This paper models conventionalisation of language structure as constitutive of processing fluency. I postulate that the difference in conventionalisation of linguistic forms used for communication significantly influences our reasoning about linguistically-expressed problems. Two studies are reported that tested this hypothesis with the use of variably conventionalised - fluent and disfluent - formulations of problem-solving tasks. Th e findings indicate that even in tasks requiring analytic reasoning, the degree to which the linguistic forms employed to communicate are conventionalised is correlated with the subjects’ performance success rate. On a more general level, this paper seeks to empirically address the nature of links between linguistic form and meaning construction.