The Use of the Verb Run in English Learner-Narratives An Analysis of Verb Constructions Influenced by Different L1s

The present paper investigates the use of the verb run in texts compiled in English by three groups of multilingual learners in the North of Italy. The participants spoke three different L1s: Ladin, Italian, and German. The main objective was to ascertain whether L1 imprints could be identified in the different groups, and whether the Ladin subjects behaved more similarly to their Italian or German counterparts. Despite considerable homogeneity found across all groups, the findings surprisingly suggested a stronger closeness between the texts of Ladin and German speakers, and not – as the typological relationship between Ladin and Italian would suggest – between Ladin and Italian speakers. The present data implicate that multilingual speakers bring to bear composite cognitive and typological systems when they construct texts in English as a third or further language, and text features are usually not easily attributable to a typologically clearly defined L1.


Introduction and Research Questions
It seems indisputable that multilingual language learners might be influenced in a dynamic way by more than one language in their acquisition process and language use (Cenoz, Hufeisen, Jessner 2001;Jessner 2006Jessner , 2008. However, the exact role of the background languages in further language (Ln) learning is controversial. According to the L1 factor hypothesis, the L1 is the privileged source of transfer at the initial state of L3 learning, while other hypotheses postulate that the L1 is favoured only when a particular property has not been acquired in the L2 yet, and that it is the L2 that takes on the strongest role in L3 learning (Bardel, Falk 2007;Falk, Bardel 2010;Hermas 2010Hermas , 2015. The typological primacy model suggests that L3 or Ln development is influenced by what the multilingual learner perceives to be the typologically closest language in his or her repertoire, whether or not such transfer is facilitative (Rothman 2011(Rothman , 2015. The scalpel model of third language acquisition argues that transfer happens property by property from any previously acquired language and that it is influenced by "additional factors pertaining to the relevant properties, such as processing complexity, misleading input, and construction frequency in the target L3" (Slabakova 2017, 655). The present study aimed to identify the level of similarity in the use of verb-phrases containing run across three learner groups representing three L1s: (i) Dolomitic Ladin, (ii) German (the Austrio-Bavarian dialect spoken in South-Tyrol), and (iii) Italian. 1 The research participants learnt English in a region where English is taught mainly as an L3 and L4, after Ladin, Italian, and German. The question was raised whether the Ladin subjects, who were proficient in both Italian and German, were closer to the Italian or to the German speakers in their use of the verb run. The possible influence of Ladin, Italian, and German on the target language -English -was investigated, and a possible rationale for a number of similarities and anomalies between the three L1 groups was proposed. The findings of this study are explained from the perspective of cross-linguistic influence. 2 Answers to the following questions were sought: 1. What are the similarities between the Ladin, Italian, and German learners of English in the use of the verb run?
EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 77 2. Which L1 group differentiates itself most? Do the Ladin speakers behave more similarly to the Italian or to the German speakers in the use of the verb run in their narratives? 3. Do the L1s Ladin, Italian, and German have an influence on the target language English?

