This work is published at *Psychonomic Bulletin & Review*: https://link.springer.com/article/10.3758/s13423-025-02668-8 We examined whether the ease of combining constituent meanings into a compound influences compound word processing. To this end, we built a compositional distributed semantic model, the CAOSS model (Marelli et al., 2017, *Cognition*) to capture this semantic composition process. Readers can implement this model themselves using the python script to build their own computational models. In the present study, measures derived from our computational model were tested across three behavioural measures: sensibility ratings for novel compounds (Study 1), lexical decision rejection latencies for novel compounds (Study 2), and lexical decision latencies for existing compounds (Study 3). In addition to these analyses reported in the manuscript, we conducted five additional analyses whose results were reported in the file "Supplementary Materials". Should you have any questions regarding the files on this site, please do not hesitate to email Cheng-Yu Hsieh (Cheng-Yu.Hsieh.2021@live.rhul.ac.uk).