Computational modeling of motivated behavior under deadline pressure in Parkinson's Disease: effects of Deep Brain Stimulation
DOI: Pending DOI
Synopsis
Parkinson's disease (PD) is characterized by non-motor features, including disrupted motivated behavior. We applied a novel deadline-pressure framework to investigate effort-based decision-making in 23 PD individuals tested with deep brain stimulation (DBS) ON and OFF. DBS selectively accelerated decisions toward the high-effort option without affecting overall motor speed, suggesting stimulation modulates motivational processing. Computational modeling showed that PD choices were best captured by models tracking time and advancement as separate signals, unlike healthy participants, where an integrated pressure signal prevailed. Future work linking neural signals to these variables may provide a foundation for adaptive DBS systems targeting non-motor symptoms.
Background
Parkinson's disease (PD) is characterized by motor symptoms alongside prominent non-motor features, including disruptions in motivated behavior such as apathy and impulsivity, which are difficult to treat and poorly understood [1]. A key hypothesis is that these deficits reflect a disruption of value-based decision-making processes shaped by altered dopaminergic signaling [2]. Previous studies have used effort discounting tasks and reward-based paradigms to characterize these deficits, revealing dopamine-dependent reductions in willingness to exert effort [3,4] - yet these approaches have not captured how patients make decisions when effort serves progress toward a time-bounded goal.
Deep Brain Stimulation (DBS) offers symptomatic motor relief in PD, but its interaction with motivated behavior remains poorly understood: apathy and impulsivity can both emerge or remit following stimulation, and their co-occurrence complicates clinical management [2].
A novel framework challenging classical effort-cost models, demonstrates that effort can be valued when it makes progress toward goals under deadline pressure, with motivation driven by the ratio of work remaining to time remaining [5]. Here, we applied this framework to a cohort of PD subjects tested with DBS ON and OFF, acquiring simultaneous intra- and extra-cranial data and behavioral data to investigate how deadline pressure shapes effort-based decision-making in PD and how DBS modulates this process.
Methods
23 PD subjects (age: 67.1 ± 9.5 years) with DBS implants performed a deadline-pressure effort-based decision-making task during simultaneous local field potentials (LFP) and electroencephalography (EEG) recording, in both DBS ON and DBS OFF conditions (counterbalanced). The task consisted of 20 blocks of 8 trials each, with a shared deadline per block (Figure 1 (a)). On each trial, participants chose between a low-effort/high-reward option (LEHR) and a high-effort/low-reward option (HELR); effort exerted via a force grip was reflected in an advancement bar progressing toward the deadline. Failing to reach the deadline by the final trial resulted in the loss of all accumulated block reward (credits). Reaction time, block success rate, trials reaching target, and credits won were recorded. Computational models of subjective value (SV) were fitted to choice data by maximizing the log-likelihood of the probability of choosing HELR, using fmincon with 20 random starts and constrained parameter bounds; each participant and condition were treated independently. Bayesian Information Criteria (BIC) was used as model comparison evaluation metric. Models tested included a classic effort discounting model (ED: SV = R − k·E²), a pressure model (PE: SV = R + γ·P·E, where P = (1−ADV)/(n+1−t)), a time-only influence model (TT: SV = R + γ₁(n+1−t) + γ₂·t), and an advancement-only model (AT: SV = R + γ₁(1−ADV) + γ₂·t), where R = reward, E = effort, ADV = advancement toward deadline, P = pressure, n = 8 trials per block, and t = current trial.
Results
Overall task performance was preserved across DBS conditions: block success rate, number of trials reaching target, and credits won did not differ significantly between DBS ON and OFF. Reaction time analysis using a linear mixed-effects model (RT ~ 1 + DBS × choice + (1|ID)) revealed a significant main effect of choice (β = −0.070, p < 0.001), with faster responses on HELR trials relative to LEHR trials (Figure 1 (c) and (d)). Critically, a significant DBS × choice interaction was observed (β = −0.087, p < 0.001): DBS ON selectively reduced reaction time on HELR choice, while RT on LEHR choice trials was unaffected (p = 0.12). This interaction was not explained by differences in overall motor speed, as the DBS main effect was non-significant, suggesting that stimulation specifically facilitates motivational processing when evaluating the high-effort option, potentially reflecting reduced decision conflict, a shift in effort-reward valuation, or increased impulsivity toward future reward.
Computational modeling revealed that PD subjects' choices were best described by models incorporating time and advancement as separate rather than as an integrated pressure signal. The TT model provided the best fit in 51% of cases, followed by AT in 33% (Figure 1 (b)). No significant differences in winning model distribution or fitted parameters were found between DBS ON and OFF.
Discussion
The classic ED model performed worst across subjects, validating the deadline-pressure framework as capturing decision-relevant computations beyond classical effort-cost models. However, in healthy subjects, the PE model was the winning model [5], thus suggesting that while both groups are sensitive to deadline pressure, PD participants decompose it into its constituent time and advancement components rather than integrating them into a unified signal.
Notably, DBS affected behavior, selectively speeding up decisions toward the high-effort option, suggesting motivational circuits are sensitive to stimulation even when the underlying computational strategy remains unchanged. Moreover, the divergence between PD and healthy subjects in the winning model provides a computational marker of the disease independent of task performance. Future work will address patient heterogeneity by stratifying across clinical profiles (apathy, anxiety, depression) and stimulation target (subthalamic nucleus and globus pallidum), which may reveal subgroup-specific computational signatures and clearer DBS-dependent effects on decision strategy.
Finally, integration of simultaneously acquired LFP and EEG signals will allow identification of neural correlates of key computational variables and how DBS modulates their encoding. Linking these biomarkers to behavior provides a foundation for adaptive DBS systems that respond to motivational state, an important step toward addressing non-motor symptoms in PD [6].
Disclosures
Nothing to disclose by any authors
References
[1] Ricciardi, Lucia, Matthew Apps, and Simon Little. "Uncovering the neurophysiology of mood, motivation and behavioral symptoms in Parkinson’s disease through intracranial recordings." npj Parkinson's Disease 9.1 (2023): 136.
[2] Le Heron, Campbell, Lee-Anne Morris, and Sanjay Manohar. "Understanding disrupted motivation in Parkinson’s disease through a value-based decision-making lens." Trends in neurosciences (2025).
[3] Zénon, Alexandre, et al. "The human subthalamic nucleus encodes the subjective value of reward and the cost of effort during decision-making." Brain 139.6 (2016): 1830-1843.
[4] Hoy, Colin W., et al. "Beta and theta oscillations track effort and previous reward in the human basal ganglia and prefrontal cortex during decision making." Proceedings of the National Academy of Sciences 121.31 (2024): e2322869121.
[5] Pisauro, M. Andrea, et al. "Neural and computational mechanisms of effort under the pressure of a deadline." bioRxiv (2024): 2024-04.
[6] Swinnen, Bart EKS, Colin W. Hoy, and Simon J. Little. "Towards Adaptive Deep Brain Stimulation for Non‐Motor Symptoms in Parkinson's Disease?." Movement Disorders 40.9 (2025): 1759-1777.