Domain review: C1 minor_issues, C2 major_issues, C3 minor_issues, C4 major_issues. Significance minor for each.

This is a useful, narrowly reproducible lexical-database result. The psychological conclusions and categorical absence language require revision. I verified SHA256 and byte sizes of all three public data files, reran the supplied scripts in isolated Python3.12/NumPy2.1.3 (different NumPy version from the declared2.5.3), and obtained identical R1/R2 JSON values. A separately written streaming CSV/ZIP implementation, using no submitted analysis functions, independently confirms the13 main pair counts and denominators, and emits actual matching forms/varieties/family IDs. It preserves the stated normalization and mapping so is independent code, not independent linguistic annotation. The report evaluates the domain interpretation rather than claiming full validation of lexical coverage or source glosses. No author identity was sought or learned.

Relevant primary literature: Tjuka et al.2025, Advancing the Database of Cross-Linguistic Colexifications with New Workflows and Data, https://aclanthology.org/2025.iwcs-main.1/ , explains that CLICS aggregates elicited word lists, distinguishes present/absent/missing colexifications in that representation, and treats colexification as covering polysemy, homophony and underspecification. Its Table1 also uses a three-family threshold when counting network links. This supports the data type and exploratory filtering convention, not interpreting three families as a calibrated statistical rejection threshold or database absence as absence of every lexical sense. The current dataset citation is verified at https://clics.clld.org/ , which lists CLICS4 v1.0,2026,doi10.5281/zenodo.16900179 andCC-BY4.0. Cite the2025 methods paper alongside the dataset. Rzymski et al.2020, doi10.1038/s41597-019-0341-x, https://pmc.ncbi.nlm.nih.gov/articles/PMC6957499/ , is the earlier CLICS methods reference.

Jackson et al.2019, Emotion semantics show both cultural variation and universal structure, doi10.1126/science.aaw8160, author repository https://openresearch-repository.anu.edu.au/items/3bf3f8f5-4f56-4bc2-9537-022817f61f19 , reports emotion-colexification structure varying across language families and associated with geography, with valence/activation also organizing networks. It does not establish that absence of selected spatial-emotion identities rules out embodiment or confines metaphor to phrases. The present selected-pair descriptive extension is narrower; I found no evidence in these sources establishing all its exact counts in advance, so I do not rate the whole contribution as already known.

C1 minor_issues: the11/50 LIGHT-EASY and7/57 HEAVY-DIFFICULT results reproduce. Treat these as observed identical forms in at least one sampled variety per family, not22% of languages or proof of related polysemous senses in each instance. Seven families appearing in the HEAVY-DIFFICULT hits are also among the LIGHT-EASY hits; the mirrors are not independent cross-family confirmations. Report examples (attached audit supplies them), verify source glosses for representative hits, and preserve the operational definition. Percentile1.0 is a rank within eligible alternative partners, not p=0 or a multiplicity-controlled test. The count claim is useful without psychological causal inference.

C2 major_issues: all five reported counts reproduce, but the title says no spatial or weight word means happy or sad, although UP-HAPPY is observed in one Austronesian family. Replace the categorical title. More fundamentally the code tests identity of forms elicited for specific glosses, not complete lexical polysemy inventories or natural usage. A zero can coexist with an unelicited secondary sense, inflection, or a different translation/segmentation. Covering both concepts is necessary, not proof of sensitivity to all senses. The limitation acknowledges missed phrases, then still concludes that emotion is kept lexically apart and metaphors are therefore phrasal/derivational. That inference is not identified by these data. Restrict the claim to these pairs and the recorded forms in this release, stating 'few or no observed identities'. Do not infer a universal distinction between lexical and phrasal realization without corpus/dictionary validation. The positive straight-correct control demonstrates detection of one association; it does not establish comparable detection sensitivity for spatial-emotion senses. DOWN-SAD has only21 eligible families, which is insufficient under the author's own coverage rule. A three-hit rule is a filter, not a power calculation or evidence for absence. No specific published embodied theory is cited that actually entails the necessary single-form identity prediction used here.

C3 minor_issues: the exact38/155,7/200,4/113,1/171 counts reproduce. Their positive-versus-negative contrast is a description of differently covered pair sets, not an established general preference for positive evaluation. The plan correctly calls this descriptive. Four families colexify CROOKED-WRONG (Indo-European,Nakh-Daghestanian,Turkic,Yana in the supplied classifications); 'mostly in Nakh-Daghestanian languages' refers to a possible concentration of varieties, not a majority of the family-level hits, and needs the matching unit. Family-any-hit aggregation is affected by how many eligible varieties each family contributes: at equal per-variety prevalence q, a family with m independent eligible varieties has chance1-(1-q)^m of a hit. Family aggregation reduces simple large-family weighting but does not remove unequal opportunity for detection or geographic contact. A common-family/common-variety comparison or rarefaction and an areal sensitivity check would support stronger asymmetry language. Retain the current precise table as descriptive.

C4 major_issues: zero HARD-ANGRY and two SOFT-QUIET family hits reproduce. 'Do not double as words' is contradicted by the two observed SOFT-QUIET families (Mayan and Wintuan IDs in audit). Failure of a three-family rule is not zero occurrences. More importantly physical hardness is not established as a valid measure of psychological tension, nor softness/quietness as calm. The database lacks the intended construct, which warrants 'not directly tested', not a negative test of tension-anger or calm mappings. The7/20 SOFT-GENTLE result, though too sparse for the chosen percentile rule, further warns against categorical absence wording. State exactly which lexical pairs were counted, show the nonzero examples, and remove claims about emotional constructs those pairs do not measure.

The staged registration and post-hoc application of the stricter rule are openly disclosed and should remain so. Reusing the same database after inspecting earlier pairs provides a planned extension, not an independent replication. The code's use of source forms/Concepticon glosses should be checked against the database's intended lexical/phonetic representation; I found no numerical contradiction under the declared normalization, and do not assert that this check would change the outputs. No source or hit validation was invented. The significance is minor for each claim: a useful bounded extraction or descriptive pattern in an existing resource, not yet a change to psychological theory.