The Impact of Age and Gender on Sensory Imagery: Insights from the IMAVIC Dataset

With the advent of Large Language Models, conversational skills are no longer exclusive to humans, but human features such as senses and imagination remain a topic of debate in conversational systems. In this paper we describe the IMAVIC (IMAgery and Vividness In Context) dataset, annotated to investigate how linguistic context modulates sensory imagery, defined as the cognitive capacity to generate mental representations of sensory experiences across modalities (vision, hearing, touch, smell, and taste) in the absence of direct external stimuli. This study focuses on how linguistic stimuli, both in isolation and in context, elicit and modulate such imagery, and how these processes vary based on demographic factors. IMAVIC includes ratings from 909 native English speakers on the vividness of mental images and the intensity of sensory experiences evoked by adjectives, nouns, and adjective-noun pairs across five sensory modalities. The analysis of data reveals significant differences in sensory imagery across genders and age groups, suggesting that biological and sociocultural factors shape our imaginative abilities and providing a foundation for the development of models targeting sensory grounding.