ArticlePDF Available

Pre-emption of L1 properties in the L2 acquisition of English wh-interrogatives: Effects of L2 proficiency and age of onset

Authors:

Abstract

This study investigates whether second language (L2) speakers can pre-empt a first language (L1) property which involves uninterpretable features, such as resumption. The Interpretability Hypothesis predicts persistent L1 effects in L2 grammars because uninterpretable features resist resetting beyond some critical period (Tsimpli and Dimitrakopoulou 2007). Unlike English, Saudi Arabic allows grammatical resumption in complex wh-interrogatives, which is highly preferred with (D)iscourse-linked wh-forms (e. g. ʔayy-NP ‘which-NP’) but disallowed with non-D-linked ones (e. g. ʔeeʃ ‘what’). The study was conducted with fifteen native English speakers and 34 (very)-advanced Saudi Arabic L2 speakers of English with age of onset (AO 1–13 years). In a bimodal, timed acceptability judgment task, their accuracy judgments of 32 (un)grammatical wh-interrogatives were tested. As predicted, results show that L2 speakers of very advanced levels inaccurately accepted resumption especially with D-linked wh-interrogatives. The results also show non-significant differences between AO 1–6 and 7–13 years in their rejection accuracy of resumption.
Kholoud A. Al-Thubaiti*
Pre-emption of L1 properties in the L2
acquisition of English wh-interrogatives:
Effects of L2 proficiency and age of onset
https://doi.org/10.1515/iral-2016-0128
Abstract: This study investigates whether second language (L2) speakers can
pre-empt a first language (L1) property which involves uninterpretable fea-
tures, such as resumption. The Interpretability Hypothesis predicts persistent
L1 effects in L2 grammars because uninterpretable features resist resetting
beyond some critical period (Tsimpli and Dimitrakopoulou 2007). Unlike
English, Saudi Arabic allows grammatical resumption in complex wh-interro-
gatives, which is highly preferred with (D)iscourse-linked wh-forms (e. g. ʔayy-
NP which-NP) but disallowed with non-D-linked ones (e. g. ʔeeʃwhat). The
study was conducted with fifteen native English speakers and 34 (very)-
advanced Saudi Arabic L2 speakers of English with age of onset (AO 113
years). In a bimodal, timed acceptability judgment task, their accuracy judg-
ments of 32 (un)grammatical wh-interrogatives were tested. As predicted,
results show that L2 speakers of very advanced levels inaccurately accepted
resumption especially with D-linked wh-interrogatives. The results also show
non-significant differences between AO 16and713 years in their rejection
accuracy of resumption.
Keywords: Age effects, L1 effects, interpretable and uninterpretable features,
resumption, discourse-linking, wh-interrogatives
1 Introduction
Explaining variability in second language (L2) grammars is a debated topic in
generative second language acquisition (L2A). The phenomenon of variability
refers to the inconsistent use of two forms in L2 where one form is allowed
(Towell and Hawkins 1994). There are two possible types of variability in L2
grammars. The first type refers to variability that is a feature of all L2 speakers
regardless of their first language (L1) at a certain stage of development. For
example, the variability in using -ed and affixes with verbs in intended past
*Corresponding author: Kholoud A. Al-Thubaiti, English Department, Umm Al-Qura University,
Makkah, Saudi Arabia, E-mail: kathubaiti@uqu.edu.sa
IRAL 2018; aop
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
tense contexts in L2 English by speakers of L1 Russian (see Ionin and Wexler
2002) and L1 Chinese (see Lardiere 1998). This variability is a characteristic of L2
development irrespective of whether the L1 has verbal affixes for tense, as in
Russian, or does not, as in Chinese. The second type refers to variability as a
feature of L1 that is transferred into the L2 where it is inappropriate. For
example, variability in allowing null and overt subject pronouns in L2 English.
This variability is a characteristic only of speakers of L1s that allow such an
alternation, such as L1 Greek (see Tsimpli and Roussou 1991) and L1 Spanish but
not L1 French (see White 1985). This study is concerned with investigating
variability as a feature of the L1 that is transferred into L2.
The Interpretability Hypothesis (IH) (Tsimpli and Dimitrakopoulou 2007)
explains variability in L2 grammars that is a feature of L1 by adopting assump-
tions of minimalism and the critical period. It focuses on the distinction
between the role of uninterpretable and interpretable features at the interfaces
of LF and PF. Interpretable features, such as animacy, are visible at the LF-
interface because they contribute to meaning. In contrast, uninterpretable
features, such as agreement and case, are required for syntactic derivations,
but do not contribute to meaning and are therefore deleted before the deriva-
tion is transferred to LF. Uninterpretable features are not visible at LF but can
be realized at PF, such as resumptive pronouns. Adopting assumptions of the
critical period, uninterpretable features, unlike interpretable ones, resist reset-
ting beyond some critical period. Interpretable features have a dual represen-
tation in the language system and the LF-interface that makes them
permanently accessible. If the L2 is acquired beyond some critical period,
properties associated with uninterpretable features will be problematic unlike
those which involve interpretable features. As far as L1 effects are concerned, if
uninterpretable features are grammaticalized in L1 but not in L2, they cannot
be pre-empted from L2 grammars. Interpretable features, however, are
assumed to constrain the persistent effects of L1 and thus improve L2
performance.
Tsimpli and Dimitrakopoulou (2007) tested the predictions of the IH with
resumptive pronouns in L2 English wh-interrogatives by L1 Greek speakers.
Whilst English only allows a gap in the wh-extraction site, Greek allows optional
grammatical resumption in the same position, as shown in (1), example taken
from Tsimpli and Dimitrakopoulou (2007: 220).
(1) Pjon fititi ipes oti (ton) aperipsan sti sinedefksi?
which student said
2SG
that him-rejected
3PL
at-the interview
Which student did you say that they rejected at the interview?
2Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
They assumed that the resumptive pronoun (ton) represents a bundle of
uninterpretable features, such as agreement and case. In Greek, the realization
of resumption increases with (D)iscourse-linked interrogatives, as shown in (1),
but is disallowed with what-interrogatives, as shown in (2), example taken from
Tsimpli and Dimitrakopoulou (2007: 220).
(2) Ti nomizis oti tha (*to) dhiavasun?
what think
2SG
that will it-read
3PL
What do you think that they will read?
Under time pressure, they elicited judgment data on (un)grammatical subject/
object wh-interrogatives with resumption and gaps. Their study sample included
intermediate and advanced L2 English groups. Their results showed persistent L1
effects, where resumption, once activated, cannot be blocked even in advanced
L2 grammars because it involves uninterpretable features. Crucially, the toler-
ance rate of resumption in advanced L2 grammars was constrained to inanimate
pronouns with D-linked wh-forms. Although animacy is not marked on Greek
wh-forms and pronouns, they were successfully acquired in English. The authors
argued that the D-linking effect was not a direct transfer from L1 because it co-
occurred only with inanimate pronouns. The constrained use of resumption was
taken as indicative of the compensatory role of interpretable features in improv-
ing L2 performance.
Although the IH assumes the critical period, it has not been tested with L2
groups by age of onset (hereafter AO). Generative L2 studies which adopt
assumptions of the critical period have extensively examined the grammars of
highly-proficient (or near-native) adult L2 speakers without comparing them by
AO (e. g. Hawkins and Hattori 2006; Sorace and Filiaci 2006; Tsimpli and
Dimitrakopoulou 2007). White and Genesee (1996) were an exception in this
respect. Although it is understood that the design of these L2 studies sought to
identify exceptional older L2 starters deemed target-like to defy the critical
period (see Long 2005), it led to overlooking young L2 starters (AO 6). In fact,
age-related studies with heritage and typical L2 speakers have started to yield
more evidence that young L2 starters also show non-target-like grammars (e. g.
Abrahamsson and Hyltenstam 2009; Montrul 2008; Nishikawa 2014). Muñoz and
Singleton (2011: 26) suggested that critical-period research should compare
highly- proficient later L2 starters with comparable early L2 starters.
This study aims to address the empirical gap in generative L2 research by
testing variability in L2 grammars when considering the effects of high L2
proficiency (very advanced and advanced) and AO (16 and 713). This paper
forms part of a larger project to examine age effects in the long run with a set of
Pre-emption of L1 properties 3
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
morpho-syntactic properties which differ between L1 Saudi Arabic and L2
English. Inspired by Tsimpli and Dimitrakopoulou (2007), this paper reports
on results from resumption in L2 English complex wh-interrogatives as a poten-
tial candidate for persistent L1 effects.
The paper is organized as follows: a descriptive account is presented for the
gap/resumptive strategy in complex wh-interrogatives in Saudi Arabic and
English (section 2). Then the research questions and predictions are presented
(section 3), followed by the method (section 4), and the results (section 5).
Finally, the paper discusses the results and concludes (section 6).
2 Complex wh-interrogatives in Arabic
and English
In the formation of complex wh-interrogatives, English and Saudi Arabic differ
in two related properties: that-trace violation and resumption.
In standard varieties of English, that-trace violations arise in subject wh-
interrogatives with an embedded that-clause, as in (3), but not in object wh-
interrogatives, as in (4). Examples are taken from Roussou (2002: 15).
(3) Which author do you think (*that) won the prize?
(4) Which author do you think (that) Bill supported?
As illustrated in (3) and (4), the presence of that is disallowed in subject
extraction but is optional in object extraction. In Saudi Arabic, on the other
hand, that-trace violations do not occur in either subject or object wh-extraction
out of an innu-clause. The presence or absence of innu is optional at both
extraction sites, as in (5) and (6), respectively (which are based on a Hejazi
variety of Saudi Arabic).
