Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs

Abstract

Research has shown that English as a second language (ESL) learners find phrasal verbs (PVs) hard to acquire and often avoid these constructions. An often-neglected aspect in avoidance literature is the uncertainty of knowing the correct verb + particle pairing. The present paper investigated this by looking at Dutch ESL learners' knowledge of verb + particle pairings as well as the influence of the L1 on L2 lexical development. To this end, over 50 Dutch learners of English at two proficiency levels were tested and compared to native speakers on English PVs that either (mis)match in form and meaning as compared to the Dutch counterpart. Results indicate that learners make more mistakes when the L1 and L2 differ in form and meaning. L1 interference is observed as learners overextend Dutch versions into English. This interference need not fossilise as advanced learners perform more like native speakers. Directions for future inquiries are provided as well.

Full text

Linguistics in Amsterdam, 13,2 (2020): 24–42
© Thomas B. Tienkamp
Help or hindrance? The influence of Dutch on the
acquisition of English phrasal verbs*
Thomas B. Tienkamp
ACLC, University of Amsterdam
Research has shown that English as a second language (ESL) learners find phrasal
verbs (PVs) hard to acquire and often avoid these constructions. An often-neglected
aspect in avoidance literature is the uncertainty of knowing the correct verb +
particle pairing. The present paper investigated this by looking at Dutch ESL
learners’ knowledge of verb + particle pairings as well as the influence of the L1
on L2 lexical development. To this end, over 50 Dutch learners of English at two
proficiency levels were tested and compared to native speakers on English PVs that
either (mis)match in form and meaning as compared to the Dutch counterpart.
Results indicate that learners make more mistakes when the L1 and L2 differ in
form and meaning. L1 interference is observed as learners overextend Dutch
versions into English. This interference need not fossilise as advanced learners
perform more like native speakers. Directions for future inquiries are provided as
well.
1 Introduction
Inquiries targeting the understanding of the bilingual lexicon typically investigate
how L2 learners access the meaning or form of L2 lexical items. A case where
syntax and lexicon might come together is the phrasal verb. Phrasal verbs
(henceforth: PVs) are defined in this paper as a verb + particle combination in
which both elements give up meaning to form one new lexical element (e.g. give
up) (Darwin & Gray 1999). Second Language Acquisition (SLA) research has
pointed out the difficulty PVs pose for L2 learners. It has been suggested that this
difficulty is in part due to them being a peculiarity of the Germanic languages,
and in part due to their idiomatic and polysemous meanings (White 2012). A
seminal study by Dagut & Laufer (1985) found that Hebrew learners of English
avoided phrasal verbs constructions; learners opted for one-word verbs instead. A
* I am indebted to Michelle Nekeman for her help in recruiting participants. I am also thankful
to Andrew Mckeown, Penelope Goemans, Parker van Nostrand, and Gautam Ottur for helping
me create the stimuli. Lastly, I am grateful for Eva van Lier’s and Jan Hulstijn’s valuable
comments on earlier versions of this paper.

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 25
Linguistics in Amsterdam 13,2 (2020)
follow up study by Hulstijn & Marchena (1989) showed that Dutch ESL learners
do not avoid PVs categorically, but only if they perceive the PV too Dutch-like.
In this case, too, Dutch-like was defined as a literal translation of the Dutch form
(e.g. opgeven → give up). However, Hulstijn & Marchena (1989) did not focus
on whether the learner knew the right verb + particle combination. This raises two
questions. First, it raises the question as to whether the avoidance can in part be
explained due to the learner’s uncertainty about the right verb + particle
combination. Second, the observation that learners deemed the construction too
Dutch-like raises the question of whether typological similarity aids or hinders
PV acquisition. To answer these questions, over 50 Dutch ESL learners were
tested on PVs that can be directly translated to English (e.g. opgeven → give up)
and PVs that have different verb + particle combinations (richten op → aim at).
The paper is structured as follows. Section 2 introduces Jiang’s model of
the bilingual lexicon (2000) and relevant studies; Sections 3 and 4 introduce the
present study and its methodology; section 5 presents the results; section 6
provides discussion; and section 7 concludes the paper and also provides
directions for future inquiries.
2 Theoretical background
2.1 Theoretical models
The Declarative/Procedural model of language (DP model) posits that language
is subserved by two memory systems (Ullman 2001, 2004). The declarative
memory system, rooted in temporal lobe structures, subserves the lexicon,
including the meaning and sounds of words, and idiosyncratic facets of language
(e.g. irregular verbs and plural forms). The procedural system, on the other hand,
which is rooted in frontal/basal ganglia structures, is important in the learning of
habits, computation, and rule-governed aspects of language (e.g. the grammar).
The model posits that late learners rely more heavily on the declarative memory
system than on the procedural system as the latter may be subject to critical period
effects. The former, on the other hand, is not.
Lexical development thus takes place within the declarative memory
system. This paper follows de Bot (1992) in assuming that phrasal verbs are
lexical in nature since lemmas are activated based on the meaning of the preverbal
message. This preverbal message has to then be encoded into words drawn from
the lexicon.
Lowie & Verspoor (2004) note that a question relevant to the bilingual
mental lexicon is whether and to what extent L2 learners make use of the lexical
knowledge from their L1 in the acquisition and use of the L2. Jiang (2000)
provides an interactive model of the bilingual lexicon and postulates that there are

