# Analysis plan (second registration): do the positive poles of bodily metaphors colexify with positive emotion and evaluation?

Registered before computing any pair below, after a first registered set of negative-pole pairs on the same data (down/low = sad, hard/tight = angry, crooked/twist = bad/evil/wrong; positive control straight = correct). Those results are known: negative-pole pairs were absent or rested on one to four families, while straight = correct was strong. The heavy = difficult rate is also known from an earlier analysis and is included only as the mirror of light = easy.

## Pairs (`code/pairs2.py`)

UP-HAPPY, HIGH-HAPPY (mirror of DOWN/LOW-SAD); LIGHT (WEIGHT)-EASY (mirror of HEAVY-DIFFICULT); LIGHT (WEIGHT)-HAPPY; SOFT-QUIET (calm proxy; mirror of HARD-ANGRY); SOFT-GENTLE (descriptive, sparse); STRAIGHT-GOOD (mirror of CROOKED-BAD).

## Measure and criterion

Family-level colexification rate and its percentile among the anchor's partners sharing at least 30 families (code reused unchanged). Because most partners have a rate of zero, a single family can reach the 95th percentile; a pair is therefore supported only if its percentile is at least 0.95 and at least three families colexify it. The same stricter criterion is applied, as a post hoc check, to the first registration's pairs.

## Comparison

For each mirrored pair, the positive- and negative-pole rates are reported side by side (descriptive: counts are too small for a formal test of asymmetry).

## Limits

As in the first registration: lexical identity only, homophony, unmodelled areal spread, sparse concepts.