Study design

The CHANSS study is a multicentre, experimental, cross-sectional study designed to investigate the cognitive underpinnings of motivation in schizophrenia. The study employs a comprehensive battery of clinical assessments, cognitive tasks and self-reported measures to evaluate motivational deficits across different illness stages, including first-episode psychosis and TRS.

Recruitment dates: recruitment started in May 2021 and is expected to continue until July 2025.

Study sites

The CHANSS study was conducted across multiple international research centres, including academic institutions and mental health centres in the UK (Cambridge, London, Humber and Doncaster), Spain (Barcelona, Bilbao, Oviedo and Santander) and China (Shanghai). In the UK, the study benefits from the support of the National Institute for Health and Care Research Clinical Research Network (NIHR-CRN), which facilitates participant recruitment, site coordination and data management.

Inclusion and exclusion criteria

Inclusion criteria:

1. (a) age between 18 and 65 years;

2. (b) diagnosis of schizophrenia according to the International Classification of Diseases, 10th Revision (ICD-10) ¹⁹ criteria, with more than one year of diagnosis;

3. (c) stable antipsychotic medication regimen for at least six weeks prior to study enrolment;

4. (d) capacity to provide informed consent.

Exclusion criteria:

1. (a) presence of neurological disorders or significant medical conditions that may affect cognitive functioning;

2. (b) history of traumatic brain injury with loss of consciousness exceeding 5 min;

3. (c) current substance use disorder (excluding nicotine) within the past six months;

4. (d) established diagnosis of neurodevelopmental disorder;

5. (e) intellectual disability (IQ < 70);

6. (f) use of anticholinergic medication, except for hyoscine in cases of clozapine-induced sialorrhea.

Ethical considerations

The study protocol has been reviewed and approved by the relevant institutional review boards and ethics committees at each participating site. In the UK, the protocol received ethical approval from the Health Research Authority (HRA; IRAS:295622; 21/WA/0056) and was registered with the NIHR Clinical Research Network Portfolio. This protocol included the other sites, but each site obtained their local ethical approval. The study complies with the principles outlined in the Declaration of Helsinki and Good Clinical Practice guidelines. Written informed consent is obtained from all participants prior to any study-related procedures. Data confidentiality and participant privacy are maintained throughout the study in accordance with the General Data Protection Regulation.

Coordination and validation

To ensure consistency and data quality across sites, standardised training sessions and inter-rater reliability assessments were conducted to validate clinical scales and ensure consistency in data collection. This process included video-based training, cross-site workshops and interrater reliability exercises facilitated by experts in psychopathology assessment before the recruitment started on each site. The same clinical research forms were used in all sites and each site had language-specific scales and tasks.

Study procedure

Participants undergo a comprehensive assessment protocol, including clinical interviews, cognitive tasks and self-report questionnaires. The assessment is conducted over one or two sessions, depending on participant preference and tolerance, each lasting approximately 2–3 h.

1. (a) Screening and Informed Consent: participants are screened for eligibility based on inclusion and exclusion criteria. Informed consent is obtained prior to any assessments.

2. (b) Clinical Assessments: administration of standardised clinical rating scales by trained researchers.

3. (c) Cognitive Tasks: participants complete a battery of computerised tasks designed to assess various components of motivation.

4. (d) Self-Report Measures: participants complete questionnaires to evaluate subjective experiences related to motivation and negative symptoms.

Clinical assessments

1. (a) Positive and Negative Syndrome Scale: a clinician-rated scale with 30 items assessing positive symptoms, negative symptoms and general psychopathology. It uses a 7-point Likert scale to evaluate symptom severity. ^{Reference Kay, Fiszbein and Opler20}

2. (b) Brief Negative Symptom Scale: a 13-item clinician-rated scale designed to assess five negative symptom domains: blunted affect, alogia, anhedonia, asociality and avolition. Each item is rated on a 7-point scale. ^{Reference Kirkpatrick, Strauss, Nguyen, Fischer, Daniel and Cienfuegos21}

3. (c) Calgary Depression Scale for Schizophrenia: a 9-item clinician-administered scale focusing on depressive symptoms specific to schizophrenia, differentiating them from negative or extrapyramidal symptoms. ^{Reference Addington, Addington and Maticka-tyndale22} In addition to the total score, we will also specifically examine the score of item 3: Self Depreciation, reflecting self-reported defeatist beliefs.