(5) miin ti-Hassib (innu) ʃaaf-at Nizaar fi-l-maTaar?
who 2MS-think that saw-3FS Nizar in-the-airport
Who do you think (*that) saw Nizar in the airport?
(6) miin ti-Hassib (innu) Nizaar ʃaaf fi-l-maTaar?
who 2MS-think that Nizar saw.3MS in-the-airport
Who do you think (that) Nizar saw in the airport?
4Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
While English and Saudi Arabic differ in whether subject wh-extraction is
possible with an overt or null complementizer, as in (3) and (5), they perform
alike in allowing object wh-extraction with an optional that/innu, as in (4) and
(6). The that-trace effect in subject extraction has been analyzed in a variety of
ways in different syntactic frameworks.
1
In minimalism, Roussou (2002) pro-
posed a parametric analysis that maintainsthe traditional position associating
that-trace violation with lack of null subjects in a language.
2
She argued that
languages differ in the way the properties of Agr (i. e. uninterpretable features)
of T can be lexicalized in conditions of subject extraction. Through the operation
of Agree, Agr as a variable is connected with the wh-phrase in the matrix clause.
In English (a language without null subjects), the Agr features of the embedded
T cannot be lexicalized in the presence of that in the C of the embedded clause.
The presence of that blocks identifying the subject agreement features for any
further syntactic operations. These Agr features are required to identify which
author with the embedded subject (3). Rather, the C of the embedded clause has
to be null so that the subject agreement features can be identified on the wh-
phrase in the matrix clause via an Agree relation that connects the embedded
and the matrix clauses. On the other hand, in null subject languages, such as
Arabic varieties, that-trace violations do not occur in subject extraction because
the Agr features are lexicalized through an agreement affix on the embedded
verb. Assuming Roussous analysis, the presence or absence of innu does not
interfere with the lexicalization of Agr because the properties of the wh-phrase
(miin) in the matrix clause in (5) can be identified through the agreement affix
on the embedded verb (-at).
Resumption is another property distinguishing English from Saudi Arabic.
Rouveret (2011: 2) defined resumption as overt pronominal elementswhich
appear in the variable position of an A-bar-dependency construction, such as
wh-interrogatives. In a minimalist account, Rouveret proposed that a resumptive
pronoun is related to its antecedent via Agree, where the functional head C is
1That-trace violations were originally taken to be a property of the null subject parameter,
permitted only in languages which allow null subjects and free subject inversion (e. g. Rizzi
1990). However, this position proves problematic for varieties of English which allow that-trace
violations but not null subjects or free subject inversion (e. g. Pesetsky 1982: 328; Sobin 2002),
and also for other languages which show a similar dissociation (Jaeggli and Safir 1989). The
general consensus in more recent times is that-trace violations are an independent phenomenon
(Andrew Radford, pc).
2For the purpose of this article, I present Roussous (2002) analysis in general terms without
involving technical details. For more recent agreement-based accounts of that-trace effects, see
Rizzi and Shlonsky (2007), and Rizzi (2015); for a PF-interface/parsing account, see Radford
(2018).
Pre-emption of L1 properties 5
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
taken to be the binder. Hence, it is plausible to assume that resumptive pro-
nouns involve the spelling-out of uninterpretable agreement and case features
(Tsimpli and Dimitrakopoulou 2007) which bind with and identify the antece-
dent. There are distinct types of resumption use across languages: Arabic
productively licenses grammaticalresumptive pronouns in the extraction site
where a gap is allowed most of the time (Aoun et al. 2010), whereas English
requires an obligatory gap in the extraction site, but exclusively resorts to
intrusiveresumptive pronouns as a saving devicein the case of syntactic
violations (Rouveret 2011). In the classic example from Ross (1967: 432), the
italicized resumptive pronoun in (7) cannot be replaced by a gap because it
alleviates movement violations.
(7) I just saw that girl who Long Johns claim that she was a Venusian made
all the headlines.
This paper examines resumption as a grammatical phenomenon which is
licensed in Saudi Arabic but not English. In what follows, the gap-resumption
distinction is described in the embedded position of a that/innu-clause in wh-
interrogatives.
In complex subject wh-interrogatives, Saudi Arabic allows the insertion of a
subject resumptive pronoun attached to innu-, as in (8) and (9). However, the
subject clitic is optional in (8) but obligatory in (9). The equivalents of these
constructions are ungrammatical in English because of that-trace and lack of
grammatical resumption.
(8) ʔayy bint
i
ti-Hassib inn-(ha
i
)aʃtar-at
i
al-fustaan?
which girl 2MS-think that-she bought-3FS the-dress
Which girl do you think (*that-she) bought the dress?
(9) ʔayy bint
i
ti-Hassib inn-*(ha
i
)miʃtarya
i
al-fustaan?
which girl 2MS-think that-she has bought-FS the-dress
Which girl do you think (*that-she) has bought the dress?
Adopting Kenstowiczs (1989: 272) analysis of similar data in Bani-Hassan
Arabic, the realization of the subject clitic seems sensitive to the verbal inflec-
tion of the embedded clause. The crucial factor is whether the verb shows person
agreement with the subject. In the case of finite verbs, such as aʃtar-at in (8), the
insertion of resumption is optional. If no resumptive pronoun is inserted, the
subject can still be extracted over innu because the inflection -at suffices to
identify the person feature of the subject. Conversely, the absence of the person
6Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
inflection in the participle form miʃtarya requires the insertion of a resumptive
pronoun following the complementizer, inn-ha (9). The clitic -ha must be present
because of lack of person agreement on the participle form required to identify
the person feature of the subject. Kenstowiczs analysis aligns well with the
position that person (unlike number and gender) is crucial for genuine subject-
verb agreement (Miyagawa 2010: 45). From the contrast between (8) and (9),
subject resumption is required whenever subject-verb agreement is not suffi-
ciently marked on the embedded verb. This contrast supports the position that
subject resumptive pronouns are [] overt manifestation of agreement features
on T []where these features are uninterpretable at LF but interpretable at PF
(Tsimpli and Dimitrakopoulou 2007: 223).
Turning to complex object wh-interrogatives, Saudi Arabic allows the inser-
tion of a resumptive pronoun in the object position regardless of the presence/
absence of innu. The resumptive pronoun is realized as an object clitic attached
to the embedded verb, as in (10) and (11). It is assumed that the realization of
object clitics involves the spelling-out of uninterpretable agreement features on
the verb (Tsimpli and Dimitrakopoulou 2007).
(10) miin
i
/ʔayy walad
i
ti-Hassib (innu) Nizaar ʃaaf-(u
i
)
who/which boy 2MS-think that Nizar saw.3MS-(him)
fi-l-maTaar?
in-the-airport
Who/which boy do you think (that) Nizar saw (*him) in the airport?
(11) ʔayy kitaab
i
ti-Hassib (innu) Nizaar Talab-(u
i
)
which book 2MS-think that Nizar requested.3MS-(it)
min-l-maktaba?
from-the-library
Which book do you think (that) Nizar requested (*it) from the library?
Following Aoun and Choueiri (1999) analysis of similar data in Lebanese Arabic,
the realization of the object clitic appears to be sensitive to the D-Linking feature
of the wh-form. D-linked wh-forms, such as which-NP, refer to a member of a
presupposed set in a shared discourse between speaker and hearer (Pesetsky
1987). D-linking is assumed to involve interpretable features at LF which affect
the semantic interpretation of wh-interrogatives (Tsimpli and Dimitrakopoulou
2007). In Saudi Arabic, ʔayy-NP which-NPis inherently D-linked by its mor-
phological form, and miin whocan be contextually D-linked in a given dis-
course. In contrast, ʔeeʃwhatcan barely have a felicitous D-linked
Pre-emption of L1 properties 7
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
interpretation.
3
Object resumption is often preferred over a gap with D-linked
wh-interrogatives (1011), but completely disallowed with non-D-linked ʔeeʃ
what(12).
(12) ʔeeʃ
i
ti-Hassib (innu) Nizaar Talab-(*u
i
) min-l-maktaba?
what 2MS-think that Nizar requested.3MS-(*it) from-the-library
What do you think (that) Nizar requested (*it) from the library?
In English, which-NP is inherently D-linked, whereas who and what are normally
non-D-linked (Pesetsky 1987). Although English has D-linking expressed in wh-
forms, it only allows a gap in object position, as in the translations of (1012).
Like English, wh-forms in Arabic are marked by animacy (miin who= animate
and ʔeeʃwhat= inanimate). Arabic pronouns, however, (unlike English) are not
distinguished by animacy (+/-animate), so there is no equivalent for it in Arabic.
Instead, Arabic pronouns (clitics and strong forms) are marked by gender as in the
transliterations of the examples. Animacy is assumed to involve interpretable
features at LF because they affect the semantic interpretation of the wh-forms and
pronouns (Tsimpli and Dimitrakopoulou 2007).
To summarize, Saudi Arabic and English differ in two respects in the forma-
tion of complex wh-interrogatives: (a) the possibility of extracting subjects and
objects from a that-clause in the presence of the complementizer, and (b) the
realization of grammaticalresumption. Unlike English, subjects in Saudi Arabic
can be freely extracted in the presence of innu. The requirement for inserting a
resumptive pronoun attached to innu- (a subject clitic) depends on whether
subject-verb agreement is sufficiently marked on the embedded verb. The subject
clitic is optional with finite verbs which show the person feature but obligatory
with participle forms which lack one. Conversely, objects in Saudi Arabic and
English can be freely extracted in the presence of innu/that. Unlike English, where
a gap is needed in the object position, Saudi Arabic licenses the insertion of an
object resumptive pronoun attached to the embedded verb (an object clitic).
