exical interference in n o n-native resolution of possessives ?

Research on non-native pronoun resolution has predominantly been concerned with (i) ‘ordinary’ 3 person pronouns/anaphors like En. he, she, they or himself, herself, themselves, (ii) language pairs involving English as the native (L1) or the foreign (L2) language, and (iii) the role that binding constraints and syntactic structure in general play in L2 versus L1 processing. The present paper – a follow-up study to Pitz et al. (2017) – deviates from this trend in all three respects: We investigate how L1-Norwegian learners of L2-German interpret the two German possessive pronouns/determiners sein (≈ his) and ihr (≈ her or their), arguing that lexical divergence between the possessive systems, and in particular the formal similarity between binding-neutral L2-German sein and the L1-Norwegian reflexive possessive sin, may enhance or interfere with L2 comprehension, depending on the structural conditions. In Section 2 we briefly present the two possessive systems. Section 3 summarizes relevant research on pronoun resolution, with a special view to possessives. Sections 4–6 present a pilot study on L1-Norwegian learners’ grammaticality judgments of sein and ihr in simple sentences (Sect. 5) and a forced-choice resolution experiment involving a group of L1-Norwegian learners with a background two or three years’ teaching of L2-German at high-school level and a control group of native speakers of German (Sect. 6). The final Section 7 provides a summary and concluding discussion of our findings.

In Section 2 we briefly present the two possessive systems. Section 3 summarizes relevant research on pronoun resolution, with a special view to possessives. Sections 4-6 present a pilot study on L1-Norwegian learners' grammaticality judgments of sein and ihr in simple sentences (Sect. 5) and a forced-choice resolution experiment involving a group of L1-Norwegian learners with a background two or three years' teaching of L2-German at high-school level and a control group of native speakers of German (Sect. 6). The final Section 7 provides a summary and concluding discussion of our findings.
[1] introduc tion Possessives -e.g. determiner-like pronominal items like English his, her, its, their, or French son, leur and German sein, ihr -vary in interesting ways across even quite closely related languages (Gunkel et al. 2017: 672-719). They consequently pose different kinds of challenges to L2 acquisition, depending on the [26] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 specific L1/L2 pair (see Sect. 3.2 below). 1 The present paper -a follow-up study to Pitz et al. (2017) -is concerned with the pair L1-Norwegian/L2-German, focusing on L1-Norwegian learners' comprehension of the German 3 rd person possessives sein* and ihr*, each of which may correspond to either the reflexive possessive si* or an irreflexive possessive in Norwegian. 2 In addition to being neutral with respect to reflexivity like sein*, ihr* is ambiguous between (possessor) singular fem. and plural. Our study aims at testing (i) in how far the similarity between sein* and si* may enhance or disturb L1-Norwegian learners' interpretation of L2-German sein* under conditions where a reflexive reading from a grammatical point of view is either demanded or excluded, and (ii) how they resolve ihr* under similar conditions, given the fact that a summative plural interpretation (in the sense of Kamp & Reyle 1993: 341-342) is also possible.
The paper is organized as follows: In Section 2 we give a brief contrastive presentation of the systems of pronominal possessives in the two languages. Section 3 summarizes relevant empirical research on pronoun and anaphor resolution, with a special view to non-native resolution and differences between possessives versus 'ordinary' pronouns. Section 4 presents the objectives and general assumption underlying two experiments that aim at testing how Norwegian 1 st year students of L2-German master the abovementioned one-tomany correspondence between possessive items in L2 grammaticality judgment (Sect. 5) and offline resolution (Sect. 6). Finally, in Section 7, we summarize the results and present some ideas for further research.
[2] c ontrastive bac kground The distribution of the German 3 rd person possessives sein* and ihr* is determined by grammatical properties of their antecedent (possessor) DP: sein* demands a singular masculine or neuter possessor DP, otherwise -i.e. with singular feminine and with plural antecedents -ihr* is used. Some examples are given in (1)-(3).
(1) a. [Peter]  hat sein-e Mütze verloren. 'The child has lost its cap.' [1] In line with a widespread practice, we use the term L2 for a language that a person starts learning after childhood and after having acquired the core properties of his/her native language(s). For native speakers of Norwegian, German will normally be at least the second foreign language (L3), following after English.
The star is meant to indicate that the lexical items in question are inflected in context. That is, sein*, ihr* and si* each represent a set of inflected forms (morphological words). The examples illustrate the typical adnominal function of German possessives. In this use, possessives act like determiners: They are confined to the initialfunctional head -position of their host DP, they inflect like determiners, agreeing with their head (possessee) noun with respect to number, gender and case; and they make their host DP semantically definite -like, e.g. English possessives (Gunkel et al. 2017: 672-719, Fabricius-Hansen et al. 2017. In Norwegian (bokmål), adnominal possessives have two positional options: They may occur DP-initially, as in German and English, making the DP semantically definite despite the indefinite form of the head noun; mostly, though, they are post-nominal, triggering overt definiteness marking of the preceding head noun (phrase).
Leaving differences of the inflectional systems aside, it is with respect to the nature of the possessive items and the conditions determining their use that the Norwegian and the German system of possessives differ most conspicuously (Fabricius-Hansen et al. 2017). Thus, Norwegian has a specific reflexive possessive si* which, roughly speaking, must be used when the possessive is locally bound, i.e. c-commanded by an antecedent inside the local clause (typically the clause subject). 3 Like the 3 rd person reflexive pronoun seg (German sich; 'his-/ herself, themselves'), si* is neutral with respect to semantic/ grammatical number and gender of its antecedent; and like articles, si* is inflected, agreeing in number and gender with its head noun. Examples are seen in (4)-(6), the Norwegian counterparts of (1)-(3): sin, si, sitt, sine are the singular masc./commune, singular feminine, singular neuter and plural forms of si*, respectively, agreeing in number and (singular) gender with the head noun, which, when preceding the possessive, exhibits corresponding forms of the definite article.
[28] cathrine fabricius-hansen, anneliese pitz, henrik torgersen With non-local possessors, Norwegian uses genitive forms (hans 'his', hennes 'her', dens/dets 'its' and deres 'their') of 3 rd person non-reflexive -or more precisely: irreflexive -pronouns as possessives. 4 As with German sein* and ihr*, the choice between these items depends solely on properties of the possessor DP/referent, albeit with respect to more syntactic-semantic dimensions (Fabricius-Hansen et al. 2017: 14-16). Thus, hans and hennes are used with singular human male and female possessors, corresponding to English his and her respectively; dens and dets with singular non-human possessors, corresponding to English its but differing in grammatical gender (commune vs. neuter); and deres demands a plural possessor, like English their. Examples with post-nominal hans, hennes and deres are given in (7); for dens and dets, which tend to be avoided in present day colloquial Norwegian, we refer the reader to Fabricius-Hansen et al. (2017 [29] OSLa volume 12(2) 2021 By contrast, the German possessives sein* and ihr* may be locally bound, as in (1)-(3) above, or not, as in (8) below, i.e. they are neutral with respect to binding conditions (reflexivity); and the choice between them depends solely on the grammatical gender and number of the possessor DP. -Note that in (8b, c), ihr* also has a plural reading ('their'); see below. In suitable contexts, then, the German possessives may be ambiguous between a reflexive (locally bound) and non-reflexive (non-locally bound) interpretation, like English possessives. This is the case with sein in (9) below, which may be bound by the singular masc. subject (Paul) of its own clause (reflexive use, local binding) or refer to the singular masc. DP Peter occurring in the preceding clause (non-reflexive use, non-local 'binding'); 5 that is, the host DP sein Auto 'his car' may refer to either Pauls's or Peter's car.