26 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
three separate stages: (1) the formal stage, (2) the L1 lemma mediation stage, and
(3) the L2 integration stage (illustrated in Figure 1, taken from Jiang 2000: 54).
Figure 1: Lexical development from the formal stage (left) to the integration stage (right).
It is important to note the difference between lexemes and lemmas. The lemma
contains the semantic and syntactic information (the top level in Figure 1) whereas
the lexeme contains the formal and morphological information (bottom level in
Figure 1). To illustrate this, the word forms (lexemes) ‘give up’, ‘giving up’, and
‘gave up’ are all part of the overarching lemma GIVE UP.
In the formal stage, a lexeme with formal specifications is established. This
lexeme is mapped onto an L1 meaning and there is no direct link from the L2
lexeme to the preverbal message. Then, in the lemma mediation stage, the
semantic and syntactic information from the corresponding L1 lemma is copied
onto the L2 lemma. The L1 thus mediates a link between the L2 lexical item and
the preverbal message. The learner uses their L1 semantic and syntactic
knowledge to use the L2 lemma which can lead to transfer mistakes. This stage
differs from the formal stage because there is a lemma. In the formal stage, there
is no information at the lemma level, only at the lexeme level. After sufficient
exposure and experience, the L2 information is integrated on the L2 lemma. Now,
the L1 information is no longer used and a direct link between the preverbal
message and the L2 lemma exists. Once the learner has done so, they have native-
like knowledge of this particular lemma.
Jiang (2000) posits that fossilisation of lexical development is indicated by
the learner not advancing beyond the second stage. Furthermore, it is postulated
that many lexical items may fossilise due to the existence of an L1 lexical and
conceptual system. Lowie & Verspoor (2004) also note how this model allows
each lexical item to be in a different developmental stage. For example, word A
might be fully developed, whereas word B still remains in the lemma mediation
stage.
2.2 Relevant studies in PV acquisition
As mentioned in the introduction, most of the research on PV acquisition has
focused on avoidance. In this line of research, avoidance was operationalised as
the learner’s preference for the PV or one-word-verb equivalent in multiple choice

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 27
Linguistics in Amsterdam 13,2 (2020)
or translation tasks (Dagut & Laufer 1985; Hulstijn & Marchena 1989; Laufer &
Eliasson 1993; Liao & Fukuya 2004). Results indicated that learners with a higher
proficiency showed less avoidance than learners with a lower proficiency level.
There have also been large scale corpus analyses investigating the rate at which
learners use PVs as compared to natives. Siyanova & Schmitt (2007) analysed
spoken and written corpora of native speakers and ESL learners on their use of
PVs. Their results suggest that ESL learners use the one-word-verb equivalent at
higher rates than the native speakers do in both spoken and written modalities.
This finding supports the previously established avoidance patterns in
experimental research as the results mirror each other.
Besides proficiency, typological distance has also been noted as a reason
for avoidance. Learners with a Germanic language as their L1 (e.g. Dutch or
Swedish) showed considerably less avoidance than learners with a non-Germanic
language as their L1 (e.g. Hebrew) (Dagut & Laufer 1985; Hulstijn & Marchena
1989). This already suggests crosslinguistic influence. Another indication of L1
influence comes from the observation by Hulstijn & Marchena (1989) that Dutch
learners avoid PVs only if they perceive them to be too Dutch-like (e.g. opgeven
→ to give up). These ‘too Dutch-like’ items are cognates, words whose form and
meaning considerably overlap across two languages (Otwinowska & Szewczyk
2019). Psycholinguistic research has found that cognates are retrieved, translated,
and recognised faster than non-cognates (Friel & Kenison 2001; Jacobs, Fricke &
Krol 2016; Lemhofer & Dijkstra 2004). Furthermore, cognates are more resistant
to forgetting, suggesting that they might be easier to learn (De Groot & Keijzer
2000). Indeed, studies have found that cognates are easier to learn than non-
cognates (cf. Otwinowska & Szewczyk 2019; for a summary, see De Groot 2011).
However, experimental data directly investigating this L1 influence on PV
acquisition is currently missing.
L1 influence and/or interference has been noted in other lines of research,
too. Lowie & Verspoor (2004) analysed Dutch ESL learners on their knowledge
of prepositions. In their experimental material, they used prepositions that were
either similar or different in meaning compared to their Dutch counterpart, and
used prepositions with varying degrees of frequency. Their results showed that up
until high proficiency (first year university students), Dutch interfered with their
usage of English prepositions. Learners overextended the Dutch preposition
meaning to English resulting in errors like ‘I am in the train’ instead of ‘on the
train’. A main effect of frequency was found, too. Higher frequency prepositions
were performed better on than lower frequency prepositions. Similar results have
been reported by Ijaz (1986), who found that advanced adult ESL learners differed
significantly from native speakers with regards to the semantic boundary they
ascribed to spatial prepositions. Crucially, the semantic boundary was influenced
by the L1 as ESL learners based the L2 semantic boundary on their L1.