4. (d) Personal and Social Performance Scale: a clinician-rated tool assessing social functioning across four domains: socially useful activities, personal and social relationships, self-care and disturbing and aggressive behaviours. ^{Reference Morosini, Magliano, Brambilla, Ugolini and Pioli23}

5. (e) Brief Assessment of Cognition in Schizophrenia (BACS): a 20-min validated cognitive battery to measure functions such as working memory, attention and verbal fluency. ^{Reference Keefe, Goldberg, Harvey, Gold, Poe and Coughenour12}

Self-report measures

1. (a) Apathy Motivation Index: an 18-item self-report questionnaire evaluating motivation across different life areas. ^{Reference Ang, Lockwood, Apps, Muhammed and Husain24}

2. (b) Autism Spectrum Quotient: a self-report measure to screen for autistic traits that may overlap with negative symptoms. ^{Reference Baron-Cohen, Wheelwright, Skinner, Martin and Clubley25}

3. (c) Apathy Evaluation Scale: a self-rated scale with 18 items designed to measure apathy, covering behavioural, cognitive and emotional domains. ^{Reference Marin, Biedrzycki and Firinciogullari26}

Cognitive and behavioural tasks

The study incorporates four computerised tasks designed to investigate the three cognitive mechanisms underlying motivated behaviour (see above). These tasks, although novel to this population, have been validated in healthy individuals. The tasks were designed to be less taxing in terms of possible fatigue and to be 5 to 15 min long to facilitate engagement. Participants are seated comfortably, interacting with a computer via keyboard, mouse or touchscreen, with clear instructions and breaks provided to minimise fatigue. All tasks were programmed in JavaScript and run on a dedicated server with Just Another Tool for Online Studies for backend. ^{Reference Lange, Kühn and Filevich27}

Task description

Participants use a stylus pen to draw paths between two red circles on a touchscreen tablet (Fig. 2). All sites used tablets of the same make and model (Samsung Galaxy Tab S6 Lite). One red circle appeared at the top and one at the bottom of the display. In the first control condition, participants were asked to draw as many paths as possible between the bottom and top red circle for 30 s. In the main experimental condition, participants were asked to draw as many different paths as possible between the red circles. Drawn paths appeared on the screen as they were drawn and remained visible throughout the task to reduce working memory demands. Participants were free to create curved or straight paths, and the paths could overlap or intersect. To ensure paths were counted as valid, they had to start at the bottom circle and end at the top circle.

Fig. 2 Illustration of the option-generation task.

Cognitive mechanism

The control condition accounts for individual differences in movement speed. The main experimental condition measures fluency (the total number of paths generated) and uniqueness (the distinctiveness of each path relative to others). Uniqueness will be computed by quantifying the geometric (Euclidean) distance between paths, as done previously, ^{Reference Ang, Manohar, Plant, Kienast, Le Heron and Muhammed28} while accounting for basic differences in movement speed. This approach provides an objective measurement of creative output and flexibility, essential for understanding goal-directed behaviour.

Task-specific hypothesis

This task has not been previously applied in schizophrenia research. However, the task has been used in people with Parkinson’s disease ^{Reference Ang, Manohar, Plant, Kienast, Le Heron and Muhammed28} and with depression. The hypothesis is that patients will show reduced fluency and reduced uniqueness, and reduced uniqueness would be related to executive dysfunction in patients, as measured using the BACS tool.

