You are on page 1of 23

Otto-Friedrich-Universität Bamberg

Lehrstuhl für Englische Sprachwissenschaft einschließlich Sprachgeschichte


Hauptseminar: Cognitive Aspects of Grammatical Variation in English
Erasmus exchange program (8 ECTS)
Dozent: Prof. Dr. Julia Schlüter
Sommersemester 2016-2017

Effects of Animacy and Information Status


on the Dative Alternation in
Serbian-English Interlanguage

Marija Milojković
Matrikelnummer: 1872678
Josef-Kindshoven-Straße 5, 96052 Bamberg
Telefon: + 49 15237180995
E-Mail: marija.milojkovic@stud.uni-bamberg.de

24. August 2017


1
Table of contents
1. Introduction .............................................................................................................. 1
2. Dative alternation in English .................................................................................... 2
3. Dative alternation in Serbian .................................................................................... 4
4. Dative alternation in SLA ......................................................................................... 6
5. Methodology............................................................................................................. 7
5.1. Aims and hypotheses ......................................................................................... 7
5.2. Data collection ................................................................................................... 8
5.3. Pilot study ........................................................................................................ 10
5.3.1. Participants ............................................................................................... 10
5.3.2. Data analyses and results .......................................................................... 10
5.4. Main study ....................................................................................................... 14
5.4.1. Participants ............................................................................................... 14
5.4.2. Data analyses ............................................................................................ 14
6. Results and discussion of the results in the main study.......................................... 14
7. Conclusion .............................................................................................................. 17
References ...................................................................................................................... 19
Appendix ........................................................................................................................ 21

0
1. Introduction

Studies of grammatical variation in the English language, such as the dative alternation
(DA), i.e. the choice of two kinds of object constructions with ditransitive verbs, have
revealed that grammatical knowledge of native speakers has a probabilistic component
(Bresnan & Hay 2008). The probability of using a particular variant in a given context
depends on different factors, such as animacy, syntactic complexity, pronominality etc.
Most studies of probabilistic constraints on the dative alternation have focused on English
native speakers, while such phenomena have been neglected in SLA research (Jäschke &
Plag 2016). A few studies deal with constraints on the DA among EFL learners of
different proficiency levels and with different native languages (Tanaka 1987; Chang
2004; De Cuypere et al. 2014; Marefat 2005; Callies & Szczesniak 2008; Jäschke and
Plag 2016), but their transfer and role in learner language still remain unclear.

The present study aims to contribute to a better understanding of the acquisition of


probabilistic constraints and their effects on the DA among advanced Serbian EFL
learners. Advanced leaners were chosen as their command of English is assumed to be
near nativelike and their choices influenced by similar factors as native speakers’ choices.

The study is mainly based on Bresnan et al. (2007) but it focuses particularly on the effects
of animacy and information status on the DA with a core dative verb to give in Serbian-
English interlanguage. Having in mind that the factors of animacy and information
structure have a strong influence on word order in the Serbian language (Piper & Klajn
2013; Jasinskaja 2016), we expect that advanced Serbian EFL learners will be sensitive
to them when choosing between the two alternative dative structures, but, since verb
argument alternations are highly language-specific (Callies & Szczesniak 2008), the
degree of their influence on Serbian EFL learners will be different compared to native
speakers. Furthermore, the study aims to determine whether Serbian EFL learners prefer
either of the two variants. Serbian is a case-marking language and does not use a
preposition to express transfer of possession with the verb dati “to give”; therefore, it can
be expected that Serbian EFL learners will tend to prefer the double object construction
due to L1 transfer. However, according to Processability Theory (Pienemann 2005) that
the structure which is easier to process will be acquired first, which is the prepositional
structure (Hawkins 1987; Mazurkewich 1984), Serbian EFL learners could show a
preference for the prepositional variant as well. These hypotheses are tested using a

1
questionnaire consisting of acceptability judgement questions administered to a group of
30 advanced Serbian EFL learners and to a control group of 30 native speakers of
American English.

Following the introductory chapter, this paper discusses the factors that influence the DA
in English in Section 2. Section 3 deals with the DA in the Serbian language while Section
4 presents an overview of the studies of the DA among EFL learners with different native
languages. Section 5 describes the methodology used in this study, and the results are
presented in Section 6. The final section offers concluding remarks.

2. Dative alternation in English

The dative alternation refers to two related argument structure realizations of ditransitive
verbs – the prepositional dative structure and the double object structure (Bresnan et al
2007). Dative verbs take two arguments whose roles are the recipient/ goal and the theme.
In the prepositional dative structure, the recipient is realized as a prepositional phrase
following the theme encoded as a noun phrase, as illustrated in sentence (1a). In the
double object construction, both arguments are realized as noun phrases whereby the
recipient precedes the theme, as in sentence (1b).

