Does comprehension (sometimes) go wrong for noncanonical sentences?

Abstract

This article addresses the question of whether the human parsing mechanism (HPM) derives sentence meaning always from representations that are computed algorithmically or whether the HPM sometimes resorts to non-algorithmic strategies that may result in misinterpretations. Misinterpretation effects for noncanonical sentences, such as passives, constitute important evidence in favour of models allowing for nonveridical representations. However, it is unclear whether these effects reflect errors in the mapping of form to meaning, or difficulties specific to the procedure used to test comprehension. We report two experiments combining two different comprehension tasks to address these alternative possibilities. In Experiment 1, participants first judged the plausibility of canonical and noncanonical sentences and then named the agent or patient of the sentence. In Experiment 2, the order of the two tasks was reversed. Both tasks require the correct identification of agent or patient/theme, but differ regarding the complexity of operations required to complete the task successfully. In both experiments, participants made a substantial number of errors with agent/patient naming, even when they had correctly assessed sentence plausibility. We conclude that misinterpretation effects do not indicate parsing errors and therefore cannot serve as evidence for non-algorithmic processing. Our results support models of the HPM that assume algorithmic processing only.

Online representations of non-canonical sentences are more than good-enough

“…Our own study and converging evidence from other work (e.g. Meng & Bader, 2021) similarly suggests that it is not the parsing of a non-canonical sentence that is good-enough, so much as the memory processes involved in retrieving information from the representation of that sentence.…”

“…The presence of such an effect would support the idea that initial interpretations of such sentences are influenced by fast-and-frugal heuristics rather than derived purely from a detailed syntactic analysis. In contrast, the absence of an effect would support accounts in which the initial parse is algorithmic, and misinterpretations only occur due to information being retrieved from this representation in response to specific cues (Bader & Meng, 2018;Meng & Bader, 2021). Participants read implausible sentences presented in canonical or noncanonical form, followed by an Algorithmically Consistent or Good-Enough Consistent sentence.…”

“…The misinterpretation of non-canonical sentences has far-reaching consequences for theories of the human parsing mechanism, and the extent to which it always arrives at a fully algorithmic parse of a sentence. As discussed above, investigations of this phenomenon have typically relied on eliciting conscious judgements from participants, with the effect being large for thematic role probes (Ferreira, 2003), small for comprehension questions (Gibson et al, 2013;2017), and absent for plausibility judgements (Bader & Meng, 2018;Meng & Bader, 2021). In the current paper, we take an alternative, more naturalistic approach to assessing the semantic propositions derived from implausible canonical and non-canonical sentences, by examining downstream processing consequences on a follow-up sentence.…”

Eyetracking while reading passives: an event structure account of difficulty

“…In our offline comprehension data, we found participants to be less accurate for passives than actives independent of predicate type. This is compatible with previous results collected in our lab and in the broader literature (Ferreira, 2003;Meng & Bader, 2021), which consistently found passives to be more errorful than actives, independent of predicate type.…”

“…In the comprehension experiments that balanced the voice of the comprehension question, only the sentences with subject-experiencers resulted in passives being less accurate than the active (Paolazzi et al, 2021). Evidence that theta-role questions demonstrate a passivisation difficulty that is unobserved in other measures is corroborated by additional studies that combined plausibility ratings with comprehension questions (Meng & Bader, 2021).…”