Milkman task

Objective

To operationalise the decision to switch in a time-constrained paradigm, as part of the motivational cognition mechanism. Specifically, the task examines reward-based decision-making in a foraging framework, evaluating tendencies to persist with diminishing returns – a behaviour hypothesised to be altered in people with affective conditions and apathy ^{Reference Bustamante, Barch, Solis, Oshinowo, Grahek and Konova29} and to be dependent on dopamine. ^{Reference Le Heron, Kolling, Plant, Kienast, Janska and Ang30}

Task description

Participants are instructed to collect as much milk as possible over a 10-min period (Fig. 3). To milk a cow, they hold down the space bar, during which a bucket on the screen visually represents the accumulating milk. Participants are told that milking would become progressively harder over time. Milk accumulation follows an exponential decay function:

$$ N_{\left(t\right)}=N_{0}e^{-\lambda t} $$

where N _(t) is the instantaneous accumulation rate at time t, N ₀ is the initial rate and λ is the rate of decline. There are four cow types, created by crossing two initial reward rates (0.01 or 0.02) with two decay rates (2e-4 or 4e-4), resulting in a 2 × 2 design: High–Slow, High–Fast, Low–Slow and Low–Fast. These conditions are designed to vary in overall profitability and optimal harvesting duration and are presented in a pseudorandom sequence such that every four-trial cycle includes one of each cow type.

Fig. 3 Schematic of foraging in the milkman task.

When participants release the space bar, they enter a fixed 4-s travel period representing the time needed to reach the next cow. During this phase, a progress bar shows the remaining travel time, and the total milk earned so far is updated. Pressing the space bar prematurely during the travel screen triggers an on-screen warning lasting 2 s. A countdown timer displays the remaining time for the task, and participants’ cumulative milk total is always visible. The task begins with a 1-min practice session to familiarise participants with the interface and response requirements.

Cognitive mechanism

The task assesses decision-making related to switching behaviours. The cognitive mechanism involves cost–benefit analysis and temporal discounting, designed and analysed using a foraging framework. The primary outcomes will be ‘stay duration’ and the related ‘exit threshold’ which is the reward rate at the decision to switch on each trial. Based on these, we will compute an ‘offset’ parameter, reflecting participant’s overall tendency to stay longer or shorter than the optimal policy; and a ‘scaling’ parameter, reflecting how much participants exaggerate differences between conditions compared to the optimal policy. ^{Reference Wolpe, Scott, Salomon, Nassar, Fletcher and Fernandez-Egea31}

Task-specific hypothesi

This task has not been previously applied in schizophrenia research. We have recently validated the task in healthy participants in an online study. ^{Reference Wolpe, Scott, Salomon, Nassar, Fletcher and Fernandez-Egea31} Other foraging tasks have been used in the context of affective disorders, ^{Reference Bustamante, Barch, Solis, Oshinowo, Grahek and Konova29} attention-deficit hyperactivity disorder and Parkinson’s disease. ^{Reference Le Heron, Kolling, Plant, Kienast, Janska and Ang30} In line with our previous findings in healthy controls, ^{Reference Wolpe, Scott, Salomon, Nassar, Fletcher and Fernandez-Egea31} we hypothesise that overstaying will be associated with general depressive symptoms, while deviation from optimal scaling would be associated with executive dysfunction, reflecting a failure to adapt appropriately to diminishing returns.

Tokens task

Objective

To examine decision-making under uncertainty and failure aversion, using a variant of a task from evidence accumulation literature, ^{Reference Thura and Cisek32} as part of motivational cognition.

Task description

Participants completed a decision-making task in which they observe the movement of tokens from a central circle to two lateral circles and are asked to make a decision as to whether the left or right circle will have the most tokens by the end of the trial (Fig. 4). At the start of each trial, 15 tokens and a central cross are presented in a central circle. These tokens move incrementally, one at a time, towards either the left or right circle at pre-defined intervals. The movement pattern is determined by a pre-generated binary sequence, ensuring variability across trials while maintaining an equal number of left- and right-winning trials. The winning side was set to receive eight or nine tokens, ensuring that trials are not overly predictable.

Fig. 4 Schematic of the decision-making process in the tokens task.