(1a) prepositional dative structure: Susan gave [toys] [to the children]. NP PP
(1b) double object structure: Susan gave [the children] [toys]. NP NP
(Bresnan et al. 2007: 70)
Different explanations for what influences the choice of the two dative structures can be
found in the literature on the topic. Early accounts of the DA state that according to the
morphological constraint alternating dative verbs tend to have Germanic stems, whereas
verbs of Latin origin with stress-final stems allow only the prepositional variant (Pinker
1989). Based on the semantic constraint proposed by Gropen et al. (1989) and Pinker
(1989), the two structural variants are associated with different meanings Namely, the
prepositional dative structure is related to a change of place (movement to a goal),
whereas the double object structure is associated with a change of state (possession)
(Bresnan et al. 2007: 71). The double object structure requires that the first object denotes
an animate entity, whereas the prepositional construction involves spatial goals. This
semantic constraint belongs to broad range (lexical) rules (BRRs), as the constraint on
animacy is considered universal (Pinker 1989). Furthermore, Wasow and Arnold (2003)

2
discuss another semantic constraint on the DA according to which idiomatic and fixed
expressions with dative verbs are usually restricted to either of the two variants, though
alternate variants are possible, but rare.

However, corpus data and evidence from the World Wide Web suggest that “semantic
explanations for the choice of dative constructions are not well founded empirically”
(Bresnan & Nikitina 2003: 2) and that the semantic constraints on the DA can be violated
with certain probabilities (Bresnan et al. 2007). The semantic restrictions can be
overridden by other factors, such as the pronominality of arguments (Green 1971) or the
principle of end weight, according to which the longer phrase follows the shorter phrase
(Wasow & Arnold 2003). Halliday (1970) identifies information structure as a factor on
the DA according to which given referents precede nongiven referents. This factor is
labeled as discourse accessibility by Collins (1995). He explains that, in the prepositional
construction, the goal contains new information while the theme carries given information
and is, at the same time, more discourse accessible. In the double object construction, the
recipient is more discourse accessible, carrying given information, whereas the theme
constitutes new information. Pronounhood is related to information structure, since
anaphoric pronouns refer to entities which are easily recoverable from the discourse, as
well as definiteness, as it serves to identify the referent based on its previous mention or
presence in the context.

Based on a corpus study using the Switchboard corpus of spoken English and with the
help of sophisticated statistical analyses, Bresnan et al. (2007) devise a model with 14
explanatory variables that can successfully predict the choice of the prepositional dative
construction. These predictors include:

semantic class1 + accessibility of recipient + accessibility of theme +


pronominality of recipient + pronominality of theme + definiteness of recipient +
definiteness of theme + animacy of recipient + person of recipient + number of
recipient + number of theme + concreteness of theme + structural parallelism in
dialogue + length difference (log scale) (Bresnan et al. 2007:78)

1
Semantic class refers to five classes of uses of verbs that occur in the dative alternation. These classes
are: abstract, abbreviates as ‘a’ (as in give it some thought), transfer of possession ‘t’ (give an armband),
future transfer of possession ‘f’ (owe, promise), prevention of possession ‘p’ (deny), and communication
‘c’(give me your name).

3
Bresnan et al. (2007) use logistic regression to determine the power of each of the
predictors. Their coefficients are given in Figure 1 from Bresnan et al. (2007: 81).
Negative coefficients signal that the predictor favors the double object construction, while
the positive coefficient favors the prepositional construction.

______________________________________________________________________
Probability{Response = 1} = 1/1+e−Xβ ,
where
Xβˆ = 0.95
− 1.34{c} + 0.53{f} − 3.90{p} + 0.96{t}
(a) + 0.99{accessibility of recipient = nongiven}
(a) − 1.1{accessibility of theme = nongiven}
(b) + 1.2{pronominality of recipient = nonpronoun}
(b) − 1.2{pronominality of theme = nonpronoun}
(c) + 0.85{definiteness of recipient = indefinite}
(c) − 1.4{definiteness of theme = indefinite}
(d) + 2.5{animacy of recipient = inanimate}
+ 0.48{person of recipient = nonlocal}
− 0.03{number of recipient = plural}
+ 0.5{number of theme = plural}
− 0.46{concreteness of theme = nonconcrete}
(e) − 1.1{parallelism = 1} − 1.2 · length difference (log scale)
and {c} = 1 if subject is in group c, 0 otherwise (and likewise for other categories).
______________________________________________________________________
Figure 1. The model for predicting the dative alternation according to Bresnan et al. (2007: 81).

The 14 factors have an irreducible effect on the DA across written and spoken modalities,
verbs senses, and different speakers of US English, while there is certain variation across
speakers of different varieties of English (Bresnan & Hay 2008). Animacy and discourse
accessibility, which are the focus of the present study, are among the factors with the
greatest influence on the DA among native speakers.

