argument placement in norwegian

This paper gives an overview of the results from three data collection sessions that took place in Norway in 2018, which specifically targeted the placement of subjects, objects and particles in main clauses. The results reveal a fairly high amount of variation in the relative linear order of phrasal subjects and negation, and phrasal objects and verb particles, while the placement of pronouns show little or no variation. We view these results in a wider context of variation within the North Germanic languages, and furthermore explicitly describe the structure of the collected data, and how to access it in the online Nordic Word Order Database.

abstract This paper gives an overview of the results from three data collection sessions that took place in Norway in 2018, which specifically targeted the placement of subjects, objects and particles in main clauses. The results reveal a fairly high amount of variation in the relative linear order of phrasal subjects and negation, and phrasal objects and verb particles, while the placement of pronouns show little or no variation. We view these results in a wider context of variation within the North Germanic languages, and furthermore explicitly describe the structure of the collected data, and how to access it in the online Nordic Word Order Database.
[1] introduction The modern North Germanic languages, Norwegian, Swedish, Danish, Icelandic and Faroese, have all developed their own argument placement patterns. In this article, we focus on the placement of subjects and objects in Norwegian, and present the results from a series of elicitation experiments that were run in 2018 with a total of 63 participants from different parts of Norway (mainly Eastern and Northern Norway). The material is available online in the database Nordic Word Order Database (NWD). The core patterns investigated were the relative linear order of (i) pronominal and NP subjects and sentence adverb (including negation), (ii) pronominal and NP objects and negation, (iii) pronominal and NP objects and verb particles, and (iv) the relative order of subjects and objects in the midfield of the clause. The pronouns to be discussed are all light first person, third person or reflexive personal pronouns.
The most striking characteristics of Norwegian argument placement is the strong influence of form, here, NP versus personal pronoun. In a string that includes a pronoun and a sentence adverb or a verb particle, the pronoun will typically precede the sentence adverb or particle. If the argument is a full phrase [32] LUNDQUIST & TENGESDAL NALS Journal, Vol. 7, 1 (NP/DP) 1 , the argument typically follows the sentence adverb or particle. In the case of pronominal arguments, there are only a few exceptions to this generalization, while there is more variation in the placement of phrasal arguments. We give examples of the core patterns in (1-3) below. The examples in (1) illustrate the unmarked placement of subjects relative to negation. In (2), the order of objects and negation is given. The examples in (3) illustrate the typical ordering of objects and particles.
(Neg-SubjNP ) yesterday took not teacher.DEF bus.DEF to town.DEF 'Yesterday, the teacher did not take the bus to town.' b. I går tok han ikke bussen til byen.
(SubjPRO-Neg) yesterday took he not bus.DEF to town.DEF 'Yesterday, he did not take the bus to town.' (2) a. Politiet arresterte ikke ranerne i går.
(ObjPRO-Part) guard.PL.DEF threw him out yesterday 'The guards threw him out yesterday.' The effect of form is most categorical in (2) above, that is, in the phenomenon standardly referred to as object shift (Holmberg 1986); NP objects never shift across sentence adverbs, and absence of pronominal object shift tends to have straightforward effects on interpretation (e.g., added contrast/focus). As for subject placement (1), NP subjects are sometimes placed before sentence adverbs, especially in cases where the sentence adverb is not a negation (Faarlund et al. 1997;Nilsen 1997;Bentzen 2007;Åfarli 2010). Similarly, NP objects quite frequently appear before verb particles, although the unmarked word order has the order in (3a), where the NP object follows the particle (see [1] We will refer to all phrasal arguments with a lexical noun as a head as NPs. Arguments realized as personal pronouns will be referred to as pronouns.

ARGUMENT PLACEMENT IN NORWEGIAN
[33] Sandøy 1976;Åfarli 1985;Svenonius 1996;Larsson & Lundquist 2014). Both with respect to NP subject placement relative to sentence adverbs, and NP object placement relative to particles, it is often unclear what governs the choice of linear order. Regarding the placement of pronominal arguments, information structure effects are quite clear when the pronoun appears to the right of the adverb or particle: pronouns following adverbs/particles are preferably interpreted as contrastive.
In all of the North Germanic languages, there is a difference in the placement preferences for pronominal and phrasal arguments, but in Swedish, Danish, Icelandic and Faroese, the difference is generally more subtle, and more restricted by syntactic function (i.e., subject and object). For example, both Swedish and Danish have strict ordering restrictions for objects and particles, independent of form (see Lundquist 2014 and references therein). Danish, Faroese and Icelandic all have similar ordering restrictions on definite phrasal subjects and pronominal subjects: subjects precede sentence adverbs (see, e.g., Svenonius 2002;Vangsnes 2002). With the data in the Nordic Word Order Database, we can investigate in more detail how the form (pronoun vs. NP) and syntactic function (subject vs. object) affect linear order.
The structure of the article is as follows. In Section 2, we give a short introduction to the relevant aspects of the syntax of Norwegian, with a focus on the linear order of constituents on the clausal level. Section 3 describes the experiment. Section 4 reports on the participants and the data collection. In Section 5, we present the results while we discuss them in Section 6.
[2] the basic clause -level syntax of norwegian Like all the modern North Germanic languages, Norwegian is a VO language, with quite systematic asymmetric V2: the finite verb is placed in second position in main clauses, while non-finite verbs surface inside the verb phrase, after sentence adverbs and the subject, but before objects and verb particles (Diderichsen 1946;Holmberg & Platzack 1995;Vikner 1995). In embedded clauses all verbs are located inside the VP. An extensive overview of the exceptions to the asymmetric V2 pattern in Norwegian is given in Westendorp (2021), but see also Julien (2007) and Wiklund et al. (2009) for overviews of patterns and variation in the North Germanic languages. In (4), we give examples of a main clause and an embedded clause, in a prototypical sentence with the subject in clause-initial position. In the main clause (4a), the subject is followed by an auxiliary in the V2 position. In the embedded clause (4b), the verb stays in the verb phrase, as diagnosed by the negation preceding the verb. In clause structures like (4), there is a strict linear order for the arguments and sentence adverbs. In the main clause in (4a), the subject and the object are separated from the midfield/TP by the finite and non-finite verb. 2 In the embedded clause in (4b), the subject and the negation are linearly adjacent, but for reasons still unknown, subject-adverb inversion is highly marked, if not impossible, unless a sharp contrastive reading of the subject is provided. This sharply contrasts with the order of subjects and negation in main clauses like (1), where an NP subject most typically follows the negation.

