Passives are not hard to interpret but hard to remember: evidence from online and offline studies

Abstract

Passives are not hard to interpret but hard to remember:Evidence from online and offline studies.Passive sentences are considered more difficult to comprehend than active sentences.Previous online-only studies cast doubt on this generalization. The current paper directly compares online and offline processing of passivization and manipulates verb type: state vs. event. Stative passives are temporarily ambiguous (adjectival vs. verbal), eventive passives are not (always verbal). Across 4 experiments (self-paced reading with comprehension questions), passives were consistently read faster than actives. This contradicts the claim that passives are difficult to parse and/or interpret, as argued by main perspectives of passive processing (heuristic, syntactic, frequentist). The reading time facilitation is compatible with broader expectation/surprisal theories. When comprehension targeted theta-role assignment, passives were more errorful, regardless of verb type. Verbal WM measures correlated with the difference in accuracy, but not online measures. The accuracy effect is argued to reflect a post-interpretive difficulty associated with maintaining/manipulating the passive representation as required by specific tasks. difficulty on offline tasks that require a judgment of a sentence interpretation (Ferreira, 2003; Street & Dąbrowska, 2010), but no difficulty or even facilitation on online ones that measure the moment-to-moment processing of sentences (Carrithers, 1989;Traxler, Corina, Morford, Hafer & Hoversten, 2014). While the offline data seem consistent with the general tenet that passives are more complex than actives, the online data question it. However, these previous studies collected either online or offline measures preventing definite conclusions to be drawn on the possible reason(s) for their contrasting data.In filling this gap, we present four self-paced reading experiments that simultaneously collected comprehension accuracy data with healthy adults. Results were replicated across 4 experiments, confirming an online vs. offline dissociation and at significance: passives were processed faster than actives at the verb and through much of the by-phrase, but induced more comprehension errors. This picture is inconsistent with the view that passives are more complex than actives. The fourth experiment supports a role for Working Memory (WM) in the accuracy effect. We argue that the complexity observed in offline data are due to postinterpretive processes required of the task and that noncanonical sentences (i.e., passives) are not complex to parse and interpret.

Does comprehension (sometimes) go wrong for noncanonical sentences?

“…In contrast, our results favour models which hold that the HPM generates veridical representations for noncanonical sentences as well, and which attribute misinterpetation effects to postinterpretive processes that operate on those representations and that possibly have very different sources. As shown here and in other studies (Bader & Meng, 2018; Christianson et al, 2010; Paolazzi et al, 2019), such processes are strongly dependent on situational variables, such as demands of the task used to assess comprehension. Assuming that the processes responsible for misinterpretation errors operate after a correct syntactic representation has been computed by the HPM is also in line with recent research into lingering effects observed for garden-path sentences such as NP/S ambiguities.…”

“…A similar account which also holds postinterpretive processes responsible for the misinterpretation of unambiguous noncanonical sentences has been proposed by Paolazzi et al (2019). This study reports the results of four experiments that examined the processing of English active and passive sentences combining an online method to tap processing difficulty (self-paced reading) and an offline method to assess interpretation (yes/no comprehension questions).…”

“…This study examined the mechanisms underlying misinterpretation effects with unambiguous noncanonical sentences, as observed for passives and other syntactic structures by Ferreira (2003), Christianson et al (2010), and others. Two different types of accounts have been proposed to explain such effects: accounts that emphasise misanalysis by assuming that misinterpretations result from systematic errors in the representations computed by the HPM for noncanonical sentences (Christianson et al, 2010; Ferreira, 2003; Karimi & Ferreira, 2016) and accounts that ascribe misinterpretation effects to postinterpretive processes which operate on sentence representations that correctly encode sentence meaning (Bader & Meng, 2018; Paolazzi et al, 2019). Based on the observation that misinterpretation effects showed up in German when using sentence materials and an experimental task (agent/patient naming) very similar to Ferreira (2003), but not when a different task (plausibility judgements) was applied to the same set of sentences, Bader and Meng (2018) proposed that misinterpretation effects are due to task-specific retrieval and interference processes induced by agent/patient naming, but not by the plausibility judgement task.…”

“…According to Ferreira (2003), these errors show that comprehenders sometimes arrive at an interpretation that conflicts with the true sentence meaning. Misinterpretation effects for English passive sentences have also been demonstrated in Christianson et al (2010) using a variant of agent/patient naming, and by Paolazzi et al (2019), who used yes/no questions to test comprehension.…”

Modeling misretrieval and feature substitution in agreement attraction: A computational evaluation

“…Although we believe that checking participants' interpretations of agreement attraction sentences is crucial to disentangling the encoding-and retrieval-based accounts, there is a significant caveat to using comprehension questions: One cannot be certain that the given answer matches the thematic relations computed during the online processing of the sentence (see also Schlueter et al, 2019). Bader and Meng (2018) have recently argued that answers to comprehension questions may be the result of what Caplan and Waters (1999) call post-interpretive processing (see also Paolazzi, Grillo, Alexiadou, & Santi, 2019). Caplan and Waters argue that the parser generates a syntactic structure and assigns meaning to the sentence online (interpretive processing), but that afterwards additional processes come into play whose output does not always match the online interpretation.…”