3. Dative alternation in Serbian

The Serbian language is a case-marking language, meaning that the syntactic functions
of arguments are determined by cases and the constituent order is flexible. Direct objects

4
are always marked with the accusative case, while indirect objects are marked with
oblique cases, such as genitive, dative, locative, or instrumental. The most frequent use
of the dative case is to indicate the indirect object of ditransitive verbs with the semantic
role of a recipient or benefactor when it refers to an animate entity, or a goal when it refers
to an inanimate entity (Piper & Klajn 2013). The following sentence illustrates the
argument structure of the ditransitive verb dati “to give”.

(2) Marko je dao Ani knjigu.

Marko-NOM give.PERF.3sgM.PAST Ana-DAT knjiga-ACC.sg

“Marko gave Ana a book.“

The verb dati “to give” does not allow the dative alternation as opposed to its English
equivalent. Although the recipient and theme arguments can change places, their case
marking remains the same. In fact, the order of arguments does not depend on
grammatical functions, but on their discourse status. In canonical sentences, the topical
constituent carrying given information comes before the focus of the sentence, that is, the
constituent carrying new information (Jasinskaja 2016). Therefore, canonical information
structure in Serbian and English appears the same, and it can be expected that Serbian
EFL learners will be sensitive to the constraints on the DA in English pertaining to
information status. Additionally, since the verb dati does not require a preposition to
denote transfer of possession in ditransitive constructions, it can be assumed that Serbian
EFL learners will tend to prefer the double object construction in English due to L1
transfer.

Further assumptions about what influences the DA among Serbian native speakers are
difficult to make since there has been no research into the DA in Serbian to date. The
study of the DA in the Croatian language by Zovko Dinković (2007) is worth considering
due to the similarity between Serbian and Croatian 2.

Based on data from the Croatian National Corpus, Zovko Dinković (2007) explains that
only three verbs, or eight, if their aspectual pairs are counted, allow the DA: darivati
(IMPF/ITER)/ darovati(PERF)/ podariti (PERF) “to give as a gift”, nuditi (IMPF)/
ponuditi (PERF) “to offer”, and služiti (IMPF)/ poslužiti (PERF)/ posluživati (ITER) “to

2
Croatian and Serbian were regarded as two components or varieties of a single language, Serbo-
Croatian, the official language of the Socialist Federal Republic of Yugoslavia until 1991.

5
serve”. In the canonical dative construction, the recipient argument is marked with the
dative case and the theme argument with the accusative case. In the alternated
construction, the recipient argument is marked with the accusative case and becomes the
direct object, whereas the theme argument receives instrumental case marking and
becomes the indirect object. The choice between the two variants mainly depends on the
affectedness of the object. Direct objects are the most affected; therefore, turning the
recipient argument into a direct object increases its affectedness. Information structure
and heaviness also influence the position of arguments, but these constraints are looser in
Croatian compared to English. Zovko Dinković (2007) concludes that alternated
constructions are much less frequent and that the effects of information structure,
heaviness, and various semantic constrains on the DA have not been fully explored.
Therefore, we now turn to studies of the DA among speakers of other languages in search
for more information on the topic.

4. Dative alternation in SLA

The influence of information status on the DA in learner language has received little
attention in SLA research and the few studies have yielded mixed results.

Tanaka (1987) investigated discourse constraints on the DA on three proficiency groups


of Japanese college students using acceptability judgements. The findings showed that
proficiency level had no effect on the results and that the subjects’ responses were not
influenced by discourse constraints. Chang (2004) examined the effects of information
structure on the DA within three groups of Chinese high- to mid-intermediate EFL
learners using written question-response tasks. The conclusion was that the learners were
not aware of information structure and that they preferred the prepositional dative
structure. Chang explained this preference based on the theory of markedness
(Mazurkewich 1984) stating that the prepositional variant was the unmarked variant,
easier to access and acquire than the double object variant. Furthermore, the study of L1
transfer on the DA in Russian-English interlanguage based on acceptability judgements
by De Cuypere et al. (2014) led to the conclusion that despite structural similarities
between L1 and L2, there was no transfer and the participants preferred the prepositional
construction, easier to process, which is in line with Processability Theory (Pienemann
2005).

6
On the other hand, based on question-answer and recognition tasks, Marefat (2005)
examined the DA in Persian-English interlanguage and found out that advanced and high-
intermediate learners were sensitive to information structure. Elementary learners
consistently preferred the prepositional variant, which he explained in terms of L1
interference, since Persian lacks the double object construction. Furthermore, based on a
learner-corpus study, Callies and Szczesniak (2008) proved that information status
influenced the DA in writing of advanced Polish and German EFL learners, while Jäschke
and Plag’s (2016) experimental study showed that German learners were influenced by
the same factors as native speakers, such as animacy of recipient, pronominality of theme
and definiteness of recipient, but to a lesser degree. The German learners also showed a
preference for the prepositional construction, though minimal.