(9)
Während Peter verreist war, reparierte Paul sein Auto. 'While Peter was away, Paul repaired his car.' Likewise, ihr in (10) may be understood reflexively, referring to Lisa, or nonreflexively, referring to Anna. Moreover, ihr in (10) -and ihren in (8b) -may receive a possessor plural reading ('their'), referring to the set made up by the two different singular referents introduced in the context: Lisa + Anna in (10) and Anna + Peter in (8b). In other words, the car in (10) may belong to Lisa (full short arrow), Anna (full long arrow) or the two together (dashed arrows). Henceforth, we shall use the term summative plural interpretation for cases like these, where the plural possessor is not introduced by a single plural DP, as in (3) and (8c), but 'construed over' two or more separate, non-coordinated DPs occurring in the context (see Kamp & Reyle 1993: 341-342). [5] In the latter case, the possessive in fact is co-referent with rather than bound by its antecedent in the strict, technical sense of the term. [30] cathrine fabricius-hansen, anneliese pitz, henrik torgersen Ambiguities like the ones illustrated by full arrows in (9) and (10) do not arise in Norwegian -at least not in simple clauses like those discussed here: 6 Coreference between the possessive (si*) in the main clause and the subject of the subordinate clause is blocked in (11a-c) (Binding Principle A); conversely, hans 'his', hennes 'her' and deres 'their' in (12a-c) cannot refer to the subject of their own clause (Binding Principle B). Note however that deres in (12c) also allows for a summative plural interpretation, combining the referents of the two (plural) possessor 'candidates' into one (dashed arrows). 'While the children were away, the parents planned their (the [6] In fact, things are not that neat in actual use. So-called long-distance binding is not ruled out, and on the whole, the principles governing the choice between si* and the irreflexive possessives are somewhat blurred, varying across dialects, sociolects and medium (spoken and written language); see e.g. -Hansen et al. (2017: 20-22) and references therein. In addition, even under hard-core conditions, 'errors' are not infrequent, in particular with plural possessors, i.e., deres occurring instead of 'correct' si* or vice versa. children's or the children's and the parents' joint) holiday.'
figure 1: German ! Norwegian lexical divergence in 3 rd person possessives with human possessor As historical cognates, German sein* and Norwegian si* exhibit a morphophonological similarity observed in many other lexical pairs, as for instance the 1 st and 2 nd pers. singular possessives mein*/mi* 'my' and dein*/di* 'your (sing.)'. Lexical divergence, then, makes sein* if not a downright 'false' so an 'unreliable friend' viewed from the perspective of L1-Norwegian users of L2-German: it may or may not be equivalent to its cognate si*. 7 German ihr*, on the other hand, bears no resemblance to any of its three lexical counterparts within the system of Norwegian possessives; in addition to being neutral with respect to binding conditions/reflexivity, however, ihr* exhibits a gender-number ambiguity without parallel in the Norwegian system. Studies conducted by Bie-Lorentzen (2012) and Pitz et al. (2017) have shown that the lack of isomorphism between the two possessive systems, as would be expected, causes problems for not very advanced Norwegian learners of L2-German when translating between the two languages, whether into or from L2. The experiments presented in Sections 4-6 aim at shedding some light on how they cope with the abovementioned divergence and unreliable morphophonological similarity in offline comprehension and grammaticality judgments of L2-German possessives. [7] According to Jarvis (2009: 107) 'false friends' may be defined as 'cross-linguistic word pairs that are (1) formally the same or similar and (2) semantically similar or dissimilar (but not the same)'. [32] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 [3] e mpiric al re se arc h on pronoun re solution

