PhD defence: Left or right of the adverb?

Date of news: 16 March 2022

‘Jan heeft waarschijnlijk de hond uitgelaten’ or 'Jan heeft de hond waarschijnlijk uitgelaten’ (Jan has probably walked the dog). In Dutch both word orders are fine. Theoretical linguistics experts claim that this variation depends on whether the direct object has already been mentioned in the conversation or whether it is new information. Gert-Jan Schoenmakers investigated this with a series of experiments and found that the distribution is more free than is often claimed. He will defend his thesis on 17 March.


Languages can differ from each other in the order of words and groups of words. For example, in English we say 'a red rose', but the French speak of 'une rose rouge'. And in Dutch we say that someone ‘een roos plukt’, but in English you should say that someone ‘picks a rose’. Words and groups of words can also appear in different positions within a language. For example, in Dutch a direct object can appear to the left or right of an adverb or an adverbial clause. For example: 'Ik zei dat Jan de film gisteren heeft gezien’ or 'Ik zei dat Jan gisteren de film heeft gezien’ (I said that Jan saw the film yesterday). This variation is called scrambling.

Why a certain (definite) direct object is to the left or right of an adverb has been the subject of earlier and more frequent research, but most theories are based on researchers' intuitions. Linguist Gert-Jan Schoenmakers used a series of experiments to map the scrambling behaviour of the definite direct object in Dutch.

Type of adverb

Does it matter, for example, which type of adverb is in the sentence? In two experiments, Schoenmakers examined sentences with negation (not) and sentences with a time clause (yesterday). ‘The results show that test subjects like the sentences better when the direct object is to the left of the negation, but not when the direct object is to the left of a time clause', says Schoenmakers. ‘The same preference occurred when the test subjects had to produce the sentences themselves.’

In two other experiments, Schoenmakers investigated another difference between types of adverbs: the difference between adverbs that relate to the whole sentence, as in 'Roos verft helaas het raam’ (Roos paints the window frame, unfortunately), and adverbs that only relate to the verbs in the sentence, as in 'Roos verft snel het raam’ (Roos paints the window frame quickly). Schoenmakers found that in the rating experiment, subjects gave all sentences high scores, regardless of order or type of adverb. There was, however, a slight preference for sentences with the direct object on the left of the adverb that relates to a verb, as in 'Roos paints the window frame quickly'. ‘That effect became even clearer in the results of the production experiment', says Schoenmakers. ‘From these four experiments we can therefore conclude that the type of adverb plays an important role in choosing one order or the other. But the sentences get high scores either way. Scrambling is in principle a free choice.’

Information structure

Schoenmakers also looked at the information structure of scrambling sentences: the order in which information is offered. For example, information that is already known to the interlocutor often appears earlier in the sentence than new information. In the case of scrambling, theoretical linguists argue that the direct object should therefore be to the left of the adverb if it has already been mentioned in the conversation, and to the right if it conveys new information.

Schoenmakers tested the information structure in combination with scrambling using two more experiments. The results of the production experiment showed that the known-for-new principle also applies to scrambling in Dutch: the definite direct object was more often placed to the left of the adverb when it was known than when it was new information. But the results of the evaluation experiment show that scrambling is optional: native speakers of Dutch found both word sequences fine. Schoenmakers: 'This shows that scrambling is influenced by the information structure, but not determined by it.’

Theoretical and empirical

Schoenmakers' results suggest that scrambling is freer than is often assumed in linguistic theory work. ‘My dissertation shows that it is necessary to empirically test a theoretical linguistic model and to look for convergent evidence, because a theoretical model that is mainly based on the intuitions and insights of language researchers does not necessarily correspond to linguistic reality', says Schoenmakers. ‘In the case of scrambling of definite direct objects, it turns out that the variation is influenced, but not determined, by the information structure. In addition, the type of adverb appears to play an important role. Without an experimental approach, these factors would not have come to light.’

Gert-Jan Schoenmakers will defend his dissertation Definite objects in the wild: A converging evidence approach to scrambling in the Dutch middle-field on 17 March from 10:30.

Photo by Josh Sorenson from Pexels