Effects of uncertainty and high stakes on performance in skilled individuals

Performance anxiety (PA) affects between 25% and 40% of professionals and students across domains (Blöte et al., 2009; Fernholz et al., 2019; Herman & Clark, 2023; Miller & et al., 2022), significantly impacting health, education, and career progression. PA is characterised by intense apprehension in performance settings involving potential scrutiny, which can impair skilled performance under high-stakes conditions such as competitions or stage performances. It elicits changes in cardiorespiratory, cognitive, and emotional states, leading to varied outcomes—some individuals maintain performance under PA, while others exhibit performance decline (Kenny, 2009). Despite its prevalence, the neurocognitive mechanisms through which PA influences skilled performance remain poorly understood. A growing body of research suggests that a core feature of clinical and subclinical anxiety is a heightened maladaptive response to uncertainty (Pulcu & Browning, 2019). While uncertainty—defined as incomplete or ambiguous information—is ubiquitous in our interaction with the environment, individuals with anxiety may exhibit a biased perception of uncertainty, altering their learning and behaviour (Pulcu & Browning, 2019). This biased perception of uncertainty is conceptually related to the disposition known as intolerance of uncertainty (IU), a trait characterised by negative beliefs about uncertain situations and widely recognised as a key factor in the development and maintenance of anxiety (Carleton et al., 2010). At the neural level, the Uncertainty and Anticipation Model of Anxiety (UAMA) identifies brain regions such as the amygdala, ventromedial prefrontal cortex (PFC), and anterior cingulate cortex (ACC) as central to maladaptive responses to uncertain threats (Grupe & Nitschke, 2013). These responses sustain anxiety by promoting hypervigilance and behavioural rigidity. Studies using electroencephalography (EEG) and magnetoencephalography (MEG) further show that state and trait anxiety affect belief updating and decision-making via altered oscillatory dynamics in the PFC and ACC. Specifically, state anxiety has been associated with increased beta-band (13–30 Hz) activity and slower belief updating, whereas trait anxiety has been linked to reduced beta and increased gamma-band (30–60 Hz) activity, facilitating faster but potentially maladaptive updating (Hein et al., 2023; Hein & Herrojo Ruiz, 2022). Despite these insights, major gaps remain in our understanding of how uncertainty affects PA in skilled performance contexts. First, although anxiety often leads to aversive interpretations of uncertainty, its specific effects on performance as a function of PA remain unknown. Second, the neurobiological basis through which uncertainty affects performance in highly trained motor tasks is not well understood. Another critical factor is the role of high-stakes contexts, which often paradoxically impair performance, with larger incentives sometimes leading to poorer outcomes—a phenomenon known as “choking under pressure” (Lee & Grafton, 2014; Smoulder et al., 2021). This effect has been associated with psychological mechanisms such as over-arousal, excessive self-monitoring, and attentional disruption (Furuya et al., 2021; Lee & Grafton, 2014; Smoulder et al., 2021). High-stakes conditions for expert performers—such as competitions, auditions, or stage performances—elevate momentary PA, particularly in individuals with high trait PA (Chanwimalueang et al., 2017; Yoshie et al., 2009b), making them more vulnerable to performance impairments under these conditions. Neuroimaging studies implicate altered connectivity between the PFC, ACC, and motor areas, and abnormal interactions between reward processing and motor preparation, in performance breakdown (Lee & Grafton, 2014; Smoulder et al., 2021). Behaviourally, this manifests as overly cautious and often unsuccessful performance (Smoulder et al., 2021). However, it remains unclear whether increased uncertainty further exacerbates the negative effects of high-stakes contexts on performance, particularly in skilled individuals with high PA. Our recent work, combining behavioural analysis, computational modelling, and EEG in skilled pianists, showed that predisposition to PA modulates the integration of reward and punishment signals based on task uncertainty (Hidalgo et al., 2025). Across three experiments involving 95 pianists learning hidden melody dynamics via graded reinforcement feedback, we found that participants learned faster from punishment in low-uncertainty conditions, but relied increasingly on reward as uncertainty increased. These learning biases were mediated by reinforcement-driven changes in motor variability and frontal theta (4–7 Hz) activity. This reversal of learning strategies with rising uncertainty suggests a key interaction between trait PA and uncertainty processing. However, a critical open question is whether the effects of task uncertainty on learning and performance are amplified under high-stakes conditions, which are expected to elevate state PA. The current study addresses this question by investigating how varying levels of uncertainty and stakes interact with trait PA to modulate learning and skilled motor execution in pianists. We aim to identify the behavioural effects and underlying neurocognitive mechanisms of performance decline in high-stakes contexts, which are expected to elevate momentary PA—particularly when combined with high uncertainty. To achieve these aims, we will conduct a study with 40 skilled pianists using a within-subject design. The study comprises two components: an online survey and an in-person lab experiment. The online survey includes three validated questionnaires assessing trait predisposition to PA, intolerance of uncertainty (IU), and sensitivity to reward and punishment (Carleton et al., 2007; Conner et al., 2018; Cox et al., 2003). In the lab-based session, participants will complete two behavioural tasks: a reinforcement learning task (Task 1) and a musical scale performance task (Task 2), each administered across four conditions—high uncertainty (HU), low uncertainty (LU), high stakes (HS), and low stakes (LS). Task performance will be recorded via MIDI on a digital piano, alongside continuous EEG and electrocardiographic (ECG) recordings. Momentary assessments of PA will be collected throughout task execution. This project is designed to address four overarching objectives: 1. Assess the influence of uncertainty on learning and performance in PA. 2. Determine the effects of high-stakes conditions on learning and performance as a function of PA. 3. Identify how high-stakes conditions and uncertainty interact to modulate performance. 4. Examine the neural processes underlying the effects identified in objectives 1–3. We hypothesise that participants will learn faster from graded reinforcement feedback (punishment) under low uncertainty and low stakes conditions, compared to high uncertainty and high stakes. The slowest learning is expected in the combined high uncertainty and high stakes condition, driven by increased momentary PA levels in response to high stakes. These effects are expected to be mediated by reinforcement-driven regulation of motor variability—specifically, increases in variability following poor outcomes, which are expected to be dampened in conditions that slow learning. In the performance task involving the execution of a highly trained musical sequence, we expect that uncertainty and high stakes will interact to impair timing accuracy, again linked to increased momentary PA. Across both tasks, we expect high-stakes conditions to elicit increased momentary PA, particularly in individuals with high trait PA. This will be reflected in elevated subjective PA ratings and reduced heart rate variability (HRV) (Chanwimalueang et al., 2017; Sporn et al., 2022). At the neural level, we anticipate that frequency-domain amplitude changes and connectivity among the prefrontal cortex (PFC), anterior cingulate cortex (ACC), and motor areas will modulate learning and performance under varying task demands, with the most pronounced alterations expected in the high uncertainty and high stakes condition.