28 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
3 The present study: Research questions and predictions
On the basis of this background section, it becomes apparent that direct
experimental data regarding the L1 influence on the acquisition of PVs is missing.
Furthermore, data regarding the learner’s knowledge of correct verb + particle
combinations is missing as well. Although previous research has suggested that
avoidance may also signal this uncertainty, it cannot be taken as fact as Hulstijn
& Marchena (1989) did not directly include this in their experimental design. The
present study will fill these gaps. The research questions can thus be stated as
follows: to what extent does crosslinguistic (dis)similarity influence the
acquisition of English phrasal verbs? Do learners reach native-like levels? Do
phrasal verbs fossilise in L2 speech? Lastly, what is the role of frequency in PV
acquisition? To this end, two groups of Dutch learners of English were tested on
PVs that can be directly translated to English, called match items (e.g. opgeven
→ give up) and PVs that have different verb + particle combinations, called
mismatch items (richten op → aim at).
Based on the model of Jiang (2000), the results of Lowie & Verspoor
(2004), and the cognate literature (De Groot 2011), it is predicted that match items
will be easier for learners than mismatch items. Initially, the L1 lemma is copied
onto the L2 lemma. As a corollary, it is predicted that mismatch items will pose
greater difficulty. This prediction thus goes against Hulstijn & Marchena (1989),
who found that match items resulted in more avoidance.
Also based on the model of Jiang (2000), as well as the results of
experimental research, it is predicted that advanced learners will score higher than
intermediate learners. The advanced learners have received more input and have
more experience with the target language. These aspects will help the lemma
move from the lemma mediation stage to the lemma integrations stage.
Based on the DP model of language (Ullman 2001, 2004), it is predicted
that learners do reach native-like levels and perform more like native speakers
than high school students. No decrease in learning ability of the declarative
memory system, the system that subserves the lexicon, has been noted and PVs
are assumed to be lexical in nature (de Bot 1992). Furthermore, it is predicted that
the initial L1 interference will not fossilise due to the absence of a decrease in
learning ability. This prediction is thus contra Jiang (2000), who predicts that
many lexical items may fossilise.
Lastly, based on Lowie & Verspoor (2004), it is predicted that participants
will score higher on higher frequency items than lower frequency items.

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 29
Linguistics in Amsterdam 13,2 (2020)
4 Methods
4.1 Participants
Three groups of participants were recruited online for the present study: (1) Dutch
high school students in their fifth or sixth year of pre-university education
(VWO); (2) Dutch university students who have followed an English taught
degree for at least three years; and (3) a group of native speakers of English. All
but one speaker was currently enrolled at a university or had received a university
degree in the past. Following Hulstijn & Marchena (1989) and Lowie & Verspoor
(2004), it is assumed that the high school learners resemble intermediate learners,
and university students advanced learners. In total, 68 participants were recruited.
Group distribution and details can be found in Table 1, where AoA stands for age
of acquisition. Participants were recruited via Facebook advertisements, as well
as contacts at high schools in Leek and Groningen, The Netherlands.
Table 1: Participant details.
Group
N
Age in Years (SD)
Range
AoA in Years (SD)
Range
Proficiency
High School
31
M = 17.1 (0.7)
R =16–19
M = 9.3 (2.1)
R = 6–12
Intermediate
University
23
M = 22.8 (2.2)
R = 20–28
M = 9.0 (2.1)
R = 6–12
Advanced
Native
141
M = 27.14 (7.6)
R = 20–45
N/A
Native
4.2 Materials
4.2.1 LEAP-Q questionnaire
A selection of the Dutch version of the LEAP-Q questionnaire
2
(Marian,
Blumenfeld & Kaushanskaya 2007) was imported into Qualtrics (Qualtrics 2019).
The LEAP-Q questionnaire provides detailed information regarding the
participant’s language background, such as age of acquisition (AoA), years of
education (YoE), and the exposure to the languages the participant speaks. The
included questions can be found in Appendix A.
1
Six were American English speakers, five were Australian English speakers, two were English
English speakers, and one was a Scottish English speaker.
2
Translated by Lisa Vandeberg, Erasmus University, adapted by Sarah von Grebmer zu
Wolfsthurn, Niels Schiller, and Leticia Pablos Robles, Leiden University.