3.2).
A key issue in the psycholinguistic discussion is the status of structuresensitive -binding -constraints as opposed to other possible constraints on pronouns, including feature agreement between pronoun and antecedent. Do binding constraints act as a filter in native comprehension in the sense that salient but structurally inaccessible (inappropriate, illicit) antecedent candidates are immediately discarded as antecedents whether or not they match the pronoun in relevant (gender, number, …) features (see e.g. Nicol & Swinney 1989;Sturt 2003)? Or is a parallel cue-based resolution mechanism more plausible -a processing strategy in which 'multiple weighted constraints (including constraints on binding) simultaneously influence the net activation of a candidate during preselection stages of antecedent evaluation' (Badecker & Straub 2002: 748)? According to Badecker & Straub (2002), Jäger et al. (2015), Patil et al. (2016), among others, evidence on the whole seems to favor models of the latter kind.
Another question concerns resolution preferences when the context offers more than one accessible and feature-matching antecedent candidate for the anaphor, i.e. when the pronoun is referentially ambiguous. Under such conditions, local resolution (binding by Principle A) is assumed to be the preferred [8] Note that we use the terms pronoun and anaphor in a wide sense, covering both reflexive and nonreflexive items. According to the theory of Government and Binding (Chomsky 1981 choice -if structurally warranted, as for instance with sein*/ihr* in (9)/(10) (Fox 1998;Heim & Kratzer 1998;Patterson et al. 2014). In any case, however, the interpretation in the end relies heavily on semantic, pragmatic and discourserelated cues, including the possibility of anaphoric competitors to the pronoun in question (see e.g. Bosch et al. 2007;Roberts et al. 2008;Arnold 2010: Baumann et al. 2014Colonna et al. 2014;Schimke et al. 2015;Bader & Portele 2019;Schulz et al. 2019).
As for non-native pronoun resolution, it is still an open question to what extent learners of a foreign language can acquire the resolution strategies and obtain the efficiency exhibited by native speakers. Studies presented by Clahsen & Felser (2006a), Felser et al. (2009), Felser & Cunnings (2012 and Patterson et al. (2014) indicate that L2-learners, even at an advanced level, more often consider (during processing) or even prefer (in the final interpretation) structurally inaccessible DPs as antecedents than do native comprehenders. Apparently, then, structure-sensitive constraints (Binding Principles A vs. B) do not play the same decisive role in non-native as in native pronoun resolution. Possibly, structural processing is impoverished in L2 comprehension in general. This is suggested in Clahsen & Felser (2006a), Felser & Cunnings (2012) and Clahsen & Felser 2006b). However, as documented e.g. by Gast & Haas (2008) and Patterson et al. (2014) (and as we shall see below), properties of L1 -transfer in a more or less strict sense -may also play a role. Taking stock, Felser (2019) argues that trying to reduce L1/L2 processing differences to a single cause might seem misguided.
What is called for instead is a more careful investigation of how different types of constraint and information sources interact during L2 comprehension, taking into account what linguistic cues need to be extracted from the input or need to be re-accessed in order for a given constraint to be applied. This should provide us with a more nuanced picture of how the relative weighting or timing of constraints or information sources might differ in L2 in comparison to L1 processing. (Felser 2019: 59) It should be added that studies on pronoun resolution predominantly concern cases where the so-called antecedent is indeed an ante-cedent, as in (1)-(15), linearly preceding the pronoun -that is, anaphoric pronouns in yet another sense of that term (cf. fn. 8). Resolution of cataphoric pronouns, which linearly precede their 'antecedent', has been studied less intensively; but see Gompel & Liversedge (2003) references there. Online cataphoric resolution involves waiting or looking forward for -perhaps predicting -a feature-matching DP that semanticpragmatically fits into the pronoun's context. However, in order for coreference to be established between the two, they must belong to the same sentence; and according to Chomsky's (1981) Binding Condition C, the candidate DP may not be structurally bound by the pronoun. Thus Principle C prevents the main clause subject he in (13a) from being resolved to the subordinate clause subject (the lord). On the other hand, as part of a DP, his in (13b) does not c-command the lord, nor can the subordinate clause subject he in (14a,b) bind a DP within the following main clause; consequently, co-reference with the lord is possible in these three cases.
(13) a. He i/*j got depressed when the lord j married the duchess. b. His i/j butler got depressed when the lord j married the duchess. Experiments conducted by Drummer & Felser (2018) indicate that condition C plays the same role in native and non-native (Russian/German bilingual) comprehension but 'does not prevent inaccessible antecedents from being considered during the initial bonding stage', its application apparently being more 'bottom-up driven than previously thought' (Drummer & Felser 2018: 112).

[3.2] 3 rd person possessives
Studies on the resolution of possessive pronouns are surprisingly scarce in view of the fact that possessives represent somewhat different processing challenges from ordinary personal pronouns: 9 First, since possessives are determiners or modifiers in DPs, binding constraints work differently under cataphoric conditions than they do for ordinary pronouns; cf. Drummer & Felser (2018) and Section 3.1 above.
Second, the two pronominal subsystems -ordinary and possessive 3 rd person pronouns -are not always isomorphic, as in Norwegian, but may differ in number of items and the nature or distribution of constraints. This is the case For native resolution see e.g. Kennison (2003)  [35] OSLa volume 12(2) 2021 in German, which has a number-and gender-neutral 3 rd person reflexive pronoun sich 'him-/her-/itself, themselves' in addition to the three number/gender-specified (irreflexive) personal pronouns (er 'he', es 'it', sie 'she/they') on the one hand and on the other hand only two possessive items (sein* and ihr*), which are specified for possessor number/gender but unspecified for reflexivity (Binding Conditions A vs. B); see Sect. 1.2. English is another case in point (Kennison 2003: 337), cf. Peter i knows himself i/*j and Peter i knows him j/*i vs. Peter i knows his i/j wife; and in French (as in other Romance languages), possessives differ from non-possessive anaphors by being underspecified not only for reflexivity but also for antecedent gender (see e.g. Helland 2017; Lago et al. 2018).
Third, in many languages, including German and e.g. French but with English as a prominent exception, adnominal possessives exhibit feature (number, gender, case, …) agreement with the head noun of their host DP (the possessee), in addition to being specified for antecedent features. Beside searching for an antecedent (the possessor), then, comprehenders must 'keep an eye' on the possessee noun (phrase) too, checking for the relevant features; as pointed out by Lago et al. (2018), however, possessee agreement is probably a bigger problem in production than in comprehension. 10 As far as challenges to L2 acquisition are concerned, then, the grammar of adnominal possessives opens up for more variation across L1/L2 pairs than ordinary pronouns -including the relative order of possessive and possessee (cf. Sect. 1.2 above). And (the few) empirical/experimental studies on L2 production and/or comprehension of possessives indicate that even quite advanced L2 learners' performance may be influenced by their L1 possessive system in ways and to a degree that, at least in part, reflect cross-linguistic (dis)similarities in the different dimensions of possessive variation. Santesteban et al. (2010), Foucart et al. (2011) and Antón-Méndez (2011), for instance, found that L1 speakers of a Romance language, where possessives encode gender agreement with the possessee alone, quite often use L2-English his and her in agreement with the possessee rather than the intended antecedent (e.g. producing Peter loves her sister when meaning 'Peter i loves his i sister'), while native speakers of Greek or Dutch master the distribution of his and her like L1-English speakers, in accordance with the fact that possessives in their L1 encode gender agreement with the possessor, like English. 11 cathrine fabricius-hansen, anneliese pitz, henrik torgersen

