Which cross-linguistic sound-meaning association biases are robust on the open ASJP database, beyond the original Swadesh-100 set?

Blasi et al. (2016) found that specific speech sounds are associated with specific basic-vocabulary meanings far more often than chance across thousands of unrelated languages (e.g. 'nose' with /n/, 'small' with /i/), challenging strong arbitrariness. Their signal was computed over ~40--100 basic concepts. QUESTION: using the open ASJP (Automated Similarity Judgment Program) word-list database, (a) reproduce the per-(concept, sound) association statistic under a genealogically and areally controlled null model (e.g. permutation stratified by family and macroarea, or a mixed-effects logistic model with family/area random effects), (b) report which associations survive multiple-comparison correction, and (c) extend the test to concepts and sound classes not in the original set, reporting any additional robust biases and their effect sizes.

Acceptance. FULLY RESOLVES: on the public ASJP release, compute per-(concept, ASJP-sound-symbol) presence statistics for the shared core concepts across languages; test each against a genealogically+areally controlled null (permutation stratified by family and macroarea, or mixed-effects model), apply multiple-comparison correction, and report the set of significant associations with effect sizes -- including at least one concept/sound not analyzed by Blasi et al. 2016; ship the reproducible pipeline from the public ASJP data. PARTIAL: a faithful reproduction of a subset of the original associations on ASJP with a stated null model, or a rigorous negative result that a claimed association does not survive control.