However, object resumption is strongly preferred whenever a D-linked interpreta-
tion is felicitous (ʔayy-NP which-NPand miin who) and disallowed in the
absence of a D-linked interpretation (ʔeeʃwhat).
3In a given discourse between a group of friends whom Nizar invited for his birthday, miin who
can easily have a D-linked interpretation and be used in a manner similar to ʔayy-NP which-NP
(i. e. contextually-D-linked). However, ʔeeʃwhatcan barely have a felicitous D-linked interpreta-
tion in a similar discourse when asking about one of a group of things which Nizar might have
bought. Since there is no prior research on D-linking in Saudi Arabic, research is required to
systematically analyze such constructions.
8Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
3 Research questions and predictions
Based on the parametric differences between Saudi Arabic and English, accu-
racy judgments of highly-proficient Saudi Arabic L2 speakers of English were
examined to answer the following three questions:
1. Can L2 speakers pre-empt an L1 property of uninterpretable features, such as
resumption?
2. If L2 speakers are found to inaccurately accept resumptive pronouns in L2
English wh-interrogatives, how is their accuracy affected by the following
factors:
a. the extraction site and presence of the complementizer,
b. the interpretable features of D-linking and animacy.
3. Can L1 effects be moderated when acquiring L2 before age six?
Following Tsimpli and Dimitrakopoulou (2007), the first two questions
seek to test predictions of the IH about variability as a feature of the L1 that
is transferred into L2. The case of variability tested concerns the L1 transfer of
theoptionaluseoftheresumptiveorgapstrategyintoL2,whichonlyallows
the gap strategy. Unlike English, Saudi Arabic licenses grammatical resump-
tion in subject/object wh-extraction sites under specific constraints. Since it is
assumed that resumption involves uninterpretable features, the prediction is
that it will not be completely pre-empted at very advanced levels. Therefore,
L2 grammars will show variability in tolerating the resumptive and gap stra-
tegies, as would be possible in the L1. The tolerance rate of resumption will
not be random but predictable under L1 influence: (a) subject resumption will
be more tolerated than object resumption because the latter is optional in L1
Saudi Arabic, and (b) neither subject nor object resumption will be affected by
the presence or absence of the complementizer that because there appears to
be no effect for innu on resumption realization in L1 Saudi Arabic. Having
been exposed to L2 English for such a long time, highly-proficient L2 speakers
will figure out constraints to restrict the range of resumption tolerance in their
L2 grammars. Therefore, the extent of variability is predicted to be constrained
by the interpretable features of D-linking and animacy involved in the deriva-
tion of wh-interrogatives.
The third question follows from the IHs explanation for variability in L2
grammars. Recall that it adopts assumptions of the critical period to explain
why properties associated with uninterpretable features resist resetting in adult
L2A. Unlike the study of Tsimpli and Dimitrakopoulou (2007), this question seeks
to examine any advantages for an earlier L2A, before age six, on the potential to
pre-empt resumption as an L1 property of uninterpretable features. Age six was
Pre-emption of L1 properties 9
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
taken to represent the period closure of peak sensitivity for morpho-syntax
(Granena and Long 2013).
4 Method
4.1 Participants
The study sample comprised 34 (very)-advanced Saudi Arabic L2 speakers of
English and fifteen English NSs. The average age at testing for the L2 speakers
was 27.18 years (2037) and 26.80 years (1956) for the NSs. The L2 speakers
were highly educated and active bilinguals in L1 Saudi Arabic and L2 English.
They were recruited among university instructors and graduate students in
Saudi Arabia. The NSs were graduate students at UK universities.
Information about the participantsEnglish learning experience was col-
lected by an extensive questionnaire. To qualify for participation, the L2
speakers had to have a unified native language (Saudi Arabic) as recom-
mended by DeKeyser and Larson-Hall (2005) for age-related studies testing L1
effects. They also had to vary in their AO to L2 English but had to have
comparably high L2 proficiency on a formal assessment of English. This
condition was crucial to tease apart effects of L2 proficiency from age effects.
They also had to have a minimum of eight years of continued experience of
using English in everyday activities. AO corresponds to the time the L2
speakers were first exposed to English in a continuous manner through
home and formal instruction in Saudi Arabia or an English-speaking country
(ESC).
4
To maximize the possibility of recruiting highly-proficient L2 speak-
ers, which is a key factor in the study, it was decided to consider participants
who showed study abroad experience in their history of English exposure.
There were fifteen L2 speakers who had been to an ESC with arrival ages
4An anonymous reviewer was concerned about the effects of not controlling context of
exposure and length of exposure. This problem is addressed through the independent measure-
ment of proficiency. Controlling for proficiency compensates both for the context in which
acquisition occurs and for variation in length of exposure, especially when length of exposure is
no longer considered a genuine measure of the quality and quantity of L2 input (see Muñoz and
Singleton 2011). In an ideal world, country of exposure should be controlled for, which I
acknowledge as a limitation. Before conducting analyses, however, I checked its effect on
accuracy judgments. Results showed no significant differences in accuracy between those
who had been first exposed to English in an ESC and those in Saudi Arabia. I interpreted
lack of differences as evidence for their comparable high L2 proficiency.
10 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
rangingfromzerotonineyears(M=3.20, SD = 2.731) and length of residence
of two to thirteen years (M=7.27, SD = 3.218). Twelve of the fifteen had
attended formal schools and three were first exposed to English at home
through older siblings who spoke English and went to school there. Those
three participants were born in an ESC. Their AO was counted from the
time they reported initial English exposure. The family profiles of the parti-
cipants showed that their parents and/or siblings had knowledge of English
to varying degrees. Siblings were rated higher than parents in their English
mastery.
The L2 speakersEnglish proficiency was measured by the Quick Oxford
Placement Test (QOPT; Max = 60) and Nations (2001) Vocabulary Levels Test
(VLT; Max = 90). Of the 48 participants, 34 fell in the (very)-advanced score
range of the QOPT (C1 and C2 in the European Framework). The VLT helped in
verifying the proficiency level of the participants who scored at the very bottom
or top of the (very)-advanced QOPT score range. Participantsscores on both
tests were tallied to form a composite score out of 150. Cluster analysis was run
on the 34 cases to form two proficiency groups. A hierarchical cluster analysis
using Wards method produced two clusters, which were labelled very
advancedand advanced. Both groups differed significantly with a large effect
size, t(32) = 11.81, p< 0.001, d= 4.08. Table 1 summarizes the L2 groupsdemo-
graphic information.
4.2 Material
ToleranceofresumptioninL2Englishwh-interrogatives was tested by an
Acceptability Judgment Task (AJT) with 164 items on a variety of structures:
Table 1: Demographic information for the L2 groups (N=34).
Proficiency score Age at testing Age at onset LOE
Groups n
Min-Max
Mean (SD)
Min-Max
Mean (SD)
Min-Max
Mean (SD)
Min-Max
Mean (SD)
Very advanced  
. (.)

. (.)

. (.)

. (.)
Advanced  
. (.)

. (.)

.(.)

. (.)
Note: Maximum proficiency score =150. LOE =Length of Exposure. Age and LOE are in years.
Pre-emption of L1 properties 11
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
half were grammatical, and half were ungrammatical. There were 32 complex
subject/object wh-interrogatives on the gap-resumption distinction. Following
Tsimpli and Dimitrakopoulou (2007), the embedded position of a that-clause
made it possible to test resumption tolerance compared with the gap require-
ment in subject and object extraction sites under the presence and absence of
the complementizer that. Unlike their design, however, the effect of the com-
plementizer was tested in both extraction sites to balance the distribution of
the variables in the design. To examine the role of interpretable features on
resumption tolerance, D-linking and animacy were included as variables (see
Table 2).
The gap-resumption distinction was tested in four conditions across extraction
site (subject and object) and complementizer that (null and overt). Within each
of these conditions, there were eight wh-interrogatives which contrasted the gap
strategy (k=4) with the resumptive strategy (k=4). In terms of D-linking and
animacy, four wh-interrogatives were constructed with non-D-linked who/what
(+/-animate), and another four with D-linked which-NP (+/-animate). All pro-
nouns were in third person singular form. Examples of test items (with who-
interrogatives) are provided in (13) to (16), where gaps are illustrated by (_) and
resumptive pronouns in bold.
5
(13) Subject extraction (null that)(k=8)
a. Gap: Who do you think __will win the tennis world championship?
b. Resumption: *Who did the police suspect he robbed the bank?
5An anonymous reviewer was concerned whether the test items of the compared conditions
formed a minimal pair. As in Tsimpli and Dimitrakopoulou (2007), they did not form strict
minimal pairs because the test items were constructed with different tense, verbs and vocabu-
lary. Varying the test items has the advantage of keeping the participants focused on meaning
without guessing the grammar under investigation. However, it is advisable to construct
minimally different pairs in future research to exclude possible confounds.
Table 2: Linguistic variables with numbers of tokens.
Extraction site Complementizer that D-Linking Animacy
Subject extraction
(k=)
null (k=)
overt (k=)
+DL (Which-NP) (k=)
-DL (Who/what)(k=)
+animate (k=)
-animate (k=)
Object extraction
(k=)
null (k=)
overt (k=)
+DL (Which-NP) (k=)
-DL (Who/what)(k=)
+animate (k=)
-animate (k=)
12 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
(14) Subject extraction (overt that)(k=8)
a. Gap: *Who did the manager suggest that __ should be promoted?
b. Resumption: *Who did the parents suspect that he had kidnapped their
child?