OSLa volume 12(2) 2021
As for non-native comprehension of possessives, Lago et al. (2018) investigated whether multilingual speakers with either L1-English/L2-Spanish or L1-Spanish/L2-English differ in their sensitivity to possessor agreement violations in German as an 'additional' non-native language ('La'), depending on whether possessor gender agreement is present in their L1 (English) or not (Spanish). Using context-free German sentences like (15a-d) in a speeded acceptability judgment and an implicit reading experiment, they found that (i) the L1-Spanish participants, as expected, showed less sensitivity than the L1-English group towards possessor-possessive mismatch as displayed in (15c-d); 12 (ii) high L2 proficiency may have helped L1-Spanish/L2-English in La-German but did not influence the performance of L1-English/L2-Spanish participants; (iii) contrary to their expectations, gender mismatch between possessive and possessee (Mutter 'mother'), as displayed in (15b,c) did not affect the L1-Spanish learners' performance in any significant way. 13 (15) a. Frau Schmidt küsste ihre Mutter bei dem letzten Familientreffen.
'Ms. Schmidt kissed her mother at the last family reunion.' b. Herr Schmidt küsste seine Mutter bei dem … Familientreffen.
'Ms. Schmidt … her mother … .' Lago et al. (2018) conclude that both the L1 grammar and the grammar of a L2 may influence multilinguals' comprehension of possessives in La and that when an agreement constraint is absent in L1, comprehenders have more difficulty deploying it in La.
The observation that L2 learners have difficulties deploying L2 constraints on possessives that are not present in their L1 is corroborated by Helland (2017), Pitz et al. (2017) and Saad et al. (2019) with respect to the structuresensitive binding constraints (Condition A vs. B) that characterize reflexive vs. irreflexive possessive markers.
sessive-possessee) agreement. But this hypothesis cannot account for the abovementioned difference between L1-Romance and L1-Dutch or Greek learners' performance; cf. Lago et al. (2018).
[12] Note that (15c) and (15d)  [37] OSLa volume 12(2) 2021 Saad et al. (2019) investigated how well multilingual speakers whose first (majority) language (Austronesian Alor Malay) has binding-neutral 3 rd person possessive marking master the distinction between reflexive and irreflexive 3 rd person possessive marking in their second (minority) language (Indonesian Abui). They found that (pre-)adolescents did not fully master the distinction, whether in production or comprehension, and that the tendency was to overgeneralize the irreflexive prefix to reflexive contexts rather than the other way around, arguably because the former, being less restricted, is considerably more frequent in natural discourse. 14 Notably, possessive marking in L1-Malay bears no morpho-phonological resemblance to either the reflexive or the irreflexive possessive marking in L2-Abui.
Turning to Norwegian versus German (Bie-Lorentzen 2012; Pitz et al. 2017) or French , things become more complicated since (i) possessive systems in Norwegian and German or French differ with respect to featureagreement constraints in addition to the structure-sensitive binding constraint, and (ii) in particular because the Norwegian reflexive s-possessive (sin/si/sitt/sine) is an 'unreliable friend' of its binding-neutral s-cognates in German (sein*; see Sect. 1.2) and French (son/sa/ses; Helland 2017). Consequently, incorrectly assigning a non-local antecedent to L2 si*, whether in production or comprehension, may be due to more or less shallow (negative) transferpriming -from L1-German sein* (or L1-French s*) rather than insufficient mastering of the binding constraints on the L2-Norwegian possessives. 15 Likewise, when L1-Norwegian learners of L2-German understand German sein* as locally bound (reflexive) under conditions that grammatically preclude such an interpretation, or erroneously use sein* instead of ihr* under local binding, this may be due to interference from si* rather than insufficient mastering of agreement constraints or nominal gender in German.
The results of translation tests (L1!L2 and L2!L1) conducted by Bie-Lorentzen (2012) and Pitz et al. (2017) indicate that the cross-linguistic divergence of sein* into si* and hans, hennes, deres and the divergence of si* into sein* and ihr*, as expected, constitute an obstacle for L1-German learners of L2-Norwegian and L1-Norwegian learners of L2-German, respectively, and that '[i]n both cases, there seemed to be a tendency to translate on the basis of the L1 system although morpho-phonological priming cannot be ruled out' (Pitz et al. 2017: 69). The study presented below, however, is the first to investigate L1-Norwegian learners' comprehension of the L2-German possessives sein* and ihr* in tasks that do not involve translation.
[14] It may be asked whether the forced choice task intended to test comprehension is indeed a comprehension task rather than some kind of restricted elicitation task. [38] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12 (2) 2021 [4] t he pre se nt study: aim and ge ne ral assumption We have seen that the German possessives sein* and ihr* (with a +human antecedent) pose at least three potential challenges to Norwegian learners: They are unspecified with respect to binding conditions and consequently may exhibit a type of referential (reflexive/non-reflexive) ambiguity that, at least under 'hard-core' conditions (see fn. 6), is excluded for Norwegian possessives (Sect. 1.2).
Our study focuses on Norwegian learners' interpretation of sein* and ihr* under conditions where sein* according to the rules of German grammar in the given (empty) context either must be or cannot be locally bound, as for instance in (1)- (2) and (8) respectively; and likewise for ihr* if understood as singular. That is, leaving locality-ambiguities caused by (i) aside, we are specifically interested in the possible effects of (ii) and (iii) on Norwegian speakers' grammaticality judgments and, in particular, (off-line) comprehension of L2 sein* and ihr*. The learners we are concerned with are 1 st year L1-Norwegian university students of German who have had two or three years' non-immersed teaching of German as a foreign language at high school level, corresponding to level B1 and B2 respectively of the CEFR (Common European Framework of Reference for Languages). They all had English as their first foreign language. Due to (ii) and in line with Pitz et al. (2017), we assume that Norwegian learners of L2-German at this level of proficiency tend to associate sein* more strongly with si* than with hans. Consequently, they should be more prone to wrongly understand and judge unambiguously non-locally bound sein* as locally bound (≈ si*) than to understand and judge unambiguously locally bound sein* as non-locally bound (≈ hans). This is one of the main assumptions underlying the experiments presented in Sections 5 and 6.
Due to (iii), we expect the pattern of interpretation and grammaticality judgment among Norwegian learners -and native speakers of German as wellto be more diffuse/ less skewed for ihr* than for sein*.
As for the inherent reflexive/non-reflexive ambiguity of ihr*, Pitz et al. (2017) speculate that the morpho-phonological dissimilarity between ihr* and its reflexive Norwegian counterpart si*, which contrasts with the sein*/si* pair, may 'push' the Norwegian learners towards associating ihr* more strongly with its irreflexive Norwegian counterparts hennes (sing. fem.) and deres (plur.) than with si*. In that case, they would be expected to misinterpret and judge unambiguously locally bound ihr* as non-locally bound (≈ hennes or deres) more often than the other way around -and more often than would be expected for sein*. It may be asked, however, whether the effect is counteracted by a default preference for the local candidate under locality-ambiguous conditions (cf. Sect. 3.1).
The singular/plural ambiguity of ihr* is restricted to contexts that offer a structurally accessible singular female and an accessible plural possessor candidate as well, as in (8c), or two or more singular candidates that can be cumulated into a plural possessor and at least one of which is female, as in (8b) and (10). Under both conditions pragmatics undoubtedly will play a major role in the resolution process. It seems plausible, however, that unless there are very good semantic-pragmatic reasons for the opposite, ihr* will preferably be resolved to a singular female possessor candidate if the alternative is a summative plural interpretation. Our resolution test (Sect. 6) also addresses this assumption.
[5] Pilot study on g rammatic ality judgme nt As a pilot study we conducted a rather simple grammaticality judgment experiment with Norwegian students of German. [

