Self-reported expressibility predicts communicative success: Open dataset, validation, and simulation

Referential communication experiments, in which participants negotiate meaning without conventional language, have provided insights into the origins of language and the modalities that accompanied its early forms. However, these experiments are resource-intensive, limiting sample sizes and the diversity of stimuli. This study provides an empirical validation of expressibility ratings - participants' subjective judgments about their ability to communicate concepts through different modalities - as an efficient alternative to laboratory-based production experiments. We collected expressibility ratings for 207 concepts across three modalities: gesture-only, vocalization-only, and combined gesture-vocalization. These ratings were empirically validated against a production experiment in which participants attempted to communicate 84 selected concepts. We measured communicative success using binary accuracy (correct/incorrect) and continuous semantic similarity between guesses and targets. Results strongly support the empirical validity of expressibility ratings. Higher expressibility ratings reliably predicted greater guessability (both measures) and required fewer corrective attempts when feedback was provided. Feedback improved exact matches but not semantic similarity. Modality comparisons revealed that gesture-only and combined modalities performed equivalently and both substantially better than vocalization-only. To further showcase the utility of the ratings, we created robust simulations to investigate potential bias in concept selection for stimuli. The analysis examined varying semantic domains and stimulus and participant numbers, finding no systematic bias toward gestural superiority, suggesting robust modality differences. The strong correspondence between expressibility ratings and actual communicative success demonstrates that humans have intuitions of their nonlinguistic communicative abilities. These findings establish expressibility ratings as a methodologically efficient tool for studying modality affordances, enabling larger-scale investigations with enhanced statistical power and broader conceptual coverage.