiest to reach and show that this comes at the expense of overall task success. The results suggest a strong tendency for humans to minimize immediate movement effort and a general failure to adapt movement strategies flexibly with changes in the task parameters.

Introduction

Imagine you are a goal keeper during a penalty kick. The decisions you have to make are complex and numerous. Before the kick, you need to choose a place to stand. Should you stand in the centre of the net or off-centre? Should you stand closer to the kicker, or inside the net? Should you start your dive before the kick and risk guessing the wrong direction, or wait to react to the kick and risk responding too slowly? When diving, which foot do you lead with? Will you turn your body toward the ball to get more distance, or face forward to create a larger blocking surface? Do you try to grab the ball, smack it away, or bounce it off another body part? The goal keeper's many dilemmas illustrate the cascade of motor decisions taking place at many different levels of deliberation and control. Some decisions can be epistemic, such as where to stand before the kick: these decisions can take into account explicit, conscious knowledge about how far you can dive in each direction and what you know about the opposing player's skills and history to increase your odds of success. Other decisions, like which foot to lead with when you start your movement, may automatically follow from an unconscious estimate of how to get your body to intersect with the ball, based on long hours of training.

Researchers have long been interested in both aspects: the efficiency of motor control and the optimality of human decision making. However, these domains have largely been investigated separately from each other, and are traditionally viewed as belonging to different fields (i.e. motor control vs. cognitive psychology). To understand how the goalkeeper's dive is executed, for example, researchers could measure the speed and trajectory of blocking movements in response to different kicks to understand visuomotor constraints on performance. In contrast, when investigating strategic decisions, such as where to stand before the kick, researchers measure deviations from optimal choices to understand the cognitive biases and heuristics that limit our success. Yet, it is clear from our goal keeper example that decisions exist on a continuum, and strong interconnections must exist between the cognitive systems that drive strategic decisions and sensorimotor control. During the last decade there has been a considerable increase in research attempting to bridge the gap between cognition and sensorimotor systems (for review see Gallivan et al., 2018; Wispinski et al., 2018).

One strand of studies that has provided initial evidence for the view that our action execution is strongly influenced by cognitive processes is based on the investigation of eye and hand movements in the presence of visual distractors (e.g., Findlay & Walker, 1999; Nakayama et al., 2007; Song and Nakayama, 2008, Song and Nakayama, 2009; Van der Stigchel et al., 2006; Walker et al., 2006). These studies demonstrated that the shape of the selected movement path toward a target is affected by the presence of distractor objects and their specific properties, suggesting that the visuomotor system initially represents both targets and distractors as potential movement locations. This demonstrates that the spatial properties of movement trajectories can provide insights into the underlying target selection process and can therefore, more generally speaking, reveal the internal cognitive processes (such as attention and decision making) that underlie our actions.

Based on these findings, numerous studies have investigated eye and hand movements in the presence of multiple potential targets and have confirmed that the motor system seems to prepare for different possible actions simultaneously, resulting in spatial changes of the observed trajectory to a final movement target. These spatial changes in the selected movement path have been considered to reflect the unsuccessful suppression of actions that were planned prior to execution but that were not called for (Cisek, 2012; Cisek & Kalaska, 2010; Kable & Glimcher, 2009). The notion that multiple potential movements are prepared by the CNS in parallel and compete for execution has been coined the “affordance competition hypothesis” by Cisek and Kalaska (2010). The behavioural effects of the competition between multiple action possibilities on subsequent movements to a target object are probably most effectively illustrated in so-called “go before you know” tasks (e.g., Chapman et al., 2010a; Gallivan et al., 2011; Gallivan & Chapman, 2014; Stewart et al., 2014). In these tasks, participants are presented with two (or more) potential action targets. They are then asked to reach to a final target as quickly as possible, but the target is only revealed once participants have initiated their movements (e.g., after the release of a start button during reaching). What is usually observed in these tasks is a movement path that is initially directed between the potential movement locations and is corrected in-flight toward the final target as soon as it has been revealed (for review see, Gallivan & Chapman, 2014). In other words, the initial movement is directed toward the average direction of all potential movement targets. This is known as spatial averaging.