[2.1] Subject placement
Most of the research on midfield subject placement in Norwegian has focused on the placement of subjects with respect to negation, rather than to other midfield adverbs, see for example Eide (2002), Østbø Munch (2013), Bentzen (2014b), Anderssen et al. (2018) and Olsen (2019) (but see also Åfarli 2010 for discussion of other adverbs). This is mainly due to the fact that negation is the most frequent midfield adverb, but also due to the fact that it is relatively easy to compare negation placement cross-linguistically. In several of the studies, the term subject shift has been used to describe the variable placement of subjects (see more below).
Above we stated that non-contrastive pronominal subjects precede negation and other adverbs in the midfield, but this is a slight simplification. As Østbø Munch (2013) shows, the opposite order is not hard to find in spoken language corpora. Østbø Munch finds as much as 17.5% post-negation pronominal subjects in her study of negation placement in the Nordic Dialect Corpus (Johannessen et al. 2009), and far from all of these are contrastive. There are a couple of factors that may explain the relatively high amount of Negation-SubjectPRO order. First, in most dialects, negation has reduced monosyllabic forms (ke/kje/tje) in [2] Following Diderichsen (1946) we will refer to the positions between the finite verb and the non-finite verb(s) in declarative main clauses as the midfield of the sentence; see also Faarlund et al. (1997: 858-860).

ARGUMENT PLACEMENT IN NORWEGIAN
[35] addition to the disyllabic ikke/ikkje/ittje. The monosyllabic form is more likely to appear with auxiliaries, copulas and modal verbs, where the negation cliticizes to the verb and gives rise to more or less lexicalized forms like ha'kke ('have not'), må'kke ('may not') and ae'kke ('is not'); see Lindstad (1999) and Torgersen & Garbacz (2020) for discussions of the limits of cliticized negation in Norwegian. A cliticized negation may intervene between the finite verb and a post-verbal subject, but, in the majority of the dialects, it may also surface after the subject, as we will see shortly. (See Østbø Munch 2013 for a study of a dialect where the negation quite strictly cliticizes to the verb.) Another factor that influences subject placement is the type of speech act. Subject pronouns following negation seem to be more common in yes/noquestions than in declarative main clauses (see Urbanik & Svennevig 2019 for discussion). This may be due partly to an increased likelihood of a contrastive interpretation of subjects in questions compared to declaratives, but it may also be related to a high frequency of auxiliaries in these type of questions, leading to higher use of a monosyllabic negation (ae-/ha-/må-/bli'kke du …?). Yet, it is important to point out that the order Subject-Negation is still unmarked even in questions with auxiliaries and truncated negation, and speakers vary between orders (5a) and (5b)  Acceptability judgment studies on the placement of subject (Anderssen et al. 2018) suggest that native speakers find the order Negation-SubjectPRO highly marked in main clauses with lexical verbs in V2 position (see also Westendorp & Lundquist 2021 for a comparison between Norwegian and Swedish).
As for NP subjects, there is a preference for the Negation-Subject order. In the study by Anderssen et al. (2018), the authors found that Norwegian speakers consistently gave a higher acceptability score for post-negation NP subjects than pre-negation NP subjects (mean 5.3 compared to 3.4 on a six-grade scale). Swedish participants, tested on similar material, found both orders equally unmarked (see Westendorp & Lundquist 2021). Yet, both orders are produced in spontaneous speech in Norwegian, and it is unclear if choice of word order reflects differences in meaning. The order preferences for subject viz. negation do not necessarily carry over to other adverbs, although the details are still not known (but see Svenonius 2002 for a more extensive discussion of subject [36] LUNDQUIST & TENGESDAL NALS Journal, Vol. 7, 1 placement in Norwegian contrasted with the other North Germanic languages). In general, longer/heavier adverbs like aldri ('never'), alltid ('always'), and muligens ('possibly') appear to be more likely to follow subjects, both pronominal and phrasal ones, while lighter, more particle-like adverbs like vel, da and jo behave more like negation. In one of the Norwegian NWD data collection sessions, other adverbs than the negation were included; the results are discussed in detail below.

[2.2] Object placement
As in the other Mainland North Germanic languages, Norwegian generally makes a strict division between pronominal and phrasal objects when it comes to placement with respect to adverbs (Holmberg 1986). When the syntactic context allows, light pronominal objects have to shift to the left of midfield adverbs (6a), while phrasal objects surface to the right of the adverbs (6b). In all North Germanic languages, object shift can only occur if the verb moves out of the VP, a pattern known as Holmberg's generalization (Holmberg 1986). Object shift can be blocked by a non-shifted subject or an indirect object; see (7) (and Section 5.3 below for discussion). In the cases where object shift can occur, it is strongly preferred (if not obligatory) with weak pronominal forms in most North Germanic languages. However, there are exceptions. First, weak (i.e., prosodically deficient) pronouns can optionally remain in situ in several Swedish and at least some Norwegian dialects (see, e.g., Holmberg 1986;Sells 1998;Thráinsson 2001;Bentzen et al. 2013;Bentzen 2014a  Tengesdal 2021; Larsson & Lundquist 2022a and references therein). Second, if an object pronoun is accented due to information structure (e.g., contrastive focus), object shift is restricted. Since the presence of stress or accentuation affects pronominal object placement, several attempts have been made to give a phonological account of object shift (Erteschik-Shir 2005;Hosono 2013;Erteschik-Shir et al. 2021;Brinkerhoff & Tengesdal 2021), but problems with phonological approaches have also been discussed (Svenonius 2005;Lyskawa et al. 2022). One crucial problem for phonological accounts is the third person neuter pronoun det: when det refers to a full clause or VP, it usually will not shift past negation, even though it is unstressed as in Jeg trur ikke det ('I don't think so') (see especially Andréasson 2008; Bentzen & Anderssen 2019 and references therein).
In addition to object shift across sentence adverbs, the experimental data in the present study also investigate the relative placement of an object in relation to a subject, so-called long object shift (LOS): An object that linearizes to the left of the subject has undergone long object shift (see, e.g., Holmberg 1986;Heinat 2007, and examples (11) and (24) below). Among the North Germanic languages, only Swedish is expected to allow LOS (in contexts with weak pronominal objects; see, e.g., Heinat 2007; Lundquist 2013). We do not expect any instance of LOS in Norwegian or the other North Germanic languages (see Larsson & Lundquist 2022a for more discussion).