(15) Object extraction (null that)(k=8)
a. Gap: Who do you think Adele would marry __ willingly?
b. Resumption: *Who do you think the boys visited her yesterday?
(16) Object extraction (overt that)(k=8)
a. Gap: Who do you imagine that the teacher is tutoring __ privately?
b. Resumption: *Who do you believe that the people will choose her as
their president?
4.3 Procedure
Following Tsimpli and Dimitrakopoulou (2007), the AJT was bimodal and
timed. The purpose of the dual modality (visual and aural) was twofold: (a)
to control the response time across participants, and (b) to reduce the proces-
sing burden of ungrammatical items. Murphy (1997) found that L2 speakers
performed better on ungrammatical items presented via dual modality.
Controlling the response time minimized their opportunities to use their meta-
linguistic knowledge (Ionin 2012). Two pseudo-randomized versions of the AJT
were presented individually by Powerpoint. Each item was read twice by a
female English NS and then displayed for eight seconds. The length of each
slide transition was fixed in the settings to control the response time.
Participants had to mark their judgments during the eight seconds of display
before the following item was automatically shown. On paper, they had to
circle their judgments on a five-point scale, with the ends labelled as 1
(definitely impossible) and 5 (definitely possible). They were instructed to
circle 3 whenever their ratings were in the middle of both extremes. They
were also asked to make judgments to the best of their understanding as there
were no right or wrong responses. Adapting task instructions from Bley-
Vroman et al. (1988), they had an extensive instruction and practice session
to ensure that they were responding based on their intuitive judgments. In
effect, no participant was unable to make a judgment.
Following Mackey and Gass (2005) guidelines, the 164 items were presented
in four blocks (41 items each) to control for fatigue and order effects. All
participants completed their judgments within 21 minutes excluding break
Pre-emption of L1 properties 13
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
times between blocks. In a different session, they were given the English
proficiency test and background questionnaire.
4.4 Analysis
Since age-related studies typically examine accuracy scores (e.g. White and
Genesee 1996), this study converted the acceptance/rejection ratings into accuracy
judgments. Tsimpli and Dimitrakopoulou (2007: 226) also analysed their judgment
data in terms of target-like and non-target-like responses after conflating the
original scale ratings and removing the mid-point zero not sure. Unlike Tsimpli
and Dimitrakopoulou, this study adopted Mackey and Gass (2005: 55) scoring
scheme without conflating the judgments: each point on the five-point scale was
assigned a correct/incorrect coding and the mid-point 3 was coded as dont
know. For ungrammatical items, the polarity of the rating scale was reversed
and then recoded for accuracy. For example, if a participant correctly rejected an
ungrammatical item by rating it as 1, that rating was recoded as 5 to signify
correctly rejected with confidence. However, if the rating for the same ungram-
matical item was 4, it was recoded as 2 to signify incorrectly accepted with some
confidence. The accuracy scale therefore ranged from 5 (correct acceptance/
rejection with confidence) to 1 (incorrect acceptance/rejection with confidence).
For (un)grammatical wh-interrogatives, the mean accuracy was computed by
strategy type (gap and resumption) within four conditions across extraction site
(subject and object) and complementizer that (null and overt). Evidence for
tolerating/rejecting resumption was based on the following criterion: if L2 speak-
ers correctly rejected ungrammatical wh-interrogatives with resumption and at
the same time correctly accepted grammatical wh-interrogatives with a gap, this
was taken as evidence that their L2 grammars disallowed resumption.
6
The
expected pattern for target-like performance was therefore comparable accuracy
judgments on the gap-resumption conditions (non-significant differences
between accuracy judgments with a small effect size). Conversely, if they cor-
rectly accepted grammatical wh-interrogatives with a gap but at the same time
incorrectly accepted ungrammatical wh-interrogatives with resumption, this was
taken as evidence that their L2 grammars showed variability (accepting gaps and
resumption) as would be possible in their L1. Since Saudi Arabic allows the use
of resumption or gaps, L1 effects were predicted through lower accuracy
6In evaluating whether specific movement constraints were acquired, White and Genesee
(1996: 245) adopted the same criterion (accurate rejection of ungrammatical structures plus
accurate acceptance of grammatical structures).
14 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
judgments on ungrammatical wh-interrogatives with resumption compared with
grammatical wh-interrogatives with a gap (significant differences between accu-
racy judgments with a large effect size).
Then, the L2 speakersaccuracy in rejecting resumption was examined
for effects of L1 and feature interpretability. Recall that ungrammatical wh-
interrogatives were constructed to include resumptive pronouns in various
positions which could potentially be allowed in Saudi Arabic. Therefore, their
mean accuracy on resumption was compared across extraction site and com-
plementizer as syntactic variables which vary in both languages. To test the role
of feature interpretability, the mean accuracy on resumption was computed by
D-linking (which-NP and who/what) and animacy (+/-animate).
Finally, the L2 speakersaccuracy judgments in rejecting resumption were
examined by AO via group (16 and 713) and individual analyses. The aim of
these analyses was to uncover any potential advantages for an early AO before
six years of age in minimizing the rate of resumption tolerance.
Following White (2003: 26), even if L2 speakers were less accurate than NSs,
what matters is whether their pattern of accuracy judgments on the paired
conditions was comparable to that produced by the NSs. Therefore, within-
group analyses were mostly performed, and where relevant between-group
comparisons were performed by proficiency and AO. The alpha was set at
p< 0.05 for all analyses and supplemented by Cohensdeffect size. The effect
size is best interpreted as follows for between-groups: d= 0.40 is small, d= 0.70
is moderate, and d= 1.00 is large, and for within-groups: d= 0.60 is small,
d= 1.0 is moderate, and d= 1.4 is large (Plonsky and Oswald 2014).
5 Results
5.1 Accuracy judgments by (very)-advanced L2 groups
For questions (1) and (2), three analyses were conducted: (a) gap-resumption
comparisons (5.1.1), (b) effects of extraction site and complementizer on resump-
tion (5.1.2), and (c) effects of D-linking and animacy on resumption (5.1.3).
5.1.1 The gap-resumption comparisons
Table 3 shows the mean accuracy on the gap and resumptive conditions by
extraction site (subject and object) and complementizer that (null and overt).
Pre-emption of L1 properties 15
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
The results from ANOVAs (2 x 2 ×2 repeated measures) showed a significant
three-way interaction effect by the NSs group, F(1,14) = 15.39, p= 0.002,
ηp2= 0.52, the very advanced group, F(1,14) = 5.33, p= 0.037, ηp2= 0.28, and
the advanced group, F(1,18) = 21.91, p= 0.001, ηp2= 0.55. The interaction effect
was examined by subsequent pairwise comparisons with Tukey-Kramer
adjustments.
In subject extraction with null that, the NSs, as expected, showed compar-
able accuracy on the gap-resumption comparison. In contrast, the very
advanced showed (marginally) non-significant lower accuracy on resumption
compared with gaps, whilst the advanced showed significant lower accuracy on
resumption. Although the magnitude of the mean difference was large for the
advanced, it was moderate for the very advanced, like that of the NSs.
These results showed that, at very advanced levels, there was a tendency
(non-statistical) for resumption to be potentially rejected in subject extraction
with null that.
Table 3: Mean accuracy (SD) on the gap and resumptive strategies by extraction site and
complementizer: results of Tukey-Kramer pairwise comparisons.
Subject extraction Null that Overt that (i. e. that-trace)
Gap
(k=)
RP
(k=)pd
Gap
(k=)
RP
(k=)pd
Native speakers . . . . . . . .
(n=)(.)(.)(.)(.)
Very advanced . . . . . . . .
(n=)(.)(.)(.)(.)
Advanced . . <. . . . . .
(n=)(.)(.)(.)(.)
Object extraction Null that Overt that
Gap
(k=)
RP
(k=)pd
Gap
(k=)
RP
(k=)pd
Native speakers . . . . . . . .
(n=)(.)(.)(.)(.)
Very advanced . . . . . . . .
(n=)(.)(.)(.)(.)
Advanced . . . . . . . .
(n=)(.)(.)(.)(.)
Note: RP =resumptive pronoun. p<0.05, d=effect size.
16 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
Turning to subject extraction with overt that, the gap-resumption compar-
ison was between ungrammatical that-trace followed by a subject gap and a
subject pronoun. As Table 3 shows, the presence of a subject pronoun seemed to
increase the rejection accuracy of all groups. For the NSs, resumption was more
severe than that-trace violation as they showed significantly higher accuracy
with a large effect size on subject resumption than subject gap following that-
trace. The very advanced showed non-significant higher accuracy on subject
resumption than that-trace with a moderate effect size, whereas the advanced
showed comparable accuracy on the same conditions with a small effect size.
The (very)-advanced groups seemed to show different L2 grammars: whilst the
very advanced showed a tendency to treat the ungrammaticality of subject
resumption to be more severe than that-trace violation, the advanced treated
the ungrammaticality of subject resumption to be as severe as that-trace
violation.
In terms of object extraction, the NSs, as expected, showed comparable
accuracy on the gap-resumption comparison with null and overt that. The
(very)-advanced groups, however, showed different performances depending
on the presence/absence of the complementizer. In the case of null that, both
L2 groups showed significant lower accuracy on object resumption than object
gap, whereas with overt that, they showed comparable accuracy with relatively
small effect size, as did the NSs (see Table 3). These results suggested that the
(very)-advanced groups managed to disallow object resumption with overt that,
but still tolerated object resumption to some degree with null that, resulting in
lower rejection accuracy.
