Choking under pressure in sensorimotor skills: Conscious processing or depleted attentional resources?

Abstract

Objectives

This study examined and compared the conscious processing hypothesis and the attentional threshold hypothesis as explanations for choking under pressure.

Design

A 2×3 (anxiety level × putting condition) within group design was employed.

Methods

Twenty experienced golfers with handicaps ranging from 0 to 12 putted using three explicit knowledge cues, three task-irrelevant knowledge cues, and a single swing thought cue under low and high anxiety to test these opposing hypotheses.

Results

Irrespective of anxiety the data revealed that putting performance was generally better in the swing thought condition requiring the mobilisation of less cognitive resources. Under increased cognitive anxiety putting performance deteriorated in the explicit knowledge condition, whereas performance did not deteriorate in the task-irrelevant and swing thought conditions, providing support for the conscious processing hypothesis.

Conclusions

These results suggest that the type and/or amount of conscious processing may influence the anxiety–performance relationship. Future research should combine qualitative and quantitative methods to gain a more complete understanding of this relationship.

Introduction

Empirical research over the last decade has heightened our understanding of the mechanisms underlying paradoxical performance, or choking under pressure. Baumeister (1984, p. 610) has defined choking as “performance decrements under pressure situations” and pressure as “any factor or combination of factors that increases the importance of performing well on a particular occasion”. It is commonly assumed that an increase in pressure is reflected in an increase in anxiety, with increases in state anxiety commonly used as an indicator of perceived pressure (e.g., Hardy, Mullen, & Jones, 1996; Jackson, Ashford, & Norsworthy, 2006; Masters, 1992; Mullen & Hardy, 2000). Two prominent attentional models, each making contrasting predictions concerning the mechanisms underlying the choking phenomenon, have been proposed to explain the debilitative effects of performance pressure and anxiety on performance. The distraction model (Carver & Scheier, 1981; Wine, 1971) maintains that performance pressure causes a distracting environment for the expert performer thereby drawing their attention away from skill execution. Conversely, the self-focus model (Baumeister, 1984) maintains that performance pressure heightens self-consciousness about performing correctly, which in turn leads to a more conscious and deliberate focus on the motor components of the performance thereby directing the expert performer's attention to consciously controlling movement with step-by-step control of the performance process. Research comparing the distraction and self-focus explanations of choking under pressure has yielded mixed results. There is support for the distraction hypothesis in tasks that load heavily on working memory (e.g., many mental tasks, such as mathematical computation) but support for a self-focus explanation in tasks that make few demands on working memory (e.g., high level motor skills, such as golf putting; Beilock & Carr, 2001; Lewis & Linder, 1997). In fact, Beilock and McConnell (2004) attribute the different mechanisms of pressure-induced performance decrements in mental and sensorimotor tasks to how such tasks are represented and implemented rather than the nature of the task. Sensorimotor skills and self-focus explanations are the focus of the present study.

The self-focus model fits well with theories of skill acquisition (e.g., Fitts & Posner, 1967). The novice performer generally begins with unintegrated explicit knowledge of the task that is overtly controlled in a step-by-step manner through working memory, which is a slow process and cognitively demanding. Through deliberate and repeated practice comes the structural refinement of this knowledge into the implicit knowledge of an expert, which is fast, effortless and covertly controlled largely without reference to working memory. As expert skills become well-learned and proceduralised (Ericsson, Krampe, & Tesch-Romer, 1993) attention to execution at the step-by-step level is thought to disrupt the automated functioning of such skills (Beilock & Carr, 2001; Langer & Imber, 1979; Lewis & Linder, 1997; Masters, 1992). Put simply, the incorporation of explicit knowledge through working memory may interfere with the execution of automatic processes, evoking the expert performer to regress to the erratic style of the novice.

It was from this learning progression that Masters (1992) elucidated the role of explicit and implicit knowledge in the disruption of the automaticity of a skill under pressure. Explicit knowledge consists of specific rules of which we are consciously aware and are able to verbalise, whereas implicit knowledge entails abstract, unconscious information we know but cannot articulate effectively (Berry & Dienes, 1993). In accordance with what is commonly referred to as the conscious processing hypothesis (CPH¹), Masters postulated that pressure situations may lead to a reversal of this learning progression whereby anxious expert performers ‘reinvest’ the explicit knowledge and control strategies characteristics of the novice performer in an attempt to maintain performance.