Materials
The test comprised a list of 24 test sentences with possessive pronouns, interspersed with 26 distractors containing different types of determiners. The test items were context-free one-clause sentences containing an occurrence of sein* or ihr* as part of a complement DP. The subject being the only antecedent candidate in the sentence, the possessive should be understood as locally bound and under that reading be judged grammatical or ungrammatical depending on whether it matches or mismatches the subject with respect to number and gender. Examples of the relevant variations are given in (16) and (17), distractors are exemplified in (18). There were 12 items with sein*, 6 grammatical and 6 ungrammatical, and 12 items with ihr, 4 of each type illustrated in (17), i.e. 8 grammatical and 4 ungrammatical. 16 [16 ] It is an interesting question -which we did not pursue -how native speakers of German would judge or react to the 'ungrammatical' items. [40] cathrine fabricius-hansen, anneliese pitz, henrik torgersen 'Kristine always orders the most expensive food.' We did not conduct a vocabulary test. However, in the formulation of test items and fillers, an attempt was made to keep both the vocabulary and the type of constructions at an everyday conversational level, i.e. a level one could expect the students to have reached.

Participants
The participants were 41 Norwegian students at the end of their first term of German studies at university level. Their L1 was Norwegian and they largely had the same background concerning German, i.e. German at high school level.

Task
The test sentences were presented in a classroom setting on paper, with the task formulated in German on the top of the page: Determine the grammaticality of the italicized determiners and pronouns in the following examples. The answer alternatives were given in Norwegian -riktig/ feil/ vet ikke ('[grammatically] correct/ [grammatically] incorrect/ don't know') -in boxes underneath each sentence. To avoid misunderstandings and prevent disturbing questions, the participants were instructed orally to read the sentence once without dwelling on the example, i.e. to make their judgment without analyzing, and then go on to the next sentence. They were told explicitly not to focus on in- [41] OSLa volume 12(2) 2021 flectional endings but to interpret the sentence in the given (empty) context. The time allotted to the whole test was 30 minutes, thus a little more than 30 seconds for each sentence. [

5.2] Hypotheses
In accordance with the assumptions outlined in Sect. 4 we had the following hypotheses: H1-1 The grammaticality of sein* is more often judged incorrectly under the (ungrammatical) mismatch than under the (grammatical) match condition. H1-2 Under the match condition, the grammaticality of ihr* is more often judged incorrectly than the grammaticality of sein*. H1-3 Under the mismatch condition, the grammaticality of sein* is more often judged incorrectly than the grammaticality of ihr*.
H1-1 relates to fact that sein* is neutral with respect to binding but morphophonologically similar to the Norwegian reflexive possessive si*; cf. (ii) in Sect. 4. The similarity is expected to enhance a correct judgment of the match cases, where a reflexive interpretation of sein* is indeed grammatically licensed, but may conversely prevent the learners from realizing that sein* cannot correctly refer to a non-matching local subject.
The rationale behind H1-2 and H1-3 is that as far as ihr* is concerned, there is no cross-linguistic similarity to trigger comparable -positive or negativeeffects on the judgment. That is, everything else being equal we would expect the ratio between correct and incorrect judgments to differ less across conditions than with sein*.

[5.3] Results
We measured the results in terms of what we shall call the error rate, i.e. the frequency (percentage) of positively wrong judgments relative to the total number of judgments that are either correct or wrong, excluding missing and don't-know judgments, i.e. uncertain responses. Under the uncertainty rate we understand the frequency of the latter relative to the total number of judgments, i.e. the number of items of the relevant type multiplied by the number of responding participants. The results are presented in Table 1.
As far as H1-1 is concerned, there were altogether 192 correct, 40 wrong and 14 uncertain judgments for grammatically correct, number-gender-matching sein*, i.e., items of type (16a). This yields an error rate of 17.2%. The uncertainty rate was 5.7%. Under the mismatch condition, i.e. with ungrammatical sein*, the
As for the ungrammatical ihr* items of the type illustrated in (17b), the results show a an error rate of 43.5% and an uncertainty rate of 10.3%. In other words, under the mismatch condition the error rate for ihr* is considerably higher than under the match condition -and significantly higher than for sein* (43.5% vs. 27.1%), χ 2 (1, N = 376) = 10.2, p = .014. H1-3, then, is not corroborated.

[5.4] Discussion
Our results may be summarized as follows: The error rate is higher for ihr* than for sein* under the match as well as the mismatch condition. This may indicate that as far as grammaticality judgment is concerned, ihr* on the whole represents a more difficult case than sein*.
(ii) For both sein* and ihr* the error rates are higher under mismatch than under match. Evidently, it is easier to correctly judge the grammaticality of a possessive occurring in a locally bound position when the possessive matches than when it does not match its (intended) local binder.

(iii)
As far as our unreliable friends sein* and Norwegian si* are concerned, their similarity may have enhanced correct judgment of [43] OSLa volume 12(2) 2021 sein* (as compared to ihr*) under the match condition, where the reflexive interpretation inherent to si* is in fact grammatically licensed for sein* ('positive transfer'). Under the mismatch condition, the reflexive reading of sein* is not grammatically licensed while si* would still be correct in Norwegian. However, we cannot tell whether or to what degree the poorer result under mismatch is due specifically to interference ('negative transfer') from si* rather than a generally lower ability to correctly judge ungrammatical possessives or structures in L2; cf. (ii).