30 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
4.2.2 Phrasal verb test
A phrasal verb test consisting of 30 items was created in Qualtrics and aimed to
test the participant’s knowledge of the correct verb + particle pairing of English
phrasal verbs. The influence of the L1 was tested by 15 match items, in which the
verb + particle can be literally translated from Dutch to English (e.g. opgeven →
to give up), and 15 mismatch items where the English construction does not mirror
the Dutch one (ophangen → *hang up, but put up). Both conditions were matched
for word frequency. Word frequency was calculated using the Corpus of
Contemporary American English (COCA; Davies 2008) using the same method
as Liu (2011). This was done by entering the verb lemma plus the particle. The
second step was a search of the tokens of transitive PVs used with their particles
separated by one or two intervening words. No queries were run for verbs that
were separated from their particle by three or more words, as Gardner & Davies
(2007) had noted that this results in many false PVs. A complete list of the stimuli
plus their frequency per million words can be found in Appendix B. The average
frequency per condition is given in Table 2.
Table 2: Average frequency of the stimuli per condition.
Condition
Frequency per million words (SD)
Match
22.25 (31.27)
Mismatch
24.17 (37.77)
Participants were given a sentence with a blank and two options regarding verb
and particle choices. An example is provided in (1).
(1) Denice is _____ posters all over town for the party she’s organising.
A. putting up
B. putting on
C. hanging up
D. hanging on
In the end, 16 match stimuli and 13 mismatch stimuli were included in the
analysis. One stimulus turned out not to be a mismatch, and was moved to the
match stimuli. A second item (chase up) was excluded as native speakers
indicated that three out of the four options were acceptable. Stimuli were
presented in six fixed blocks of five questions. Blocks were presented in a
randomised order and items within blocks were randomised as well. Lastly, all
the possible answers were presented in a randomised order as well to prevent any
ordering effects.

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 31
Linguistics in Amsterdam 13,2 (2020)
4.3 Procedures
All participants participated online at a time of their choosing and gave their
informed consent. They were first presented with the LEAP-Q questionnaire
(approx. 5 minutes) and then with the PV-test (approx. 10–15 minutes).
Participants were not rewarded with the correct answers. The answers of the
native speakers were assumed to be the correct answers to the questions. In cases
where more than 10% of the natives chose for an unintended answer, it was
marked as correct as well. This led to the exclusion of the item chase up as natives
were almost equally divided over three possible answers. There were no
ambiguities regarding the correct answer for all other items. The tests of the ESL
learners were marked afterwards. Their answers were scored as correct or
incorrect. Additionally, when the answer was incorrect, it was indicated whether
the mistake was transfer-based or non-transfer based. Transfer mistakes were
regarded as literal translations of mismatch items (e.g. hang up instead of put up).
In this example, the transfer results from an overextension of the lexical verb.
However, transfer could also result from an overextension of the particle (e.g.
come over as instead of come off as).
4.4 Analyses
All statistics were carried out using R and the lme4 and lmerTest packages (Bates,
Maechler, Bolker & Walker 2015; Kuznetsova, Brockhoff & Christensen 2017;
R-core team 2018). In order to test whether there were group differences for the
reported scores for proficiency and exposure to English, independent samples t-
test were run with group as the independent variable and reported proficiency and
exposure as the dependent variable.
In order to test where advanced learners are as compared to intermediate
learners and natives on the PV-test, a linear mixed effects model was fitted with
percentage correct as a function of the ternary predictor group (High school,
University, and Native), and the binary predictor condition (Match and
Mismatch).
3
A second linear mixed effect model was fitted with percentage
correct as a function of the binary predictors group (High school and University)
and condition (Match and Mismatch) to investigate group differences. Categorical
predictors were coded with sum-to-zero orthogonal contrasts. A linear mixed
effects model was opted for instead of a generalised linear mixed effects model
due to the difference in condition sizes (13 items in mismatch, 16 items in match).
3
Other models were run as well with predictors from the LEAP-Q questionnaire. These
predictors included: age of acquisition, years of education, reported proficiency, reported
exposure. None of these predictors improved the models, as predictors were strongly correlated.
They were therefore left out.

32 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
The statistic that will answer the research question is the interaction between
group and condition.
To still analyse the effect of frequency, as this was not possible using
aggregated scores in the linear model, a generalised mixed effects model was
fitted with correct as a function of the binary predictors group (High school and
University) and condition (Match and Mismatch), and the continuous predictor of
frequency per million words
Next, an error analysis was carried out. An independent samples t-test was
run to check for a group difference in amount of transfer mistakes. Lastly, two
paired t-tests were carried out, one for each group, to test whether participants
made more transfer mistakes than non-transfer mistakes.
5 Results
Descriptive results of both the LEAP-Q questionnaire and the PV-test can be
found in Table 3. These results (in percentages) are visualised in a boxplot in
Figure 2.
Table 3: Descriptive statistics Experiment 1.
Group
Exposure to
English (in %)
Proficiency
(From 1-10)
Total score
(Max 29)
Match score
(Max 16)
Mismatch score
(Max 13)
High School
24.84 (16.2)
7.16 (0.94)
20.5 (5.2)
12.9 (2.8)
7.6 (2.8)
University
39.9 (14.9)
8.97 (1.67)
27.6 (1.1)
15.7 (0.57)
12.0 (1.0)
Native
N/A
N/A
28.7 (0.61)
15.86 (0.53)
12.86 (0.36)
The independent samples t-tests investigating group differences in reported
exposure and proficiency were both significant. The university students in the
experiment reported significantly higher English proficiency than the high school
students in the experiment (estimated group difference = 1.8 points; 95% CI =
1.26…2.36 points; p < 0.001). They also reported significantly higher exposure
to English than the high school students (estimated group difference = 14.86%;
95% CI = 8.89…20.83%; p < 0.001).