Previous Investigations and Theories
The high-frequency verb run has been classified as a manner-of-motion verb, since its stem includes information about how the figure physically moves (i.e. move fast on foot). Unlike manner-of-motion verbs, path verbs (e.g. exit, enter, get) and their adnominal and adverbial encodings (e.g. into, away) describe the trajectory the figure follows. Talmy (1985Talmy ( , 1991Talmy ( , 2000Talmy ( , 2016 considers languages to be verb-framed (V) when they habitually express paths of motion in the main verb and manner optionally outside it with prepositional phrases, adverbs, or gerunds (e.g. Italian Entrò di corsa / correndo, 'He entered running').
Languages are called satellite-framed (S) when they characteristically encode manner-of-motion information in the verb root and paths in directional particles (e.g. English Peter ran out or German Peter rannte heraus). Although most Romance languages are seen as V-framed, Ladin displays S-framed characteristics. It possesses a restricted manner verb lexicon but a wide repertoire of path particles, which is arguably due to its mainly oral nature, as suggested by Irsara (2015). The verb run can occur as a bare verb, providing no further elaboration of path beyond its inherent directionality, e.g. He started running (Slobin 1996a). V-framed languages have been shown to use bare motion verbs more often than S-framed languages (Berthele 2006;Slobin 1996a). Moreover, run can appear in a minus-ground or in a plus-ground clause: "Minus-ground clauses" consist of bare verbs or verbs with satellites indicating direction of movement; "plus-ground clauses" have, in addition, one or more prepositional phrases encoding source and / or goal. (Slobin 1996a, 201) In Slobin's (1996a) data, both the English (S-framed) and Spanish (Vframed) speakers use more plus-ground than minus-ground clauses. Moreover, Slobin (1996a) shows that English narratives tend to use more ground adjuncts than Spanish narratives. Similarly, Berthele (2006) shows that the number of plus-ground clauses exceeds the number of minus-ground clauses. However, his figures do not permit to draw conclusions about path elaboration with ground elements in S-framed as opposed to V-framed languages. In line with the above, Spreafico (2009) shows that all the languages considered by him fa-EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 78 vour the adnominal locus, represented mainly by prepositional phrases, to express paths of motion.
Complements of source and goal are often found with run. Endpoints are typically marked by the prepositions to, into, and onto, while sources are normally expressed by prepositional phrases headed by away, from, out, and off. The source marker away can occur alone (He ran away) or with a from phrase as complement (He ran away from him) (Huddleston, Pullum 2002). It is maintained by Von Stutterheim, Carroll, and Klein (2009) that German speakers refer to endpoints more often than speakers of other languages, e.g. English and Dutch. In general, goals are expressed more often than sources. Evidence has been found for the goal-over-source principle by Verkerk (2017) in 17 Indo-European languages.
Leaving a source or reaching a goal can involve crossing an enclosure or boundary. In S-framed languages (e.g. English and German), manner verbs like run can be used in boundary-crossing situations (He ran into the house / Er rannte ins Haus), whereas they are normally not employed to indicate the crossing of enclosures in V-framed languages (e.g. Italian), where path verbs seem to be preferred (Entrò in casa [correndo], 'He entered the house [running]') (Alonso-Alonso 2015 ;Filipović 2007;Slobin, Hoiting 1994).
The verb run can be preceded by the inceptive verbs start and begin. Ladin and Italian have developed a rich repertoire of aspectual verbal periphrases, among which a number of periphrases for start / begin, e.g. Ladin mëte man a, de / scomencè a, de / se mëte a or Italian (in)cominciare a / mettersi a / prendere a / venire a. Gallmann, Siller-Runggaldier, and Sitta (2007) note that German normally prefers adverbial expressions (Plötzlich regnete es, 'suddenly it rained') over periphrastic start / begin expressions (e.g. Es fing an / begann zu regnen 'It started / began to rain'). Although adverbial phrases can replace start and begin in Ladin and Italian as well, this seems to be the case in German in particular.

Hypotheses of the Current Study
A high degree of similarity between the groups was expected in the present study, due to the subjects' analogous multilingual biographies. Because of the nature of the stimulus material used for the analysis, it was expected that all the groups would use run mainly in the sense of 'fast motion' and 'escape' (Glynn 2014;Gries 2006). It was foreseen that the sense of 'escape' would be expressed frequently with the phrasal verb run away and that run would therefore be found more often with sources than with goals. The speakers of German (S-framed) were expected to use more locative adverbs and prepositions with the verb run than the speak-EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 79 ers of Italian (V-framed). Similarly, the English texts by the Ladin speakers were expected to exhibit a large number of locative particles, due to the large use of these in Ladin. Vice versa, the narratives by the Italian subjects were expected to have a higher percentage of run as a bare verb, in line with Slobin's (1996a) and Berthele's (2006) findings.
Germans were expected to express targets more often than Italians, in line with findings by Von Stutterheim, Carroll, and Klein (2009). In the expression of the meaning of 'escape', the German speakers were expected to use the phrasal verb run away more often than the Italian and Ladin speakers. While German has forms that are similar to the English one, weglaufen, wegrennen, 'away + run', corresponding verbs in Italian are normally fuggire, scappare, 'escape' rather than the analytical form correre via, 'run + away'. 3 Similarly, the Ladin equivalent is a reflexive form of sciampè, 'escape', and not saltè demez, 'run + away'. In this respect, the Ladin participants were therefore expected to behave like the Italian speakers.
If manner verbs (such as correre, 'run') are normally not used in Italian to describe the crossing of enclosures, the Italian participants were supposed to use the manner verb run less often than the German subjects when the path involved crossing a boundary. Also due to an L1 influence, the Italian and Ladin speakers were assumed to have a stronger tendency than the German speakers to use start and begin in their English texts.

