# Analysis plan: do words for bodily or spatial states double as words for emotions or moral evaluation across language families?

Registered before computing any of the pairs below (except DOWN-SAD and LOW-SAD, whose family counts with SAD or GRIEF combined were reported by an earlier registered analysis on the same data: zero families each).

## Question

Embodied accounts of affect propose that low or downward position, physical tension and twisting are the bodily templates of sadness, anger and badness. If so, languages should often use one word for both (colexification), more often than the bodily word pairs with arbitrary concepts.

## Data and measure

CLICS4 v1.0 (CC-BY-4.0), fetched by digest; the family-level colexification measure of the earlier registered analysis, code reused unchanged (`code/colex.py`): identical normalised forms in a variety; share of families, among families with forms for both concepts, in which some variety colexifies them. Before registering we looked only at which concepts exist and how many languages and families cover them.

## Pairs and tests (`code/pairs.py`)

Down/low = sad: DOWN-SAD, LOW-SAD. Tense = angry: no TENSE concept exists; TIGHT covers only six families and is reported descriptively; HARD is used as the nearest well-covered proxy (HARD-ANGRY, HARD-ANGER). Twisted = bad: CROOKED-BAD, CROOKED-EVIL, CROOKED-WRONG, TWIST-BAD, TWIST-EVIL. Positive control: STRAIGHT-CORRECT (RIGHT), and STRAIGHT-RIGHT (descriptive, since RIGHT also means the side).

For each pair, the percentile of its family-level rate among the anchor's rates with every other concept sharing at least 30 families with the anchor. A pair is supported if its percentile is at least 0.95. The method is validated if the positive control reaches that level; if it does not, all null results are uninformative.

## Limits stated in advance

Lexical identity only (no phrases such as "feeling down"); homophony counts; areal spread is not modelled; sparse concepts (EVIL in about 28 families, ANGER 67, DOWN 83) give imprecise rates.