Based on these studies, it can be inferred that the factors influencing the DA among native
speakers have different or no effects on nonnative speakers, but it seems that nonnative
speakers consistently tend to prefer the unmarked, prepositional construction. The
following sections of the present study deal with these issues among Serbian advanced
EFL learners.

5. Methodology

5.1. Aims and hypotheses

The aim of the present study is to examine the dative alternation in Serbian-English
interlanguage. The main research questions are as follows:

• Do Serbian advanced EFL learners prefer either of the two alternative dative
constructions and which one?

• Do the factors animacy of recipient, discourse accessibility of recipient, and


discourse accessibility of theme affect Serbian advanced EFL learners?

• Are the effects of these factors the same as on native speakers of American
English?

As previous studies of the DA in SLA show that nonnative speakers tend to prefer the
prepositional variant, our hypothesis is that Serbian learners will follow this trend. Since
constituent order in the Serbian language is influenced by animacy of recipient and
information structure, we expect that Serbian learners will be sensitive to the factors when

7
choosing the dative structure in English. However, as verb argument alternations are
highly language-specific phenomena, we assume that the investigated factors will affect
Serbian learners to a lesser degree. The effects of the factors are measured in terms of
coefficients obtained via ordinal logistic regression and compared with those given in
Bresnan et al. (2007).

5.2. Data collection

The data collection process in the present study was mainly based on Bresnan (2007) with
certain modifications. The experiment included 24 acceptability judgement tasks
consisting of a short passage with both dative constructions and a 1-9 scale. The
participants were asked to choose which of the two variants sounded more natural in the
given context by ticking a point on the scale, whereby 1-4 favored the double object
construction, 5 indicated that both variants were equally acceptable, and 6-9 favored the
prepositional construction.

The passages were initially excerpted from the COCA corpus and then modified to
manipulate the 14 constraints on the DA listed in Bresnan et al. (2007). The three
investigated factors were varied as to make all possible factor combinations, while the
remaining 11 predictors were kept constant. There were eight factor combinations and
three examples for each of them. The predictors were coded in the manner of Bresnan et
al. (2007):

• 1 if animacy of recipient = inanimate


• 1 if accessibility of recipient = nongiven
• 1 if accessibility of theme = nongiven
• 0 in all other cases

Recipients were considered animate only when they referred to human beings. If the
recipient or the theme were mentioned before in the passage, they were coded as given,
and if they were mentioned for the first time in the sentence containing the DA, they were
coded as nongiven.

Table 1 shows the factor combinations and their coding, as well as the total value of the
coefficients of these factors according to Bresnan et al. (2007). The coefficients were
exponentiated to calculate odds ratios in favor of the prepositional construction and sorted
in an ascending order.
8
Code Animacy Accessibility Accessibility Sum of Odds
of of recipient of theme coefficients ratios
recipient
1 001 0 0 1 -1.1 0.33
2 011 0 1 1 -0.11 1.32
3 000 0 0 0 0 1.43
4 010 0 1 0 +0.99 2.42
5 101 1 0 1 +1.4 2.83
6 111 1 1 1 +2.39 3.82
7 100 1 0 0 +2.5 3.93
8 110 1 1 0 +3.49 4.92
Table 1. Factor combinations, their coding, and their propensity to the prepositional construction in terms of
coefficients and odds ratios according to Bresnan et al. (2007).

All the tasks contained the verb to give in the sense of transfer of possession. The recipient
and the theme arguments consisted of two words – a determiner, that is, the definite article
or a possessive adjective like my, his, and a concrete noun in the singular form. This way,
the factors of length, measured in graphemic words, pronominality, definiteness, number
and person of recipient, number of theme, and concreteness of theme were kept constant.
All the passages contained only one instance of the DA as to eliminate the effect of
structural parallelism.

The search strings used to look for excepts in the COCA corpus involved the lemmatized
verb to give followed either by two noun phrases consisting of a determiner and a singular
noun or by a noun phrase, the preposition to and another noun phrase, both consisting of
a determiner and a singular noun. The excepts were shortened and modified. The final
form of the tasks can be found in the Appendix.

The tasks were then randomized and organized in the form of a questionnaire using
Google Forms. The questionnaire also included questions about participants’ background
information such as age, sex, native language, English proficiency level, number of years
of learning English, and time spent in an English-speaking country.

Prior to distributing the questionnaire to the target group, a pilot study was conducted to
confirm that the questionnaire was comprehensible and that it addressed the investigated
problem effectively.

