Research proposal in cognitive science
This is a proposed programme, not a report of completed studies or an institutional commitment. Sample sizes, measures and work packages describe intended research. The proposal’s behavioural operationalisation of agency is narrower than the philosophical account.
The full proposal follows below. The source PDFs with their original layout and tables are preserved in the project’s research archive.
Assessment documents
The Assessment Protocol v1.1 — EN and Assessment Protocol v1.1 — RU preserve the full proposed instrument. They are development documents, not validated diagnostic tools. No scores or empirical effects are reported here.
Full proposal
PHD RESEARCH PROPOSAL Perceptual Differentiation and Behavioural Agency
How Perceptual Resolution Shapes the Capacity for Adaptive Choice
A cognitive-science investigation of whether the resolution of perception constrains the
resolution of adaptive choice.
CURRICULUM Cognitive Science
INTERVENTION Aesthesys Embodied Perceptual Training (AEPT)
CORE CONSTRUCTS Capacity for Perceptual Differentiation (CPD); Behavioural Agency
STATUS Draft research proposal
Central research question
Does the resolution with which a person perceives differences constrain the resolution
with which that person can choose and adapt actions?
1. State of the art and research problem
Adaptive behaviour requires an organism to detect differences in its environment, distinguish relevant from irrelevant variation, and adjust action accordingly. Perception is therefore not only a stage preceding cognition and action; it helps constitute the action possibilities available to an agent. This premise is central to ecological and embodied approaches to cognition, in which perception is understood in relation to affordances, sensorimotor capacities, and ongoing interaction with the environment (Gibson, 1979; Varela, Thompson, & Rosch, 1991).
Perceptual-learning research shows that discrimination can improve with practice, but the extent of transfer remains unresolved. Improvements are often specific to trained features, tasks, or modalities, although broader generalization can occur under particular conditions. Recent experiments have demonstrated transfer across visual tasks and from crossmodal temporal training to unimodal auditory and visual temporal discrimination, while other work continues to show strong specificity. This tension raises a fundamental question: should perceptual differentiation be treated only as a collection of modality-specific skills, or does performance across different sensory domains also express a partially domain-general capacity?
The proposed research introduces capacity for perceptual differentiation (CPD) as the capacity to detect, organize, and use fine-grained differences in perceptual information across sensory and sensorimotor domains. CPD is not assumed a priori to be a unitary faculty. Its dimensional structure is itself an empirical question. The project will test whether performance across temporal perception, proprioception, touch, taste, olfaction, vision, and audition is explained entirely by modality-specific factors or whether a higher-order factor accounts for meaningful shared variance.
The second construct is behavioural agency. Here it is distinguished from the phenomenological “sense of agency” and from metaphysical notions of free will. It is operationally defined as the capacity to detect changes in a situation, maintain access to more than one viable response policy, select among alternatives, and update behaviour when previously successful responses cease to be adaptive. This overlaps with cognitive flexibility but is not intended as a synonym for it. Reversal learning, task switching, and adaptive choice under changing contingencies provide established paradigms for measuring relevant components; probabilistic reversal-learning tasks can yield both behavioural and computational indices of updating and perseveration.
The central claim to be tested is that behavioural agency is partly constrained by perceptual differentiation. If two environmentally relevant states are not perceptually distinguishable to an agent, they cannot reliably support different actions on the basis of that distinction. Conversely, increasing the resolution with which relevant differences are perceived may increase the number and precision of distinctions upon which action selection can operate. In this restricted sense, perceptual differentiation may constitute one condition of adaptive choice.
This claim does not imply that finer perception is sufficient for agency. Attention, working memory, motivation, executive control, prior learning, and task goals also constrain behaviour. The empirical question is whether CPD explains unique variance in adaptive choice after these factors are considered, and whether experimentally increasing CPD produces measurable changes in behavioural flexibility.
2. Research questions and hypotheses
RQ1. Is perceptual differentiation best characterized as a set of modality-specific abilities, or is there evidence for a partially domain-general capacity expressed across sensory modalities?
RQ2. Does CPD predict behavioural agency, operationalized through adaptive choice, reversal learning, task switching, and response updating?
RQ3. Can a multisensory embodied training programme improve CPD beyond simple practice and expectancy effects?
RQ4. If training changes behavioural agency, is that change mediated by improvement in CPD?
The principal hypotheses are:
H1. Measures of discrimination, structural ordering, and perceptual-to-action calibration will show reliable individual differences, containing both modality-specific variance and a higher- order shared component.