In summary, from the gap-resumption comparisons, accuracy results
showed variability in (very)-advanced L2 grammars in subject position more
than object position. In subject position, there was a tendency (non-statistical),
at very advanced levels, for accurately rejecting subject resumption (with null
and overt that), but it was not comparable with their accuracy in accepting the
grammatical equivalent with subject gap. In contrast, in object position, the
(very)-advanced groups performed with comparable accuracy in rejecting object
resumption (with overt that) and accepting the grammatical equivalent with an
object gap. The same level of comparable accuracy was not achieved with null
that.
5.1.2 Effects of extraction site and complementizer on resumption
Next, the effects of extraction site (subject and object) and complementizer that
(null and overt) are reported on accuracy in rejecting resumption. Based on the
Pre-emption of L1 properties 17
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
mean accuracy in Table 3, pairwise comparisons were conducted with Tukey-
Kramer adjustments.
For the extraction site, L1 effects were predicted through lower accuracy on
subject versus object resumption. This prediction was based on the optional use
of object resumption in Saudi Arabic compared with subject resumption that is
required whenever the embedded verb does not show full subject-verb agree-
ment.
7
The results from the advanced group showed a tendency for
lower accuracy on subject than object resumption with null and overt that (see
Figure 1). The observed effect of the extraction site, however, was non-statistical.
Figure 1: The effects of extraction site and complementizer on accuracy in rejecting resumption.
7An anonymous reviewer asked if there was a control of presence vs. absence of person agreement
in the experiment. In the relevant condition of subject extraction with overt that, two of the four
items had an embedded verb in the 3rd person form (requires and is), and the other two did not
show person agreement (will win and had kidnapped). However, the author admits that this factor
was not built in the design of the experiment. This prediction was mainly based on an assumption
that while object resumption is optional, subject resumption can be obligatory irrespective of all the
details. As shown in the description of the Saudi Arabicdata, the distribution of subject resumption
is quite complex. Although subject resumption was not tested on the fine details, the current
prediction is the closest to the facts of the Arabic data. Most important, the L2 speakersjudgments
did not contradict the prediction. In Greek, another null subject language, Tsimpli and
Dimitrakopoulou (2007) predicted the same on the grounds that subject-verb agreement can have
a resumptive function. The current findings are just the start, and more research is required to
uncover the complexities involved in the licensing of subject resumption in L1 and L2 data.
18 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
In the condition of null that, the effect was (marginally) non-significant with a
moderate effect size (p= 0.053, d= -0.80) and non-significant with overt that
(p= 0.82, d= -0.33). The observed effect of the extraction site significantly
decreased at very advanced levels. Like the NSs (ps > 0.05, ds -0.39), the very
advanced showed comparable accuracy on subject and object resumption with
null that (p= 1.00, d= 0.08) and overt that (p= 0.92, d= -0.31).
Regarding the complementizer, no effect was predicted because in Saudi
Arabic the insertion of subject and object resumption is not contingent on the
presence/absence of innu. As predicted, there was no effect for the complemen-
tizer (null or overt) on accuracy in rejecting subject or object resumption by the
(very)-advanced groups (see Figure 1). Like the NSs (ps > 0.05, ds -0.35), the
very advanced showed comparable accuracy by complementizer on subject
(p= 1.00, d= -0.06) and object resumption (p= 0.52, d= -0.52). Likewise, the
advanced showed comparable accuracy by complementizer with subject
(p= 1.00, d= -0.01) and object resumption (p= 0.12, d= 0.33).
The results also showed a significant improvement in accuracy in rejecting
resumption with increased proficiency. In independent samples t-tests, the very
advanced showed significant higher accuracy judgments than the advanced in
most conditions of resumption (tranging from 2.30 to 2.07,ps0.05, ds 0.79),
except for object resumption with a null that, where no significant differences
were attested, t(32) = 0.70, p= 0.49, d= -0.24 (see Figure 1 and Table 3).
5.1.3 Effects of D-linking and animacy on resumption
The final analysis examined the role of interpretable features of D-linking and
animacy on accuracy in rejecting resumption. As Table 4 shows, the mean
accuracy for each group was recomputed by D-linking (+/-DL) and animacy
(+/-animate). A D-linking effect was predicted through lower accuracy
on which-NP because in Saudi Arabic resumption is preferred with D-linked
Table 4: Mean accuracy (SD) on resumption by D-linking and animacy.
Groups/Conditions +Animate -Animate
-DL (k=) +DL (k=) -DL (k=) +DL (k=)
Native speakers (n=).(.).(.).(.).(.)
Very advanced (n=).(.).(.).(.).(.)
Advanced (n=).(.).(.).(.).(.)
Note: who =(+Animate, -DL); which teacher =(+Animate, +DL); what (-Animate, -DL); which book
(-Animate, +DL)
Pre-emption of L1 properties 19
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
wh-interrogatives. Moreover, if animacy regulates the realization of resumption,
lower accuracy was predicted with animate wh-forms because the equivalent of
inanimate what cannot be resumed in Saudi Arabic.
As predicted, results from ANOVAs (2 x 2 repeated measures) showed a
significant main effect for D-linking on the accuracy of the very advanced,
F(1,14) = 13.47, p= 0.003, ηp2= 0.49, and advanced, F(1,18) = 35.81, p< 0.001,
ηp2= 0.67, but not the NSs, F(1,14) = 3.10, p= 0.10, ηp2= 0.18. Subsequent pair-
wise comparisons confirmed significant lower accuracy on + DL than -DL by the
very advanced (p= 0.003, d= -0.95) and advanced (p< 0.001, d=1.37) (see
Figure 2). In contrast, the results showed no significant main effect for animacy
on the accuracy of the NSs, F(1,14) = 0.960, p= 0.34, ηp2= 0.06, very advanced,
F(1,14) = 0.052, p= 0.82, ηp2= 0.00, and the advanced, F(1,18) = 2.39, p= 0.14,
ηp2= 0.12 (see Figure 2).
The results also showed no significant interaction effect between D-linking and
animacy on the accuracy of the NSs, F(1,14) = 0.351, p= 0.56, ηp2= 0.02, very
advanced, F(1,14) = 4.61, p= 0.050, ηp2= 0.25, and the advanced, F(1,18) = 2.53,
p= 0.13, ηp2= 0.12. Although the very advanced achieved a (borderline) non-
significant interaction effect, pairwise comparisons confirmed they gave lower
Figure 2: Main effects of D-Linking and animacy on accuracy in rejecting resumption.
20 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
accuracy judgments to +DL than -DL in animate (p= 0.016, d=0.73) and
inanimate (p= 0.003, d=1.02) conditions.
As Figure 2 shows, the attested D-linking effect decreased with increased
proficiency. The accuracy rejection of +DL by the very advanced was signifi-
cantly higher than that of the advanced, t(32) = 2.20, p= 0.035, d= 0.75, whereas
the accuracy rejection of -DL by the very advanced and advanced groups did not
differ significantly, t(32) = 1.72, p= 0.096, d= 0.59.
Since the effect of D-linking in Saudi Arabic is observed in object position, it
was crucial to check whether it was restricted to object position in the L2
grammars. Table 5 shows the mean accuracy on D-linking (+/-DL) by extraction
site (subject and object).
The results from ANOVAs (2 x 2 repeated measures) again confirmed a signifi-
cant main effect for D-linking by the (very)-advanced groups unlike the NSs. The
results showed an interaction effect for D-linking with the extraction site by the
advanced, F(1,18) = 8.59, p= 0.009, η2
p= 0.32, but not the very advanced,
F(1,14) = 0.25, p= 0.63, ηp2= 0.02. As presented in Table 5, the advanced showed
a significant D-linking effect in the object position as would be possible in L1,
and a borderline (non-significant) effect in the subject position. Conversely, the
very advanced showed a significant D-linking effect in subject and object posi-
tions alike.
In summary, the results on L1 effects (5.1.2) and interpretable features
(5.1.3) revealed three main findings as follows: (a) there was a tendency for
an effect of extraction site that disappeared in very advanced levels (subject
and object resumption become indistinguishable); (b) there was no effect for
the complementizer that (null and overt) on the rejection accuracy of subject or
object resumption; (c) there was a clear effect for D-linking but no effect for
Table 5: Mean accuracy (SD) on resumption by D-linking for each extraction site.
Groups/Conditions Subject Extraction Object Extraction
-DL +DL pd-DL +DL pd
(k=)(k=)(k=)(k=)
Native speakers . . . . . . . .
(n=)(.)(.)(.)(.)
Very advanced . . . . . . . .
(n=)(.)(.)(.)(.)
Advanced . . . . . . <. .
(n=)(.)(.)(.)(.)
Note: -DL (who/what); +DL (which-NP).
Pre-emption of L1 properties 21
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
animacy on the rejection accuracy of resumption. The D-linking effect
decreased with increased proficiency levels.
5.2 Accuracy judgments by AO L2 groups
To answer question (3) on age effects, the L2 speakers were divided into two AO
groups: 16 years (n= 19) and 713 years (n= 15). Age six was chosen to mark the
groupsboundary division because it is assumed to mark the closure of peak
sensitivity for morpho-syntax (Granena and Long 2013). Although the AO groups
were comparable in age at testing, t(32) = 1.33, p= 0.19, d= 0.46, they varied in
length of exposure, t(32) = 5.13, p< 0.001, d= 1.80. Crucially, they showed com-
parable high L2 proficiency levels, t(31.81) = 1.36, p= 0.18, d= 0.47 (see Table 6).
