Explaining the paradox of grammatical innovation with relative frequency

Abstract I set out to develop a better understanding of why speakers use lexical expressions innovatively for the expression of grammatical meanings instead of using a conventional grammatical expression and why they continue to do so to such an extent that the construction indeed grammaticalizes. I argue that this phenomenon is somewhat paradoxical. Grammatical expressions are more frequent and conventionalised than lexical expressions, so it is not clear why speakers would choose a lexical expression in their stead for the expression of a grammatical meaning. I argue that none of the existing usage-based solutions for this paradox — expressivity, extravagance, cognitive activation — can fully explain the phenomenon. In this paper, I propose a new explanation for innovation based on relative frequency that complements the cognitive activation approach. I argue that a lexical expression can have a selective advantage in contexts in which analogical models (i.e., existing uses of the lexical expression that are similar to the novel grammatical meaning) have a higher relative frequency than the competing grammatical expressions. Using the semantic innovation of the meaning intention in be going to as a case study, I show that almost all early instances of be going to with intention occur in contexts in which its analogical model (motion-with-intention be going to ) has a significantly higher relative frequency than competing constructions such as will, ’ll , and shall . The paper highlights the importance of considering relative frequency and considering competing constructions for understanding why lexical expressions get used innovatively with grammatical meaning.