While previous work has examined the ability of language models (LMs) to capture pragmatic patterns, it remains understudied to what extent these abilities are shaped by training conditions. To address this gap, we test several models using an existing evaluation paradigm targeting LMs’ sensitivity to conversational norms (Askari et al. 2025). Across three experimental settings, we investigate the role of model size, training dynamics, and data composition in shaping pragmatic competence. Our results show that different pragmatic abilities exhibit distinct learning dynamics across models with varying sizes. While size or data scaling lead to inconsistent gains, exposure to conversational data causes moderate improvements in a modest training setting. Together, our findings suggest that pragmatic competence in LMs is heterogeneous and highlight the importance of targeted training conditions over scale alone for developing pragmatically sensitive models.

Implicit meanings are integral to human communication, making it essential for language models to be capable of identifying and interpreting them. Grice (1975) proposed a set of conversational maxims that guide cooperative dialogue, noting that speakers may deliberately violate these principles to express meanings beyond literal words, and that listeners, in turn, recognize such violations to draw pragmatic inferences.Building on Surian et al. (1996)’s study of children’s sensitivity to violations of Gricean maxims, we introduce a novel benchmark to test whether language models pretrained on <10M and <100M tokens can distinguish maxim-adhering from maxim-violating utterances. We compare these BabyLMs across five maxims and situate their performance relative to children and a Large Language Model (LLM) pretrained on 3T tokens.We find that overall, models trained on <100M tokens outperform those trained on <10M, yet fall short of child-level and LLM competence. Our results suggest that modest data increases improve some aspects of pragmatic behavior, leading to finer-grained differentiation between pragmatic dimensions.