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 33
Linguistics in Amsterdam 13,2 (2020)
Figure 2: Percentage correct on the PV-test across conditions and groups.
Dots represent the individual scores of participants.
The linear mixed effects model investigating whether advanced learners are more
like native speakers than intermediate learners was significant. Advanced learners
scored significantly, on average, 10.66% higher than the average of native
speakers and intermediate learners combined (95% CI = 4.26%…17.06%; p =
0.002). It is therefore concluded that advanced learners are more like native
speakers than intermediate learners. As a sanity check, it was checked whether
native speakers performed better than intermediate learners. Native speakers
performed significantly better than intermediate learners (estimated group
difference: 29.5%; 95% CI: 21.7%…37.3%; p < 0.001).
The linear mixed effects model investigating group differences on the PV-
test was also significant. There were significant effects for group, condition, and
a significant interaction between group and condition. The university group
scored significantly higher than the high school group (estimated group difference
= 25.42%; 95% CI = 17.91%…32.93%; p < 0.001). Participants averaged across
groups scored significantly higher on match trials than mismatch trials (estimated
condition difference = 14.09%; 95% CI = 10.45%…17.74%; p < 0.001). The
interaction indicated that high school students scored an additional 16.48% lower
on mismatch trials than on match trials when compared to university students
(95% CI = 9.19%…23.77%; p < 0.001). It is therefore concluded that the group
difference is larger in the mismatch condition than in the match condition.

34 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
As for the effect of frequency, the generalised mixed effects model showed
a significant main effect. Participants across groups were 1.033 times more likely
to pick the correct verb + particle combination on PVs with a higher frequency
than PVs with a lower frequency (95% CI = 1.003…1.064 times more likely; p <
0.05).
4
The descriptive statistics for the error analysis can be found in Table 4 and
is visualised in a boxplot (in percentages) in Figure 3. There was a significant
group difference. High school students made 2.4 more transfer mistakes on
average as compared to university students (95% CI = 1.48…3.32 points; p <
0.001).
Table 4: Error distribution of the learner groups.
Group
Transfer mistakes
Non-transfer mistakes
High School
3.55 (1.70)
2.07 (1.81)
University
1.13 (0.64)
0.4 (0.51)
Figure 3: Error distribution of the learner groups.
Dots represent the individual scores of participants.
4
Generalised mixed effect models yield a log-odds outcome. Therefore, the outcome was
exponentionalised and interpreted as X- times more/less likely than.

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 35
Linguistics in Amsterdam 13,2 (2020)
Paired t-tests indicated that both groups were more likely to make transfer
mistakes than non-transfer mistakes. High school students made an estimated 1.47
more transfer mistakes than non-transfer mistakes (95% CI = 0.62…2.31
mistakes; p < 0.001). University students made an estimated 0.73 more transfer
mistakes than non-transfer mistakes (95% CI = 0.24…1.22 mistakes; p < 0.001).
6 Discussion
The aim of the present study was to answer two questions. The main question was
whether typological similarity aids or hinders PV acquisition. This question was
addressed from a developmental perspective by looking at intermediate learners
(high school students) and advanced learners (university students). Based on the
model of Jiang (2000) and the DP model of language (Ullman 2001, 2004), it was
predicted that (1) university students should perform better than high school
students in both conditions, (2) that mismatch items should pose the largest
difficulty for both groups, (3) that there should be more transfer than non-transfer
mistakes, (4) that lexical overextension need not fossilise provided that the learner
receives enough input, and (5) that learners would perform better on higher
frequency items than lower frequency items.
Differences were found in overall ability to correctly pick the right verb +
particle pairing, with university students performing better than the high school
students, and participants from both groups performing better on match trials than
mismatch trials. There was also considerably less variation in the scores of
university students as compared to high school students as evidenced by the small
standard deviations. Additionally, participants from both groups were more likely
to overextend a Dutch form into English than making a different kind of mistake
in mismatch trials. These results confirm predictions (1–3). It can be concluded
that L1 Dutch hinders the acquisition of L2 PVs where there is a form meaning
mismatch. The error analysis and the discrepancy between match and mismatch
trials suggest that learners assume, albeit unconsciously, that the semantics of the
L1 and L2 forms align. The results thus support the notion of lemma mediation in
Jiang’s (2000) model. Input and experience is needed to learn and integrate the
L2 semantic and syntactic information for the L2 lemma. This is further reflected
by the significant effect of frequency. Learners have more experience with and
come across higher frequency items more often, thus having more opportunity to
integrate the L2 specific information. This main finding, albeit a small one,
supports prediction (5).
Lemma mediation need not be a hindrance in all cases, though. Dutch seems
to aid PV acquisition in cases where form and meaning match. In these instances,
integration is not necessary as target-like production does not depend on L2
lemma integration. This is an interesting finding, as Hulstijn & Marchena (1989)