[2.3] Particle placement
The North Germanic languages differ with respect to the placement of verb particles in relation to objects. Norwegian and Icelandic overall pattern together with English (see, e.g., Collins & Thráinsson 1996;Svenonius 1996), where weak pronouns and reflexives linearize before the particle, while NP objects and accented pronouns can appear on either side of the particle; see (8a-b). Swedish, on the other hand, only allows direct objects following the particle; cf. (8c). Danish only allows objects preceding the particle; cf. (8d), irrespective of object form (see, e.g., Toivonen 2003;Lundquist 2014;Larsson & Lundquist 2022b). Faroese is most similar to Danish; see Lundquist (2020) for data and discussion. In this article, we will refer to the variable placement of the particle with respect It should be noted that there is dialectal variation in terms of linearization of pronouns in Norwegian, for instance in the Trønder dialects, which also allow post-particle weak pronominal objects (as indicated by the percentage sign in (8a); see, e.g., Larsson & Lundquist 2014;Aa 2015;Tengesdal & Lundquist 2021 for discussion).
We can also note that variable linearization of particles and objects is known to be influenced by a variety of factors in the North Germanic languages. These include not only the object type, but also the particle and the semantics of the verb and particle in combination (see, e.g., Larsson 2022). Regarding Norwegian, corpus studies have revealed that directional particles more often linearize after the verb and object, while metaphorical (non-directional/non-transparent) particles more often linearize before the object (see Tengesdal et al. 2018, and discussion below). Objects that take the semantic role of GROUND as opposed to FIGURE seem to be preferably linearized after the particle, irrespective of the form (see, e.g., Svenonius 2003;Aa 2015). 3 Moreover, information structure and prosody can most likely influence the order.
As we saw above, there is variation within and between the North Germanic languages in particle placement with regard to objects. However, the order of particles relative to subjects is stable: subjects precede verb particles. On the other hand, what we call long particle shift (LPS), that is, a particle that precedes the subject, has been observed in Norwegian child and teenage language (see Lundquist et al. 2019, p. 23, for discussion). For this reason, we also include long particle shift as a variable in our study. [3] Particles are often analysed as intransitive prepositions, see, e.g., Emonds (1976) and Svenonius (2003), where the argument of the particle typically is a FIGURE, i.e., an entity that is moved or located in relation to a location or an object, referred to as the GROUND argument. Typically, the ground has to be introduced with the help of a preposition, e.g., He threw the dog (FIGURE) out of the room (GROUND). Sometimes, the figure argument is implicit, or realized as the subject, and the ground argument appears to be an argument of the particle, e.g., He jumped off the train (GROUND).

ARGUMENT PLACEMENT IN NORWEGIAN
[39] The experiment included several conditions that test five types of verb particle constructions and the argument linearization of these in relation to pronominal and NP objects. More details are given in Section 5.4 below.
[3] materials Data collection took place in three different sessions, the first one in Oslo, followed by two sessions in Tromsø, one at the university and one at a high school. For the two first sessions, the same elicitation material was used, but for the last one, two items were removed, and four items were added. The first two sessions included 92 experimental items, and the last one 94. The general elicitation method, and the design of the experiment, is described in detail in Lundquist et al. (2019). Below we will only describe the linguistically relevant properties of the experiments; readers who are interested in the visual and temporal information about stimuli presentation are referred to Sections 3-4 in Lundquist et al. (2019).
The argument placement experiment consists of three parts, which all make use of different word order manipulations and target different word order phenomena. The phenomena investigated each concern the linear order of two elements within the midfield (subject shift, object shift, long object shift) or the VP (particle shift). In each part of the experiment, participants first have to read a sentence with only one element in the midfield or VP. They then have to transform the sentence in some way (described in detail below), so that it would now contain two (or three) elements in the relevant domain.
In the first part of the experiment, the participants are first presented with a subject-initial sentence, with a main verb in second position. This sentence is referred to as the background sentence in Lundquist et al. (2019). The subject is either a light pronoun (han 'he' or hun 'she') or an NP (in most cases a nonmodified definite noun). The verb is followed by a light pronoun (reflexive seg or first person meg), negation (ikke 'not') or a verb particle. In addition, each sentence contains some typical VP-material, for example a selected PP, and a temporal adverb, usually in sentence-final position; see (9) for an example. The participant is asked to read the background sentence aloud, and when this is done, a cue for a new sentence appears on the screen. The cue is the temporal adverb from the background sentence, which now appears as the start of a new sentence. When presented with this cue, the participant produces a target sentence, which should contain the same lexical material as the background sentence, but now with the subject in a postverbal position, either preceding or following the negation/object/particle. An example with negation is shown in (9) below: (9a) is the background sentence, and (9b) is the target, with the cue In (9), the word order choice made by the participant in the target sentence corresponds to one of the core variables in the study, namely subject shift. In some items the verb is followed by a light object (e.g., laereren hjalp meg med leksene i går, 'the teacher helped me with the homework yesterday', see (24) below), and here we test another variable, namely long object shift, by checking if the object in surfaces before or after the subject in the target sentences. In items with postverbal particles (e.g., løperen ga opp under siste runde i går, 'the runner gave up during the final lap yesterday'), we investigate if the particle surfaces before or after the subject in the target sentence, i.e., if it has undergone long particle shift (see example (25)).
In one of the data collection sessions (Tromsø high school), we additionally included four items that targeted subject shift with respect to other adverbs than negation, namely alltid 'always' and ofte 'often'. An example is given in (10) The second part of the experiment also targets the three variables subject shift, long object shift and long particle shift, but the focus in this part is on object shift, i.e., the linear ordering of an object and an adverb (here, negation). The objects tested were either a first person singular pronoun (meg) or a simple reflexive (seg). In this part, the background sentence has a periphrastic verb ARGUMENT PLACEMENT IN NORWEGIAN [41] form, usually the future tense. The background sentence is subject-initial, just like in the first part. The auxiliary is sometimes followed by negation, in order to test object shift, as in (11). The main verb is followed by either a pronominal object or a verb particle. The cue for the target sentence is a temporal adverb and a simple past tense form of the main verb used in the background; see (11) for an example of the background and the cue.
(11) a. Advokaten kommer ikke til å barbere seg (Background) Lawyer.DEF comes not INF shave REFL med barberhøvel. with razor 'The lawyer will not shave (himself) with a razor.' b. I går barberte … (Cue) yesterday shaved 'Yesterday, … shaved …' In the target sentence, there are now three elements that have to be linearized with respect to each other: the subject (advokaten), the object (seg) and the negation (ikke); this gives rise to six logically possible word orders. These word orders can all be attested in Swedish (see Larsson & Lundquist 2022a). In Norwegian, an object cannot be placed before a subject in the midfield (see further below); this restricts the possible orders to the following three: As was discussed in Section 2.2, light pronominal objects almost obligatorily shift past sentence adverbs in Norwegian, if the syntactic context allows it; (12c) is therefore not expected to be very frequent. In the present study, object shift is tested with either reflexive or first person object pronouns. In the experimental items, reflexive objects always co-occur with an NP subject, and first person objects always co-occur with a pronominal subject. When the subject is pronominal, we expect obligatory subject shift, and thus only the order in (12b).
The third part of the experiment targets object shift and particle shift. In these items, the object is either a non-modified noun phrase or a third person pronoun. In items targeting particle shift, several syntactically relevant parameters were manipulated. A subset of the items contained semantically transparent directional particles, like ut ('out') combined with the verb kaste ('throw') in (14). In these cases the particle was either followed by a directional preposition phrase, or just by a temporal adverb. Since we manipulated the object as well, that is, it was either a pronoun or a noun, we ended up with sets of four sentences, as in (15). In total four sets of the type illustrated in (15)  Another subset of test sentences contained non-directional particles. Again, the objects were either pronominal or phrasal in these cases, and we included in total three different verb-particle combinations: skjelle ut 'scold' (lit. 'bark out'), kjøpe opp 'acquire' (lit. 'buy up'), and sjekke opp 'hit on' (lit. 'check up'). An example of possible target orders is given in (16): (16) Rektoren skjelte {henne/studenten} ut {henne/studenten} principal.DEF yelled her/student.DEF out her/student.DEF i går. yesterday 'The principal scolded her/the student yesterday.' In addition, one particle classified as GROUND-selecting was included, i.e., an item where the direct object is not interpreted as the FIGURE (see footnote 3). This is the particle verb phrase rydde av bordet 'clear the table' (lit. 'clean off the table'). Finally, one item contains the particle verb bygge om 'rebuild' (lit. 'build about/around'), which is classified as having a prepositional particle.
A summary of the experimental items, with number of items per part and condition, is given in Table 1. For conditions where the number of items changed during the period of the field work, we give the number of items in both settings; an asterisk points to the number of items in the session at the Tromsø high school.
[  [4] participants and experimental setup As mentioned above, data collection took place during three different sessions in 2018. 5 Data was collected from 20 students at a high school in the Oslo area, who all identified as speakers of Urban East Norwegian (Kristoffersen 2000) or simply 'Oslo dialect', and 23 students from a high school in the Tromsø area, who identified as speakers of Troms dialect (Northern Norwegian). Finally, we collected data from 20 participants at the University of Tromsø, and these participants were from various parts of the country, from the northernmost parts of Finnmark to Kristiansand in the south. The geographical spread of the participants are shown in the map in Figure 1 below: the home town/self-defined dialect of the participants are marked with red dots. Tromsø and Oslo are marked with blue dots, indicating both fieldwork locations and dialectal background. Although we cover several dialects, we still lack data from most of the Norwegian dialects. Dialect speakers from Trøndelag and Western Norway are currently few in the database; we return to this in the discussion in Section 6. The participants from the high schools were all between 16 and 18 years old, but the age range for [5] In November 2022, 19 additional recordings were made of students at a high school in Fosen (Ørland municipality). The results from Fosen are not presented in the current article.