Masters (1992) and Hardy et al. (1996) first examined the CPH using a learning paradigm in which novice golfers were required to learn a golf-putting skill, either explicitly (with knowledge of rules) or implicitly (without knowledge of rules). In an attempt to suppress the acquisition of explicit knowledge, implicit learners were required to carry out Baddeley's (1966) random letter generation task in parallel with the putting task during learning trials. Consistent with the CPH, under conditions of heightened state anxiety the explicit learners’ performance deteriorated while the implicit learners’ performance continued to improve. However, the random letter generation task employed to suppress the accumulation of explicit knowledge during learning could have served to distract attention from the anxiety-related cognitions that were meant to be engendered by the anxiety manipulation. Consequently, participants may have become desensitised to self-generated verbal distractions during learning rendering them at least partially immune to the effects of state anxiety (Hardy et al., 1996; Lewis & Linder, 1997).

Hardy, Mullen, and Martin (2001) extended this early work using a performance paradigm designed to more closely replicate the breakdown of performance under pressure thought to occur when expert performers regress to conscious processing. Twelve female, national-level trampolinists performed their voluntary competition routines with and without shadowing coaching cues in both high and low anxiety conditions. Shadowing coaching involved the coach calling out technical points as task-relevant cues (i.e., explicit knowledge) designed to induce conscious processing. The results, which Hardy, Mullen et al. (2001) viewed as consistent with the CPH, showed that performance deteriorated in the high anxiety condition when shadowing was present. No performance decrements were observed under low anxiety with or without shadowing cues, or without shadowing cues when anxious.

Not completely satisfied with the conscious processing interpretation, Hardy, Mullen et al. (2001) offered an alternative hypothesis arguing that a threshold of attentional capacity may have been exceeded when increased state anxiety was experienced while an individual was exposed to shadowing cues. According to this attentional threshold hypothesis (ATH), a ‘chunk’ of attentional capacity is taken up by both anxiety-related cognitions (e.g., worry) and the coaching instructions, and while individually these components do not appear to diminish performance “together they may have operated additively and depleted the attentional resources available to maintain performance (Hardy, Mullen et al., 2001, p. 946).”

Although not discussed by Hardy, Mullen et al. (2001), working memory research offers plausible support for their attentional threshold explanation. Both explicit and implicit knowledge are kept in long-term store for future retrieval (Reber, 1993). While the application of explicit knowledge requires the availability of working memory, the application of implicit knowledge does not (Berry & Broadbent, 1988; Maxwell, Masters, & Eves, 2003). Working memory is also unsuited to processing information comprising implicit knowledge but is required to make executive decisions based on explicit knowledge (Curran & Keele, 1993). Therefore, the participants exposed to the coaching cues incorporated by Hardy, Mullen et al. (2001) would have required working memory to process these explicit instructions. If it is the case that anxiety reduces working memory capacity (Eysenck, 1992; Wine, 1971) then it may be that the deterioration in performance in the shadowing condition was a result of an anxiety-induced reduction in working memory capacity thereby limiting its ability to store and process explicit knowledge.

Only two investigations to date have examined the opposing explanations of the CPH and the ATH. Mullen and Hardy (2000) asked 18 experienced golfers to putt while verbalising explicit putting instructions, while generating random letters, and to putt as they would normally in conditions of low and high state anxiety. According to the CPH, the random letter generation task should prevent performers using explicit knowledge, while verbalising explicit putting cues should induce lapses into conscious control. In contrast, the ATH suggests that both the shadowing task and the random letter generation task consume attentional resources as they both require the assistance of working memory, and that these tasks would operate additively with anxiety to exceed the attentional capacity required for successful performance. The results supported the CPH as the performance of participants putting with explicit knowledge deteriorated under increased state anxiety, whereas the performance of those in the random letter generation condition did not, suggesting that participants maintained sufficient attentional resources for the dual-task performance.

More recently, Mullen, Hardy, and Tattersall (2005) replaced the random letter generation with a tone counting task due to verbalisation difficulties experienced by participants. Mullen et al. (2005) found that, in contrast to Mullen and Hardy (2000), participants’ putting performance was impaired by both explicit knowledge cues and the task-irrelevant (tone counting) secondary task under increased levels of state anxiety. While this evidence appears to support an attentional threshold explanation, a conscious processing interpretation cannot be totally ruled out as attending to explicit cues also resulted in performance decrements under high anxiety. It could be argued, therefore, that the performance decrements observed did not occur via a single mechanism such that conscious efforts to maintain performance and reductions in available attentional resources both lead to deterioration in performance (Mullen et al., 2005). The present study sought to address this limitation.