9
5.3. Pilot study

5.3.1. Participants

The pilot study included two groups of participants. Group 1 consisted of 10 Serbian EFL
learners, 6 female and 4 male, whose average age was 28.2 and English proficiency level
C1 to C2. All of them obtained a bachelor’s degree or a master’s degree in English
language at the University of Niš, Serbia, and none of them spent more than 5 months in
an English-speaking country. Group 2 included 10 German EFL learners, 9 female and 1
male, whose average age was 28 and English proficiency level C1 to C2. They were either
students of English at the University of Bamberg, Germany, or English language
graduates who obtained their degree in Germany. None of the participants spent more
than 9 months in an English-speaking country.

5.3.2. Data analyses and results

The data analysis process was done in three stages. In the first stage, we obtained mean
scores for the 24 questions and the eight combinations of factors. The scores of both
groups are given in Table 2.
The findings point to certain differences in the mean scores per factor combination
between the Serbian and the German group, which implies that the observed factors do
not have the same effect on the two groups. In addition, the scores of both groups do not
align with the odds ratios determined according to Bresnan et al. (2007), meaning that the
influence of these factors on native speakers differs from that on the two groups. Our
findings regarding the German participants are in accordance with Jäschke and Plag
(2016) who claim that German EFL learners are influenced by fewer factors than native
speakers, one of them being animacy of recipient. Their study also demonstrates that
German EFL learners slightly prefer the prepositional dative construction. The mean
scores of the Serbian group are higher than those of the German group, except the final
two factor combinations; therefore, we can conclude that Serbian EFL learners will also
tend to prefer the prepositional variant.

10
Factor Odds ratios PP according to
Question Serbian German
combination Bresnan et al. (2007)

code001 Q5 5.9 4.1


Q17 4.5 5.7
Q20 5.3 4.3
Average 5.23 4.7 0.33
code011 Q8 6 4.4
Q13 4.8 5.1
Q22 5.8 3.4
Average 5.53 4.3 1.32
code000 Q2 4.9 4.7
Q10 5.6 4.6
Q12 6.2 7.6
Average 5.67 5.63 1.43
code010 Q4 6.9 6.7
Q15 6.5 7.2
Q19 7 5.7
Average 6.8 6.53 2.42
code101 Q3 4.7 5
Q11 6.3 7
Q23 5.8 4.3
Average 5.6 5.43 2.83
code111 Q1 6.4 6.7
Q9 4.9 3.3
Q14 7.1 7.9
Average 6.13 5.97 3.82
code100 Q6 7.4 7.2
Q18 6.9 8
Q21 7.2 7.4
Average 7.17 7.53 3.93
code110 Q7 5.4 5.2
Q16 6.5 7.9
Q24 7.1 7.5
Average 6.33 6.87 4.92
Table 2. Mean scores per question and factor combination for the Serbian and the German group in the pilot study.

In the second stage, we conducted a logistic regression analysis, using the statistical
software R, modeled according to Bresnan et al. (2007) to analyze the effects of the
investigated factors in greater detail. Since logistic regression requires that the dependent
variable be dichotomous, we converted the participants’ responses into binary values. All
the responses from 1 to 4 were converted to 0, the responses from 6 to 9 to 1, and the
responses 5 were excluded. The independent variables were coded in the same manner as

11
in Bresnan et al. (2007). The results of the analysis for the Serbian group are presented in
Table 3.

Predictor Coefficient p value Odds ratio PP


{accessibility of theme = nongiven} -1.02 < .05 0.36
{animacy of recipient = nongiven} 0.56 > 0.5 1.75
{accessibility of recipient = nongiven} 0.45 > .05 1.57
Table 3. Results of the logistic regression analysis for the Serbian group in the pilot study.

The results show that the coefficients and the odds ratios in favor of the prepositional
construction differ significantly between the Serbian group and Bresnan et al. (2007).
Only the predictor {accessibility of theme = nongiven} appears to be statistically
significant (p < .05), with a negative coefficient value, thus favoring the double object
construction. The other two factors do not seem to be statistically significant, but
{animacy of recipient = inanimate} has a greater effect size of 0.56 compared to
{accessibility of recipient = nongiven} 0.45, both in favor of the prepositional variant.
The direction of the effects of the three factors corresponds to that given in Bresnan et al.
(2007).

The results of the logistic regression analysis for the German group are considerably
different compared to the Serbian group, as can be seen in Table 4.

Predictor Coefficient p value Odds ratio PP


{animacy of recipient = inanimate} 1.04 < .05 2.82
{accessibility of theme = nongiven} - 1.33 < .05 0.26
{accessibility of recipient = nongiven} 0.17 > .05 1.18
Table 4. Results of the logistic regression analysis for the German group in the pilot study.