Participants are instructed to make a decision as early as they felt, using the left or right arrow key to indicate their response. Once a decision is made, the cross moves to the chosen circle, and the remaining tokens are quickly distributed from the middle circle. If the choice is correct, the cursor turned green, a positive auditory cue is played and the text ‘X number of tokens earned’ appears in green, where X is the number of tokens earned on that trial. If incorrect, the cursor turns red, a negative auditory cue is played and the text ‘no tokens earned’ appears in red. If participants do not make a choice before all tokens had moved, a ‘timeout’ warning appears, and participants receive incorrect feedback (red circles and negative auditory cue).

Participants earn the number of tokens remaining in the middle circle when a correct decision is made. This speed accuracy trade-off is emphasised in the instructions given to participants, while they are encouraged to collect as many points as possible. Unlike previous studies, ^{Reference Derosiere, Thura, Cisek and Duque33} we chose not to add an explicit penalty for incorrect choices and focused on the effect of implicit-loss feedback. Each participant completes 100 trials, with an equal number of left- and right-winning trials presented in a randomised order.

Choice behaviour will be modelled using a Bayesian decision framework in which participants update beliefs about the eventual majority pile on each token draw. At each state, the model evaluates the value of choosing left, right or waiting, with choice values determined by the posterior probability of each pile being the majority. Two free parameters were estimated: a subjective-punishment parameter capturing the subjective cost of making an incorrect choice and a cost-of-waiting parameter representing the individual’s opportunity cost of time. Actions are generated probabilistically using a softmax rule over the three action values, and parameters will be fit by maximising the log-likelihood of the observed sequences of waits and choices.

Cognitive mechanism

The task evaluates the decision of when to act and which action to choose. The main model free outcome measures are number of tokens drawn until decision and response (‘right’ or ‘left’). The computational model parameters will compute each participant’s loss aversion and opportunity cost of time, with the hypothesis that people with schizophrenia will require more evidence before committing to a choice due to loss aversion, over and above general slowness.

Task-specific hypothesis

This task has not been previously applied in schizophrenia research. It is expected to reveal increased loss aversion and reduced opportunity cost of time in patients. While increased loss aversion is hypothesised to correlate with meta-cognition measures, sensitivity to the opportunity cost of time is expected to correlate with executive dysfunction.

Clock stopping task (performance beliefs task)

Objective

To explore meta-cognition deficits in schizophrenia, particularly defeatist performance beliefs, which suggest that patients generalise negative thoughts about their abilities, which leads them to avoid goal-directed behaviours. ^{Reference Grant and Beck16}

Task description

The Performance Beliefs Task (Fig. 5) involves participants estimating their ability to stop a moving clock hand at a target location (‘time’). The task has two phases: First, during the action (hand-stopping) phase, participants observe a moving red clock hand that rotated clockwise or counterclockwise and are asked to stop it as close as possible to a predesignated target time (marked in colour red). They press the space bar to stop the red clock hand. Second, after stopping the clock hand, during the estimation phase, the clock hand disappears, and the target time turns black again. Participants are prompted to estimate where they stopped the red hand by manually adjusting a second blue clock hand using the arrow keys. The experiment consists of 50 trials, with pseudo-randomised hand movement directions and stopping targets. Participants are also given a time limit for making their estimates. If they respond too late, they receive a ‘timeout’ message and repeat the trial.

Fig. 5 Schematic of the meta-cognition beliefs in the clock-stopping task.

Cognitive mechanism

The task was designed to assess meta-cognition beliefs about performance, testing how well participants could predict and evaluate their own actions. Each trial has two main outcome measures: performance error, calculated as the difference between the target position and where the red hand was stopped, and estimation error, calculated as the difference between their estimated and actual stopping position. These measures can be fit using a Bayesian model that assumes that belief about the stopping position combines noisy sensory input with prior expectations about success (target position). ^{Reference Wolpe, Wolpert and Rowe34} To account for non-integration trials, ^{Reference Nassar, Waltz, Albrecht, Gold and Frank35} the model will be extended with a non-updating function that probabilistically converts moderate prescribed updates into either non-updates (reliance on the prior) or full updates (reliance on sensory evidence). ^{Reference Nassar, Waltz, Albrecht, Gold and Frank35} In this framework, small discrepancies between prior expectations and sensory evidence increase the probability of relying on the prior, large discrepancies increase the probability of relying on sensory evidence and intermediate discrepancies favour Bayesian integration of both sources of information. The parameters will measure prior precision (expectation of success), sensory noise and non-updating thresholds. We hypothesise that prior precision will be associated with self-depreciation.

