# Analysis plan: is distress lexically "heavy" across language families?

Registered before any colexification involving HEAVY, LOW, DOWN or any other physical-property concept with SAD, GRIEF, DIFFICULT or TIRED was computed or looked up.

## Question

Accounts of depression that treat felt weight as more than metaphor rest on the premise that distress is described in gravitational terms across unrelated languages. If so, languages should often use a single word for "heavy" and for affective or burden concepts (colexification), more often than "heavy" shares a word with arbitrary concepts, and gravity-related property words should colexify with sadness more than other physical-property words do.

## Data

CLICS4 v1.0 (CLDF; CC-BY-4.0; https://github.com/clics/clics4, tag v1.0), files cldf/forms.csv, cldf/concepts.csv, cldf/languages.csv, fetched by digest. Concepts are identified by Concepticon gloss. Before registering we looked only at concept coverage (number of languages and families per concept) and checked our colexification counting against CLICS's own colexifications.csv on four unrelated pairs (ARM/HAND 50 vs 53 families, TREE/WOOD 58 vs 58, MOON/MONTH 59 vs 60, FINGERNAIL/CLAW 65 vs 65; differences reflect form normalisation).

## Measure (`code/colex.py`)

Two concepts are colexified in a variety if it lists an identical form (NFC, case-folded, trimmed) for both. A pair's family-level rate is the share of families, among families with at least one variety having forms for both concepts, in which at least one variety colexifies them. Families are Glottolog top-level families; isolates count as their own family.

## Hypotheses and tests

**H1 (primary).** For each target T in SAD, GRIEF, DIFFICULT, TIRED, the HEAVY-T family-level rate lies in the top 5% of HEAVY's rates with all other concepts that share at least 30 families with HEAVY (the targets excluded from that reference distribution). Reported: counts, rate, percentile. H1 is supported for a target if its percentile is at least 0.95; the overall premise is supported if this holds for at least two of the four targets.

**H2 (specificity).** Among 15 physical-property concepts (HEAVY, LIGHT (WEIGHT), LOW, DOWN, COLD, HOT, DARK, BITTER, SWEET, HARD, SOFT, THICK, DEEP, WEAK, SLOW), the family-level rate of colexification with SAD or GRIEF (either) is higher for the gravity-related set (HEAVY, LOW, DOWN) than for the other twelve: difference of means, one-sided permutation p over all 3-of-15 relabellings (10,000 permutations, seed 20261007). Also reported: HEAVY's rank among the 15.

## Limits stated in advance

Colexification is lexical identity, not metaphor: it misses phrasal or derived expressions ("heavy-hearted") and includes homophony. Coverage of affective concepts is much thinner than of HEAVY (SAD in about 160 families), so rates rest on fewer families. Contact and areal spread are not modelled beyond counting families once. A positive result would support the cross-linguistic premise, not the mechanism behind it.