Abstract
This paper presents three acceptability experiments investigating German verb-final clauses in order to explore possible sources of sentence complexity during human parsing. The point of departure was De Vries et al.'s (2011) generalization that sentences with three or more crossed or nested dependencies are too complex for being processed by the human parsing mechanism without difficulties. This generalization is partially based on findings from Bach et al. (1986) concerning the acceptability of complex verb clusters in German and Dutch. The first experiment tests this generalization by comparing two sentence types: (i) sentences with three nested dependencies within a single clause that contains three verbs in a complex verb cluster; (ii) sentences with four nested dependencies distributed across two embedded clauses, one center-embedded within the other, each containing a two-verb cluster. The results show that sentences with four nested dependencies are judged as acceptable as control sentences with only two nested dependencies, whereas sentences with three nested dependencies are judged as only marginally acceptable. This argues against De Vries et al.'s (2011) claim that the human parser can process no more than two nested dependencies. The results are used to refine the Verb-Cluster Complexity Hypothesis of Bader and Schmid (2009a). The second and the third experiment investigate sentences with four nested dependencies in more detail in order to explore alternative sources of sentence complexity: the number of predicted heads to be held in working memory (storage cost in terms of the Dependency Locality Theory [DLT], Gibson, 2000) and the length of the involved dependencies (integration cost in terms of the DLT). Experiment 2 investigates sentences for which storage cost and integration cost make conflicting predictions. The results show that storage cost outweighs integration cost. Experiment 3 shows that increasing integration cost in sentences with two degrees of center embedding leads to decreased acceptability. Taken together, the results argue in favor of a multifactorial account of the limitations on center embedding in natural languages.
Highlights
One of the few features of natural languages for which there is general agreement is the existence of non-local dependencies (Tallerman et al, 2009)
Structural integration cost was greater in sentences with one embedding than in sentences with two embeddings, whereas storage cost was greater in 2-embedding sentences than in 1-embedding sentences
Since center-embedded sentences with one embedding were judged as more acceptable than center-embedded sentences with two embeddings, Experiment 2 allows the conclusion that storage cost is more important than integration cost
Summary
One of the few features of natural languages for which there is general agreement is the existence of non-local dependencies (Tallerman et al, 2009). Non-local dependencies play a key role in several theories of the human parser, including the Syntactic Prediction Locality Theory (Gibson, 1998), the Dependency Locality Theory (Gibson, 2000), the Efficiency Theory (Hawkins, 2004, 2014), and the Minimize Dependencies Theory (Temperley, 2007; Gildea and Temperley, 2010). These theories have focused on two properties of syntactic dependencies, their number and their length, but syntactic dependencies have other properties which may be relevant too.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