Task-specific hypothesis

This task has not been previously applied in schizophrenia research. Similar paradigms have been used in Parkinson’s disease and healthy populations, ^{Reference Wolpe, Nombela and Rowe36,Reference Hezemans, Wolpe and Rowe37} We hypothesise that patients will exhibit diminished expectation of success, compared with the typical optimistic expectations people show in the task. ^{Reference Wolpe, Wolpert and Rowe34}

Cross-task analysis plan

Data from the four experimental tasks will be pooled for cross-task analyses. Specifically, as per one of the main aims of CHANSS, we sought to capture clusters or subtypes of motivational deficits across patients. As set out at the beginning of this paper, we expect these clusters to overlap and to have a hierarchical structure, with executive cognition influencing both motivational and meta-cognition, and with motivational cognition influencing meta-cognition. To capture this, we will use Gaussian Mixture Modelling (GMM), which is ideally suited for this nested structure, as it models clusters as overlapping Gaussian distributions and provides posterior probabilities of cluster membership rather than ‘hard’ cluster assignments. ^{Reference McLachlan and Peel38} First, as GMMs use full covariance matrices which capture the interdependencies inherent in our hierarchical model, where deficits in lower-order functions (executive) systematically affect higher-order functions (motivational, meta-cognition). This covariance structure allows GMM to model the cascading effects that define our theoretical hierarchy. Second, the ‘soft’ clustering approach of GMMs aligns with our aim to compute the relative contribution of each cognitive mechanism to apathy.

As we expect patients to vary significantly in processing speed, which could confound some of the results, we will account for individual differences. To this end, we will compute a composite processing speed measure across tasks, derived from the first principal component of the following indices: (a) overall score in the tokens motor task from BACS; (b) mean cross-trial maximum speed of path drawing in the control condition of the option-generation task; (c) median latency in the milkman task; and (d) median reaction time in the estimation phase of the clock-stopping task.

Another possible confounder we will consider is the testing site. To address this, we will test for differences in clustering results between sites. Specifically, we will test whether the different site countries (the UK, Spain and China which would recruit a roughly similar number of participants each) will have different clustering probabilities in the GMMs.

Power calculation

As our principal aim was to investigate the cognitive mechanisms of negative symptoms, our study was powered for the within-patient group analyses. These included (a) task-specific correlations of the task parameters and clinical or cognitive scale; and (b) cross-task subgroup identification. As a secondary aim and for the comparison with normative data in the tasks, we sought to collect a similar sample size of age- and gender-matched healthy controls.

For the task-specific correlations, considering we had six a priori correlations of interest across all tasks (uniqueness parameter in option-generation task; offset and scaling parameters in the milkman task; loss aversion and opportunity cost of time in the tokens task; and prior precision in the clock-stopping task), a sample size of 162 individuals would provide 90% power to detect correlations with medium effect size (r = 0.3) at the level of α = 0.05/6 = 0.008. Moreover, based on previous literature ^{Reference McLachlan and Peel38} and simulation for Gaussian Mixture Models with six features and three hierarchical clusters with a good (d = 1) separation, this sample size would provide moderate power (70–80% from simulations) to detect the three hypothesised clusters (equal number of participants per cluster assumed).

Data management

All data are collected using secure, encrypted electronic data capture systems. Data is pseudo-anonymised on each site, assigning a unique identifier (e.g. CC001) to each volunteer. Data are stored in compliance with local regulations and specifically local data protection regulations. Quality control measures include regular audits and inter-rater reliability assessments for clinical ratings.