Research suggests that the negative relationship between performance and attention to task-related information may not extend to all types of task-related attention. Kanfer and Ackerman (1989), for example, demonstrated that self-regulatory activities enhance experts’ skill performance, whereas they are detrimental to novices’ skill performance. Similarly, Jackson and Wilson (1999) posited that some level of conscious processing might actually facilitate performance under pressure. They argued that instead of minimising the amount of explicit knowledge accumulated during learning as other authors have suggested (e.g., Liao & Masters, 2002), the focus should be on preventing individuals reinvesting that knowledge in pressure situations. Jackson and Wilson (1999) hypothesised that by initiating skill execution using a higher-level (i.e., global) cue word that collectively represents the explicit rules of skill execution, the potential for regression to conscious control could be reduced. If conscious processing can be limited to a global cue word this holistic representation of the performance process may reduce the risk of reinvesting more detailed and intrusive explicit knowledge (Hardy et al., 1996; Kingston & Hardy, 1997). Essentially, global cue words represent limited conscious processing that may attract the benefits associated with self-regulatory techniques without requiring a level of conscious processing likely to impede performance.

Jackson and Wilson (1999) examined experienced golfers’ putting performance under low and high state anxiety when using a variety of different attentional strategies, including swing thoughts, process goals during the set-up phase, and perceptual focusing techniques. Additional support was revealed for the CPH, as participants performance deteriorated under increased state anxiety when verbalising several explicit knowledge cues while the negative impact of competitive anxiety on putting performance was significantly reduced when participants verbalised a single global cue word, referred to as a swing thought in golf.

There are several aspects of the experimental manipulations employed by Mullen and colleagues (Mullen & Hardy, 2000; Mullen et al., 2005) that need to be considered when interpreting their findings. First, the conflicting results between the studies might be attributable to the differing task-irrelevant manipulations employed. For example, Mullen and Hardy (2000) required participants to generate random letters, whereas Mullen et al. (2005) employed a tone counting task, and it is possible that these two tasks make different demands on attentional resources. Second, the ecological validity of these manipulations must also be questioned. It would seem rather peculiar to observe an expert golfer verbalising explicit cues or random letters when putting during a competition. Mullen and Hardy (2000) noted that participants had problems with the verbal tasks and that throughout the task-relevant conditions “participants expressed difficulty with speech while verbalising their coaching points” (p. 790). This is not surprising given that the majority of athletes employ self-talk as a method for gaining control of their thoughts (Hardy, Gammage, & Hall, 2001). Importantly, self-talk is central to cognitive and cognitive-behavioural interventions for anxiety reduction and behavioural change (Beck, 1970; Ellis, 1979; Meichenbaum, 1977) and it is considered the key to cognitive control in athletic performance as it allows the body to act automatically by keeping the mind occupied (Zinsser, Bunker, & Williams, 2001).

Given the equivocal findings in the research literature (Mullen & Hardy, 2000; Mullen et al., 2005) the aim of the present study was to examine and compare the ATH (Hardy, Mullen et al., 2001) and the CPH (Masters, 1992) using more realistic experimental manipulations by requiring participants to think rather than verbalise different attentional cues in an attempt to enhance ecological validity. In three independent conditions participants were asked to: (a) focus on three task-irrelevant thoughts (e.g., three colours); (b) focus on three explicit cues (e.g., arms, weight, and acceleration) related to their own putting technique; and (c) focus on a single swing thought cue word (e.g., smooth) representative of their own putting technique.

If the CPH is valid then putting performance under heightened anxiety should deteriorate in the explicit knowledge condition but not in the task-irrelevant and swing thought conditions, as both the task-irrelevant and swing thought cues are hypothesised to prevent attention to step-by-step components of performance while the explicit knowledge cues should induce lapses into conscious control. However, if the ATH is valid, deterioration of putting performance should be evidenced in the explicit knowledge and task-irrelevant conditions under increased levels of state anxiety but not in the swing thought condition, as the former conditions are hypothesised to load more heavily on working memory resources than the single swing thought cue. The present study sought to examine these experimental predictions.