H2. Higher CPD will predict faster adaptation after contingency changes, lower maladaptive perseveration, and more efficient switching between response policies.
H3. Participants receiving multisensory embodied perceptual training will improve more than an active control group on untrained perceptual tasks. This is the critical test distinguishing generalized perceptual change from rehearsal of specific exercises.
H4. Gains in CPD will predict gains in behavioural agency; mediation models will test whether CPD accounts for part of the intervention effect on adaptive behaviour.
3. Conceptual framework
The project lies at the intersection of perceptual learning, embodied cognition, ecological approaches to action, and computational models of adaptive decision-making.
From an embodied perspective, perception and action form a coupled system rather than independent processing stages. An agent responds not to an exhaustive representation of the environment but to distinctions that become relevant in relation to its capacities, goals, and history. Ecological theories describe possibilities for action as affordances. The present proposal asks whether the granularity of perception influences the granularity with which action possibilities can be discriminated and selected.
The project distinguishes three levels of perceptual performance:
• Difference: the detection of whether stimuli differ. • Structure: the perception of ordered relations among multiple stimuli, such as gradients, sequences, or relative magnitudes. • Action calibration: the capacity to reproduce or act upon a perceived magnitude or relation.
These levels are analytically distinct and their relation will be tested rather than assumed. A participant may detect that two pressures differ yet reproduce neither pressure accurately; may discriminate individual tones but fail to order a sequence; or may reproduce a target with moderate accuracy without possessing an unusually low discrimination threshold. Recent work on
perceptual learning also indicates that changes can arise at different stages of the sensorimotor decision process rather than solely in sensory encoding.
A further distinction concerns near and far transfer. Improvement on a practiced task is not evidence for a general increase in CPD. The strongest evidence would be improvement on tasks, stimulus sets, and modalities not directly trained. Existing findings make both outcomes plausible: perceptual learning can remain highly specific, but cross-task and crossmodal transfer have also been observed. The project therefore deliberately separates training exercises from primary outcome measures.
4. Methodology
The doctorate will consist of three linked work packages.
Work Package 1 — Measurement model of perceptual differentiation
Year 1 will develop and validate a laboratory battery derived from a simpler preliminary behavioural protocol. Seven domains will be sampled: temporal perception, proprioception, touch, gustation, olfaction, vision, and audition. Each domain will contain tasks representing difference, structure, and action calibration.
Difference tasks will use psychophysical forced-choice procedures such as temporal-interval discrimination, weight discrimination, tactile texture discrimination, gustatory and olfactory discrimination, visual length or contrast discrimination, and auditory frequency discrimination. Structure tasks will require ordering or reconstructing relations among stimuli. Action-calibration tasks will require reproduction of perceived targets, for example temporal duration, applied pressure, taste or odour composition, spatial distance, or auditory frequency.
The preliminary home protocol uses fixed-difficulty trials and hit/miss scoring because its purpose is feasibility and repeated field assessment. The PhD version will instead employ adaptive staircase procedures, continuous reproduction errors, psychometric functions, and carefully standardized stimulus presentation.
A pilot will examine floor and ceiling effects, fatigue, sensory adaptation, and task duration. A larger sample will then complete the battery on two occasions. Confirmatory factor analysis and hierarchical models will compare competing structures: modality-specific factors only; process factors corresponding to difference, structure, and calibration; and hierarchical models containing modality-specific factors plus a higher-order CPD factor. Test-retest reliability and measurement stability across sessions will be assessed.
Repeated testing itself can produce perceptual learning; alternate stimulus sets, counterbalancing, and explicit modelling of retest effects will therefore be required. Gustatory and olfactory tasks will use controlled exposure duration, standardized containers and quantities, and sufficient inter-trial intervals to minimize adaptation.
The provisional target is 120–160 healthy adults, with the final sample determined by simulation- based power analysis for the planned latent-variable models.
Work Package 2 — CPD and behavioural agency
The second phase will test whether individual variation in CPD predicts adaptive behaviour in tasks independent of the perceptual battery.
Behavioural agency will be measured using a preregistered set of paradigms. A probabilistic reversal-learning task will quantify adaptation when stimulus–outcome contingencies change, including trials-to-criterion after reversal, perseverative choices, learning rates, and sensitivity to prediction errors. A task-switching paradigm will quantify switch costs and response updating. A dynamic choice task will introduce unsignalled changes in environmental contingencies and measure how rapidly participants detect and exploit new response policies.
