# Analysis plan: does depression dilate time inside the experienced moment and accelerate it beyond? A meta-analysis of time production

Registered before any pooled estimate was computed, and before extraction of most primary studies.

## Prediction under test

A temporal account of depression holds that depressed time is dilated at the scale of the experienced (present) moment, about 1-3 s, and accelerated at longer scales (working-memory presence and beyond). In a time production task (produce a named duration), dilation means producing intervals that are too long (overproduction) and acceleration means producing intervals that are too short (underproduction), relative to controls. Boundary fixed in advance: the experienced moment is any target interval of 3 s or less; acceleration is predicted for every interval above 3 s, including minutes.

## Studies

Eligibility: a group currently depressed (major depressive episode, bipolar depression, dysthymia, or depressive group defined by a depression inventory) compared with healthy controls on a time production task with a stated target duration; means and SDs (or convertible statistics) per target. Excluded: manic or remitted groups, groups pooling anxiety and depression, bipolar samples not in a depressive episode, bisection/estimation/reproduction tasks.

Sources: the eight production samples of Thones & Oberfeld (2015, J Affect Disord 175:359), and studies published since 2014 found by PubMed searches run on 2026-10-07 (queries in `plan/search.md`): Mioni et al. 2016, Ahmadi et al. 2019, Caceda et al. 2020, Medvedeva et al. 2023. Each study's numbers are extracted from the primary paper (exact target durations, means, SDs, Ns) when the full text is available before analysis; otherwise the study enters with the effect sizes of Thones & Oberfeld's Table A2, assigned to zones from the target durations reported in the primary paper. Kornbrot et al. 2013 and Oberfeld et al. 2014 enter from Table A2 because their primary reports give no per-interval group means (Oberfeld: 0.5 and 2 s in the moment zone, 60 s beyond; Kornbrot: 5, 10, 18, 34, 65 s, all beyond).

## Already seen before registration

Thones & Oberfeld's text and Table A2 (including their pooled zone estimates under their own interval ranges); full extraction of Kitamura & Kumar 1983 (30 s; noted that Thones & Oberfeld coded it as >10 min), Tysk 1984 (10, 20, 30 s) and Caceda et al. 2020 (8, 35, 43, 109 s); target durations of Oberfeld 2014 and Kornbrot 2013; abstracts of the post-2014 studies.

## Effect sizes and rules (`code/meta.py`)

- Hedges' g with small-sample correction, signed positive for underproduction by the depressed group, negative for overproduction; measures given as true/produced ratios are signed accordingly.
- Depressed subgroups sharing one control group are pooled (weighted mean; SD including between-subgroup variance), as in Thones & Oberfeld; repeated sessions of the same participants are averaged.
- Within a study, effects are averaged per zone (moment / beyond), giving at most two effects per study.

## Tests

Random-effects meta-regression of g on zone (cell means for moment and beyond), REML between-study variance, standard errors clustered by study (CR1), t with (studies - 2) df.

- H1: moment-zone pooled g < 0 (one-sided).
- H2: beyond-zone pooled g > 0 (one-sided).
- H3 (primary): beyond minus moment > 0 (one-sided).

Sensitivity: leave-one-study-out difference; primary-extraction-only studies. Secondary: slope of g on log10 target duration across primary per-interval effects; Caceda et al. with only the non-suicidal depressed group; description of melancholic versus non-melancholic subgroups where reported (Tysk 1984).

## Limits stated in advance

Few studies, especially in the moment zone; heterogeneous samples (inpatients, outpatients, students by inventory), medication, counting allowed or not; production tasks are motor as well as temporal, and psychomotor slowing could lengthen productions independently of timing. A null H3 with few moment-zone studies would be inconclusive rather than a refutation.