36 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
argued that Dutch ESL learners refrain from using PV constructions that have
identical Dutch counterparts because they are too ‘Dutch-like’. So, while L1-L2
similarity might hinder the actual use of the construction, the results of the present
study suggest that similarity is helpful in the acquisition stage. The study thus
supports the prediction made by De Groot & Keijzer (2000) that cognates might
be easier to learn.
Results are also in line with previous studies that found that groups with a
higher proficiency performed more native-like (Hulstijn & Marchena 1989; Liao
& Fukuta 2004; Lowie & Verspoor 2004). Furthermore, Lowie & Verpoor (2004)
reported a similar discrepancy between form meaning (mis)matches in L2
prepositions for Dutch learners of English. They reported higher accuracy when
there was L1-L2 similarity. This effect was not found for the group with the
highest proficiency, who performed well in both conditions. The present study
mirrors this finding with the significant interaction between group and condition.
The results also provide support for prediction (4). The significant
interaction between group and condition suggests that advanced learners managed
to bridge the initial gap between conditions. Advanced learners performed almost
equally well on both conditions. Jiang (2000) predicts that many lexical items
fossilise, meaning that the learner remains in stage two, the lemma mediation
stage. This prediction was not borne out by the data in the present study. The DP
model (Ullman 2001, 2004), on the other hand, posits that the lexicon is subserved
in the declarative memory system. This memory system is not constrained by
maturational constraints like the procedural system that subserves grammar and
computation. Thus, the DP model predicts that there may be L1-L2 differences in
the level of attainment of grammar, but not necessarily in the lexicon. The results
of the present study support the DP model in this aspect as the advanced learners
performed significantly more like native speakers than intermediate learners.
However, it should be noted that the few mistakes advanced learners made, were
mostly transfer-based. It should also be noted that this significant difference was
based on a mean difference of only 0.7 as advanced students made, on average,
1.14 transfer mistakes versus 0.4 non-transfer mistakes (see Table 4 and Figure
3). So, although there is no considerable fossilisation in terms of number of
mistakes, it is hypothesised that the mistakes that do remain show that the learner
is still in stage 2 for those specific items. This also reflects the flexibility of Jiang’s
(2000) model, as different lexical items may be in different developmental stages.
It has been suggested in the literature that compositional PVs, where the
meaning of the PV equals the combined meaning of its components, are built
online (González 2010). In this instance, the PV would not be stored as a whole
in the lexicon but put together by means of computation. As computation is
subserved by the procedural memory, it could be the case that these items are

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 37
Linguistics in Amsterdam 13,2 (2020)
prone to fossilisation. However, it was beyond the scope of the paper to
investigate this more closely. Future inquiries could look into this.
The second question was whether PV avoidance can in part be explained by
the uncertainty of knowing the correct verb + particle pairing. As mentioned in
the introduction, most PV studies focused on preference, not performance (see
Dagut & Laufer 1985; Laufer & Eliasson 1993; Hulstijn & Marchena 1989). They
posit that the learners know these constructions, either by a short test of whether
the learner understands the PV, or by presupposing their knowledge. However,
Jiang (2000) notes that knowing a word(string) need not imply that the learner is
also able to use the construction correctly. Very often, learners’ receptive
vocabulary is larger than their productive vocabulary (Schmitt 2010). The results
of the present study indicate that the intermediate learners have difficulty with
choosing the right verb + particle pairing, failing to pick the correct pairing in
30% of the cases. This is also the group where significant differences in
preference have been reported when compared to native speakers (Hulstijn &
Marchena 1989). The advanced learners in Hulstijn & Marchena (1989) did not
differ significantly from native speakers in their preference. In the present study,
the advanced learners showcased extensive knowledge of the correct verb +
particle pairings, scoring over 90% in both conditions. It could thus very well be
the case that part of the avoidance that is reported is due to the uncertainty of not
knowing the right construction. Although this hypothesis is validated by cross
experiment comparison, within participant experimental data is needed in order
to accept or reject this hypothesis. Experimental designs of Hulstijn & Marchena
(1989) and the present study may be combined in order to do so in future inquiries.
Although the present study has shed light on the influence of the L1 in L2
lexical development, some limitations must be noted, as well. Two kinds of
limitations will be discussed: (i) design limitations and (ii) data limitations.
As for the design limitations, the present study consisted of one multiple-
choice task in a in a cross-sectional, non-longitudinal design. Previous studies
comprised the administration of multiple tasks, thus reducing the chance of task-
bias. The results of the present study might have been influenced by the task
choice. Furthermore, evidence, or the lack thereof, of fossilisation ideally comes
from longitudinal research. However, due to time-constraints, a cross-sectional
design was opted for instead.
Secondly, even though the present study improved on previous work by
including a proficiency measure, it can still be seen as a limitation. Self-reported
proficiency is limited in the sense that it remains doubtful whether non-experts
can correctly assess their proficiency. It is therefore recommended that future
studies employ an objective measure of proficiency and use multiple tasks using
a longitudinal design that probe both language knowledge and use.

