![Sample our Behavioral Sciences journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/110801/TOC-Subject-Sample-2021-Behavioral-Sciences.jpg"})
ABSTRACT
We compare the processing of relative clauses in comprehension (self-paced reading) and production (planned production). We manipulated the locality of two syntactic dependencies: filler-gap (subject vs object gap) and subject–verb (centre-embedded vs right-branched). The non-local filler-gap dependency resulted in a longer embedded predicate duration, across domains, consistent with memory-based accounts. For the non-local subject–verb dependency, we observe longer reading times at the main verb, but in production a greater likelihood and duration of a pause preceding the main verb. We argue that this result stems from the cost of computing the restriction, which manifests as a prosodic break. In the context of the subject–verb dependency manipulation, we also revisit the source of interpretation break-down in multiple centre-embedding. Generally, our findings imply that memory-based accounts are adequate for filler-gap, but not subject–verb, dependencies and production studies can aid in understanding complexity effects.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 See also Santi, Grillo, Grodzinsky, and Wagner (Citation2011).
2 There is also some evidence that object relative clauses are more difficult because they are less expected than subject relatives, in accordance with expectation-based parsing. The evidence in support of these models, seems to be dependent on the paradigm, with evidence arising from eye-tracking Staub (Citation2010). Here we use self-paced reading and focus on memory mechanisms, thus, we will not elaborate further on expectation-based parsing and long-distance dependencies despite it also playing a role.
3 The relation between verb and the subject head is thematic. However, given the entire subject (not just the head) is assigned a thematic role, it is hard to understand the special status given to the head of the subject in order to provide a distance effect in this relation.
4 (Gibson & Fedorenko, Citation2013) argue a variation of (2) with the nested relative clause attached to the object is equally hard to process, which is unexpected if double-center-embedding is what creates the problem. We will return to this observation in the main discussion.
5 Experimental sentences appeared without the first word capitalized and a final period. As this is true of all experimental conditions, it should not affect results and if anything should lead participants to think that is the interest of the experiment.
6 Figures of the self-paced reading data present raw RT (msec), as this is more intuitive to understand. The analysis of raw RT produced the same pattern of results.
7 Model would not converge with full random effect structure. Analysis is based on random effects without the interaction term.
8 Model would not converge with full random effect structure. Analysis is based random effects with intercepts only.
9 Model would not converge with full random effects structure. Analysis is based on random effects including intercept and extraction.
10 Apart from the proportion of pauses, we also looked at the duration of pauses, but we found no no effect of Embedding, Extraction or their interaction on duration of silence after the head of the relative clause.
11 The model would not converge with random slopes. This model was therefore based on random intercepts alone.
12 Why the effect on duration of silence was not observed in Experiment 2A is not clear and may simply be a Type II error. It is hard to conceive an explanation in terms of the presence/absence of the preposition phrase.
13 Under the information-flow account, the lack of such an effect could be due to the incorrect flow of information in object-modified relative clauses equally reducing their acceptability.
14 At this point we have not addressed the longer duration of the RC in right branching than center-embedding, but we will attend to this later on in the discussion.
15 These predictions clearly need to be tested further. In an acceptability rating study, not reported here (based on a 6-point Likert scale), we found no difference when comparing the ratings between structures like (7) and (8), but with the lexical semantics better controlled, replicating Gibson & Fedorenko (Citation2013). Further, in object-modifying environments, an object-extracted relative that nested a subject-extracted relative was rated more difficult than a subject-extracted relative modified by an object-extracted relative (all else equal). This indicates a critical factor for the integration cost. We could have included these rating studies in the final manuscript, but thought that this might divert too much from the main thread of this paper with its focus on self-paced reading and production.