Participants and Writing Task
The data were provided by 94 learners of English, who formed three groups according to their L1s: Ladin (41 participants); Italian (21 participants); German (32 participants). All participants were multilingual and could speak Italian, German, and English. The Italian and German subjects learnt English as an L2 and L3, whereas the Ladin speakers learnt it as an L4 after Ladin, Italian, and German. The participants were estimated to be at a B1-B2 level of English according to the Common European Framework of Reference. At the moment of data collection, the study participants were attending the final year of upper secondary school, or they had just started their English courses at university.
The data were collected through a shortened version of Mayer's (1969) picture-story Frog, where are you?, which depicts the adven-3 See Iacobini, Masini (2007, 2009 for the aspectual meaning of correre via, 'run away'. EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 80 tures of a boy who undertakes a search for his lost pet frog. The writing task was carried out on the basis of 18 black-and-white pictures that had been selected out of 24. During the task, which was anonymous and unrehearsed, collaboration and support materials were not permitted.

Procedure
After the digitization of the texts, the sentences that contained instances of run were isolated for ease of analysis. Answers to the questions in Section 1 were sought by observing: (i) the frequency of occurrence and distribution of the verb run; (ii) the locative prepositions and adverbs accompanying run; (iii) the occurrence of the inceptive verbs start and begin; (iv) the nominal locative specification in the run-events.
Quantitative and statistical analyses were carried out on the basis of the raw data, and qualitative reflections were made in order to broaden understanding of the data and answer why-questions.
To test whether Ladin children produced constructions in written language production that were more similar to the constructions used by learners with German L1 than to those occurring in participants with Italian L1, we fitted generalized mixed effects logistic regression models (e.g. Baayen 2008;Baayen, Davidson, Bates 2008).
For the mixed effects analysis we used R and the 'lme4' package (Bates et al. 2015). We first fitted generalized mixed effects models for the overall occurrence of the verb run, before we calculated models for the occurrence of different constructions with run, like started running, as dependent variables. In each model, we included the predictor Language group with the levels Ladin, German, and Italian as fixed effect, where we defined the level Ladin as baseline.
In the following, we report z-and p-values for the mixed effects models if factors contribute significantly or marginally significantly to the models. In order to keep participant variation under statistical control, the factor participant was included as random effect.

