Red, green, corpus! A usage-based analysis of word order variation in Dutch bipartite clause final verbal clusters


PhD defence

I defended my PhD thesis at the University of Leuven (8 June 2005)

Board of examiners:


Summary of the PhD

In my PhD dissertation, one of the most intriguing types of syntactic variation in Dutch is studied, viz. word order variation in bipartite verb clusters, consisting of a past participle and the auxiliary verb zijn "to be", hebben "to have" or worden "to be(come)":
  1. [...] dat moordende chauffeurs van de weg gehaald worden (green word order)
    [...] that murderous drivers from the road removed are.
    [...] that murderous drivers are removed from the road.
  2. [...] dat moordende chauffeurs van de weg worden gehaald (red word order)
    [...] that murderous drivers from the road are removed.
    [...] that murderous drivers are removed from the road.
The goal of the research is to assess the mechanisms that influence the choice for green (a) or red (b) word order in contemporary Dutch, and to answer the question why both word order variants are available. On the basis of language data, extracted from two representative corpora, the effect of a set of language-internal and language-external factors on the choice of word order is studied empirically and quantitatively.

Chapter 1 is devoted to an overview of the relevant general and scientific literature of the past decades. The overview shows that language scientists, starting from the early 1950's, have focused on the word order variation from very different angles (diachronic versus synchronic, language standards/attitudes versus language use, formal grammatical models versus usage-based models). The observation that the past participle and the auxiliary verb can trade places has also led to different recommendations in grammars and advisory books. It appears, more particularly, that some of the grammars / advisory books have a preference for either the green (a) or the red (b) variant, but none of the variants is disapproved of. Chapter 2, which concludes the introductory part of the dissertation, presents the methodological fundaments of the research, viz. the corpora, the selection criteria and the statistical analyses.

In chapter 3 to chapter 6, the results of the empirical investigation are presented and discussed. In order to organize things properly, the different language-internal and language-external factors are classified along 4 dimensions: the contextual dimension, the prosodic dimension, the semantic dimension and the psycholinguistic dimension.

Along the contextual dimension (chapter 3), the influence of the factors region and register is studied. The results show that the green word order (a) is used more frequently in Belgian Dutch than in Netherlandic Dutch, and that within Belgian Dutch, people living in Brabant use the red (b) variant more often than people living in the other provinces, while people living in East-Flanders use green word order (a) more frequently than people living in the other provinces of the Belgian Dutch area. These results were interpreted in terms of the specific standardization process in Belgium, whereby the province of Brabant functions as the linguistic center. Furthermore, the choice of word order is influenced by the language mode, the degree of interaction and the degree of editorial control: in spoken language, green word order (a) is used more often than in written language; in communicative situations with a high degree of interaction (such as dialogues) green word order (a) is preferred more often in comparison with situations with a low(er) degree of interaction (monologues); in situations with a high degree of editorial control (news papers) red word order (b) is used more often than in situations with a low degree of editorial control (internet dialogues). These results show that language users prefer green word order (a) more often in circumstances when temporal restrictions on the phonetic, prosodic and/or grammatical encoding of the language product are high.

Along the prosodic dimension (chapter 4), the effect of the factor accent distribution before, in and after the verbal cluster is discussed. More specifically, it was investigated whether the choice of word order is influenced by the avoidance of an accent clash, the avoidance of a long interval with unaccented syllables and the protection of the so-called 'flat hat' pattern (which is one of the basic intonation contours in Dutch). All variables that were scrutinized against this background (both directly and indirectly) confirm the existence of such a prosodic principle (the variables analysed were morphology of the participle, definiteness of the previous NP, informationality of the previous word, presence of a fixed expression, length of the middle piece, presence of a postverbal constituent, grammatical status of the postverbal constituent and distance between two accented syllables).

Along the semantic dimension (chapter 5), the effect of the factor status of the past participle is studied, i.e. the distinction between adjectivally used participles and (different kinds of) verbally used participles. The results show that adjectival participles occur more often in green word order (a) than verbal participles, but, contrary to what is traditionally assumed, it is not impossible that adjectival participles show up in red word order (b). Moreover, it could be discerned that verbal clusters containing the auxiliary verb zijn "to be" are used in green word order (a) more often than clusters containing the auxiliaries worden "to be(come)" and hebben "to have", and that passive constructions are more frequently used in green word order (a) than active constructions.

Along the psycholinguistic dimension, the role of syntactic priming and participial frequency is scrutinized (chapter 6). The results show that word order is determined by the word order chosen in a previous verbal cluster: if a preceding verbal cluster appears in red word order (b), the same word order is adopted by the next cluster, and, vice versa, if a green verbal cluster (a) precedes, a green word order (a) cluster is chosen in the next cluster. Finally, the results reveal that highly frequent participles occur in red word order (b) more often than participles that are less frequent.

The analyses in chapters 3 to 6 were intended to answer the question whether a certain factor has an influence on the choice of word order, and, if it does, how strong the influence is. A global statistical analysis, taking into the account the effect of all separate factors, is performed in chapter 7. The global analysis weighs the effect of all the factors against each other and computes the explanatory and predictive power. In particular, a classification tree analysis and logistic regression analysis were used. The first technique reveals that almost 80% of the observed variation can be predicted on the basis of 5 factors. The second technique shows that all investigated factors, except for the purely prosodic factor, have a statistically significant effect on the explanation and prediction of the choice of word order. Based on these two global techniques, chapter 7 also seeks a linguistic explanation for the coexistence of the green (a) and red (b) word order. It is argued that at least part of the observed variation can be explained in terms of production pressure: the higher the production pressure, the more green word order (a) is chosen, i.e. the higher the temporal restrictions on the phonological and prosodic encoding of the intended message and/or on the grammatical encoding, the more green word order (a) is chosen.

Download

> The full version of my PhD (in Dutch) can be downloaded here.

Official reference: De Sutter, Gert (2005). Rood, groen, corpus! Een taalgebruiksgebaseerde analyse van woordvolgordevariatie in tweeledige werkwoordelijke eindgroepen. Unpublished PhD K.U.Leuven. http://hdl.handle.net/1979/98