The German participants appear to be sensitive to the factors of animacy of recipient and
accessibility of theme, while accessibility of recipient is not statistically significant (p >
.05). The predictor {animacy of recipient = inanimate} has the greatest effect with the
coefficient of 1.04 and the odds ratio of 2.82 in favor of the prepositional variant. The
predictor {accessibility of theme = nongiven} favors the double object construction,
reducing the odds ratio for the alternative construction by 74%. The direction of the
effects for the German group also corresponds to that in Bresnan et al. (2007).

12
Finally, upon completing the logistic regression analyses, the data were further analyzed
using ordinal logistic regression in R, which was not applied in Bresnan et al. (2007).
Ordinal logistic regression was used as it is more appropriate than logistic regression
when the dependent variable is ordinal (Baayen 2008), such as our dependent variable
that ranges from 1 to 9. The predictors were coded in the same way as in the previously
conducted logistic regression analyses. Table 5 and 6 present the results of the ordinal
regression analysis for the Serbian and the German group, respectively.

Predictor Coefficient p value Odds ratio PP


{accessibility of theme = nongiven} -0.67 < .05 0.51
{animacy of recipient = nongiven} 0.38 > 0.5 1.47
{accessibility of recipient = nongiven} 0.24 > .05 1.28
Table 5.Results of the ordinal logistic regression analysis for the Serbian group in the pilot study.

Predictor Coefficient p value Odds ratio PP


{animacy of recipient = inanimate} 0.76 < .05 2.14
{accessibility of theme = nongiven} - 1.03 < .05 0.36
{accessibility of recipient = nongiven} 0.003 > .05 1.003
Table 6. Results of the ordinal logistic regression analysis for the German group in the pilot study.

The results of the ordinal logistic regression analyses do not differ significantly from the
results of the previously applied logistic regression analyses. Statistical significance
expressed by the value of p remains identical as well as the direction of the effects as
reflected in positive and negative coefficients. The values of individual coefficients are
slightly different but the relations between them are quite similar. Regarding the Serbian
group, the ordinal logistic regression analysis shows that the predictor {accessibility of
theme = nongiven} is the only statistically significant, favoring the double object
construction. Concerning the German group, the predictors {animacy of recipient =
inanimate} and {accessibility of theme = nongiven} are also statistically significant in the
ordinal regression analysis. The former favors the prepositional construction, whereas the
latter favors the other variant.

Since both types of analyses produced quite similar results, we decided to perform only
ordinal logistic regression on the data in the main study since this type of analysis is more
suitable considering the nature of our data.

13
5.4. Main study

5.4.1. Participants

The main study also included two groups of participants. Group 1 consisted of 30 Serbian
EFL learners, 23 female and 7 male, whose average age was 23.8, and English proficiency
level C1 to C2. The participants were third or fourth year of bachelor studies at the
University of Niš, Serbia, and none of them spent more than 4 months in an English-
speaking country. Group 2 was a control group consisting of 30 native speakers of
American English, 16 female and 14 male, whose average age was 22.5 and who studied
at the University of Bamberg, Germany, when the research was conducted.

5.4.2. Data analyses

The data analysis was carried out in two stages. In the first stage, we calculated and
compared the mean scores for the 24 questions and the 8 factor combinations for both
groups. In the second stage, we conducted ordinal regression analysis in R, following the
same method used in the pilot study. An additional analysis was performed to test whether
native language was a predictor that influenced the DA. Again, we used an ordinal logistic
regression analysis, and the predictor was coded as 0 if the native language was English
and 1 if the native language was Serbian.

6. Results and discussion of the results in the main study

The mean scores per question and factor combination for the Serbian and the American
group are given in Table 7. It can be noticed immediately that that the mean scores of the
Serbian participants are higher than the mean scores of the control group, except for the
factor combination labeled as code111. The fact that the Serbian learners rated the
prepositional construction higher in the acceptability judgement tasks than the native
speakers implies that they have a preference for this variant. In line with other SLA
studies of the DA, nonnative speakers in our study also preferred the unmarked
construction despite structural similarities between the double object construction and the
canonical argument realization of ditransitive verbs in the Serbian language. We believe
that Processability Theory (Pienemann 2005) explains this preference.

14
Factor Odds ratios PP according to
combination Question Serbian American Bresnan et al. (2007)
code001 Q5 6.13 4.13
Q17 5.63 6.1
Q20 5.6 5.3
Average 5.79 5.18 0.33
code011 Q8 5.73 5.1
Q13 6.9 4.9
Q22 4.63 5.1
Average 5.76 5.04 1.32
code000 Q2 5.73 4.5
Q10 4.43 4.1
Q12 6.27 5.1
Average 5.48 4.58 1.43
code010 Q4 6.63 6.43
Q15 7.23 6.36
Q19 6.3 5.2
Average 6.72 6 2.42
code101 Q3 5.9 3.77
Q11 5.77 5.53
Q23 6.13 6.23
Average 5.93 5.18 2.83
code111 Q1 6.33 7.23
Q9 4.3 4.1
Q14 7.4 7.43
Average 6.01 6.26 3.82
code100 Q6 7.43 6.73
Q18 6.87 6.2
Q21 6.53 6.77
Average 6.94 6.6 3.93
code110 Q7 6.97 5
Q16 6.53 6.1
Q24 6.9 6.03
Average 6.8 5.71 4.92
Table 7. Mean scores per question and factor combination for the Serbian and the American group in the main study.