Similarities between the Groups
In all the groups, run was used around twice per learner and corresponded to approximately 4-5% of the total number of verb-phrases. 4 The verb was used most often to describe the events depicted in the EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 81 images given below as figures 1 and 2, showing a deer and a dog fast approaching a high riverbank and the protagonist's dog escaping from a swarm of bees. Table 1 shows that, in all the groups, more than 40% of the occurrences of run referred to figure 1. Figure 2 came next in terms of run-occurrences. The verb run was most often followed by locative prepositions and adverbs, whereas it was a bare verb in less than half of its occurrences. The locative particle that accompanied the verb run most often was away. Table 2 illustrates that run was followed by away in more than 40% of its occurrences in the German and Ladin groups, whereas the percentage was lower in the Italian group. However, there was no statistically significant difference between the groups in the use of away with the verb run, although we saw differences when we compared the use of bare vs. non-bare verbs in general (see § 5.2).  In the majority of cases, run away was not post-modified by further elements, as in (1), where run away does not need an overt source adjunct, as the entity the subject is fleeing from is mentioned previously, namely an owl, referred to by the superordinate expression a bird of prey.
(1) The boy hid from the bird of prey and ran away. (Ladin L1) run away was accompanied by a prepositional or adverbial element in less than half of its occurrences, as table 3 shows. The percentage was the highest in the Italian group. Hence, when the Italian subjects used the phrasal verb run away, they often added further information. However, the difference between the groups did not reach statistical significance in the use of run away with further elements. The verb-phrase run was most often followed by directive particles, but these were not always accompanied by noun-phrases specifying the run-trajectory further. In (2), run is followed by the preposition of direction to and a noun-phrase indicating the deer's destination. The sentences in (3) and (4) exemplify the use of the verb run without any nominal locative specification. (2) The deer ran to a river with the boy on its head. (Ladin L1) This animal began to run very fast. (Ladin L1) (4) The dog ran away, so fast he was able to run. (Ladin L1) Noun-phrases accompanied run in one quarter of the total number of occurrences of the verb, whereas in 75-76% of its occurrences, run was not accompanied by any nominal or pronominal locative specification. Table 4 shows similar percentages for each group, with a dif-EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 83 ference which is far from being statistically significant. The minusground clauses therefore exceeded the plus-ground clauses. The verb run was found more often with sources than with targets [tab. 5]. In all the groups, run was used to describe a subject leaving a source location in more than 40% of its occurrences. Explicit and implicit references to sources were found especially in the Ladin and German groups, where the percentages were similar. However, neither group distinguished itself from the others significantly. As with the sources, the verb run was followed by targets especially among the German and Ladin speakers, whereas this was not often the case in the Italian group, as illustrated in the third column of table 5. However, the rather low figures in this column did not reach statistical significance. The source was marked mainly by away (from) and out of. While in (5) the dog has escaped from the bees, in (6) the boy has just left his house.
(5) The dog was afraid and ran away from the bees. (Ladin L1) (6) He ran out of the house to help the dog. (German L1) A frequent use of the lexical aspectual verbs start and begin came to light in the analysis of the learner data. In the texts written by the Ladin and Italian speakers, run was preceded by start and begin in one quarter of its occurrences. The verb run was accompanied by start EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 84 and begin less often in the German group [tab. 6]. However, the difference between the groups turned out to be statistically non-significant. Table 6 Start / begin running / to run(100% = the total number of occurrences of the verb-phrase run)

Differences between the Groups
The Ladin group was closer to the German than to the Italian group in the distribution of the verb run through the narratives. In the story sections accompanying 6 images out of 18, a similar concentration of run verb-phrases was found in the Ladin and German groups, whereas the Ladin group was closer to the Italian group only in 3 images. Table 1 above showed that the concentration of the verb run in figures 1 and 2 was strongest among the German speakers, who differentiated themselves from the Ladin speakers most (80% in the German group vs. 69% and 65% in the Ladin and Italian groups). However, a consideration of figure 2 on its own made the Italian group stand out (16% in the Italian group vs. 26% and 29% in the Ladin and German groups). Unlike the German and Ladin groups, the Italian speakers showed a reluctance to describe the dog's escape from the swarm of bees with a run verb-phrase.
The unwillingness among the Italian speakers to use a run verbphrase to describe an act of fleeing was confirmed in another story section. Ladin and German speakers used run verb-phrases in the escaping scene depicted in figure 3 below, which shows the story protagonist bracing himself against an owl with wings spread wide apart above him, whereas none of the Italian speakers used a run expression in this story section.
Vice versa, a stronger use of run verb-phrases among the Italian than among the Ladin and German speakers was ascertained in the story section referring to figure 4 below, which shows the story protagonist standing outside his bedroom window with his pet dog in his arms. The dog has just fallen from the window, smashing a jar where its head was stuck. In the Italian group, 10% of the run occurrences were used in this story section, as exemplified in (7), whereas in the Ladin and German groups only 4% and 5% of the run instances were found in this scene.

