Intonation influences processing and recall of left-dislocation sentences by indicating topic vs. focus status of dislocated referent

Abstract

We tested the effects of two intonation contours on the processing and cued recall of German sentences with a left-dislocated subject vs. object: (i) a rising accent on the dislocated phrase, followed by a rising-falling hat contour on the main clause; (ii) a falling accent on the dislocated phrase, followed by a falling accent plus subsequent deaccentuation. The contours had differential effects depending on the grammatical function of the dislocated phrase (subject/object) and, for the recall, on the cue type for the recall (subject/object), in certain conditions overriding the subject-before-object preference normally found in processing. To account for the findings, we propose: (1) Contour (i) signals the topic status of the referent of the dislocated phrase. Contour (ii) signals that referent's focus status. (2) Topics are referents that serve as an address in a structured discourse representation in working memory under which information about that referent is stored. (3) Subjects are default topics, whereas objects are not, so that topic-marking an object is motivated, which results in an object-before-subject preference for sentences with topical objects during processing. (4) Retrieval of information from an address incurs a lower processing load if the appropriate address is cued than if some other referent is cued.

Keywords:

• left dislocation
• intonation
• topic
• focus
• sentence processing
• cued recall
• discourse representation

Acknowledgements

We would like to thank Berry Claus, Aria Adli, Andreas Haida, Manfred Krifka, Tue Trinh for comments on earlier versions of this paper, as well as audiences at AMLaP 2011, at the Romance Department at Freiburg University and at the Department of English and American Studies of the Humboldt-Universität zu Berlin for fruitful discussion. The discussion of left dislocation as potentially marking foci is based on joint work of the first author with Stefan Hinterwimmer. We thank Frank Kügler for help with the stimulus recordings, and Constantin Freitag, Kristin Bünnemeyer and Kristina Ruthe for helping with the preparation of the stimuli and with the annotation of the participant recordings.

Notes

1. If we assume that focus marks alternatives, one might wonder what the role of contrast marking – which also indicates the presence of alternatives – is. The precise delimitation of contrast from narrow focus is highly problematic. Contrast is often associated with a limited set of explicit alternatives, with an exhaustive (i.e. exclusive) interpretation, or with discourse contexts like corrections (see Repp, 2010 for a review). All these contexts can also be said to involve narrow focus – they all involve alternatives. Narrow focus, however, can be found in a wider range of contexts, for instance, in answers to wh-questions. In experimental studies, contrast is often implemented as either a limited set of explicit alternatives (e.g. Fraundorf et al., 2010; Ito & Speer, 2008; Sedivy; Tanenhaus, Chambers, & Carlson, 1999) or as correction (e.g. Breen, Fedorenko, Wagner, & Gibson, 2010). In other studies, contrast and narrow focus are considered the same (e.g. Sanford et al., 2006). Contrast does not play a role in the current study because we tested isolated sentences so there is no context that would provide explicit alternatives or licence a correction reading. We will say more about contrast in the Discussion section.

2. Cowles and Ferreira (2012) also tested whether participants actually classified the referent introduced by the about-phrase as a topic. They found that overall the subject of the matrix clause was classified as the topic of the sentence but that the presence or absence of an about-phrase modulated this result. This is evidence that topichood is not entirely dependent on subjecthood in English.

3. Left dislocation is also possible with prepositional phrases, complementiser phrases or adjective phrases, where the resumptive pronoun is an adverbial proform. We do not investigate such cases here.

4. Note, however, that these examples have an intonation contour in the main clause which is different from the intonation contour that we use in our experimental materials. (7) has two falling accents – one on the resumptive pronoun and one on gleich (‘straight away’). (8) has a rather flat contour throughout the main clause. These differences are due to the fact that intonation is influenced by context, and the context in the two examples differs – different things are given. For our experimental materials, we used the intonation contour that is appropriate for (9). See the Stimuli and design section for specification.

5. Most descriptions of the German tone inventory assume an H*+L accent but German ToBI (Grice et al., 2005) assumes that the corresponding pitch movements are better analysed as H* followed by an L–phrase accent, which is aligned with the post-stressed syllable.

6. The low tone of the H*+L was actually only reached on the next stressed syllable, which corroborates the assumptions in the GToBI model (Grice et al., 2005), according to which there is no H*+L bitonal accent in German but that such accents are better described as H* tones followed by a low phrase accent. We gloss over this here as this is immaterial for the present investigation.

7. At the moment, it is not possible to calculate p values for models with random correlation patterns. Therefore, we present the p values for the simpler models with random slopes but without correlation patterns (e.g. Hofmeister, Jaeger, Arnon, Sag, & Snider, 2011).

8. This was suggested to us by an anonymous reviewer.

9. This assumption needs quantitative backup. We are not aware of any studies that have investigated this issue.

10. The explanation was suggested to us by an anonymous reviewer.

Additional information

Funding

This work was supported by the German Research Foundation DFG as part of the Collaborative Research Centre (Sonderforschungsbereich, SFB) 632 Information Structure at the Humboldt-Universität zu Berlin and the University of Potsdam.