Another important point to be made regarding the mean scores is that the scores of the
American group do not follow the ascending trend of the odds ratios calculated based on
the coefficients in Bresnan et al. (2007). Ideally, as the odds ratios in favor of the
prepositional construction increase, the mean scores for this variant should increase as
well, which is not the case in our study. This may point to a methodological problem
related to the modification of the passages used in the acceptability judgement tasks.
Namely, most of the original excerpts from the COCA corpus were shortened to keep the

15
questionnaire reasonably long, and some words were changed to keep certain factors
constant. In some cases, we replaced the indefinite article within the recipient or theme
noun phrase to manipulate the definiteness constraint, which might have led to a less
naturally sounding sentence and an unexpected rating from the native speaker. Also, some
of the recipient arguments coded as inanimate in our analyses could have been interpreted
as animate due to a slightly modified context and metonymy, such as school, university,
country. To prevent such problems, we suggest using authentic examples in their original
form without any modification. Since finding examples with a particular factor
combination might be time-consuming, it is more effective to test all the predictors
simultaneously, while focusing on their individual effects, although such procedures
could require more complex statistical analyses.

Regarding the effects of the investigated predictors in the present study, the results of the
ordinal logistic regression analyses indicate significant differences between the Serbian
and the American group. Tables 8 and 9 contain the results for the two groups
respectively.

Predictor Coefficient p values Odds ratio PP


{animacy of recipient = inanimate} 0.33 < .05 1.39
{accessibility of theme = nongiven} -0.38 < .05 0.69
{accessibility of recipient = nongiven} 0.20 > .05 1.23
Table 1. Results of the ordinal logistic regression analysis for the Serbian group in the main study.

Predictor Coefficient p values Odds ratio PP


{animacy of recipient = inanimate} 0.49 < .05 1.63
{accessibility of recipient = nongiven} 0.23 > .05 1.26
{accessibility of theme = nongiven} -0.19 > .05 0.82
Table 2. Results of the ordinal logistic regression analysis for the American group in the main study.

The findings indicate that the Serbian learners are influenced by the predictors {animacy
of recipient = inanimate} and {accessibility of theme = nongiven}, while {accessibility
of recipient = nongiven} is not statistically significant (p > .05). Inanimate recipients are
1.39 times more likely to occur in a prepositional construction, whereas a nongiven theme
decreases the odds for a prepositional variant by 31%. On the other hand, the native
speakers of American English in our study appear to be influenced only by the predictor
{animacy of recipient = nongiven}, whereby the coefficient is higher compared to the

16
Serbian group, while the other two examined factors are statistically insignificant (p >
.05). Nonetheless, the direction of the effects of the factors corresponds to those listed in
Bresnan et al. (2007) and the absolute value of the coefficient of animacy of recipient is
more than twice as high as those of accessibility of recipient and theme, which is also in
accordance with their study (see Figure 1.). A larger dataset might provide more precise
values of the coefficients which would reach statistical significance and would
correspond better to the findings in Bresnan et al. (2007).

Having these results in mind, it can be inferred that animacy of recipient influences the
DA in Serbian-English interlanguage, but the effect is lesser compared to the native
speakers. Regarding information status of recipient and theme, our study could not
confirm any effects on the native speakers, but the results show that accessibility of theme
has a significant effect on the Serbian learners. More research into information structure
in the Serbian language could help determine whether L1 interference might account for
the different influence of the investigated factors.

Finally, we conducted another ordinal regression analysis to test whether native language
was a predictor of the DA and the findings showed that, in fact, it was the strongest
predictor of those analyzed, as can be seen in Table 10.

Predictor Coefficient p values Odds ratio PP


{native language = Serbian} 0.42 < .05 1.53
{animacy of recipient = inanimate} 0.40 < .05 1.50
{accessibility of theme = nongiven} -0.29 < .05 0.75
{accessibility of recipient = nongiven} 0.22 < .05 1.25
Table 3. Results of the ordinal regression analysis with the additional predictor {native language = Serbian}.

Even though the Serbian participants are advanced EFL learners, the present study shows
that they have not fully acquired nativelike command of the dative alternation. Serbian
EFL learners should be instructed more thoroughly in the factors that influence the DA if
their use of English is to be closer to nativelike.

