Example queries and generated hypotheses across psychology subfields

Query

"mechanisms of relapse prevention in cognitive behavioral therapy for major depressive disorder"

Representative hypothesis

Patients who show greater within-session reductions in negative self-referential processing during the middle phase of CBT (sessions 6–10) will exhibit lower depressive symptom relapse rates at 12-month follow-up, independent of post-treatment symptom severity.

Suggested design

n=120 adults with recurrent MDD entering 16-session CBT. Self-referential processing measured weekly via Self-Referent Encoding Task. Outcome: time to relapse over 12 months. Cox regression controlling for baseline severity, medication, and prior episodes.

Commentary

Targets a within-treatment mechanism rather than treatment-versus-control. The Self-Referent Encoding Task is well-validated. Risk: the proposed mediator is correlated with symptom change, so causal claims require careful temporal modeling.

Query

"neural mechanisms of attention restoration after exposure to natural environments"

Representative hypothesis

Exposure to natural visual scenes (≥40 min) will increase posterior cingulate cortex resting-state connectivity with the dorsal attention network by ≥10%, mediating improved performance on the Attention Network Test compared to matched urban-scene exposure.

Suggested design

Within-subjects crossover, n=40. Two 60-min exposure sessions (natural vs urban scenes) separated by ≥1 week. Pre/post resting-state fMRI and ANT. Multilevel mediation analysis.

Commentary

Operationalizes Attention Restoration Theory in a testable network-neuroscience frame. Effect size estimate (10%) drawn from prior connectivity studies cited in the output. Key risk: novelty effects of the scanner environment may dampen between-condition differences.

Query

"early predictors of executive function trajectories from preschool through early adolescence"

Representative hypothesis

Variability in parent-reported effortful control at age 4, measured by the CBQ-VSF, will predict between-person differences in executive function growth slopes from ages 6 to 12 (measured by NIH Toolbox EF battery) more strongly than concurrent SES or vocabulary, after controlling for child IQ.

Suggested design

Longitudinal cohort n=400 followed annually from age 4 to 12. Latent growth curve modeling. Predictors entered hierarchically.

Commentary

Frames a longitudinal prediction question with explicit comparisons against established predictors (SES, vocabulary). Useful template for grant aims targeting development. Risk: attrition over 8 years; the design should include multiple imputation and FIML strategies.

Query

"moderators of the effect of perspective-taking interventions on outgroup prejudice in polarized political contexts"

Representative hypothesis

The reduction in implicit outgroup negativity following a written perspective-taking intervention will be moderated by participants' perceived overlap with the outgroup, such that effects are largest at low baseline overlap (Cohen's d ≥ 0.30) and attenuated or reversed at high baseline overlap.

Suggested design

Pre-registered between-subjects experiment, n=600, recruited to balance political identification. Measures: Inclusion of Other in Self scale, IAT, feeling thermometers. Pre/post within-subjects component.

Commentary

Directly addresses replication-crisis concerns by specifying a moderator and an effect-size threshold. The 'reversal' prediction is risky and falsifiable, which is exactly what well-designed social hypotheses should be.

Query

"reinforcement learning parameters and treatment response in adolescent depression"

Representative hypothesis

Adolescents with depression who show a Q-learning learning-rate asymmetry favoring negative-prediction-error updates (α− > α+) at baseline will show smaller depressive symptom reduction over 12 weeks of fluoxetine treatment compared to those with symmetric or positive-biased learning rates.

Suggested design

n=80 adolescents with MDD beginning fluoxetine. Probabilistic reversal learning task at baseline; weekly PHQ-A. Q-learning model fits compared by AIC. Mixed-effects modeling of symptom trajectories with α-asymmetry as a baseline predictor.

Commentary

Connects a computational mechanism to a clinically actionable outcome. The learning-rate-asymmetry literature is real and growing. Risk: model identifiability with small samples — output flags this and suggests parameter recovery simulations.

Query

"ecological momentary assessment of stress and physical activity in adults with type 2 diabetes"

Representative hypothesis

Within-person increases in daily perceived stress (measured 5x/day via EMA) will predict same-day reductions in objective physical activity (accelerometer step count) of ≥15%, with the effect amplified on days with poor self-reported sleep quality the prior night.

Suggested design

14-day EMA + actigraphy study, n=100 adults with T2D. Multilevel modeling with stress, sleep, and their interaction as time-varying predictors.

Commentary

Strong use of within-subject ecological design. Avoids the common trap of between-subject correlations. Replicates and extends prior EMA-T2D work cited in the supporting papers. Risk: EMA compliance below 70% can bias estimates.

Query

"amygdala-prefrontal coupling during emotion regulation in PTSD"

Representative hypothesis

Veterans with PTSD will show reduced negative functional coupling between the basolateral amygdala and ventromedial prefrontal cortex during a cognitive reappraisal task compared to combat-exposed controls without PTSD, with the magnitude of reduced coupling correlating with CAPS-5 reexperiencing severity (r ≥ 0.30).

Suggested design

Task-based fMRI, n=60 (30 PTSD, 30 combat-exposed controls). Psychophysiological interaction analysis. Bayesian regression for the coupling-symptom correlation.

Commentary

Operationalizes a widely discussed circuit-level claim with explicit effect-size and direction predictions. The PPI approach is appropriate. Risk: PTSD samples are heterogeneous; the design should include sensitivity analyses by trauma type.