ARGUMENT PLACEMENT IN NORWEGIAN
[45] the participants recorded at UiT was larger (approximately 20-50 years).
figure 1: Overview of the NWD fieldwork locations (blue dots) and the birth place, home town or county of the participants (red dots) in Norway. For some of the participants, there is only information about which county they grew up in (following the old county division), see the yellow, purple and green areas on the map. 6 [6] The map was created by customizing version 1.0.0 of a map drawing script in R (R core team 2020) from GitHub, archived by Zenodo (see Tengesdal 2022, andGADM 2022).
The experiment was run on a laptop with the open-source software OpenSesame (Mathôt et al. 2012). The recordings were made with different types of sound equipment, but in most cases, handheld digital audio recorders (Zoom H2n/H4/H4n Pro Handy Recorder). A limited number of recordings were made with an external lapel microphone (Audio-Technica ATR3350/AT8532), but most recordings were made using the recorder microphone(s). The recordings were made in WAV-format at 44,1 kHz audio sampling rate, with a bit depth of 16. In some cases, we recorded directly onto our computers. The sound files were subsequently segmented and annotated in ELAN. [

5] results
This section is divided into four subsections, focusing in turn on subject placement with respect to negation and adverbs (5.1), subject placement with respect to objects and particles (5.2), object placement with respect to negation (5.3) and object placement with respect to particles (5.4). Each subsection ends with a short instruction of how to access the results and sound files in the NWD online interface. [