Currently, there are two different explanations for the observation that the visuomotor system seems to take into account the location of all possible movement targets when planning the initial direction of the movement. The first is based on evidence from neurophysiological studies indicating that neurons in sensorimotor regions represent multiple potential targets and actions before a final decision is made (e.g., Cisek & Kalaska, 2005; Cisek & Kalaska, 2010). In line with the affordance competition hypothesis specified above, it has been suggested that the averaging behaviour is a consequence of the simultaneously specified competing motor plans and can therefore be considered unintentional behaviour (e.g., Gallivan et al., 2016; Stewart et al., 2013). The second, more recent, explanation is that these averaged movements are actually deliberate and may represent one single optimal movement plan (e.g., Haith et al., 2015; Hudson et al., 2007; Nashed et al., 2017; Wong & Haith, 2017). In other words, the observed averaging behaviour in “go-before-you-know” tasks may represent a deliberate and optimal strategy that participants implement to deal with the uncertainty of the task (Wong & Haith, 2017). The strategy can be considered optimal as the movement-related costs are minimised by initialising a movement toward the average location of all potential movement targets, which ensures that the required in-flight corrections are minimal once the final target is revealed.

These explanations are difficult to disentangle as the predicted movement trajectories are identical. A few recent studies have tried to address this issue by experimentally creating situations in which averaging behaviour is no longer beneficial (e.g., Haith et al., 2015; Wong & Haith, 2017). For example, Wong and Haith (2017) showed that participants abandoned averaging behaviour if they had to execute extremely fast movements. That is, averaged movements were found to be produced only at slower speeds, when there was sufficient time to correct them in flight. Similarly, Haith et al. (2015) found no averaging behaviour if the separation between potential targets was very large or a barrier was put between them. The observation that a change in high-level task requirements can eliminate averaging behaviour is clearly at odds with the idea that these movements are a result of competing motor plans and are thus automatic and unintentional. Based on these findings, it has been argued that averaged movements in decision tasks can be best understood within an optimal control theory, according to which a single motor plan is specified with the aim of minimising movement costs while at the same time increasing the chances of task success (Haith et al., 2015).

The suggestion that humans show nearly optimal behaviour when presented with a visuomotor decision task presents an interesting contrast with findings demonstrating clearly suboptimal performance in similar, but more deliberative, decision making tasks that require people to specify and select a movement strategy prior to movement onset. This seems to be true for a range of different tasks (Clarke & Hunt, 2016; Morvan & Maloney, 2012; Nowakowska et al., 2017). For example, Clarke and Hunt (2016) manipulated the distance between two potential targets in a beanbag throwing task based on participants' baseline throwing performance. Participants' task was to choose a standing position prior to the experimenter revealing which of the two targets participants had to hit with their beanbag. The rationale of this task is simple: If the two potential targets are separated by a short distance, participants should choose to stand midway between them (similar to spatial averaging in reaching) as this would ensure that they could easily hit either target after it is revealed by the experimenter. If the two targets are separated by a large distance, such that standing in the middle would make it difficult to successfully hit either target (as established through participants' baseline performance), participants should choose a standing position close to one of the potential targets, thus ensuring at least a task success of 50%. However, it was repeatedly found that very few participants ever select this (optimal) strategy with changes in distance between the targets, thereby falling far short of the maximum throwing success they could have achieved. Instead, participants seem to show highly variable standing position choices, resulting in inferior task performance (see Clarke & Hunt, 2016). The same failure to adjust choices with changes in task difficulty was observed in two other task contexts (memorizing digits and detecting targets), suggesting a general tendency that cuts across many contexts, from eye and hand movements to more complex decisions. Thus, in tasks that require participants to select movements in the presence of multiple potential goals there is a clear mismatch between more abstract and deliberative decisions on the one hand, and more implicit sensorimotor decision behaviour on the other, with the former appearing to be far from optimal, and the latter close to optimal.