The central prediction is not that high-CPD participants will simply respond faster. Rather, they should be more sensitive to task-relevant changes and less likely to continue a previously rewarded policy after its usefulness has declined.
Analyses will test whether latent CPD predicts behavioural outcomes after controlling for age, processing speed, working memory, and sustained attention. Computational modelling will be used where appropriate. In reversal learning, reinforcement-learning models can distinguish outcome sensitivity, updating rate, choice stochasticity, and perseveration. This distinction is important: an apparent increase in flexibility may reflect faster learning, reduced decision noise, greater exploration, or reduced habitual repetition. The project aims to identify which mechanism, if any, is related to CPD.
Work Package 3 — Intervention and causal test
The third phase will test whether CPD can be trained and whether changes in CPD alter behavioural agency.
Participants will be randomly assigned to a multisensory embodied perceptual training condition or an active control condition matched, as far as possible, for duration, social contact, instructor attention, novelty, and expectancy. The experimental intervention will be a manualized research version of Aesthesys Embodied Perceptual Training (AEPT), specified in advance as a reproducible protocol and separated from its commercial presentation.
Training will repeatedly engage perceptual differentiation, ordering, comparison, and calibrated action across multiple modalities. Exercises may include temporal discrimination, haptic and proprioceptive comparison, tactile texture and pressure, taste and odour comparison, visual ordering, auditory discrimination, and reproduction of perceptual targets. The exact stimuli and tasks used as primary laboratory outcomes will not be practiced during training.
Participants will be assessed at baseline, immediately after training, and at follow-up. Primary outcomes will be change in untrained CPD measures and change in preregistered behavioural- agency measures. Secondary outcomes will examine modality-specific learning, crossmodal transfer, retention, and action-calibration accuracy.
The primary causal analysis will test Group × Time interactions. Mediation analysis will examine whether intervention-related change in CPD accounts for part of the change in behavioural agency. Where feasible, intermediate measurements will strengthen temporal ordering and permit longitudinal modelling rather than reliance on simple pre-post difference scores.
A provisional sample of 90–120 participants is envisaged, with final recruitment based on pilot estimates, preregistered power analysis, and the smallest effect considered theoretically meaningful.
5. Researcher independence and open-science safeguards
Aesthesys is included as a source of the intervention, not as an object whose efficacy the dissertation is intended to confirm. This distinction is essential because the applicant has been involved in developing the method.
The intervention will therefore be manualized before the main trial; hypotheses, primary outcomes, exclusion criteria, and analyses will be preregistered; scoring and analysis will be blinded to group allocation where feasible; protocol deviations will be recorded; and null or negative results will be treated as theoretically informative. Researcher allegiance will be explicitly disclosed. Independent supervision and, where feasible, independent assessment of primary outcomes will reduce bias. Materials, anonymized data, and analysis code will be made openly available where ethical and legal requirements permit.
6. Expected results and scientific contribution
Three broad outcomes are possible.
First, performance may prove predominantly modality-specific. This would reject a strong domain-general account of CPD while still allowing investigation of whether specific perceptual abilities predict specific forms of behavioural adaptation.
Second, a higher-order CPD factor may emerge but show little relation to behavioural agency. This would indicate that perceptual differentiation and adaptive choice are separable capacities and would constrain stronger embodied interpretations.
Third, the predicted pattern may occur: a reliable higher-order component of perceptual differentiation, transfer after multisensory training, and a prospective relationship between CPD change and behavioural adaptation. This would support the proposal that perceptual resolution contributes to the conditions under which adaptive choice is possible.
The intended contribution is therefore not the claim that “better senses produce better decisions.” It is a more precise account of whether the resolution of perceptual distinctions constrains behavioural updating. The project also contributes methodologically by developing a multimodal battery that distinguishes difference detection, structural organization, and perceptual-to-action calibration within a common framework.
7. Innovation
The project is innovative in four respects. First, it treats perceptual differentiation as a testable multidimensional capacity rather than presupposing either a unitary “sensitivity” trait or complete modality specificity. Second, it links perceptual differentiation to behavioural agency through objective adaptive-choice paradigms rather than self-report. Third, it tests causal direction through a multisensory intervention and active-control design. Fourth, it makes transfer
central: improvement on practiced exercises will be separated from changes on untrained perceptual and behavioural tasks.
8. Feasibility, ethics, and limitations
The project primarily uses low-risk behavioural and psychophysical procedures. Ethical approval will be obtained before data collection and all participants will provide informed consent. Taste and odour tasks will use food-safe stimuli with allergy and sensitivity screening. Touch and passive-positioning procedures will be non-invasive and explicitly consented to.