7. Conclusion

Inspired by Bresnan et al. (2007), this study examined the dative alternation and its
probabilistic constraints in Serbian-English interlanguage. The aim was to test whether
Serbian advanced EFL learners favored either the prepositional dative construction or the

17
double object construction, and whether animacy and information status influenced their
choice and to what degree compared to a control group of native speakers of American
English. The results showed that the Serbian learners preferred the prepositional
construction which can be explained by Processability Theory (Pienemann 2005).
Animacy of recipient proved to affect the Serbian learners but to a lesser degree compared
to the native speakers. The effects of discourse accessibility of recipient and theme on the
native speakers could not be confirmed, but the Serbian learners appeared to be influenced
by discourse accessibility of theme. Moreover, the study showed that the native language
had the strongest influence on the DA in Serbian-English interlanguage, but more
research in the Serbian language is necessary to confirm and describe L1 interference.
Since our findings suggested that even advanced EFL learners lacked near-nativelike
command of the DA, we suggest that this phenomenon should receive more attention
when it comes to English language teaching.

(6.109 words)

18
References

Baayen, R. Harald. 2008. Analyzing linguistic data: A practical introduction to statistics


using R. Cambridge University Press.
Bresnan, Joan & Tatiana Nikitina. 2003. The gradience of the dative alternation. Ms.,
Stanford University, http://www-lfg.stanford.edu/bresnan/download.html.
Bresnan, Joan. 2007. Is syntactic knowledge probabilistic? Experiments with the English
dative alternation. Roots: Linguistics in search of its evidential base. 75-96.
Bresnan, Joan & Jennifer Hay. 2008. Gradient grammar: An effect of animacy on the
syntax of give in New Zealand and American English. Lingua 118(2). 245–259.
Callies, Marcus & Konrad Szczesniak. 2008. Argument realisation, information status
and syntactic weight: A learner-corpus study of the dative alternation. In: P.
Grommes & M. Walter (eds.), Fortgeschrittene Lernervarietäten.
Korpuslinguistik und Zweitspracherwerbsforschung (Linguistische Arbeiten,
Band 520). Tübingen: Niemeyer. 165-187.
Chang, Lan-Hsin. 2004. Discourse effects on EFL learners’ production of dative
constructions. Journal of National Kaohsiung University of Applied Sciences 33
(2004). 145-170.
Collins, Peter. 1995. The indirect object construction in English: An informational
approach. Linguistics 33(1). 35-50.
De Cuypere, Ludovic, Evelyn De Coster & Kristof Baten. 2014. The acquisition of the
English dative alternation by Russian foreign language learners. PHRASIS
(GENT): Studies in language and literature (2).187-212.
Green, Georgia. 1971. Some implications of an interaction among constraints. Seventh
Regional Meeting, Chicago Linguistic Society. Vol. 7.
Gropen, Jess, Steven Pinker, Michelle Hollander, Richard Goldberg & Ronald Wilson.
1989. The learnability and acquisition of the dative alternation in English.
Language. 203-257.
Halliday, Michael A.K. 1970. Language structure and language function. In: John Lyons
(ed.), New horizons in linguistics. Harmondsworth: Penguin Books. 140–165.
Hawkins, Roger. 1987. Markedness and the acquisition of the English dative alternation
by L2 speakers. Interlanguage studies bulletin (Utrecht) 3(1). 20-55.
Jasinskaja, Katja. 2016. Information structure in Slavic. The Oxford handbook of
information structure.

19
Jäschke, Katja & Ingo Plag. 2016. The dative alternation in German-English
interlanguage. Studies in second language acquisition 38(3). 485-521.
Klajn, Ivan & Predrag Piper. 2013. Normativna gramatika srpskog jezika. Novi Sad:
Matica srpska.
Marefat, Hamideh. 2005. The impact of information structure as a discourse factor on the
acquisition of dative alternation by L2 learners. Studia linguistica 59(1). 66-82.
Mazurkewich, Irene. 1984. The acquisition of the dative alternation by second language
learners and linguistic theory. Language learning 34(1). 91-108.
Pienemann, Manfred (ed). 2005. Cross-linguistic aspects of Processability Theory. Vol.
30. John Benjamins Publishing.
Pinker, Steven.1989. Learnability and cognition: The acquisition of argument structure.
MIT press.
Tanaka, Shigenori. 1987. The selective use of specific exemplars in second-language
performance: The case of the dative alternation. Language learning 37. 63-88.
Wasow, Thomas & Jennifer Arnold. 2003. Post-verbal constituent ordering in English.
In: Günther Rohdenburg & Britta Mondorf (eds.), Determinants of grammatical
variation in English. Berlin: Mouton de Gruyter. 119-154.
Zovko Dinković, Irena. 2007. Dative alternation in Croatian. Suvremena lingvistika 63
(1). 65-83.

20
Appendix

21

You might also like