5.1] Placement of subject with respect to negation and adverbs
The variable subject shift, i.e., the placement of a midfield subject with respect to an adverb, was tested in several conditions in part 1 and 2 of the experiment.
In the first part, we tested the linear order of the elements subject and negation/adverb, and in the second part, we tested the order of the elements subject, negation and object, where the object is either a reflexive or a first person pronoun. In total, we recorded and annotated 1 354 elicited sentences that contain both a subject and an adverb in the midfield. As mentioned above, the relevant midfield adverb in the test is the negation ikke. Other adverbs were only tested in the Tromsø high school data collection. We will first discuss the ordering of subject and negation, and thereafter discuss the other adverbs. From part 1, we have in total 735 observations. 303 of these have pronominal subjects: an example of background and target pair is given in (17) In the material, all the 303 elicited sentences have the order Subject-Negation; there is not a single instance of a pronominal subject following negation.
The placement of pronominal subjects with respect to negation is also tested in part 2 of the experiment, where the subject and negation in addition have to be linearized with respect to a light pronominal direct object. The placement of the object will be discussed in Section 5.2 and 5.3; in this section we focus on the placement of the subject with respect to negation. In total, 309 examples of this kind were elicited, and not a single one contains a pronominal subject preceding negation. There were 18 production errors in total; they involved dropping of negation or object.
As for the items with NP subjects, we will treat the ones with negations separately from the sentences with other adverbs. From part 1 of the experiment, we have 306 elicited sentences that directly test SubjectNP-Negation order. Of these, 276 (90%) have the order Negation-SubjectNP, 28 (9%) have the opposite order and two sentences contained a production error or irrelevant response (NP was substituted with a pronoun, or negation left out). The dominant Negation-Subject order is exemplified in (18b), and the corresponding background sentence is provided in (18a).
(Background) teacher.DEF took not bus.DEF to job yesterday 'The teacher did not take the bus to work yesterday.' b. I går tok ikke laereren bussen til jobb. (Target) yesterday took not teacher.DEF bus.DEF to job 'Yesterday, the teacher did not take the bus to work.' We find no clear pattern with respect to interspeaker variation for the Subject-Negation order. 17 out of the 63 participants produce the order Subject-Negation at least once. There seems to be a slightly higher likelihood of finding this word order in Oslo (8/20 participants, 40%) than in Northern Norway (5/23, 22%, of the high school students, 4/20, 20%, of the UiT participants).
The placement of NP subjects was also tested in part 2, in two different versions: one with sentences that contained both negation and a light object (20), and one that only contained negation. The second version was only tested at the We elicited in total 42 sentences of this kind, and 15 of them (36%) have the order Subject-Negation (items no. 1251 and 1252). Note the two main alternatives in the target sentences were identical to the alternatives in two items in part 1 of the experiment, although the background sentences were different. However, the results from the two parts differ strikingly for the same two target sentences in part 1 (items no. 1122 and 1124), the same participants did not produce a single instance of this word order. We discuss this finding in Section 6.
All participants were also tested for subject shift in the context of a light reflexive object. The background sentence and a plausible target is given in (20).
(20) a. Advokaten kommer ikke til å barbere seg (Background) lawyer.DEF comes not INF shave REFL med barberhøvel. with razor 'The lawyer will not shave (himself) with a razor.' b. I går barberte ikke advokaten seg (Target) yesterday shaved not lawyer.DEF REFL med barberhøvel. with razor 'Yesterday, the lawyer did not shave (himself) with a razor.' We have elicited 310 sentences in this condition, out of which 31 sentences contain some error, usually replacement of the NP subject by a pronoun or dropping of negation. Of the 279 produced sentences without task errors, we find 170 (61%) instances of the Negation-Subject order, and 109 (39%) of the Subject-Negation order. This word order distribution is clearly different from the distribution we found in part 1, where only 10% of the elicited sentences had

ARGUMENT PLACEMENT IN NORWEGIAN
[49] Subject-Negation order, but similar to the two sentences discussed illustrated in (19). Again, we will discuss this discrepancy in Section 6.
As in part 1 of the experiment, there seems to be some differences in the results from part 2 in the three recording locations. In the Tromsø high school, 29% of the sentences had the word order Subject-Negation, while the corresponding number in Oslo is 41%. In the mixed UiT group, we found 48% Subject-Negation order. A closer investigation of possible dialect patterns within the mixed group is outside the scope of this study.
There are also some apparent item effects in part 2, seen most notably in the difference between item 1217 (21a), for which the subject (and the reflexive) precedes negation in 55% of the elicited sentences, and 1220 (21b), which only has 27% Subject-Negation order (note the intervening reflexive object in both cases).
(21) a. I går følte studenten seg ikke trøtt (1217) yesterday felt student.DEF REFL not tired etter skolen. after scool.DEF 'Yesterday, the student did not feel tired after school.' b. I går vasket løperen seg ikke (1220) yesterday washed runner.DEF REFL not etter løpet. after race.DEF 'Yesterday, the runner did not wash (himself) after the race.' It is not clear what triggers this difference: either it is the predicate -the highly lexicalized verb-reflexive combination føle seg ('feel') compared to the transparent vaske seg ('wash oneself'), or it is the status of the element following the predicate -a selected predicative adjective compared to a non-selected temporal adverbial. Another possible explanation is the stress pattern of the subject NP: studenten has stress on the second syllable, while løperen has stress on the first. More research, and more targeted data collection, is required for a better understanding of this variation.
To summarize the results regarding the placement of NP subjects with respect to negation, we find considerable variation in our data. Overall, the order Negation-SubjectNP is by far the most common: considering all subconditions discussed above together, we find 75% Negation-Subject order. Elicitation method, i.e., the form of the background sentence, appears to be the factor that most influence word order choice. We also see effects of location: there is slightly more Subject-Negation order in Oslo compared to Tromsø. There is also an item effect. Altogether, a majority of the participants produce at least one shifted NP subject throughout the course of the experiment, but 15 out of 63 participants consistently place all NP subjects after negation. Of course, if the experiment had included more items, these participants might have produced shifted orders as well. We show the variation between participants in the histogram in Figure 2. As we can see, there are not two groups of speakers with different grammars, for instance, shifters vs. non-shifters. No one produces only shifted NP-subjects (max number of items = 11), and most participants shift only in a very small number of items (median = 2, mean = 2.5).
figure 2: Histogram of interindividual variation in NP-subject shift.
In the Tromsø high school data collection, the experiment included four items targeting NP subject placement with respect to the adverbs alltid 'always' and ofte 'often', exemplified in (22)  In total, we have 84 elicited sentences in this condition, of which two contain a task error. Of the remaining produced sentences, only 11 (13%) have the adverb placed before the subject. That is, for adverbs like alltid ('always') and ofte ('often'), the most common placement with respect to the subject is the opposite of the order with negation.
In sum, we find (i) completely categorical shifting of pronominal subjects around negation, (ii) variable shifting of NP subjects across negation, with a rather strong preference for non-shifting but with a clear effect of elicitation method, and (iii) a strong preference for shifting NP-subjects around the midfield adverbs alltid and ofte. The exact numbers of the produced orders are given in Table 2   The interested reader can further investigate and listen to the sets of sentences that target subject placement with respect to negation and other adverbs in the NWD online interface, by selecting 'Subject -Adverb' in the Pairs dropdown menu, and furthermore either 'PRO' or 'NP' in the TypeElement1 menu, and 'Neg' or 'Adv' in the TypeElement2 menu (alternatively, select 'SS' in the Exact Category menu, and use the other menus for further subsetting the dataset). See also Section 2 and figure 2 in Lundquist et al. (2019) for a more in-depth description of the search interface. [52] LUNDQUIST & TENGESDAL NALS Journal, Vol. 7, 1