A similar mismatch between visuomotor and deliberative/cognitive decisions has been revealed and discussed in the visuomotor literature using a paradigm investigating decision making under risk using eye and hand movements (e.g., Jarvstad et al., 2014; Maloney et al., 2007). In this paradigm, participants are asked to reach within a tight time constraint to a target region that incurs a small (usually monetary) reward. What makes this task a motor decision task is that the target can be overlapped with a penalty region that, when touched, incurs a small (monetary) loss. While it would be optimal to hit the centre of the target in the absence of a penalty region, in its presence participants should shift their movement end-point away from the target centre and the penalty region. The beauty of this task lies in the fact that the optimal hitting point (that is, the location that yields the maximum expected gain in points and money) can be precisely determined for each participant using their own motor precision and cost functions. Consequently, participants' performance can be easily compared to that of an optimal decision maker (Trommershäuser et al., 2003a, Trommershäuser et al., 2003b, Trommershäuser et al., 2008). The initial studies consistently found that participants select strategies that are close to optimal when making these speeded and risky visuomotor decisions, in contrast with the sub-optimal behaviour usually observed in traditional economic decision making tasks (Trommershäuser et al., 2008). More recent studies have, however, identified a number of boundary conditions for optimal behaviour to occur, specifically suggesting that consistent feedback and experience are essential (Neyedli & LeBlanc, 2017; Neyedli and Welsh, 2013, Neyedli and Welsh, 2014, Neyedli and Welsh, 2015) as well as relatively constant and simple gain landscapes (Jarvstad et al., 2014; Wu et al., 2006). Based on these findings, it has been argued that the perceptuomotor systems may not be as superior to higher-level cognitive systems as initially thought when it comes to optimal decision making (Jarvstad et al., 2014; Neyedli & LeBlanc, 2017).

What distinguishes this task from the go-before-you-know tasks outlined previously is the presence of externally imposed payoffs. Neyedli and Welsh (2014) varied both the distance between the target and the penalty region, as well as the associated pay-off values, on a trial-by-trial basis. Interestingly, they found that while participants seem to optimally shift their movement endpoints as the distance between the two circles varied, this was not the case for trial-by-trial variations in the associated penalty values. Based on this finding, they argued that spatial parameters, as an intrinsic property of the visual stimuli, are more tightly linked to the motor system than pay-off parameters, which need to be interpreted by the cognitive system. If spatial parameters are indeed easier for the motor system to optimize, then in go-before-you-know tasks, changes in appropriate task strategies depending on target distance might be more readily applied to optimally meet the demands of the task at hand.

In our current study, we combined the logic of Clarke and Hunts' choice task of pursuing one goal versus two, with a “go-before-you-know” reaching paradigm as employed in previous studies on sensorimotor decision making. Specifically, we varied the distance between the potential movement targets whilst limiting the amount of time available to participants to reach the final pointing target specified after movement onset. We thereby selected the available time to reach the final movement target such that averaging behaviour was feasible for targets separated by a short distance but not for targets separated by a large distance. The primary aim of Experiment 1 was to determine the approximate movement time limit that would allow participants to apply averaging behaviour for near distance targets but not for far distance targets. In addition, this experiment also allowed us to examine reaching behaviour in sensorimotor decision making under quite liberal timing conditions. We then applied a tight response deadline (as determined by Experiment 1) in Experiment 2 to assess whether participants would be able to flexibly adapt their movement strategies on a trial-by-trial basis based on the expected likelihood of reaching the target before the deadline. Participants were told that trials where the wrong target was hit or the correct target was not reached in time would have to be repeated and were regularly reminded how many trials they had successfully completed and how many they had left to do. We defined optimal performance as the strategy that would allow participants to maximize their probability of success on any single trial, and thereby also to finish the experiment in the smallest possible number of performed trials.1