(iv)
The fact that ihr* seems to present a more difficult case could be ascribed to several factors. First, although the informants were told not to focus on morphological endings, it is impossible to tell in how far their judgments actually were influenced by the latter. Expressions such as ihr Studium, ihr Pensum might appear odd without a morphological ending. Furthermore, the form ihr is homophonous with the singular dative form of the number-ambiguous 3 rd person pronoun sie 'she/they'. Thus, the combination of ihr and a neutral noun, as for instance ihr Studium, might seem wrong altogether. In addition, some test persons may have associated ihr* with irreflexive hennes or deres, disregarding the possibility of a reflexive interpretation (cf. Sect. 4).
Some of the uncertainties concerning the judgments might have been resolved by following up the test with an interview. Still, we consider the results to give some indication of the type of problems facing Norwegian learners in relation to L2-German possessives. While there appear to be numerous challenges concerning ihr*, the results for sein* are compatible with our assumption of transfer from L1-Norwegian.
[6] R e solution e xpe rime nt

[6.1] Preliminaries
Whereas the previous section was concerned with L1-Norwegian participants' grammaticality judgments on locally bound occurrences of ihr* and sein*, the test presented in this section investigates their ability to identify the possessor under unambiguously local or non-local binding of the possessive (in the somewhat sloppy sense of the term binding; se fn. 5). Our main assumption concerning sein* still encompasses the notion of transfer -or so-called crosslinguistic influence (CLI) -in that we expect the L1-Norwegian participants to [44] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 wrongly understand non-locally bound sein* reflexively, in the sense of locally bound si*, rather than to interpret locally bound sein* as non-locally bound (irreflexive) hans. For the correct/incorrect judgment of ihr* we presented in Sect.

a list of factors that might interfere; the one of importance to interpretation
is the assumed association with the irreflexive Norwegian possessive hennes (singular) or deres (plural) even in cases of local binding. A related matter, suited for comparison with native speakers, is the Norwegian learners' interpretation of ihr* as either singular or (summative) plural.

Material
The test material comprises a set of 32 experimental items interspersed with 32 fillers. Each experimental item is a context-free complex sentence consisting of a main clause and a subordinate adverbial clause (with während 'while') in that order; both clause subjects are singular gender-specific proper names but differ in gender; and one of the clauses contains a 3 rd person possessive as part of a complement DP. 17 The experimental items come in eight types, representing every combination of three binary conditions: (i) POSSESSIVE: sein*, demanding a male singular possessor, vs. ihr*, demanding a female singular or a plural possessor; (ii) MATCH/BINDING: The setup ensures that the possessive either matches or mismatches the subject of its own clause -the local subject -in (singular) gender. If the two match, a locally bound, i.e. reflexive, interpretation of the possessive is warranted (Local Match/Binding), since there is no pre-context; if not, the possessive will match the clause-external subject (Non-local Match/Binding), thus demanding a non-reflexive interpretation -cataphoric (forward-looking) or anaphoric (in the restricted sense of backward-looking, see Sect. 3.1) depending on its position in the sentence; cf. (iii).
(iii) POSITION: The possessive either occurs in the first, main clause (PossC1) or in the second, subordinate clause (PossC2).
[17] The design is inspired by the offline comprehension experiment described in Pitz et al. (2017: 58-69), which concerned L1-German speakers' comprehension of the L2-Norwegian possessives si* and hans. However, the order of subordinate and main clause is reversed; and the experimental items are context-free, primarily because a pre-context introducing the competing possessor candidates would blur the distinction between anaphoric and cataphoric use of the possessive. Note also that under all conditions, ihr* in principle is ambiguous between a singular (locally or non-locally bound) reading and a summative plural interpretation ('their'), referring to the two subject DPs.
The 32 fillers served to test reading comprehension at a general level and to distract the participants from the possessives. They all consisted of two clauses -mostly a main and a während-clause -with proper name subjects differing in gender; and the majority contained the locally bound (reflexive) pronoun sich or a non-locally bound 3 rd person personal pronoun referring to the non-local subject; cf. (23). The rest were of the contrasting type illustrated in (24) With some exceptions among the fillers, the vocabulary and syntactic complexity of the test items were adapted to the proficiency level of the L2participants.
Participants 31 Norwegian learners of L2-German (L2 group) and a control group of 16 native speakers of German (L1 group) participated in the experiment. As in our pilot study (Sect. 5), the Norwegian participants were students of introductory-level German courses at Norwegian universities at the time of testing; 18 (group L2-a) had had two years' and 13 (group L2-b) three years' teaching of German at high school level. The L1-participants were exchange students from a Germanspeaking country, with an average age of 26 years. The majority of the participants were female.

Procedure
The test items (experimental items and fillers) were shown on a computer screen to one participiant at a time, being presented one by one in random order. All participants were confronted with all 64 items. For each item, they had to presss a button to indicate that they had finished reading. Having done so, they were prompted to answer a wh-question in German relating to the content of the item. With experimental items they were asked to identify the referent of the possessive pronoun, i.e. the 'possessor' of the entity denoted by the DP containing the possessive. The answer was given by pressing a specific button: M for the male referent, F for the female referent and B ('both') for the (summative) plural referent. For instance, (19a) is followed by (25), the correct response being M. For each participant, the reading time and the reaction timethe time taken to submit an answer -were recorded. Only a single answer was permitted per question; in the case of ihr*, then, the participants had to choose between a female singular and a plural interpretation. In other words: the coreference resolution task presented to the participans involved forced choice interpretation. (25)

Wem gehört der Korb? ('Who owns the basket?') Press F for Siri
Press M for Jakob Press B for Both

[6.3] Hypotheses and questions
Our central hypotheses concern the possible effects of POSSESSIVE, MATCH/BINDING and LANGUAGE (L2-vs. L1-German) on the resolution of sein* and ihr* in complex sentences of the kind specified in the previous section: H2-1 On average, the Norwegian learners resolve the possessives correctly less often than the German-speaking control group. [48] cathrine fabricius-hansen, anneliese pitz, henrik torgersen

OSLa volume 12(2) 2021
H2-2 Specifically the Norwegian learners more often resolve the possessives incorrectly under Non-local than under Local Match/Binding, the difference being more substantial for sein* than for ihr*.
H2-3 The Norwegian learners resolve sein* to the subject of its own clause (i.e. reflexively) significantly more often than they do with ihr*.
We also wanted to see (i) whether POSITION might have an effect on the two groups' resolution performance, and (ii) whether there might be a general preference for resolving ihr* to the accessible singular female (F button) rather than understanding the possessive as a summative plural (B button). Throughout, it should be borne in mind that since the items with ihr* in principle allow both a singular and a plural interpretation while there is only one correct answer (the M button) to the sein* items, the chances of answering correctly when choosing randomly are twice as high for ihr* as for sein* items. Accordingly, we would expect the proportion of questions answered correctly to be generally higher for ihr* than for sein*.
In the analysis, a χ 2 -test was used to test combinations of conditions against each other in 2x2 contigency tables with one dimension representing the given condition and the other the number of successfully and unsuccessfully answered items. Every condition was tested across the other conditions, with primary focus on POSSESSIVE (sein* vs. ihr*), MATCH/BINDING and LANGUAGE.