[5.2] The placement of subjects with respect to objects and particles (long object shift and long particle shift)
In the previous section, we saw that a post-verbal subject does not necessarily directly follow the verb, as adverbs may intervene between the verb and the subject. Here, we focus on two other elements that in principle could intervene between the finite verb and an inverted subject: light object pronouns and verb particles.
The placement of a light pronominal object with respect to a midfield subject is tested in part 1 and part 2 of the experiment. In the first part, reflexive objects (seg) are tested against both pronominal and NP subjects (23), and first person objects are tested against NP subjects (24) We have elicited in total 300 sentences with a reflexive object and an NP subject, 305 sentences with a reflexive object and a pronominal subject, and 309 sentences with a first person object pronoun. Out of these, there were 7 production errors (substitution of NP with pronoun, or dropping of object pronoun). We find only one case of an object preceding a subject (reflexive object, NP subject), but this contains a long hesitation by the speaker. We are quite sure that this can be classified as a production error.
In the second part of the experiment, the linear order of subject and object is also tested, but now in the presence of a negation. Reflexive objects are tested against NP subjects (see, e.g., examples (20-21) above), and first person objects ARGUMENT PLACEMENT IN NORWEGIAN [53] are tested against pronominal subjects. For the reflexive objects, we have 310 elicited examples, of which 31 contain some production error (pronoun instead of NP, dropping of negation). We do not find a single attestation of a reflexive pronoun preceding a subject. For the first person objects, we find one example of a shifted object out of 309 elicited sentences, and this sentence also contains considerable hesitation and self-correction.
In total, out of 1 525 elicited sentences, there are only two instances of objects shifted over the subject, and these contain hesitations. We feel quite confident in concluding that shifting of objects over subjects is not a part of the Norwegian grammar. This is in line with previous descriptions of Norwegian (see, e.g., Faarlund 2019, p. 202).
The placement of subjects with respect to particles is tested in part 1 and 2 in a similar fashion as with subjects and adverbs. We give examples from the two parts in (25-26) below. The subjects were either pronominal or phrasal. In two cases, the NP subjects were indefinite, and they were combined with unaccusative verbs (dette ned 'fall down', brenne ned 'burn down') in order to increase the chances of post-particle subjects, as subjects of unaccusative verbs are more likely to show object-like syntactic properties (see Perlmutter 1978 We have elicited in total 1 205 sentences that test for the order of subjects with respect to particles, 716 with NP subjects and 489 with pronominal subjects. Of these, we find two examples with a particle preceding the subject. Both are produced by the same speaker, and involve the two examples given in (25-26), both with NP subjects. There are in total 43 production errors, mainly in items with NP subjects. Of certain interest are five cases where an expletive subject has been inserted, and the original subject (or "pivot") surfaces after the particle, as in (27) Still, the number of expletive insertions is relatively small, and participants seem to have no problems separating the verb from the particle, even when the subject is a complex NP. Despite the two exceptions, we conclude that long particle shift is not a part of the grammars of the participants tested in this study. We summarize the data for long object shift and long particle shift in Table 3.  The sound clips can be accessed in the NWD interface by choosing 'Subject -Object' or 'Subject -Particle' in the pairs menu, optionally followed by a specification of the type of subject in the TypeElement1 menu.

[5.3] Object shift
Regular object shift across a sentence adverb is tested in part 2 and part 3 of the experiment. Four types of direct objects are tested: simple reflexive pronouns (seg), first person singular pronouns (meg), third person singular or plural pronouns (ham, henne and de), and definite NPs. In this section, we present the results for the different types in turn, starting with reflexive objects. Reflexive objects are tested in combination with negation and NP subjects in part 2, as already discussed in Section 5.1 (subject shift) and Section 5.2 (long objects shift) above. We repeat an example of background and target sentences in this condition in (28) In (28b), the object follows both the negation and the subject. This cannot be counted as absence of object shift, since object shift is dependent on subject shift, which has not applied here; in other words, (28b) is not a context where object shift could apply (see also Larsson & Lundquist 2022a). We find three different word order patterns in this condition in the Norwegian data: 170 instances of the order Negation-Subject-Reflexive, 106 instances of Subject-Reflexive-Negation, and three instances of the order Subject-Negation-Reflexive. This last word order has subject shift, but not object shift. Two of the three examples are produced by the same speaker, and all three examples seem prosodically wellformed. In short, we find 106 instances of object shift, three instances of absence of objects shift, and 170 produced sentences where the structural configuration does not allow object shift (since the non-shifted subject blocks object shift).
First person pronominal objects were tested in part 2 of the experiment, but together with pronominal subjects. As was reported in Section 5.1, pronominal subjects shift around negation without exception, and therefore never block object shift. In the results, there are 267 (92.4%) elicited sentences with the shifted order Subject-Object-Negation, and 22 (7.6%) sentences with the unshifted order Subject-Negation-Object. In addition, the results contained 19 production errors, and one instance of the order Object-Subject-Negation; see Section 5.4.
Third person object pronouns were tested in the third part of the experiment, where the elicitation method consisted of a transformation from a passive sentence to an active sentence, as illustrated in (29) We have elicited 456 sentences in this condition, out of which seven involve a production error. 428 (95.3%) sentences have the expected shifted word order Object-Negation, and 21 (4.7%) sentences have an unshifted word order.
figure 3: Histogram of total number of unshifted objects in contexts with expected shifted objects.
In sum, the number of unshifted pronominal objects is quite small. In total, there are 802 shifted pronominal objects compared to 46 unshifted (5.4%). Distinguishing different pronouns, there are three (2.8%) unshifted and 106 shifted reflexives in our data set, 22 (7.4%) unshifted and 267 shifted first person objects, and 21 (4.7%) unshifted compared to 428 shifted third person pronominal objects. In other words, first person pronouns are more likely to stay in an unshifted position than the other types. There seems to be some effect of recording location: unshifted objects make up less than 2% in the Oslo material ( effect of participant, i.e., there is not a small set of participants that are responsible for all the unshifted objects; rather, many participants occasionally deviate from the typical object shift pattern, as shown in the histogram in Figure  3. No one produces more than 4 unshifted objects out of up to 17 possible contexts for (non-)object shift. Still, the majority of the participants (n=36) do not produce a single unshifted pronominal object. Again, the interested reader is encouraged to listen to the sound files in NWD, where it is possible to investigate the stress patterns of shifted and unshifted object pronouns. NP object shift was tested in part 3 of the experiment, in the same type of context as pronominal object shift. An example is given in (30). The database contains 247 elicited target sentences of this type, of which seven have a production error (usually substitution of NP with a pronoun ). 236 sentences have the expected word order Negation-Object, and four have the opposite order (i.e., they are instances of NP object shift). Two of these involve considerable hesitation, and should probably be considered production errors. The other two examples are pronounced with a natural intonation pattern.