The main methodological risks are practice effects, fatigue, expectancy, sensory adaptation, and construct overlap with attention or executive function. These will be addressed through alternate forms, counterbalancing, active controls, standardized exposure, breaks, covariate measurement, and preregistered exclusion criteria.
A conceptual limitation concerns behavioural agency itself. No single laboratory task exhausts the construct. For this reason, the project uses multiple paradigms and will retain the higher-order term only if the measures demonstrate sufficient coherence. Otherwise, conclusions will be stated at the level of specific components such as reversal learning, task switching, or adaptive response updating.
Ecological validity is a further limitation. Laboratory tasks deliberately simplify real-world choice. If the core hypotheses are supported, subsequent research could examine whether CPD predicts adaptive behaviour in richer interpersonal, motor, or naturalistic settings. Such extension is beyond the minimum scope of the doctorate.
9. Work plan and outputs
Year 1: Theoretical refinement; ethics and preregistration; development and piloting of the laboratory CPD battery; reliability and latent-structure study. Planned output: manuscript on the measurement structure of perceptual differentiation.
Year 2: Study of the relation between CPD and behavioural agency using reversal learning, task switching, dynamic choice, and computational modelling. Planned output: manuscript on perceptual differentiation as a predictor of adaptive behavioural updating.
Year 3: Randomized intervention study; follow-up; mediation and transfer analyses; integration of findings and dissertation writing. Planned output: manuscript on the causal effects and transfer profile of multisensory embodied perceptual training.
10. Expected outcome
The project asks a simple but consequential question:
Does the resolution with which a person perceives differences constrain the resolution
with which that person can choose and adapt actions?
The answer is not presupposed. Perceptual differentiation may be largely modality-specific and weakly related to behavioural flexibility; it may contain a domain-general component without
causal relevance to agency; or it may constitute a trainable capacity that contributes to adaptive choice. By experimentally distinguishing these possibilities, the project aims to transform an intuitively compelling proposition — that choice depends partly on what differences an agent can perceive — into a falsifiable programme of cognitive-science research.
References
Banai, K., Karawani, H., Lavie, L., & Lavner, Y. (2022). Rapid but specific perceptual learning
partially explains individual differences in the recognition of challenging speech. Scientific Reports, 12, 10011.
Ernst, M. O., & Banks, M. S. (2002). Humans integrate visual and haptic information in a
statistically optimal fashion. Nature, 415, 429–433.
Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Houghton Mifflin.
Harris, H., Gliksberg, M., & Sagi, D. (2012). Generalized perceptual learning in the absence of
sensory adaptation. Current Biology, 22, 1813–1817.
Hauser, T. U., Iannaccone, R., Walitza, S., Brandeis, D., & Brem, S. (2015). Cognitive flexibility in
adolescence: Neural and behavioral mechanisms of reward prediction error processing in adaptive decision making during development. NeuroImage, 104, 347–354.
Ivanov, V., Manenti, G. L., Plewe, S. S., Kagan, I., & Schwiedrzik, C. M. (2024). Decision-making
processes in perceptual learning depend on effectors. Scientific Reports, 14, 5644.
Kahalani-Hodedany, R., Lev, M., Sagi, D., & Polat, U. (2024). Generalization in perceptual learning
across stimuli and tasks. Scientific Reports, 14, 24546.
McGovern, D. P., Roudaia, E., Newell, F. N., & Roach, N. W. (2016). Perceptual learning shapes
multisensory causal inference via two distinct mechanisms. Scientific Reports, 6, 24673.
Varela, F. J., Thompson, E., & Rosch, E. (1991). The Embodied Mind: Cognitive Science and Human
Experience. MIT Press.
Virsu, V., Oksanen-Hennah, H., Vedenpää, A., Jaatinen, P., & Lahti-Nuuttila, P. (2008). Simultaneity
learning in vision, audition, tactile sense and their cross-modal combinations. Experimental Brain Research, 186, 525–537.
Waltmann, M., Schlagenhauf, F., & Deserno, L. (2022). Sufficient reliability of the behavioral and
computational readouts of a probabilistic reversal learning task. Behavior Research Methods, 54, 2993–3014.
Zhao, X.-N., Guan, S.-C., Xiong, Y.-Z., & Yu, C. (2024). Crossmodal to unimodal transfer of temporal
perceptual learning. Perception, 53, 753–762.