Two analyses were performed: (a) group-based analyses on the effect of AO on
accuracy in rejecting resumption in various conditions (5.2.1), and (b) individual-
based analyses by AO on the gap-resumption conditions (5.2.2).
5.2.1 Group results by AO on accuracy in rejecting resumption
At first, the effect of AO on L2 group accuracy in rejecting resumption was
tested in four conditions crossing extraction site (subject and object) and
complementizer (null and overt). As shown in Table 7, the group of AO 16
gave slightly higher rejection accuracy ratings than the group of AO 713 did.
The results from independent samples t-test confirmed non-significant differ-
ences between the rejection accuracy of both AO groups on the tested condi-
tions (see Table 7).
Table 6: Demographic details for the L2 speakers by AO (N=34).
Age at onset Age at testing Proficiency score LOE
Min-Max Min-Max Min-Max Min-Max
AO groups n Mean (SD) Mean (SD) Mean (SD) Mean (SD)
AO    
. (.). (.). (.). (.)
AO      
. (.). (.). (.). (.)
Note: Maximum proficiency score =150. AO =Age of onset. LOE =Length of exposure. Age and
LOE are in years. LOE is the sum of years using English up to time of testing.
22 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
Next, the effect of AO on L2 group accuracy in rejecting resumption was tested
by D-linking and animacy. As shown in Table 8, the rejection accuracy of the
group of AO 16 was higher than that of the AO 713 on the D-linking condi-
tions but slightly higher on the animacy conditions. The results from indepen-
dent sample t-test confirmed no significant differences between both AO groups
in their rejection accuracy of resumption in D-linking and animacy conditions
(see Table 8). When testing the judgments of each AO group on paired condi-
tions of D-linking (+/-DL) and animacy (+/-animate), it was shown that both AO
groups rated resumption in D-linked wh-interrogatives less accurately than non-
D-linked ones, but gave comparable judgments on animate and inanimate
ones (see Table 8). In paired samples t-tests, the results confirmed a significant
D-linking effect in the judgements of AO 16, t(18) = 5.13, p<0.001, d= 1.25, and
AO 713, t(14) = 4.45, p= 0.001, d= 1.17. The results also confirmed no signifi-
cant animacy effect in the judgments of AO 16, t(18) = 0.949, p=0.36, d=0.23,
and AO 713, t(14) = 0.807, p=0.43, d=0.21.
Table 8: Mean accuracy (SD) by AO groups on resumption by D-linking and animacy.
Conditions AO groups t-test
AO (n=)AO (n=)t()pd
+DL (k=). (.). (.). . .
-DL (k=). (.). (.). . .
+Animate (k=). (.). (.). . .
-Animate (k=). (.). (.). . .
Note: AO =Age of onset. +DL (which-NP); -DL (who/what); +Animate (who/which teacher);
-Animate (what/which book). p<0.05, d=effect size.
Table 7: Mean accuracy (SD) by AO groups on resumption by extraction site and
complementizer.
Conditions AO groups t-test
AO (n=)AO (n=)t()pd
Subject extraction (null that)(k=). (.). (.). . .
Subject extraction (overt that)(k=). (.). (.). . .
Object extraction (null that)(k=). (.). (.). . .
Object extraction (overt that)(k=). (.). (.). . .
Note: AO =Age of onset. p<0.05, d=effect size.
Pre-emption of L1 properties 23
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
5.2.2 Individual results by AO on the gap-resumption distinction
In addition to group results, individual performance was examined within AO
groups (16 and 713) to uncover any potential exceptional cases which could
pass as target-like in their judgments on the gap-resumption distinction. For this
analysis, responses on the five-point scale were reduced into binary responses
so that the number of correct responses for each participant could be counted
per (un)grammatical conditions. Unlike the ungrammatical items, if a gramma-
tical item were incorrectly rejected, it is not obvious why that item was rejected.
Therefore, it was decided to set up a criterion of absolute accuracy only with the
grammatical items (4/4 items per condition correctly accepted = 100 %). In con-
trast, absolute accuracy with ungrammatical items was not realistically expected
with L2 data. Therefore, 75 % (3/4 items per condition correctly rejected) was
chosen as the cut-off point for judging that a participant had pre-empted
resumption in L2 English wh-interrogatives. If the participant had scored abso-
lute accuracy (100 %) on the grammatical condition, and at the same time
achieved high accuracy (75 %) on the ungrammatical equivalent, they were
judged to have successfully made the distinction between the non-target
resumptive strategy and the target gap strategy. As a benchmark, the English
NSs were checked individually. In object extraction (null and overt that), all NSs
made the distinction (100 %), and in subject extraction over null that, all but one
made the distinction (93 %). In subject extraction with overt that, the NSs were
divided: those who made the distinction (47 %) and those who did not (53%).
This was expected due to dialect variations on English that-trace. Research has
shown that English NSs of different dialects are not often categorical in rejecting
that-trace violations (see Sobin 2002, 2009).
Table 9 shows the number of L2 speakers who made the gap-resumption
distinction by AO. Using Pearsons chi-square, the results from all conditions
consistently showed non-significant differences by AO in the number of L2
speakers who successfully made the gap-resumption distinction (see Table 9).
Hence, the proportion of L2 speakers who could perform target-like did not
differ by AO. These individual results support the findings from group results
(as in Tables 7 and 8). It is important to note that among the AO 16group
there were 13 participants who had their first exposure to L2 in an ESC (with an
averageofsevenyearsoflengthofresidence)andyettheydidnotallsuccess-
fully make the distinction between grammatical and ungrammatical wh-
interrogatives.
In summary, the results showed no main effects for AO on the L2 groups
accuracy in rejecting resumption in subject and object wh-interrogatives (null
and overt that). Regardless of AO, the results showed an effect for D-linking (but
24 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
not animacy) on the rejection accuracy of resumption. Individual results also
revealed no differences by AO on the gap-resumption comparisons.
6 Discussion
This study examines whether L2 speakers can pre-empt an L1 property which
involves uninterpretable features, such as resumption. The tolerance of resump-
tion was tested in the L2A of English complex wh-interrogatives by L1 Saudi
Arabic speakers. Unlike English, which requires a gap in the wh-extraction site,
Saudi Arabic licenses grammatical resumption or a gap in the extraction site.
Under the IH (Tsimpli and Dimitrakopoulou 2007), if uninterpretable features
are grammaticalized in L1 but not in L2, they cannot be pre-empted from L2
grammars. Therefore, L2 grammars will show variability in tolerating the
resumptive and gap strategies, as would be possible in the L1. However, inter-
pretable features of D-linking and animacy are assumed to constrain variability
in tolerating resumption. Since the IH adopts assumptions of the critical period,
the effects of high L2 proficiency (very advanced and advanced) and AO (16
and 713) are considered in the investigation of persistent L1 effects.
A bimodal timed AJT was used to measure accuracy ratings of (un)gramma-
tical subject/object wh-interrogatives with resumption and gaps. Importantly,
the L2 speakersratings were elicited under time pressure to minimize use of
metalinguistic knowledge. Variability was measured in terms of incomparable
accuracy judgments on the gap-resumption distinction. As predicted by the IH,
the L2 grammars of (very)-advanced L1 Saudi Arabic speakers showed a degree
of variability in their accuracy judgments of resumption and gaps, which was
Table 9: Number of individuals (%) by AO L2 groups who made the distinction on the gap and
resumptive strategies.
Conditions AO groups Chi-square
AO (n=)AO (n=)χ2p
Subject extraction (null that)(k=)/ ( %) / ( %) . .
Subject extraction (overt that)(k=)/ ( %) / ( %) .
a
.
Object extraction (null that)(k=)/ ( %) / ( %) . .
Object extraction (overt that)(k=)/ ( %) / ( %) . .
Note: AO =Age of onset. χ2(N =34, df =1).
a
Fishers Exact Test was used for measuring signifi-
cance because of violating the expected sample size per cell. p<0.05.
Pre-emption of L1 properties 25
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
more prevalent with D-linked wh-interrogatives. Interestingly, there were no
significant differences due to AO. Evidence for persistent L1 effects on resump-
tion tolerance, which involves uninterpretable features, is discussed in terms of
the parametric differences between Saudi Arabic and English, and the role of
interpretable features of D-linking and animacy.
First, the effects of extraction site and complementizer that on resumption
tolerance are considered. For extraction site, variability in (very)-advanced
L2 grammars was found in subject more than object positions. Recall, in Saudi
Arabic, the realization of subject resumption is obligatory whenever agreement
is not fully marked on the embedded verb, whilst object resumption is optional
in most conditions. Comparing rejectionsaccuracy of subject and object
resumption, the results showed a tendency for subject pronouns being more
tolerated than objects. The effect of extraction site decreased significantly at very
advanced levels, by which subject and object pronouns become indistinguish-
able. Nonetheless, the remnant of the effect still surfaced in the gap-resumption
comparisons. Unlike their performance on subject wh-interrogatives, the (very)-
advanced groups managed to show target-like comparable accuracy with object
wh-interrogatives (gap and resumption), at least with overt that. Therefore, their
accuracy judgments from both analyses are taken as showing a tendency for pre-
empting subject resumption at very advanced levels. On the other hand, their
accuracy showed no effect for the complementizer (null or overt) on the toler-
ance of subject and object resumption. This finding was expected due to the
optional presence of innu with subject/object wh-extraction in Saudi Arabic. In
fact, the Saudi Arabic L2 groups correctly rejected ungrammatical subject wh-
interrogatives where that was followed by a subject pronoun more often than
when it was followed by a subject gap. Assuming L1 transfer, if resumption were
obligatory, lower accuracy would have been expected in the presence of subject
pronoun following that.