38 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
As for the data limitations, there was a lack of experimental control as the
data was collected online. It could have been the case that participants (1) looked
up the right answers or (2) did not try and filled in random answers. Although the
first cannot be entirely ruled out, the results do indicate that participants filled in
the test in earnest. As for latter, participation was voluntary and therefore seems
unlikely. Additionally, the time it took the participants to fill in the online test did
not suggest that participants rushed through it. The time further suggests that
participants did not look up the right answers, as this would have increased their
time significantly.
The second data limitation is the frequency data. Some of the used verbs
were polysemous, and the corpus that was used is not tagged for semantic content.
Moon (1998) suggests that around five percent of phrasal lexemes are
polysemous. Thus, the attained frequency numbers are for all the different
meanings that each lexeme has. This means that the frequencies reported in
Appendix A do not correspond one-to-one with the meaning of the PV, but rather
give an indication of the verb’s frequency in general. However, this limitation is
only secondary, as frequency was not of primary interest in the present study.
7 Conclusion
This paper has made a contribution to the avoidance literature on PVs by
examining the learner’s knowledge of the correct verb + particle pairing.
Additionally, this paper examined the role of the L1 in L2 PV acquisition. It has
been argued that the L1 may be of help when L1 and L2 align in form-meaning
mappings. When they do not, the L1 may be a hindrance. The results reflect this
as participants scored higher on match items than mismatch items. However, the
results also indicate that this hindrance need not be long term. Initial transfer
mistakes do not fossilise in L2 speech. The results can thus be explained by the
model of Jiang (2000), who proposed lexical development in terms of L1 lemma
mediation and L2 information integration. However, contra Jiang (2000), the
present paper showed that lexical items do not fossilise as often as initially
proposed. The expected non-fossilisation was motivated by the division of lexicon
and grammar in the DP model instead (Ullman 2001, 2004).
Two fruitful directions for future inquiries have been noted in the
discussion. First, inquiries may tackle the distinction between compositional and
non-compositional PVs in terms of lexical fossilisation. Second, the
methodologies of traditional avoidance research and the present study may be
combined to directly assess the role of not knowing the verb + particle pairing in
avoidance through the administration of multiple tasks that probe language
knowledge as well as language use. This would ideally be investigated in a
longitudinal design. Future inquiries will further elucidate why ESL learners find

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 39
Linguistics in Amsterdam 13,2 (2020)
them so challenging. Understanding the nature of the experienced difficulties will
guide the field in helping ESL learners acquire these constructions better and
make them more confident in using them.
8 References
Bates, D., M. Maechler, B. Bolker & S. Walker. 2015. Fitting linear mixed-effects models using
lme4. Journal of Statistical Software 67: 1–48.
Dagut, M. & B. Laufer. 1985. Avoidance of phrasal verbs – A case for contrastive analysis.
Studies in Second Language Acquisition 7: 73–79.
Darwin, C.M. & L.S. Gray. 1999. Going after the phrasal verb: An alternative approach to
classification. TESOL Quarterly 33: 65–83.
Davies, M. 2008. The Corpus of Contemporary American English (COCA): One billion million
words, 1990-2019. https://www.english-corpora.org/coca/.
de Bot, K. 1992. A bilingual production model: Levelt’s ‘speaking’ model approach. Applied
Linguistics 13(1): 1–24.
de Groot, A.M.B. 2011. Language and Cognition in Bilinguals and Multilinguals: An
Introduction. New York: Psychology Press.
de Groot, A.M.B. & R. Keijzer. 2000. What is hard to learn is easy to forget: The roles of word
concreteness, cognate status, and word frequency in foreign language vocabulary
learning and forgetting. Language Learning 50(1): 1–56.
Friel, B.M. & S.M. Kennison. 2001. Identifying German-English cognates, false cognates, and
non-cognates: Methodological issues and descriptive norms. Bilingualism: Language
and Cognition 4(3): 249–274.
González, R.A. 2010. Making sense of phrasal verbs: A cognitive linguistic account of L2
learning. AILA Review 23(1): 50–71.
Hulstijn, J. & E. Marchena. 1989. Avoidance: Grammatical or semantic causes? Studies in
Second Language Acquisition 11(3): 241–255.
Ijaz, I.H. 1986. Linguistic and cognitive determinants of lexical acquisition in a second
language. Language Learning 36(4): 401–451.
Jacobs, A., M. Fricke & J.F. Kroll. 2016. Cross-language activation begins during speech
planning and extends into second language speech. Language Learning 66: 324–353.
Jiang, N. 2000. Lexical representation and development in a second language. Applied
Linguistics 21(1): 47–77.
Kuznetsova, A., P.B. Brockhoff & R.H. Christensen. 2017. lmerTest package: tests in linear
mixed effects models. Journal of Statistical Software 82(13): 1–26.
Laufer, B. & S. Eliasson. 1993. What causes avoidance in L2 learning: L1-L2 difference, L1-
L2 similarity, or L2 complexity? Studies in Second Language Acquisition 15(1): 35–48.

