Grammatical acceptability is affected by the fit between the construction and noun phrase/contextual meaning:...

Human language allows for virtually limitless productivity. Why, then, are certain perfectly-comprehensible sentences (e.g., *The boy dragged the girl the box) deemed ungrammatical by native speakers? Here, we test the prediction of Construction Grammar accounts (c.f., Categorical Grammar accounts) that grammatical acceptability is probabilistically affected by the compatibility between the meaning (semantics) of the construction (e.g., [AGENT] [ACTION] [RECPIENT] [PATIENT]) and (a) the nouns (e.g., The boy sent the girl the package is more grammatical than *The boy sent Chicago the package; Studies 1,3,5) and (b) the speaker's intended meaning in context (e.g., the team in the Chicago office > the geographic location; Studies 2,4,6). Across a series of six large (N = 100) pre-registered studies, we find that (in all but one case), both of these factors significantly affect English speakers' rated acceptability of sentences, across the transitive and periphrastic causative (e.g., The clown made the man laugh vs *The clown laughed the man); the to- and double-object dative (e.g., The girl said something to the boy vs *The girl said the boy something) and the active/passive (e.g., The book cost $10 vs *$10 was cost by the book). These findings constitute support for Construction Gramar approaches, under which this type of grammatical acceptability results from a probabilistic semantic clash, and against Categorical Grammar accounts, which can explain graded (as opposed to categorical) ungrammaticality only via some kind of add-on.