Having seen resumption still being tolerated with varying degrees in (very)-
advanced L2 grammars, the role of interpretable features is now discussed. Under
the IH, the extent of variability is predicted to be constrained by the interpretable
features of D-linking and animacy involved in the derivation of wh-interrogatives.
Although D-linking is marked in English and Saudi Arabic wh-forms, D-linking in
Saudi Arabic (unlike English) triggers the use of resumption. Recall, in Saudi
Arabic, object resumption is highly preferred with D-linked ʔayy-NP which-NP
but disallowed with non-D-linked ʔeeʃwhat. It was also assumed that object
resumption is not sensitive to animacy contrasts. Unlike miin who,ʔeeʃwhat
cannot be resumed because it cannot be contextually D-linked, but not because of
its inanimate feature. The results showed a strong D-linking effect on resumption
tolerance and no animacy effect. The L1 effects of D-linking are mapped into L2
26 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
English wh-interrogatives along with the resumption strategy. Therefore, the
which-NP interrogatives triggered lower accuracy judgments than who/what-inter-
rogatives. The D-linking effect crucially decreased in the L2 judgments of the very
advanced compared with the advanced. Although the D-linking effect
was restricted to object position by the advanced as would be possible in L1, its
effect was generalized with subject and object resumption by the very advanced.
The D-linking effect seen in very advanced grammars cannot be taken as an effect
of direct transfer from L1. It is perhaps more indicative of the compensatory role of
interpretable features assumed by Tsimpli and Dimitrakopoulou (2007). To com-
pensate for the persistent L1 effects of uninterpretable features, they formed a rule
that resumption is restricted to D-linked conditions across the board. It could be
argued, however, that the generalization of the D-linking effect across both
extraction sites expanded rather than constrained the range of resumption toler-
ance. Since the compensatory role of interpretable features is not clearly defined,
it is hard to evaluate the directionality of their presumed role. Different from
Tsimpli and Dimitrakopoulou (2007), animacy did not constrain resumption tol-
erance by the Saudi Arabic L2 speakers. In their study, with increased proficiency,
resumption was prevalent in inanimate D-linked conditions. It was not explained
why animate pronouns did not trigger higher rates of resumption tolerance. They
argued that animacy, bearing interpretable features, was successfully acquired
with English wh-forms and pronouns even though it was not marked on the Greek
equivalents. Having acquired animacy helped the L1 Greek speakers restrict
resumption tolerance to inanimate pronouns, and thus improved their L2 perfor-
mance. The effect of interpretable features deserves more research to understand
the directionality of its effect and how it could explain successful L2 performance.
The overall results from the Saudi Arabic L2 groups are compatible with
those of Tsimpli and Dimitrakopoulou (2007). Nonetheless, L2 data from near-
natives is needed, especially with the observed tendency in this study to pre-
empt resumption in subject wh-interrogatives. There is also recent evidence for
successful rejection of resumption in English wh-interrogatives by advanced
Spanish L2 speakers (Leal-Méndez and Slabakova 2014) and Kuwaiti Arabic L2
speakers (Leal-Méndez et al. 2016). In a contextualized but untimed AJT, they
reported that none of the factors of extraction site, complementizer, D-linking
and animacy were significantly attested in their advanced L2 data. Lack of time
pressure and provision of context are two differential factors in their AJT that
could explain the different findings. In future research, the effect of provision of
context but under time pressure on L2 judgments deserves testing to compare
with the effect of lack of context under time pressure on L2 judgments.
Turning to age effects, the results from group and individual analyses showed
no advantages for a younger start. This outcome could be due to the lack of
Pre-emption of L1 properties 27
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
variation in the first place among a group of (very)-advanced L2 speakers even
when they vary by AO (see DeKeyser and Larson-Hall 2005; White and Genesee
1996). The results from the AO 16 group showed a D-linking effect where which-
NP was rated less accurately than who/what. This finding is evidence for persistent
L1 effects regardless of AO at a very subtle detail of L2 grammar. It is interesting to
note that 13/19 participants in the AO 16 group had their first exposure to L2
English in an ESC (with an average of seven years of length of residence) and yet
they did not all perform target-like. A comparable outcome of no advantages for a
younger start on a set of morpho-syntactic properties including resumption was
obtained from another group of instructed early L2 starters of English in Saudi
Arabia with no study abroad experience (see Al-Thubaiti 2010). Regardless of type
of exposure and high proficiency, these results indicate that target-like attainment
of morpho-syntactic properties is often not guaranteed even before the AO of six
years (see Granena and Long 2013). From a generative L2 study, the current
findings of AO are consistent with those of age-related studies that are non-
generative (e. g. Abrahamsson and Hyltenstam 2009; Hyltenstam 1992).
To conclude, as grammatical resumption involves uninterpretable features,
once activated in L1, it is often hard to pre-empt in L2 grammar regardless of AO
and high L2 proficiency. It could be argued, however, that the L2 speakers in this
study were not tested in a naturalistic setting. Although this reasoning is partly
correct, it cannot be ignored that the D-linking effect was found among the young
L2 starters who had study-abroad experience. This piece of evidence is promising
and needs validation with further research. Admittedly, evidence from near-natives
would be more compelling. I also acknowledge the need for L2 data from a larger
sample of young L2 starters in naturalistic and instructed settings. In generative L2
research, the L2 grammar of young starters (AO 6) is under-researched and often
not studied in detail as much as older starters. This paper, however, is one step
forward to better understanding the extent to which variability in L2 grammars is a
characteristic of younger starters as much as we know of older starters.
Acknowledgments: I would like to thank first and foremost two anonymous
IRAL reviewers for their valuable and meticulous feedback that helped in
improving the paper in many ways. I would also like to thank Roger Hawkins
and Rex Sprouse for their advice and comments on an earlier draft of the paper.
I am so grateful to Heather Marsden for advice on handling reviewerscom-
ments. I am so thankful as well to Andrew Radford for his valuable feedback on
that-trace effects, and Lina Choueiri for her thought-provoking discussion and
input on resumption. Finally, I thank Jang Dong Seo and Samah Al-Thubaiti for
tips and advice on statistics.
28 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
References
Abrahamsson, Niclas & Kenneth Hyltenstam. 2009. Age of onset and nativelikeness in a second
language: Listener perception versus linguistic scrutiny. Language Learning 59. 249306.
Al-Thubaiti, Kholoud. 2010. Age effects in a minimal input setting on the acquisition of English
morpho-syntactic and semantic properties by L1 speakers of Arabic. Unpublished doctoral
dissertation, University of Essex, UK.
Aoun, Joseph, Elabbas Benmamoun & Lina Choueiri. 2010. The syntax of Arabic. Cambridge:
Cambridge University Press.
Aoun, Joseph & Lina Choueiri. 1999. Modes of interrogation. In Elabbas Benmamoun (ed.),
Perspectives on Arabic linguistics XII,726. Amsterdam: Benjamins.
Bley-Vroman, Robert W., Sascha W. Felix & Georgette L. Ioup. 1988. The accessibility of
Universal Grammar in adult language learning. Second Language Research 4. 132.
DeKeyser, Robert & Jenifer Larson-Hall. 2005. What does the critical period really mean?
In Judith Kroll & Annette De Groot (eds.), Handbook of bilingualism: Psycholinguistic
approaches,88108. Oxford: Oxford University Press.
Granena, Gisela & Michael Long. 2013. Age of onset, length of residence, language aptitude,
and ultimate L2 attainment in three linguistic domains. Second Language Research 29.
311343.
Hawkins, Roger & Hajime Hattori. 2006. Interpretation of English multiple wh-questions by
Japanese speakers: A missing uninterpretable feature account. Second Language Research
22. 269301.
Hyltenstam, Kenneth. 1992. Non-native features of near-native speakers: On the ultimate
attainment of childhood L2 learners. Advances in psychology 83. 351368.
Ionin, Tania. 2012. Formal theory-based methodologies. In Alison Mackey & Susan Gass (eds.),
Research methods in second language acquisition: A practical guide,3052. Malden, MA:
Wiley-Blackwell.
Ionin, Tania & Kenneth Wexler. 2002. Why is iseasier than -s?: Acquisition of tense/
agreement morphology by child second language learners of English. Second Language
Research 18. 95136.
Jaeggli, Osvaldo & Ken Safir. 1989. The null subject parameter. London: Kluwer Academic
Publishers.
Kenstowicz, Michael. 1989. The null subject parameter in modern Arabic dialects. In Osvaldo
Jaeggli & Ken Safir (eds.), The null subject parameter, 263275. Dordrecht: Kluwer.
Lardiere, Donna. 1998. Case and tense in the fossilizedsteady state. Second Language
Research 14. 126.
Leal-Méndez, Tania & Roumyana Slabakova. 2014. The interpretability hypothesis again:
A partial replication of Tsimpli and Dimitrakopoulou (2007). International Journal of
Bilingualism 18. 537557.
Leal-Méndez, Tania, Roumyana Slabakova, Ivan Ivanov & Marta Tryzna. 2016. Uninterpretable
features in L2A again: Interrogatives in the L2 English of Kuwaiti Arabic speakers. Paper
presented to the 13th Generative Approaches to Second Language Acquisition Conference,
Indiana Bloomington University, 2016.
Long, Mike. 2005. Problems with supposed counter-evidence to the critical period hypothesis.
International Review of Applied Linguistics in Language Teaching 43. 287317.
