Successes and shortcomings of phonological accounts of Scandinavian object shift

Paulina Łyskawa, Jade Sandsted, Eline Visser, Nathan Young, Björn LundquistProceedings of the Linguistic Society of America 2022 · DOI

Abstract

Object shift (OS) is a word order phenomenon in Scandinavian languages where under some circumstances the object appears before a sentential adverb. Despite the frequent assumptions that word order is determined in syntax, and despite the link of OS and syntactic phenomena like V2, there is no consensus that OS is a syntactic phenomenon. Particularly, it has been observed that OS targets specifically prosodically weak elements. This motivated recent analyses of OS as a prosodic phenomenon. We focus on two proposals that look for a synchronic motivation for OS in a correlation between its distribution and some prosodic property: (i) Erteschik-Shir et al. (2020) posit that OS is motivated and modulated by prosodic incorporation, and (ii) Hosono (2013) hypothesizes that shifted pronominal objects help facilitate downstep. We identify concrete predictions from both proposals (default prosodic incorporation, and no downstep in unshifted OS-context sentences, respectively) and test them using novel data. The results show that neither of the proposals can be maintained in its original form. In addition to the empirical shortcomings of the prosodic proposals, we explore a missed syntactic generalization regarding the role objecthood plays in OS.

Introduction

In Scandinavian languages, speakers often exhibit a word order variation known as Object Shift (OS). In Swedish, for instance, a speaker might say the equivalent of "She helped me not" (shifted) rather than "She helped not me" (unshifted). Traditionally, this was viewed as a syntactic rule. However, because OS primarily affects "weak" pronouns—short, unstressed words—recent scholarship has pivoted toward phonological explanations. This paper evaluates two prominent prosodic theories: the Prosodic Incorporation model and the Downstep Facilitation hypothesis.

Evaluating the Downstep Hypothesis

Hosono (2013) proposed that Object Shift is driven by the need to facilitate a specific melodic pattern known as "downstep." Downstep is a phonological phenomenon where the pitch of successive high tones decreases in a staircase-like fashion throughout a sentence.

Schematic representation of a downstep pattern featuring successive pitch drops (H1, L1, H2, L2...).

If OS were motivated by the facilitation of this pattern, one would expect a strong correlation between shifted word orders and the presence of downstep. However, our analysis of the Nordic Word Order Database (NALS) reveals a significant disconnect. In Oslo Norwegian, 57% of shifted sentences showed no downstep. Furthermore, in Stockholm Swedish, the frequency of downstep was nearly identical regardless of word order: 45% in shifted sentences versus 43% in unshifted ones. These results suggest that downstep is an independent melodic choice rather than a driver of word order.

Testing Prosodic Incorporation

The second theory, proposed by Erteschik-Shir et al. (2020), suggests that pronouns shift because they are prosodically weak and need to "incorporate" or lean on a preceding verb for support. Under this view, adverbs are considered "bad hosts" for these weak sounds.

To test this, we conducted acoustic analyses using spectrograms to look for physical evidence of merging, such as the deletion of initial consonants (e.g., the 'h' in henne).

Spectrogram of a Swedish sentence showing a shifted pronoun ("henne") that remains prosodically independent with a clear pitch contour.

Our findings contradict the incorporation requirement. We observed numerous instances where pronouns shifted but remained prosodically independent—retaining clear 'h' sounds and distinct pitch contours, as seen in the spectrogram above. Conversely, we found cases where pronouns did not shift but still successfully incorporated into the following adverb. This indicates that shifting is neither a prerequisite for nor a guarantee of prosodic merging.

The Persistence of Syntax

Beyond the acoustic failures of prosodic models, there is a significant syntactic generalization that these theories overlook: the role of "objecthood." In Norwegian, certain articles or phrase components are phonologically identical to weak pronouns and are equally "weak" in terms of stress. However, these elements do not undergo shift. If the phenomenon were purely driven by sound patterns, we would expect all weak elements to shift. The fact that only objects do so suggests that the brain is following a grammatical (syntactic) rule rather than a purely rhythmic one.

Conclusion

While prosodic factors certainly correlate with Object Shift, they do not appear to be its primary motivation. The failure of both the Downstep and Incorporation hypotheses to account for naturalistic production data suggests that OS remains a fundamentally syntactic phenomenon. Future research should re-examine classic syntactic constraints, such as Holmberg’s Generalization, and explore how syntactic "phases" might interact with prosody in more nuanced ways than current phonological models allow.

Summary updated

Abstract

Object shift (OS) is a word order phenomenon in Scandinavian languages where under some circumstances the object appears before a sentential adverb. Despite the frequent assumptions that word order is determined in syntax, and despite the link of OS and syntactic phenomena like V2, there is no consensus that OS is a syntactic phenomenon. Particularly, it has been observed that OS targets specifically prosodically weak elements. This motivated recent analyses of OS as a prosodic phenomenon. We focus on two proposals that look for a synchronic motivation for OS in a correlation between its distribution and some prosodic property: (i) Erteschik-Shir et al. (2020) posit that OS is motivated and modulated by prosodic incorporation, and (ii) Hosono (2013) hypothesizes that shifted pronominal objects help facilitate downstep. We identify concrete predictions from both proposals (default prosodic incorporation, and no downstep in unshifted OS-context sentences, respectively) and test them using novel data. The results show that neither of the proposals can be maintained in its original form. In addition to the empirical shortcomings of the prosodic proposals, we explore a missed syntactic generalization regarding the role objecthood plays in OS.
BibTeX
@inproceedings{yskawa2022successes,
  author = {Paulina Łyskawa and Jade Sandsted and Eline Visser and Nathan Young and Björn Lundquist},
  title = {{Successes and shortcomings of phonological accounts of Scandinavian object shift}},
  booktitle = {Proceedings of the Linguistic Society of America},
  year = {2022},
  doi = {10.3765/plsa.v7i1.5261},
}
Figure