We were particularly interested in the types of errors individual participants would make in Experiment 2: specifically, we determined the proportion of trials ending at the wrong target, and the proportion of total trials that failed to reach the target in time. Previous studies usually excluded movement errors from further analysis, which amount to the exclusion of a very large number of trials, and even entire participants who were unable to complete enough trials correctly within the given time-constraints (see Gallivan & Chapman, 2014 for detailed discussion). These exclusions may potentially lead to an incomplete picture of the (unsuccessful) strategies that participants employ when doing the task. For example, one could assume that for speeded movements as required in Wong and Haith's (2017) experiment, only straight movement paths would actually allow participants to reach the targets in time. Therefore, it is unsurprising that the trajectories of the correct movements are straight. To understand how participants solve the problem, it would be necessary to look at the frequency and types of errors they make. Optimal performance in our Experiment 2 requires participants to employ an averaging strategy for targets presented close to each other, but a pre-selection strategy (i.e. straight movement path) for targets separated by a large distance. If participants fail to adapt their strategies to the distance between the targets, we should see consistent use of a single strategy across all distances; this would mean more time-out errors at the far distances (if they over-apply an averaging strategy), and more movements to the wrong target at the close distances (if they over-apply a pre-selection strategy).

By examining error trials, we revealed, similar to Clarke and Hunt (2016), that none of our participants (N = 20) flexibly adapted their behavioural strategy to the different task demands. However, they did show a strong tendency to minimize effort on a trial-by-trial basis. To confirm these observations, in Experiment 3 we tested participants' decision-making behaviour in a visuomotor choice task by encouraging them, through the wording of our instructions, to select and move to one of the two potential targets, while still reinforcing speed and accuracy by repeating movement errors and time-outs. We found that participants readily adopted a pre-selection strategy and consistently chose targets that were associated with reduced biomechanical effort.

Section snippets

Participants

Twelve undergraduate and postgraduate students of the University of Aberdeen (aged between 18 and 28 years, mean age: 20.3 years, 6 males) who were naïve to the purpose of the experiment participated in the study. Undergraduate students received course credits for their participation. They all had normal or corrected to normal visual acuity and were right-handed by self-report. The experiment was approved by the School of Psychology Ethics Committee at the University of Aberdeen and all

Participants

Twenty-three undergraduate and postgraduate students of the University of Aberdeen took part in this experiment. Three participants had to be excluded from data analysis – one who was unable to execute movements fast enough and two who turned out to not be naïve to the purpose of the experiment. Thus, the final sample consisted of 20 participants (aged between 19 and 29 years, mean age: 22 years, 8 males). All participants had normal or corrected-to-normal visual acuity and were right-handed by

Participants

Twenty undergraduate and postgraduate students of the University of Aberdeen took part in this experiment (age-range 18 to 37 years, mean age: 21 years, 8 males). All participants were right-handed by self-report and had normal or corrected-to-normal visual acuity. They all provided informed consent prior to participating. Participants either received course credits or were reimbursed with £5 after participation. The study was approved by the School of Psychology Ethics Committee at the

General discussion

The aim of this study was to investigate which strategies humans would employ when the optimal movement strategy depended on the exact stimulus arrangement and varied from trial to trial. We demonstrate that participants can quickly and consistently select the more expedient of two targets, but on a broader level, their behaviour does not flexibly adjust to changes in the spatial configuration by shifting between averaging and pre-selection strategies on a trial-by-trial basis to enhance

CRediT authorship contribution statement

Constanze Hesse:Conceptualization, Methodology, Software, Validation, Formal analysis, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration.Karina Kangur:Conceptualization, Methodology, Validation, Investigation, Data curation, Visualization.Amelia R. Hunt:Conceptualization, Methodology, Writing - original draft, Writing - review & editing, Supervision, Project administration, Funding acquisition.

