PP and SVP word order in child- and adult-directed Dutch: a corpus investigation

Aleksei Nazarov

Utrecht University

In Dutch, verb-final word order has standardly been seen as underlying since Koster (1975), even if V2 word order is used in main clauses. From a learnability perspective, it is important to gauge how often exceptions against surface verb-finality occur outside main clauses. For instance, PPs and separable verb particles (SVPs) may be placed to the right of a verb, (1). Verb raising, (1b), is required in Dutch and will always provide exceptions to verb-finality.

(1) a. PP-left: … dat zij [[in Utrecht] doceerde]. / PP-right: … dat zij [doceerde [in Utrecht]].

that she in Utrecht taught that she taught in Utrecht

‘that she taught in Utrecht’

b. SVP-left: … dat zij [[door]-t_i kon] lopen_i. / SVP-right: … dat zij [t_i kon] [door]-lopen_i.

that she SVP could walk that she could SVP- walk

‘that she could keep walking.’

In this paper, the frequency of these orders in adult- and child-directed (AD, CD) and child-produced texts (both written and spoken) is gauged using GrETEL 4 (Odijk et al. 2018), a tool to search treebanks in Alpino (Bouma et al. 2001) format. Corpora previously uploaded to GrETEL 4 were used, in addition to Dutch-spoken subcorpora of CHILDES (MacWhinney 2000) that contain multi-word adult-child interactions.

Generally, PP-left order is preferred, but not SVP-left order. The rates at which each order occurs depend on the modality of the corpus and whether it is adult- or child-directed. Table 1 indicates the percentage of PP/SVP-left word orders in each subcorpus along with the total occurrence of PP/SVP-left/right sentences per 1000 sentences in that same subcorpus. There is a continuum from written adult-directed to spoken child-directed/child-produced texts where the former have the least preference for PP/SVP-left orders and the latter, the strongest.

Table 1.

WritAD WritCD SpokAD SpokCD SpokenChild-produced

% PP-left 51% 74% 71% 77% 80%

Total PP (/1000 sentences) 260 128 98 10 4

% SVP-left 12% 26% 30% 51% 48%

Total SVP (/1000 sentences) 11 4 6 <1 <1

PP word order also fluctuates based on PP function; the most popular of these are location/direction (ld), modification (mod) and predicate complement (pc). As can be seen in table 2, location/direction more strongly prefers PP-left order than the other two functions. However, the mod and pc functions still show a clear continuum from less PP-left order in written adult-directed texts to more PP-left order spoken child-directed/produced texts.

Table 2.

WritAD WritCD SpokAD SpokCD SpokenChild-produced

% PP-left, ld 79% 92% 88% 88% 86%

% PP-left, mod 50% 65% 64% 63% 70%

% PP-left, pc 37% 70% 60% 75% 77%

Summarizing, PP-left is preferred across the board, which leads to fewer exceptions to verb-finality, while SVP-left is not preferred across the board, but there are few attested SVP tokens, making it an insignificant source of exceptions, especially in child-directed language, which has more PP-left/SVP-left orders and fewer occurrences of PPs/SVPs in general (table 1).

CLIN33
The 33rd Meeting of Computational Linguistics in The Netherlands (CLIN 33)
UAntwerpen City Campus: Building R
Rodestraat 14, Antwerp, Belgium
22 September 2023
logo of Clips