Pre-emption of L1 properties 29
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
Mackey, Alison & Susan Gass. 2005. Second language research: Methodology and design.
New Jersey: Lawrence Erlbaum Associates.
Miyagawa, Shigeru. 2010. Why agree? Why move: Unifying agreement-based and discourse-
configurational languages. Cambridge, MA: MIT Press.
Montrul, Silvina. 2008. Incomplete acquisition in bilingualism: Re-examining the age factor.
Amsterdam: Benjamins.
Muñoz, Carmen & David Singleton. 2011. A critical review of age-related research on L2 ultimate
attainment. Language Teaching 44. 135.
Murphy, Victoria. 1997. The effect of modality on a grammaticality judgement task. Second
Language Research 13. 3465.
Nation, Ian. 2001. Learning vocabulary in another language. Cambridge: Cambridge University
Press.
Nishikawa, Tomomi. 2014. Nonnativeness in near-native child L2 starters of Japanese: Age and
the acquisition of relative clauses. Applied Linguistics 35. 504529.
Pesetsky, David. 1982. Complementizer-trace phenomena and the nominative island condition.
The Linguistic Review 1. 297344.
Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In Eric Reuland & Alice
Ter Meulen (eds.), The representation of (in)definiteness,98129. Cambridge, MA: MIT
Press.
Plonsky, Luke & Frederick Oswald. 2014. How big is big? Interpreting effect sizes in L2
research. Language Learning 64. 878912.
Radford, Andrew. 2018. Colloquial English: Structure and Variation. Cambridge: Cambridge
University Press.
Rizzi, Luigi. 1990. Relativised minimality. Cambridge, MA: MIT Press.
Rizzi, Luigi. 2015. Cartography, criteria, and labeling. In UR Shlonsky (ed.), Beyond functional
sequence: The cartography of syntactic structures, 314338. Oxford: Oxford University
Press.
Rizzi, Luigi & Ur Shlonsky. 2007. Strategies of subject extraction. In Uli Sauerland & Hans-
Martin Gärtner (eds.), Interfaces +recursion =language? Chomskys minimalism and the
view from syntax-semantics., 115160. Berlin: de Gruyter.
Ross, John. 1967. Constraints on variables in syntax. MIT unpublished doctoral dissertation.
Roussou, Anna. 2002. C, T, and the subject: That-t phenomena revisited. Lingua 112. 1352.
Rouveret, Alain. 2011. Some issues in the theory of resumption: A perspective on early and
recent research. In Alain Rouveret (ed.), Resumptive pronouns at the interface,162.
Amsterdam: Benjamins.
Sobin, Nicholas. 2002. The Comp-trace effect, the adverb effect and minimal CP. Journal of
Linguistics 38. 527560.
Sobin, Nicholas. 2009. Prestige case forms and the Comptrace effect. Syntax 12. 3259.
Sorace, Antonella & Francesca Filiaci. 2006. Anaphora resolution in near-native speakers of
Italian. Second Language Research 22. 339368.
Towell, Richard & Roger Hawkins. 1994. Approaches to second language acquisition. Clevedon:
Multilingual matters.
Tsimpli, Ianthi-Maria & Maria Dimitrakopoulou. 2007. The interpretability hypothesis: Evidence
from wh-interrogatives in second language acquisition. Second Language Research 23.
215242.
Tsimpli, Ianthi-Maria & Anna Roussou. 1991. Parameter-resetting in L2. UCL Working Papers in
Linguistics 3. 149169.
30 Kholoud A. Al-Thubaiti
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
White, Lydia. 1985. The pro-drop parameter in adult second language acquisition. Language
Learning 35. 4762.
White, Lydia. 2003. Second language acquisition and universal grammar. Cambridge:
Cambridge University Press.
White, Lydia & Fred Genesee. 1996. How native is near-native? The issue of ultimate attainment
in adult second language acquisition. Second Language Research 12. 233265.
Pre-emption of L1 properties 31
Authenticated | kathubaiti@uqu.edu.sa author's copy
Download Date | 7/6/18 6:31 PM
... Since Ulrich Weinreich first introduced the concept of "interference" as "those instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language" [1] in 1953, there has been extensive research into two fields: the forward transfer from a first language (L1) to a second language (L2), and the backward or reverse transfer from L2 to L1. The former, the more obvious and practical of the two interferences, has been studied for longer in the field of second language acquisition (SLA) [2][3][4][5][6]. Backward transfer, from L2 to L1, came into focus much later and is usually explored from two dimensions: (1) The attrition or loss of L1 in a migration setting or L2-dominant context; and (2) The effect of L2 on L1 in an L1-dominant context, where L2 is learned as a foreign language (FL) and used mainly for academic purposes. ...
Article
Full-text available
The advocates of multi-competence theory argue that the L2 learners’ language system is unique because of the crosslinguistic influences of both languages. However, the influence of a foreign language on the learner’s L1 has not been extensively investigated. In order to address the gap, the present study sought to investigate the effects of EFL learning on written L1 Chinese at the lexical level. Two studies were conducted on 200 abstracts of MA theses written in Chinese, half on English literature written by Chinese-L1 English majors (EMs), and half on Chinese literature written by Chinese-L1 Chinese majors (CMs). The first study investigated the differences between the two groups in terms of the frequencies of nouns, verbs, adjectives, adverbs, pronouns, prepositions, and conjunctions in the abstracts. The second study examined the differences in the lexical complexity and diversity between the two groups. The results reveal 12 significant differences in 27 investigated word classes and subclasses, as well as significant differences in lexical complexity, but no significant difference in lexical diversity. The identified differences are discussed from a multi-competence perspective.
Chapter
Until recently, cognitive science virtually ignored the fact that most people of the world are bilingual. During the past ten years this situation has changed markedly. There is now an appreciation that learning and using more than one language is the more natural circumstance of cognition. As a result, there is a wealth of new research on second-language learning and bilingualism that provides not only crucial evidence for the universality of cognitive principles, but also an important tool for revealing constraints within the cognitive architecture. In this volume, Judith Kroll and Annette de Groot have brought together the scientists at the forefront of research on second-language learning and bilingualism to present chapters that, rather than focusing simply on their own research, provide the first comprehensive overviews of this emerging field. Bilingualism provides a lens through which each of the central questions about language and cognition can be viewed. The five sections of this book focus on different facets of those questions: How is language acquired when infants are exposed to multiple-language input from birth, and how is it acquired when adults are required to learn a second language after early childhood? How do adult bilinguals comprehend and produce words and sentences when their two languages are potentially always active and in competition with one another? What are the neural mechanisms that underlie proficient bilingualism? What are the general consequences of bilingualism for cognition and for language and thought? This handbook will be essential reading for cognitive psychologists, linguists, applied linguists, and educators who wish to better understand the cognitive basis of bilingualism and the logic of experimental and formal approaches to language science.
Book
Vocabulary is now well recognized as an important focus in language teaching and learning. Now in its third edition, this book provides an engaging, authoritative guide to the teaching and learning of vocabulary in another language. It contains descriptions of numerous vocabulary learning strategies, which are supported by reference to experimental research, case studies, and teaching experience. It also describes what vocabulary learners need to know to be effective language users. This new edition has been updated to incorporate the wealth of research that has come out of the past decade. It also includes a new chapter on out of-classroom learning, which explores the effect of the Internet and electronic resources on learning. This vital resource for all vocabulary researchers shows that by taking a systematic approach to vocabulary learning, teachers can make the best use of class time and help learners get the best return for their learning effort.
Book
This authoritative 2003 textbook provides an overview and analysis of current second language acquisition research conducted within the generative linguistic framework. Lydia White argues that second language acquisition is constrained by principles and parameters of Universal Grammar. The book focuses on characterizing and explaining the underlying linguistic competence of second language learners in terms of these contraints. Theories as to the role of Universal Grammar and the extent of mother tongue influence are presented and discussed, with particular consideration given to the nature of the interlanguage grammar at different points in development, from the initial state to ultimate attainment. Throughout the book, hypotheses maintaining that second language grammars are constrained by universal principles are contrasted with claims that Universal Grammar is not implicated; relevant empirical research is presented from both sides of the debate. This textbook is essential reading for those studying second language acquisition from a linguistic perspective.
Book
An argument that not only do movement and agreement occur in every language, they also work in tandem to imbue natural language with enormous expressive power. An unusual property of human language is the existence of movement operations. Modern syntactic theory from its inception has dealt with the puzzle of why movement should occur. In this monograph, Shigeru Miyagawa combines this question with another, that of the occurrence of agreement systems. Using data from a wide range of languages, he argues that movement and agreement work in tandem to achieve a specific goal: to imbue natural language with enormous expressive power. Without movement and agreement, he contends, human language would be merely a shadow of itself, with severe limitation on what can be expressed. Miyagawa investigates a variety of languages, including English, Japanese, Bantu languages, Romance languages, Finnish, and Chinese. He finds that every language manifests some kind of agreement, some in the form of the familiar person/number/gender system and others in the form of what Katalin É. Kiss calls “discourse configurational” features such as topic and focus. A key proposal of his argument is that the computational system in syntax deals with the wide range of agreement types uniformly—as if there were just one system—and an integral part of this computation turns out to be movement. Why Agree? Why Move? is unique in proposing a unified system for movement and agreement across language groups that are vastly diverse—Bantu languages, East Asian languages, Indo-European languages, and others.
Book
Cambridge Core - Sociolinguistics - Colloquial English - by Andrew Radford
Book
Cambridge Core - ELT Applied Linguistics - Learning Vocabulary in Another Language - by I. S. P. Nation