Constraints on syntactic adaptation from a failure to generalise to the subject-object ambiguity

<i>Adaptation</i> to garden path processing has been observed when processing immediately follows a nonsyntactic conflict (Stroop) trial, demonstrating domain-general conflict adaptation. Likewise, adaptation to garden path processing has been observed progressively over an experiment with incremental exposure. Two high-powered experiments tested generalisability of both forms of adaptation to the subject/object garden path within the same dataset. Each trial included a Stroop task followed by self-paced sentence reading. Experiment 1 had 50% co-occurrence between a subordinate-first clause and early closure, which then increased to 100% in Experiment 2, making the subordinate clause a reliable cue. While Stroop and ambiguity effects were observed, neither adaptation effects were. Trial-level conflict adaptation may be constrained to particular paradigms (e.g. visual world) or across conflicts of the same representational format (e.g. syntactic-to-syntactic conflicts). Absence of experiment-level adaptation suggests constraints on changes to our expectations for late closure and/or a less probable but frequent parse.