Histograms show that stimuli from the same material class exhibited a wide distribution of gloss levels, which persisted even when only the strongest loading stimuli on each dimension were considered (Supplementary Fig. 10 ). Although some dimensions have a narrower range of gloss levels than others, stimuli from very visually distinct material dimensions like ceramics and (gold and uncoloured) metals have completely overlapping distributions of gloss ratings. Subjacent scatter plots show correlations between perceived gloss and material score, for each material class. Red coefficients indicate statistically significant correlations (uncoloured metals: r = 0.58, P < 0.001; ceramics: r = 0.55, P < 0.001; rubber-like: r = −0.31, P = 0.001; gold metals: r = 0.47, P < 0.001; glazed porcelain: r = 0.61, P < 0.001; plastic: r = 0.31 P = 0.004; melted chocolate: r = 0.49, P < 0.001). Black coefficients indicate correlations are not statistically significant (all P > 0.05). The results hold for other factor solutions (Supplementary Fig. 11 ).