5.4] Particle placement
In the experimental design, we separated five types of verb-particle constructions, and these were tested both with pronominal and NP objects. We give examples of the five types in (31)  The last two categories were only tested with one item each, and since the results from these categories are indistinguishable from the metaphorical/nondirectional particles, we collapse these three categories into one, which we will refer to as non-directional particles below. In total, there are 1 525 elicited sentences that target the order of verb particles and objects (particle shift). 806 of these contain an NP object and 719 have a pronominal object. There are in total 66 unexpected responses in the data, which we treat as production errors. The errors usually occur in sentences with the inanimate pronoun det, which participants often misanalyse as an expletive subject (e.g., Det ble bygget om … 'It was rebuilt…'). We find clear effects of both type of argument (pronoun vs. NP) and type of verb particle (directional + PP, directional, and non-directional). The general trends are shown in Figure 4, where we have excluded the production errors.

ARGUMENT PLACEMENT IN NORWEGIAN
[59] figure 4: Proportions of particle placement w.r.t. particle and object types.
The results show that directional particles are more likely to follow NP objects (50.8%) compared to non-directional particles (11.5%), and among the directional particles, the ones with following PPs are most likely to follow the object (68.3% compared to 33.6%). However, there is no difference between the three classes of particles when the object is a pronoun: pronouns generally precede particles. In total, there were only 4 cases of pronominal objects following a particle. Three of these come from the same item (skjelle ut henne 'scold her'). The raw numbers for the particle shift data are given in Table 5.  We encourage the readers to do a closer investigation of possible effects of dialects and items in the NWD database. The particle shift data can be found by either choosing 'Object --Particle' in the pairs dropdown menu, and then select 'PRO' or 'NP' in the TypeElement1 menu, or alternatively by choosing one or several of the different particle types in the ExactCategory menu.
[5.5] Summary The experimental results that are presented in this paper reveal both categorical word order restrictions and gradient variation. We have confirmed the expected strict precedence relation between subjects and objects, and subject and particles. Furthermore, we found no variation in the placement of pronominal subjects; all pronominal subjects preceded negation, objects, and particles. NP subjects, on the other hand, showed a highly variable placement pattern with respect to negation, but this variation was to some extent modulated by the elicitation method, as we will discuss further below. With regard to objects, the majority of pronominal objects shifted over negation (OS), while the number of shifted phrasal objects is so small that they may be considered production errors. Lastly, regarding particles, we found that non-directional particles more often shifted over phrasal objects compared to directional particles. For the directional particles, the presence of a PP further increased the likelihood of the order Object-Particle. For the pronominal objects, on the other hand, there was no effect of particle type. Here, the order Object-Particle was near-categorical. As expected, particles did not shift over pronominal or phrasal subjects (long particle shift), with the exception of two instances of a particle shifting over a phrasal subject, which most likely should be treated as production errors.
[6] discussion and conclusion The goal of this paper was to give an overview of the Norwegian data on argument placement in NWD, and to further place the results in a comparative frame with the other North Germanic languages. In this discussion, we briefly compare the results with those from the other North Germanic languages, as covered in the papers in the NALS special issue on Nordic Word Order Database. In addition, we discuss different factors that influence word order choice in syntactic contexts where variation is found. As for the non-variable patterns, the results from the Norwegian data collection is similar to the patterns found in the other Mainland North Germanic languages. The following three categorical patterns are found in Swedish and Danish as well: (i) subject pronouns always precede negation and other adverbs, (ii) all subjects, independent of form, precede verb particles, and (iii) NP objects ARGUMENT PLACEMENT IN NORWEGIAN [61] follow negation. The only potentially surprising finding here is (i): the strict order of subject pronouns with respect to adverbs. As mentioned in the background section, Østbø Munch (2013) found quite a high proportion Negation-SubjectPRO order in her corpus material. The reason we do not find this pattern in our data might be due to the dialects covered by our participants: clitic negation is mainly found in Western Norwegian (e.g., Nordfjord, Sunnmøre). In other places, Negation-SubjectPRO is more restricted to contrastively focused subjects, and in our elicitation paradigm, all pronouns were non-contrastive. Furthermore, we did not elicit any sentences where the finite verb was an auxiliary or a copula verb, which also reduces chances for cliticization of negation.
The fourth categorical pattern is (iv) the placement of subjects with respect to objects (long object shift). Here, we find that the Norwegian-speaking participants consistently place subjects before objects, independent of form of the both subject and object. These results look identical to the results found in the Danish (Larsson & Tengesdal 2022), Faroese (Lundquist 2020) and Icelandic (Larsson 2022) data. It is only Swedish that displays variation with respect to long object shift (see Larsson & Lundquist 2022a).
The last (near-)categorical pattern concerns the order of (v) pronominal objects with respect to verb particles. Here again, Norwegian shows the same pattern as Danish, Faroese and Icelandic, with object pronouns preceding particles, and again, Swedish is the odd one out with a categorical Particle-Object order. The Norwegian data contain four exceptions to the standard ObjectPRO-Particle order, and two of these are produced with some hesitation. It is not clear if these four attestations should be regarded as production errors, or if there is genuine variation in Norwegian. As pointed out in the background section, the Particle-ObjectPRO order is found in parts of Trøndelag, and also in dialects spoken in areas close to the Swedish border in Eastern Norway (see Tengesdal & Lundquist 2021 and references therein). 7 We find variation in the word order in three of the investigated phenomena: NP subject shift, particle shift in contexts of NP objects, and (pronominal) object shift. Out of these, we find the least variation for object shift; close to 95% of the pronominal objects shift. Similarly to the other North Germanic languages, we find a near categorical (97.2%) shifting of the simple reflexive object seg). The highest proportion of unshifted objects (7.6%) was found with the first person object pronoun meg. The proportion of unshifted meg is still relatively small, but it is still higher than in the results from the Danish, Faroese and Icelandic data [7] We can confirm from preliminary analysis of the Fosen NWD data collected in November 2022 that we find verb-particles preceding pronominal objects. In addition, we also find several cases of NP subjects following particles (long particle shift) in this material.