Acknowledgement

This work was supported by the James S. McDonnell Foundation (Scholar Award to ARH).

References (58)

  • B.M. Stewart et al.

    Motor, not visual, encoding of potential reach targets

    Current Biology

    (2014)
  • J. Trommershäuser et al.

    Decision making, movement planning and statistical decision theory

    Trends in Cognitive Sciences

    (2008)
  • S. Van der Stigchel et al.

    Eye movement trajectories and what they tell us

    Neuroscience & Biobehavioral Reviews

    (2006)
  • D.M. Wolpert et al.

    Motor control is decision-making

    Current Opinion in Neurobiology

    (2012)
  • J.F. Ackermann et al.

    Choice of saccade endpoint under risk

    Journal of Vision

    (2013)
  • E. Brenner et al.

    Fast responses of the human hand to changes in target position

    Journal of Motor Behavior

    (1997)
  • A.-M. Brouwer et al.

    Hitting moving targets

    Experimental Brain Research

    (2003)
  • D.P. Carey et al.

    Reaching to ipsilateral or contralateral targets: Within-hemisphere visuomotor processing cannot explain hemispatial differences in motor control

    Experimental Brain Research

    (1996)
  • P. Cisek et al.

    Neural mechanisms for interacting with a world full of action choices

    Annual Review of Neuroscience

    (2010)
  • A.D.F. Clarke et al.

    Human search for a target on a textured background is consistent with a stochastic model

    Journal of Vision

    (2016)
  • A.D.F. Clarke et al.

    Failure of intuition when choosing whether to invest in a single goal or split resources between two goals

    Psychological Science

    (2016)
  • A.D.F. Clarke et al.

    The saccadic flow baseline: Accounting for image-independent biases in fixation behavior

    Journal of Vision

    (2017)
  • I. Cos et al.

    The influence of predicted arm biomechanics on decision-making

    Journal of Neurophysiology

    (2011)
  • I. Cos et al.

    The modulatory influence of endpoint controllability on decisions between actions

    Journal of Neurophysiology

    (2012)
  • B. Day et al.

    Voluntary modification of automatic arm movements evoked by motion of a visual target

    Experimental Brain Research

    (2000)
  • J.M. Findlay et al.

    A model of saccade generation based on parallel processing and competitive inhibition

    Behavioral and Brain Sciences

    (1999)
  • J.P. Gallivan et al.

    Three-dimensional reach trajectories as a probe of real-time decision-making between multiple competing targets

    Frontiers in Neuroscience

    (2014)
  • J.P. Gallivan et al.

    Decision-making in sensorimotor control

    Nature Reviews Neuroscience

    (2018)
  • J.P. Gallivan et al.

    One to four, and nothing more: Nonconscious parallel individuation of objects during action planning

    Psychological Science

    (2011)
  • Cited by (12)

    • Card posting does not rely on visual orientation: A challenge to past neuropsychological dissociations

      2021, Neuropsychologia
      Citation Excerpt :

      Consequently, we expect that biomechanical effort plays a greater role in posting as compared to matching. The observation that during posting participants aim to reduce biomechanical effort, is consistent with a large amount of evidence from the motor control literature suggesting that humans aim to select and perform movements that are safe and efficient and minimise the biomechanical and energetic costs and thus the immediate effort involved in the action (Cohen et al., 2010; Hesse et al., 2020; Rosenbaum and Gaydos, 2008; Rosenbaum and Gregory, 2002; Ross et al., 2014). Furthermore, the finding that directional errors increase with the amount of rotation required is also consistent with the assumption that the posting task does not necessarily rely on orientation information but rather constitutes an obstacle avoidance task thereby supporting the strategy hypothesis (Hesse et al., 2011).

    • Embodied decisions during walking

      2022, Journal of Neurophysiology
    View all citing articles on Scopus
    View full text