The Relation-Sensitive Acceptability of Novel NN Compounds: An Ordinal Mixed-Effects Study of Japanese and English

This paper examines the acceptability of novel noun–noun (N–N) compounds expressing shape, colour, and material relations in Japanese and English, two languages that permit Generalized Modification (GM). The present study analyzes ratings from 125 Japanese and 103 English speakers using cumulative link models and cumulative link mixed models that retain all four ordered response categories and account for participant-level variation and overall item differences. This investigation was motivated by my 2025 pilot study, which reported higher ratings for shape-based compounds and cross-linguistic differences in material-based compounds but collapsed responses on a four-point rating scale into two categories and which relied on a small, highly unbalanced English sample (Japanese: n = 125; English: n = 13). The results show that English speakers gave lower acceptability ratings than Japanese speakers to shape-based compounds but higher ratings to both color- and material-based compounds. The language effect for shape differed significantly from those for color and material. Although the estimated language effect was descriptively larger for color than for material, the two effects did not differ significantly when compared directly. These findings show that GM accounts for the structural availability of novel N–N compounding but does not by itself determine relation-specific acceptability. They also illustrate the value of retaining ordinal information rather than reducing graded judgments to binary categories.