Verb-particle constructions in translated Dutch: a pilot study

This paper presents a pilot study on Dutch verb-particle constructions (VPCs) and their use in translated Dutch. The study falls within the framework of a larger project on how spatial elements become applicatives. In Dutch, VPCs are formed by combining a base verb such as werken ‘work’ with a preverbal element such as uit ‘out’. This type of word formation is quite productive in Dutch, for example: uitwerken ‘elaborate’, overwerken ‘work overtime’, afwerken ‘finish’. These preverbal elements usually have a spatial origin and may alter the valency of the base verb. English, closely related to Dutch, also has an abundance of VPCs, also known as particle verbs (e.g., look up, look for) (Thim 2012). In contrast, French is a verb-framed language, and does not use VPCs (Herslund 2005, Hijazo-Gascón & Ibarretxe-Antuñano 2013).
While VPCs in Dutch and English have been extensively researched (Blom, 2005; Cappelle, 2023; Thim, 2012; Van Kemenade & Los, 2003), their translation has not received equivalent attention, except for select studies (for instance, Cappelle and Loock, 2017; and Claridge, 2002 – focusing on English-German translations). To tackle this underexplored area, this study examines Dutch VPCs in translated subtitles, building on Cappelle and Loock's (2017) hypothesis that typological differences between Romance and Germanic languages ‘shine through’ in translations. Concretely, the following research questions will be addressed: First, do Dutch subtitles translated from French feature fewer VPCs than Dutch subtitles translated from English, in line with Cappelle and Loock’s (2017) study on VPCs in English translations? Second, what source-language structures most commonly give rise to VPCs in Dutch translations?
This pilot study draws on the The OpenSubtitles subcorpus of the OPUS corpus available on Sketch Engine containing subtitles in multiple languages, including English, French, and Dutch. The dataset includes approximately 31 million aligned English-Dutch sentences and 22 million French-Dutch sentences, collected in 2011 (Tiedemann, 2011). Queries were conducted to identify both separable and inseparable forms of the VPCs containing the particles op, af, and uit. For example, oplossen ('to solve') appears as Ik los het probleem op ‘I solve the problem’ in separable contexts and as Het probleem is opgelost ‘The problem is solved’ in inseparable ones. I included all English or French source language sentences (<s/>) that corresponded to a VPC with the preverbal elements op / af / uit in the target language: ([word="op.*" & tag="verb.*" | ([tag="verb.*"] []{0,2} [word="op"] ); Filter by aligned corpus:opus2_en;Shuffle:]). Each identified instance was manually verified to ensure that that sentences where the preposition does not form a VPC with the verb were excluded. For instance: Ik koop wel iets op het vliegveld ‘I will buy something at the airport’. For this small-scale pilot study, I extracted a random sample of 10 000 cases, which I analyzed case per case until ending up with 100 relevant VPCs. This process was conducted for Dutch translated from English, Dutch translated from French, as well as for non-translated Dutch and non-translated English.
The analysis followed three steps, a frequency analysis, calculating the relative frequency of VPCs in Dutch translations from English and French, normalized by corpus size. Followed by correspondence mapping, which consisted of identifying whether Dutch VPCs were translated from VPCs, simplex verbs, or other constructions in the source language. Finally, I conducted a qualitative analysis: examining specific examples of translation choices to uncover patterns, with particular attention to typological influences.
Preliminary findings suggest that the translation of VPCs in Dutch involves a complex interplay of factors, including source-language structures, target-language norms, and contextual considerations. Note first of all that the distribution of VPCs with different particles in non-translated Dutch subtitles varies: VPCs with op account for 0.42%, VPCs with af for 0.18%, and VPCs with uit for 0.26% (proportions of VPCs out of the total number of aligned sentences per subcorpus).
Additionally, the preliminary results of the pilot study show that a VPC in the source language does not automatically correspond to one in the target language. Only 46% of Dutch VPCs with op align with English VPCs, while the rest involve simplex verbs or other structures. Similarly, just 29% of af VPCs and 32% of uit VPCs are translations of English VPCs, indicating weak correspondence.
Despite this, the findings support the hypothesis that crosslinguistic differences influence translation choices. Dutch subtitles translated from English contain relatively more VPCs (op 0.50%, af 0.51%, and uit 0.39% ) than those translated from French (op 0.39%, af 0.20%, and uit 0.32%). Moreover, Dutch translations from English feature more VPCs than non-translated Dutch (op 0.45%, af 0.18%, and uit 0.26%), reflecting the high frequency of VPCs in English, further supporting Cappelle and Loock’s (2017) “shining through hypothesis”. However, this pilot study is an initial exploration, and further research with larger samples, including other Germanic and Romance languages and covering additional particles, is needed.