Using Perspectival Words Is Harder Than Vocabulary Words for Humans - and Even More So for Multimodal Language Models

Multimodal language models (MLMs) increasingly demonstrate human-like communication, yet their use of everyday perspectival words remains poorly understood. To address this gap, we compare humans and MLMs in their use of three word types, which we predict impose increasing cognitive demands: vocabulary (e.g., 'boat' or 'cup'), possessives (e.g., 'mine' vs. 'yours'), and demonstratives (e.g., 'this one' vs. 'that one'). Testing seven MLMs against human participants, we find that perspectival words are harder than vocabulary words for both groups. The gap is even larger for MLMs: while models approach human-level performance on using vocabulary, they exhibit clear deficits with possessives and even greater difficulties with demonstratives. Ablation analyses point to limitations in perspective-taking and spatial reasoning as key sources of these gaps in MLMs. Instruction-based prompting helps close the gap for possessives but still leaves demonstratives far below human performance. These results show that, unlike vocabulary, perspectival words pose a greater challenge in human communication-and this difficulty is further amplified in MLMs, revealing a crucial shortfall in their pragmatic and social-cognitive abilities. * Equal contribution † Work has been done as an student assistant at MPI Codes available at https://github.com/Beckinetic/ VLMIndexical * 'Perspectival words' is a non-technical label for indexicals, specifically personal, possessive, and demonstrative pronouns.