[6.4] Results
Overview (H2-1) All participants responded adequately to all 32 experimental items, yielding a total of 512 responses from the L1 and 992 from the L2 group. the two referents together] -across the different item types (combinations of conditions; see Sect. 6.2). Numbers/ percentages of grammatically licensed ('correct') resolutions are blue, numbers/ percentages of grammatically illicit ('wrong') resolutions in italics and red.
Note that, abstracting from the plural option for ihr*, the condition Local Match/Binding demands a reflexive interpretation of the possessive, whether the latter occurs in the first or second clause of the complex sentence (PossC1 vs. PossC2). With Non-local Match/Binding, on the other hand, the possessive is non-reflexive and cataphoric (PossC1) or anaphoric (PossC2); see Section 6.2.  The two tables also show the success rate (henceforth also: SR), i.e. the number of correct responses as percentage of the total number of responses under the given condition. Note that the success rate is based on M-responses in relation to sein* but on the sum of F-and B-responses in relation to ihr.
Note, however, that while the L1-participants with very few exceptions resolve ihr* in accordance with the rules of German grammar, whether the possessive is locally bound or not, their 89% success rate for sein* deviates significantly from the 100% one would expect (p < .001 according to Fischer's Exact test). It is difficult to say what may have caused the various wrong answersapart from lack of attention (fatigue) or a preference for top-down reading [52] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 based on pragmatic expectations rather than linguistic form. 18 Notably, though, it is the non-local cataphoric condition that causes most problems, exhibiting a success rate of only 80%; cf. Table 5, row 2b. We return to the cataphoricity issue below and in Section 5.5.

above).
Obviously, the lower frequency of reflexive/local L2-resolutions of ihr* may have to do with the fact that the summative plural interpretation is a licensed alternative to the female singular. Thus if we leave the B responses out of the counting for both possessives, their reflexivity rates (sein* 93%, ihr* 96%) do not differ significantly under Local Match/Binding, where the reflexive/local resolution is correct. Under Non-Local Match/Binding, however, ihr* still exhibits a significantly lower rate of (incorrectly) reflexive/local resolutions than sein*: 35% (50/141) vs. 54% (125/230), χ 2 (1, N = 371) = 11.8, p < .001. Altogether, then, we may consider H2-3 corroborated.

Effects of POSITION and the singular-plural ambiguity of ihr*
In our experimental items, the possessive either occurs in the second clause (PossC2), as in (26) (28) a. Jakob legt das Gemüse in seinen Korb, während Siri an der Kasse wartet.
'Jakob puts the vegetables in his basket while Siri is waiting at the checkout.' b. Siri legt das Gemüse in ihren Korb, während Jakob an der Kasse wartet.
As shown in Table 5 above, the success rates in both the L2-and the L1-group are somewhat higher with PossC2 than with PossC1 under Local Match/Binding (rows 1a vs. 1b) than under Non-local Match/Binding (rows 2a vs. 2b). On the whole, however, the differences are too small to be significant. The greatest numerical contrast is found for L1 with respect to cataphoric vs. anaphoric sein* (SR 80% vs. 95%), Fischer's Exact test: p = .14.
Turning to the question of how the L1 and L2 participants performed with respect to the singular-plural ambiguity of ihr* (Sect. 4), Table 6 shows the distribution of (female) singular (F) and summative plural (B) resolutions in relation to the total number of grammatically licensed (F or B) resolutions of ihr*.  As can be seen, the L1-participants clearly prefer the F response unless ihr* is used cataphorically. By contrast, the L2-participants show a only moderate bias towards the female singular resolution (69%, χ 2 (1, N = 471) = 16.7, p < .001) under Local Match/Binding, where it amounts to a reflexive interpretation. Under the anaphoric condition, where the matching female singular candidate occurs in the preceding clause, the 61% bias towards that candidate is insignificant because of the lower number of correct (F+B) responses (χ 2 (1, N = 207) = 2.2, p = .14). Under the cataphoric condition, however, only 29% of the resolutions fell on the female singular candidate; that is, here the preference is significantly inverted to the summative plural option (71%, χ 2 (1, N = 191) = 7.6, p < .01).