in NWD, where unshifted first person pronominal objects are virtually absent (0-1% in these three languages). In both Danish and Faroese, third person pronominal objects are more likely to remain in an unshifted position (around 10%), while unshifted third person objects are rare in the Norwegian data (4.7%, 21 unshifted objects, 428 shifted objects). In the Swedish data, third person pronouns are the least likely to shift (as much as 40% unshifted) compared to 20% unshifted first person objects, and 3% unshifted reflexives (see Larsson & Lundquist 2022a). In other words, Norwegian is the odd one out in having a higher proportion of unshifted first person compared to third person object pronouns. We do not know why this is so, but it suggests that object shift is governed by slightly different factors in the five North Germanic languages (see also Bentzen et al. 2013 for differences in corpus frequencies with respect to object shift in the Mainland North Germanic languages). It should be mentioned that the material used for elicitation for the five different languges is near-identical, with only small changes to accommodate lexical differences in subcategorization patterns.
As far as we can tell, the differences in object shift patterns cannot be explained by differences in elicitation material. Despite some variation in object shift, pronominal objects adhere to formbased generalizations stated in the introduction of the article: If an argument (a subject or an object) is realized as a pronoun, and is in the same syntactic domain (VP or TP/midfield) as an adverb or particle, the argument will in the unmarked case precede the adverb or particle. In the results from the current study, the only exception to this is object shift, where we find a small number of weak pronominal objects that follow negation. The pattern for NP arguments is less strict; subjects can precede or follow adverbs, and objects can precede or follow verb particles. The variation we find is not straightforwardly explained in terms of information structure, e.g., whether they are new or given in the discourse (see further below).
For placement of NP objects with respect to particles, we find an effect of the type of particle. When the particle and the verb make up one semantic unit, e.g., skjelle ut ('scold') and sjekke opp ('make a move on'/'hit on'), an NP object rarely precedes the particle, in contrast to cases where the verb and the particle both make a transparent semantic contribution (e.g., kaste ut, 'throw out'). This pattern is not related to the givenness of the object, but seems rather to be an effect of a preference to have elements that form a tight potentially nontransparant semantic unit next to each other (here, the verb and the particle). Note, however, that object pronouns precede particles even when verbs and particles form a close semantic unit. This may suggest that the preference is rather to realize the verb and its particle within the same prosodic unit, rather ARGUMENT PLACEMENT IN NORWEGIAN [63] than as linearly adjacent. Still, it is important to remember here that inverted phrasal subjects always intervene between a main verb in V2 position and a particle, thus breaking up both the linear adjacency and the prosodic unity, e.g., I går skjelte laereren ut eleven (lit. Yesterday yelled the teacher out the student, 'Yesterday, the teacher scolded the student'). Thus, a factor like semantic unity could only affect the linear order when the argument and the particle/adverb are inside the same syntactic domain (VP or TP/midfield).
For particles, the presence of a PP that expresses the ground of the particle has an effect on linear order, e.g., ut av puben ('out of the pub'), compared to ut ('out'). It is not clear if the particle has a different syntactic status when followed by a PP, e.g., if the particle in this case is a modifier of the PP rather than a modifier or complement of the V/VP (see Larsson & Lundquist 2022b for discussion). The presence of a PP does not have an effect on word order in the other languages in NWD, but that might be due to the fact that Swedish, Danish and Faroese have no variation in particle placement (Larsson & Lundquist 2022a, Larsson & Tengesdal 2022, Lundquist 2020, and that the Object-Particle order is more general in Icelandic compared to Norwegian (see Larsson 2022).
For the placement of phrasal subjects with respect to negation and other adverbs, we find that factors other than information structure have strong effects on linear order. First, the choice of adverb has a large effect: negation typically precede NP subjects, while the other adverbs typically follow NP subjects. It is not clear if this pattern can be explained by making reference to the clausal position of the adverb, if it is a frequency/collocation effect (negation being the most frequent adverb), or if it is just an idiosyncratic pattern for negation. Importantly, though, variation in placement is found both with negation and the other adverbs, suggesting that there are no fixed templatic positions for the adverbs with respect to subjects. Second, we find a strong effect of elicitation method. In the second part of the experiment, we elicited a much higher proportion of Subject-Negation order than in the first part, even in the cases where the target sentence was potentially the same; cf. the background examples in (31a-b) below, with (31c) as the intended target.  Considering information structure, it is unclear why a reader/speaker would infer a different information structure role of the subject in (31a) compared to (31b); changing the tense should not affect information structure. The Swedish NWD results show the same effect of elicitation method (Larsson & Lundquist 2022a). A possible explanation is that this is an effect of syntactic or prosodic priming. In the first part of the experiment (31b), the participant reads the main verb directly preceding the negation in the background sentence. It is likely that the reader temporarily stores the verb-negation sequence either as a prosodic word, or as an ordered/linearized pair, that can be re-used in the target sentence. In the second part of the experiment, the main verb and the negation are not linearly adjacent in the background sentence, and the participant thus cannot re-use the verb-negation sequence in the target sentence. Still, a more detailed prosodic study of the relation between verb and negation in the variable cases is needed in order to better understand the effect of priming (syntactic or prosodic) and information structure.
The effect of information structure is also limited or completely absent in other cases, since there is no word order flexibility to begin with. For example, there is a categorical lack of NP object shift in Norwegian, even in cases when the NP object is given, in contrast to Icelandic, and West Germanic languages like German and Dutch, where scrambling could be more straightforwardly linked to information structure (see, e.g., Hinterhölzl 2012; Andréasson 2013, but also Struckmeier 2017 for problems with information structure-driven approaches to scrambling). The strict midfield ordering of objects with respect to subjects, i.e., the absence of long object shift, also restricts potential word order shifts: a noncontrastive object pronoun will end up after sentence adverbs as long as the subject surfaces after the sentence adverb. In short, the effect of information structure on the ordering of midfield or VP elements is highly limited (and possibly non-existent).
The effects of syntactic function (subject vs. object vs. adverb) is on the other hand highly salient, and so is the effect of form, i.e., light pronouns contra NPs. Still, the variation that we find within and across speakers in this tightly controlled production experiment suggests that word order is not fully determined by either syntax or information structure. The role of prosody has not been investigated in this article, but the database enables prosodic analysis directly linked to syntactic variation, and it is likely that it will carried out in the near future. As we mentioned in Section 4, at present we only have a few speakers from Trøndelag and Western Norway in NWD. This means that we cannot rule out more variation, especially in certain areas of the country (see Footnote 5). The database is open to be expanded, and new material from Trøndelag and Tromsø will be added in 2023.