see for further details

(Non)factivity influences Prosody in Russian Complex Sentences

Ina Mleinek (University of Leipzig)

Within Russian complex sentences with object complements I observed a tendency to mark information structural distinctions (old vs. new) both within right/left – peripheral embeddings; a phenomenon which in fact is independent of the information structure of the matrix clause. I observed that known information will be de-accented. In a further step I investigated the backgrounding strategies for whole propositions embedded under a matrix clause. The question arises if there is de-accentuation and if so, what does it mean to a complement clause to be de-accentuated. To answer this question a reading study was carried out with strong contexts triggering different, but unambiguous focus readings. The result of this experiment clearly shows that de-accentuation induced by different focus readings is possible, but it is not accompanied by a low tone (L) spreading to the rest of the sentence like in left focus constructions. Instead, it will be indicated by an H+L* - pitch accent plus a low boundary tone in the matrix clause and a reset for the de-accentuated clause showing declination on a lower level. Here I extensively refer to pitch tracking data. Since I used strong triggers for backgrounding (non)factivity did not play a role within this experiment. Instead, (non)factivity interferes with information structure when triggers are weakened.

In order to disentangle potential information structural and factivity induced effects I conducted a second production study where old and new propositions were embedded under both factive and non-factive matrix verbs, this time without enforcing backgrounding by strong triggers. One hypothesis then would be that factives and non-factives can both take advantage in the same way of the contextual input. According to the results of the first experiment this would have the following consequence: In case the material of the secondary clause is previously mentioned, I expect a fall on the matrix verb. If on the other hand the secondary clause is new a rising tone within the matrix clause will be expected indicating that the main stress will follow in the embedded clause. So in a sense, falls and rises are dependent on the IS status, but independent of the (non)factivity of the matrix verb. If factivity would interfere falls and rises could be distributed in a special manner over conditions. The zero hypothesis, instead, would claim random distribution of falls and rises within all four conditions.

Interpreting the results forced me to reject the zero hypothesis. Instead, the experiment revealed significant contrasts 1) between factives with old vs. new information, 2) between non-factives with old vs. new information and 3) between factives vs. non-factives with old information, but not for factives vs. non-factives with new information. In the latter case new information is both for factives and non-factives represented by a rise within the matrix clause. This is counter-evidence against the claim that factive verbs assign the embedded sentence a background feature. If this would be so, the main accent of the whole clause would be always on the matrix verb itself, to mark it the exclusive focus of the whole sentence while the embedded clause is de-accented. This means that factives can embed new information. The contrast under 1) I take as evidence that factives mark old and new information differently. Although 2) shows that this difference is also present within non-factives, the percentage of falls within the condition ‘non-factives with old information’ is much smaller (30%) than within factives with old information (75%). This explains the contrast in 3). It gives rise to the assumption that non-factives do not de-accent their complements when old but only if the matrix verb is old, too. Factives, on the other hand, clearly de-accent if the embedded clause represents old accessible information. In general it means, Russians make choices of tonal configurations which differ according to factivity clearly showing that semantics affects prosody. But apart from preferences for certain tonal configurations within the matrix clause the embeddings, too, show interesting effects. A non-de-accented but old embedded clause will have lower F0-values than one representing new information.