Correct resolutions/ responses for ihr*
In short: with the exception of the cataphoric condition, where the possessive precedes the only matching singular candidate, the structure of our experimental items generally favours resolving ihr* to the only accessible female singular candidate rather than to the sum of both candidates, in particular among the L1-participants.
From the results presented in tables 5 and 6 we conclude that POSITIONwhether the possessive occurs in the first or the second clause of the experimental structure -has an effect only in combination with Non-local Match/Binding.
[6.5] Summary and discussion First, our three central hypotheses (Sect. 6.3) were all supported: In accordance with H2-1, the Norwegian learners of L2-German altogether were less successful in resolving the possessives correctly than the L1-German control group. However, their performance varied considerably across conditions. Like the L1group, they achieved higher success rates with ihr* than with sein* across conditions -not surprisingly in view of the fact that the ihr* items have two correct resolution options against only one with sein*. Different from the L1-group, however, they had significantly lower success rates when the possessive referred to the subject of the preceding or following clause (Non-local Match/Binding) than when a reflexive interpretation was licensed (Local Match/Binding); and in accordance with H2-2, the contrast was significantly more marked for sein* than for ihr*; see Table 5. More specifically: Under Nonlocal Match/Binding, sein* showed a significantly higher rate of (incorrectly) reflexive/local resolutions than ihr* (see Sect. 6.4 for details). Thus, H2-3 too is supported.
Second, the L1-group preferred the female singular to the summative plural interpretation of ihr*, the preference being very pronounced under Local Match/Binding, somewhat less so under the anaphoric and insignificant under the cataphoric condition ( Table 6). Singular was preferred by the L2-group, too, but the bias was less marked and under the cataphoric condition the plural resolution prevailed (Table 6). Now, since plural is a generally acceptable interpretation of ihr* in our experimental items, it may be chosen so as to avoid deciding between the local and the non-local singular candidate. This would explain why it has a higher frequency in the L2 than in the L1 responses.
Third, POSITION had a substantial effect on the resolution only under Nonlocal Match/Binding, i.e. when the possessive according to the rules of German grammar cannot be understood reflexively: In both the L2-and the L1-group, the cataphoric structure triggered more plural resolutions of ihr* than the anaphoric structure (Tab. 6); and the L1-group showed a lower success rate with cataphoric than with anaphoric sein* (Tab. 5). In view of the possible fallback function of the summative plural, these findings indicate that our cataphoric condition represents a resolution challenge (even) to native speakers of German. 19 The L2-participants, too, perform worse with cataphoric than with ana- [19] In fact, some L1-participants commented on cataphoric (sein*-) items as being 'incorrect'.
[56] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 phoric sein* but their success rate in the latter case is already very low (Tab. 5). 20 Note here that the experimental items all consisted of a main and a subordinate clause in that order, with the possessive occurring in either the first or the second clause (Sect. 6.2). Consequently, our anaphoric and cataphoric conditions differ in two respects: The anaphoric possessive e.g. in (30) refers back from a subordinated to a matrix clause DP while the cataphoric possessive in (31) refers forward from the matrix to the subordinate clause. And as mentioned in Section 3.1, cataphoricity of that kind is more severely constrained than forward dependencies from a subordinate clause to its matrix clause. On this background, the abovementioned results may not be too surprising. Accordingly, one would expect a more successful performance when the cataphoric dependency goes from a preposed subordinate clause to its matrix clause, i.e. under conditions that license cataphoric use of ordinary pronouns (see Sect. 3.1). A natural follow-up to our resolution test, then, would be based on experimental items that differ from the present ones with respect to the order of subordinate and main clause but otherwise follow the same pattern of variation (see Sect. 6.2), i.e. pairs like (32) and (33) for anaphoric vs. cataphoric use of the possessive. 21 (32) a. Während Daniel an der Kasse wartet, legt Brigitte Gemüse in seinen Korb. b. Während Brigitte an der Kasse wartet, legt Daniel Gemüse in ihren Korb.
[21] We originally intended to include these variations in our experiment but for practical reasons had to give up the idea. Fourth, concerning L2-PROFICIENCY/LEVEL, we found a significant difference in success rates between L2-learners with 2 years of high-school German (group L2-a) and those with 3 years' training (group L2-b). This indicates the fruitfulness of formal training both with regards to mastery of possessive systems that are in very nuanced ways different from one's L1 and with regards to handling the negative cognacy effects of false friends.
[7] Conc lusion and outlook The primary objective of our enterprise was to find out how L1-Norwegian learners of L2-German at a low or moderate level of profiency (see Sect. 5.1 and 6.2) cope with the 3 rd person possessives sein* and ihr* in comprehension, given the lexical divergence holding between the two binding-neutral German possessives on the one hand and their respective binding-sensitive counterparts in Norwegian on the other hand and, in particular, the morpho-phonological similarity between sein* and the Norwegian reflexive possessive si* (see Sections 1 and 3).
As far as sein* is concerned, our results seem to corroborate the assumption that the Norwegian L2-learners due to the sein*/si* similarity are more biased towards a reflexive interpretation than warranted by the general preference for locality that has been observed in prior research on pronoun resolution (see Sect. 3.1): While our non-native comprehenders by and large resolved unambiguously reflexive, i.e. locally bound occurrences of sein* correctly, almost half of their resolutions of anaphoric or cataphoric sein* were erroneously reflexive; in comparison, the German-speaking control group did not have serious resolution problems under either condition (Sect. 6.4; Tables 3, 4 and 5). Also, the (Norwegian) participants in the grammaticality judgment test had a higher error rate when sein* did not than when it did match the subject of its local clause, i.e. its intended local binder (see 5.3-4). Being unspecified for possessor number and gender, the Norwegian reflexive possessive si* would be grammatically licensed not only when sein* matches but also when it does not match the subject of its local clause in gender, i.e. under the conditions where our L2groups significantly often incorrectly chose a reflexive interpretation or erroneously judged sentences containing a mismatching sein* to be correct. It seems reasonable, then, to ascribe the non-native resolution failures and misjudg- [58] cathrine fabricius-hansen, anneliese pitz, henrik torgersen OSLa volume 12(2) 2021 ments of sein* to interference from the unreliable possessive L1-friend si* rather than to a general preference for local binding. 22 In line with this explanation, the L1-Norwegian comprehension of L2-German ihr* offers quite a different picture: Under conditions licensing a reflexive (female singular) interpretation of ihr*, our L2-comprehenders showed only a modest (65%) preference for that option -in contrast to the L1-group; and when the local subject did not match the possessive, their rate of (erroneous) reflexive resolutions was considerably lower for ihr* than for sein* even when we disregard the summative plural alternative (Sect. 6.4, Table 4). In addition, the L2-learners generally had more difficulties judging the grammaticality of ihr* correctly than they had with sein* (Sect. 5.3-4). Altogether these findings may indicate that our L2-learners, as suggested in Section 4, tend to lexically associate ihr* with the irreflexive Norwegian possessives hennes 'her' and deres 'their' rather than with the reflexive si* but that corresponding interference effects on L2-comprehension to a certain degree are levelled out by a general preference for local resolutions -and the context-dependent possibility of a (summative) plural interpretation.
The more general lesson to be drawn from our observations could be that L1/L2-specific relations of divergence and morpho-phonological similarity at the lexical level, at least at non-advanced stages of L2-learning, might play a more important role in non-native resolution of pronouns than has hitherto been acknowledged.
It might be objected that our study for various reasons does not allow any far-reaching conclusions. The number of experimental items is rather restricted in both experiments; some of the items in the resolution experiment are pragmatically biased towards one or the other licensed interpretation; the number of L1-participants is too small; and the experimental procedures may not have been optimal. Still, we find the results interesting enough to warrant follow-up studies in different directions, in part along the lines suggested in Section 6.5 but also -and perhaps most importantly -investigations into L1-Norwegian learners' online processing of L2-German possessives compared to ordinary 3 rd person pronouns (cf. Sect. 3.2 and Pitz et al. 2017: 58-69) and their resolution preferences under locality-ambiguous conditions, where the referential issue cannot be settled by grammatical cues (cf. Sect. 4).