85
Despite the preference for non-bare verbs in all the groups, the Ladin and German subjects used locative prepositions and adverbs with the verb run more often than their Italian counterparts [tab. 7]. Vice versa, the Italian speakers had a higher percentage of run as a bare verb. For the occurrence of bare verbs vs. non-bare verbs, there was a trend for the Italian group to use run as a bare verb more often than the Ladin group [tab. 8]. Table 7 Percentages of run followed by locative particles and percentages of run occurring as a bare verb (100% = the total number of occurrences of the verb-phrase run)

L1
Percentage of run + locative particle

Percentage of run as a bare verb
Ladin 66% 34% Italian 51% 49% German 66% 34% The Italian group also differentiated itself in the analysis of boundary-crossing situations, as in (6) Table 9 Boundary-crossing in the English learner-texts (100% = the total number of occurrences of the verb-phrase run)

L1-Influence on English
The slightly larger use of directional particles by the German and Ladin speakers (66%) than by their Italian counterparts (51%) might be an L1 effect. The figures in table 7 seem to confirm the hypothesis that speakers of S-framed L1s (German) have a stronger tendency to employ locative particles than speakers of V-framed languages (Italian), who appear to favour bare motion verbs (Berthele 2006;Slobin 1996a). The statistically marginal difference between the Ladin and Italian groups of learners in their use of directional particles might find an explanation in the wide use of locative particles in the Ladin language. Crosslinguistic influence from the learners' L1 was detected in the extent of the participants' use of run away. Although the group differences in the use of run away could not be statistically confirmed [tab. 2], the lower percentage in the Italian (29%) than in the Ladin (41%) and German (44%) groups might nonetheless be interpreted as a possible L1 effect. Unlike the German and Ladin learners, the Italian speakers were unwilling to use run in the description of the dog escaping from the bees [ fig. 2] and never used it to describe the boy hiding from a preying bird [ fig. 3]. The one-word forms fuggire and scappare, 'escape' in Italian, might have restricted the use of the two-word phrase run away in the narratives written by the Italian speakers. Vice versa, the German prefixed verbs weglaufen and wegrennen, 'away + run', might have prompted German participants to choose the similar phrase run away in their English texts. Since Ladin normally uses the form sciampè to describe an act of fleeing, like Italian, the frequent use of run away in the Ladin group might be due to a German influence, or to the strong tendency to use syntagmatic verbs and locative adverbs in Ladin in general.

87
Possible influence from their source language was assumed in the German learners' use of the verb run with targets. Low figures resulted and did not reach statistical significance [tab. 5], but the hypothesis based on findings by Von Stutterheim, Carroll, and Klein (2009) that German speakers tend to indicate the target or goal of motion more often than speakers of other languages could be supported to some extent.
However, the minus-ground clauses exceeded the plus-ground clauses in each group [tab. 4]. The results obtained by Berthele (2006), Slobin (1996a), and Spreafico (2009) could thus not be supported in the analysis presented here. It could not be confirmed that speakers of S-framed languages tend to add more ground adjuncts, as the separation of the German group from the Italian and Ladin groups was only 1%.
Counter to expectations, the highest percentage of run-instances in boundary-crossing situations was found in the Italian group, with a statistically significant difference between the Italian and Ladin groups [tabs 9-10]. If it is true that verb-framed languages (including Italian) prefer main verbs of path (e.g. uscire, 'exit') rather than of manner (e.g. run) to indicate the crossing of enclosures, the frequent use of run in boundary-crossing situations in the Italian group is not explainable as an L1 effect.
As was foreseen, the Ladin and Italian research participants used start and begin more often than the German-speaking subjects, which might be due to the stronger use of aspectual verbal periphrases in Ladin and Italian than in German and, therefore, explainable as an L1 effect. However, this could not be proved statistically, since the difference between the groups turned out to be non-significant [tab. 6].