40 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
Lemhofer, K. & T. Dijkstra. 2004. Recognizing cognates and interlingual homographs: Effects
of code similarity in language-specific and generalized lexical decision. Memory &
Cognition 32: 533–550.
Liao, Y. & Y. Fukuya. 2004. Avoidance of phrasal verbs: The case of Chinese learners of
English. Language Learning 54(2): 193–226.
Liu, D. 2011. The most frequently used English phrasal verbs in American and British English:
a multicorpus examination. Tesol Quarterly 45(4): 661–688.
Lowie, W. & M. Verspoor. 2004. Input versus transfer? The role of frequency and similarity in
the acquisition of L2 prepositions. In M. Achard & S. Niemeier (eds.), Cognitive
Linguistics, Second Language Acquisition, and Foreign Language Teaching. Berlin:
Mouton de Gruyter, 77–93.
Marian, V., H.K. Blumenfeld & M. Kaushanskaya. 2007. The language experience and
proficiency questionnaire (leap-q): assessing language profiles in bilinguals and
multilinguals. Journal of Speech, Language, and Hearing Research 50(4): 940–67.
Moon, R. 1998. Frequencies and forms of phrasal lexemes in English. In A.P. Cowie (ed.),
Phraseology: Theory, Analysis, and Applications. Oxford: Oxford University Press, 79–
100.
Otwinowska, A. & J.M. Szewczyk. 2019. The more similar the better? factors in learning
cognates, false cognates and non-cognate words. International Journal of Bilingual
Education and Bilingualism 22(8): 1–18.
Qualtrics. 2019. Version 20.05. Utah, USA. https://www.qualtrics.com.
R Core Team. 2018. R: A Language and Environment for Statistical Computing. R Foundation
for Statistical Computing, Vienna, Austria. https://www.R-project.org.
Schmitt, N. 2010. Researching Vocabulary: A Vocabulary Research Manual. London: Palgrave
Macmillan.
Siyanova, A. & N. Schmitt. 2007. Native and non-native use of multi-word vs. one-word verbs.
International Review of Applied Linguistics 45: 119–139.
Ullman, M. 2001. The neural basis of lexicon and grammar in first and second language: The
declarative/procedural model. Bilingualism, Language and Cognition 4: 105–122.
Ullman, M. 2004. Contributions of memory circuits to language: The declarative/procedural
model. Cognition 92(1): 231–270.
White, B.J. 2012. A conceptual approach to the instruction of phrasal verbs. Modern Language
Journal 96(3): 419–438.
Thomas Tienkamp
Amsterdam Center for Language and Communication
Spuistraat 134
1012 VB Amsterdam
thomas.tienkamp@student.uva.nl

Help or hindrance? The influence of Dutch on the acquisition of English phrasal verbs 41
Linguistics in Amsterdam 13,2 (2020)
Appendix A: LEAP-Q questions included in the study:
1. What is your gender?
2. What is your age?
3. How many years of formal education do you have since primary school?
4. What is your current education level?
5. Please list all the languages you know in terms of dominance
6. Please list all the languages you know in terms of order of acquisition
7. Please list what percentage of the time you are currently and on average to all your
languages.
8. When you could read a text in your specified languages, in what percentage of cases
would you choose each language?
9. When you could speak to a person who speaks all languages equally well as you, in
what percentage of cases would you choose each language?
The following questions were about English specifically
1. At what age did you start learning English?
2. Please specify your knowledge of English on a 1-10 scale on speaking, reading, and
understanding English
3. On a scale from 1-10, please select how much the following factors contributed to you
learning English (interaction with friends, interaction with family, reading,
(self)instruction, watching tv, music)
4. On a scale from 1-10 please specify to what extent you are currently exposed to
English in the following contexts (interaction with friends, interaction with family,
reading, (self)instruction, watching tv, music)

42 Thomas B. Tienkamp
Linguistics in Amsterdam 13,2 (2020)
Appendix B: Stimuli used in the experiment and their frequency per million
words
Match trials
English PV
Dutch translation
Frequency per million words
bite off
afbijten
1.393
count on
rekenen op
13.922
call back
terugbellen
10.881
blow off
afblazen
5.163
carry out
uitvoeren
26.225
give away
weggeven
13.92
shut off
uitdoen
4.492
tear out
uitscheuren
1.768
stick out
uitsteken
8.002
take back
terugnemen
23.622
watch out
uitkijken
8.029
focus on
focussen op
119.999
stand for
staan voor
20.671
come off
afkomen
14.289
try on
aanpassen
9.166
get back
terugkomen
74.4
Mismatch trials
English PV
Dutch translation
Frequency per million words
talk over
doorheen praten
2.355
come off as
overkomen als
14.289
doze off
indutten
0.925
keep to
houden voor
12.785
think of
denken aan
139.136
wander off
weglopen
1.427
be dependent on
afhankelijk zijn van
45.683
aim at
richten op
18.981
jump at
springen op
2.561
put up
ophangen
26.105
be angry with
boos zijn op
1.879
vote for
stemmen op
42.54
strike out
doorstrepen
5.485