Constraints on ontology changing complexation processes: Evidence from event-related brain potentials

Abstract

This paper investigates complex anaphoric reference (i.e., when an anaphor refers to a propositionally structured referent). Complex anaphors (e.g., this process, this event) differ in their ontological feature setup, and the ontological type assigned to a referent can change due to the lexical meaning of the complex anaphor. Previous research has proposed that such changes have to comply with an ontological ‘abstractness constraint’ restricting the direction of ontological change. We present an event-related potential study that provides evidence that violations of the abstractness constraint result in processing costs. The data reveal that violating this constraint by shifting the referent towards a less abstract ontological type elicits an enhanced N400, while reduction of ontological features towards a more abstract type exerts no extra processing demands. The data indicate that the abstractness constraint affects real-time sentence comprehension and that different ontological types are implicationally related.

Keywords:

• Complex anaphors
• Ontology
• Reference
• Language comprehension

Acknowledgements

This research was supported by Deutsche Forschungsgemeinschaft (DFG) research grants (BU 1853/2-1 to the first author, and SCHW509/6-1,2,3 to the second and third author). We are grateful to Manuel Dangl, Elisabeth Dietz, Jona Sassenhagen and Jan Patrick Zeller for their assistance at various stages in data preparation, collection and analysis.

Notes

¹Lyons (1977) defines ‘first order entities’ as physical objects that exist continually in space and time independent from their verbalisation. ‘Second order entities’ are spatio-temporal events, situations, or processes (that take place, happen) which have no continuity, but a temporal duration. These referents can be mere verbal constructions (cf. Lyons, 1989). ‘Third order entities’ (concepts, propositions) are always verbal constructions and independent from space and time.

²It is somewhat surprising that the average ratings for the critical conditions are not more spread out along the 6-point scale. However, the common rating behaviour of the participants indicates that they were generally reluctant to assign extreme ratings. This is also indicated by a mean rating of 4.86 for the infelicitous items. This tendency might have to do with the complexity of the stimulus material and the subtleness of the ontological manipulation. Nonetheless, statistical analyses indicate a clear divergence for the event anaphors.

³We carried out the same rating study with the participants from the ERP study reported in this paper (N = 24), who were asked to complete the questionnaire following the ERP recording. The general pattern described above was confirmed by this study with a main effect of ONTOLOGY, F ₁(2, 36) = 6.82, p < .003; F ₂(2, 178) = 8.37, p < .001, and for the (more meaningful) items analysis, reliable differences between both process and state anaphors vs. event anaphors, as well as no significant differences between process and state anaphors. Yet, the analysis by subject yielded slightly different results, with significant differences between process vs. event and process vs. state, and only a marginal difference (p < .09) for state vs. event. However, two caveats led us to carry out an independent study (as suggested by anonymous reviewers). First, data from five participants had to be discarded from further analysis because their average ratings of the (unacceptable) filler items were lower than 3, indicating that these participants were not paying attention in a satisfactory manner. Second, the overall rating behaviour indicated a bias toward the acceptable end of the scale (with mean ratings of 1.81 for process anaphors, 2.17 for state anaphors, 2.58 for event anaphors, and 4.81 for the incorrect filler items). This could be due to prior exposure of the critical constructions leading to biased rating behaviour or it could be caused by the participants' weariness following the ERP session.