Discussion and Conclusions
The comparative evaluation revealed considerable resemblance between the Ladin, Italian, and German groups of learners. The verb run was used to a similar extent and had an analogous distribution in the learner-narratives. Locative prepositions and adverbs most often followed run, which occurred as a bare verb in less than half of its occurrences. The directive particle away was the most common satellite that accompanied run, which most frequently occurred in a minus-ground clause, without further nominal locative specification. A preference for sources over goals was observed in all the groups, but neither sources nor targets were made explicit very often. The study participants spoke different L1s, but they all learnt English in a multilingual context, so that the high degree of similarity between the groups might not be surprising. It was expected that EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 88 learners at an intermediate level of proficiency would use the verb run widely to narrate a story based on a succession of events with numerous changes of place. The verb run is a basic verb that normally appears at pre-A1 Starters level and seems to be employed extensively at higher levels as well. Learners of English have been shown to overuse elementary verbs of motion and to ignore more specific manner-of-motion verbs that have a lower frequency of occurrence, and suggestions have been made to pay more explicit attention to motion-event descriptions in classrooms (Irsara 2017).
Although the research groups used run more often with sources than with targets, the findings cannot deny the sound goal-oversource principle revealed by Verkerk (2017). They rather confirm that language patterns are highly context-dependent, since chasing and escaping scenes that depict the protagonists fleeing from hazardous animals are central in Mayer's (1969) Frog, where are you?. The findings also show that learners often leave information unexpressed when they narrate stories based on pictures, where a number of details are left to be inferred, e.g. sources and endpoints of motion events.
The research group that differentiated itself most was the Italian group, while the Ladin group of learners generally tended to be closer to the German speakers. The Italian participants showed a higher percentage of run as a bare verb (without locatives), confirming that speakers of V-framed languages (i.e. most Latin-based languages, e.g. standard Italian) have a stronger preference for bare verbs than speakers of S-framed languages (e.g. German) (Berthele 2006;Slobin 1996a). As expected, the Italian group of learners used the verb run less often than the other groups in escaping scenes. Despite the lack of statistical confirmation in this respect, the Italian speakers' reluctance to employ a run verb-phrase to indicate an act of fleeing was interpreted as a possible L1 effect, since it might be due to the formal dissimilarity between the English run away and the Italian one-word form scappare. Against predictions, Italians used run more often than other speakers in the crossing of a boundary, where a statistically significant difference between the Ladin and Italian groups was found. However, the use of run in boundary-crossing situations was not always target-like in the Italian group. While running out of a house is possible and grammatically correct, the verb run is not appropriate to describe someone's exit from a window (e.g. run out of the window).
There might be a number of reasons why the Ladin speakers turned out to be closer to the German than to the Italian group of learners. Although Ladin belongs to the Romance language group, like Italian, it has also a number of characteristics in common with German. While certain Old Romance traits that are also present in current German varieties were lost in a number of later Romance EL.LE e-ISSN 2280-6792 9(1), 2020, 75-94 89 languages, they survived in Ladin, probably due to linguistic contact with German-speaking communities (Benincà 1994). The widespread use of syntagmatic verbs in Ladin has for instance been suggested to be the result of different factors, i.e. the existence of similar constructions in vulgar Latin; language contact with German; and the alpine landscape, which seems to invite detailed descriptions of directions (Hack 2011). Referring to Rhaeto-Romance, Ascoli spoke of "materia romanza e spirito tedesco" (1880-83, 556). Ascoli's (1880-83) description of Ladin traits as Romance substance with a Germanic spirit points to the similarities that Ladin shares with both Italian and German. Moreover, Ladin speakers are proficient in Italian and German, so that they can be influenced by both languages in their production of English texts. However, German seems to be most often used for comparisons in English language teaching at school, so that this might encourage learners to draw upon German more than upon Italian. The Ladin speakers generally behaved more similarly to the German speakers, but they were closer to the Italian participants in the use of the inceptive verbs start and begin, which they employed slightly more often in conjunction with run than the German speakers. Although statistically not confirmed, this was explained as a possible influence from the Ladin and Italian L1s, which appear to make wide use of verbal periphrases to express the beginning of events (Gallmann, Siller-Runggaldier, Sitta 2007).
Hence, few L1 effects could be identified in the multilingual groups of learners at an intermediate level of English proficiency. Although the L1 appeared to be somehow involved, it could not be argued to be the privileged source of transfer, in which case more inter-group differences would have been found. While it might be unproblematic to detect various forms of L1 influence in L2 learning, this becomes increasingly complex in L3 or Ln acquisition, particularly if the background languages are typologically related. In the acquisition of English as an L4 by speakers of Ladin, psychotypology might play a central role. Learners might perceive Ladin to be typologically distant from English because of its native status and regional nature, thus favouring German as the main source of influence. Given the high degree of similarity between the L3 and L4 groups of learners in this analysis, it might be ventured that L3 and L4 learners share more commonalities than sometimes assumed.