Bilingualism: Language and Cognition

http://journals.cambridge.org/BIL

Additional services for Bilingualism: Language and Cognition:

Email alerts: Click here

Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here

Second language sentence processing in reading for comprehension and

translation

JUNG HYUN LIM and KIEL CHRISTIANSON

Bilingualism: Language and Cognition / Volume 16 / Issue 03 / July 2013, pp 518 - 537
DOI: 10.1017/S1366728912000351, Published online: 06 September 2012

Link to this article: http://journals.cambridge.org/abstract_S1366728912000351

How to cite this article:

JUNG HYUN LIM and KIEL CHRISTIANSON (2013). Second language sentence processing in reading for comprehension
and translation. Bilingualism: Language and Cognition, 16, pp 518-537 doi:10.1017/S1366728912000351

Request Permissions : Click here

Downloaded from http://journals.cambridge.org/BIL, IP address: 130.126.255.100 on 31 Oct 2013

 Bilingualism: Language and Cognition 16 (3), 2013, 518–537 
C Cambridge University Press 2012 doi:10.1017/S1366728912000351

JUNG HYUN LIM

Second language sentence University of California, San Diego
processing in reading for KIEL CHRISTIANSON
University of Illinois, Urbana–Champaign & Beckman
comprehension and Institute for Advanced Science and Technology

translation∗
(Received: October 22, 2011; final revision received: June 21, 2012; accepted: June 22, 2012; first published online 6 September 2012)

A self-paced reading and translation task was used with learners of English as a second language (L2) to explore what sorts
of information L2 learners use during online comprehension compared to native speakers, and how task (reading for
comprehension vs. translation) and proficiency affect L2 comprehension. Thirty-six Korean native speakers of English and 32
native English speakers read plausible and implausible subject relative clauses and object relative clauses. Reading times,
comprehension accuracy, and translations were analyzed. Results showed that L2 learners were able to use syntactic
information similarly to native speakers during comprehension, and that online L2 processing and offline comprehension
were modulated by reading goals and proficiency. Results are interpreted as showing that L2 processing is quantitatively
rather than qualitatively different from first language processing, i.e. strategically “good enough”.

Keywords: L2 sentence processing, proficiency and task effects, Good-Enough processing, translation, integration of syntactic and
semantic information

Several decades of research in second language investigation of processing mechanisms that non-native
acquisition (SLA) have revealed much information about speakers employ during L2 comprehension.
WHAT can be acquired in second language learning Much of the recent research on second language (L2)
(knowledge). The study of SLA also focuses on the processing has explored how non-native speakers process
issue of HOW a second language is used (performance), language in real time when they read sentences in the L2,
because both the sufficient knowledge of grammatical employing psycholinguistic techniques such as self-paced
rules in a target language and the ability to apply the reading or eye-tracking paradigms (Felser, Roberts, Gross
proper processing routines are necessary for successful & Marinis, 2003; Frenck-Mestre & Pynte, 1997; Jackson,
second language (L2) mastery (Clahsen & Felser, 2006a; 2008; Juffs, 1998; Marinis, Roberts, Felser & Clahsen,
Gregg, 1996). Accordingly, a question of whether 2005; Papadopoulou & Clahsen, 2003). One of the key
differences in ultimate attainment are in part attributable questions in this literature is to what extent the behavior
to underlying processing mechanisms in adult L2 learners of L2 speakers varies from adult native speakers. In
has been raised in recent years. While there have particular, many studies have examined L2 speakers’ use
been many experimental studies on how native speakers of lexical and morphosyntactic information during online
process their first language (L1) in real time using reading comprehension compared to native speakers.
psycholinguistic techniques, it is only recently that the Some of these studies have indicated that there are
same psycholinguistic methods have been applied to the characteristic differences between adult L2 learners and
adult native speakers in grammatical processing (Felser
et al., 2003; Marinis et al., 2005; Papadopoulou
* An earlier version of this research was presented at the Second & Clahsen, 2003), while others have suggested that
Language Research Forum (University of Maryland, 2010). The their respective underlying processing mechanisms are
authors thank the members of the Educational Psychology
Psycholinguistics Lab for helping with data collection and Steve
basically the same (Frenck-Mestre & Pynte, 1997;
Luke for advice about data analysis. The authors also thank three Jackson, 2008; Juffs, 1998). The issue of how language
anonymous reviewers for their valuable comments. This research learners process L2 input still needs further investigations
was funded in part by a UIUC Campus Research Board grant, a in part due to the discrepancies in existing research. In
summer GA support award from the UIUC Bureau of Educational addition, a variety of structures, different populations,
Research, and NSF BCS-0847533 to Kiel Christianson and a
Cognitive Science/Artificial Intelligence (CS/AI) award from the
individual differences, proficiency level, and task
Beckman Institute to Jung Hyun Lim. demands should be considered (Clahsen & Felser, 2006a;

Address for correspondence:

Jung Hyun Lim, Department of Educational Psychology, Education Building, MC-708 University of Illinois, 1310 S. 6th St., Champaign,
IL 61820, USA
junglim@illinoisalumni.org

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 519

Dussias, 2001; Jackson, 2008; Juffs, 1998) to fully information. These results were replicated by Felser et al.
understand L2 processing mechanisms. (2003) with temporarily ambiguous relative clauses,
The present study adds to research exploring L2 suggesting that native and non-native speakers differ in
processing mechanisms by examining how Korean that the L2 comprehension mechanisms do not always
learners of English read subject/object relative clauses integrate phrase-structure information immediately.
(SRCs/ORCs) that are manipulated with respect to the Marinis et al. (2005) provided additional evidence that
semantic properties of the events they depict. This L2 speakers lack detailed syntactic representations during
study further compares processing patterns in two online comprehension, particularly in using intermediate
reading tasks (reading for comprehension vs. reading syntactic gap information involving long-distance wh-
for translation). A few studies report the roles of phrases. In a self-paced reading task, L2 learners of
secondary tasks on influencing L2 speakers’ reading English (Chinese, Japanese, Greek, and German L1)
time patterns (Jackson & Bobb, 2009; Leeser, Brandl and native English speakers read sentences involving
& Whiteglass, 2011; Williams, 2006), demonstrating long-distance wh-dependencies. It was found that the L1
that secondary tasks such as an explicit grammaticality English speakers made use of the intermediate syntactic
judgment or a memory task increase L2 learners’ attention gap to retrieve the filler to integrate with its subcategorizer,
to grammatical structures. Given these observations, whereas the L2 speaker groups failed to postulate such
the current study also investigates how different tasks intermediate gaps during processing. These results were
modulate learners’ awareness of syntactic information also replicated in Felser & Roberts (2007) using a cross-
during online comprehension. L2 proficiency is also modal picture-priming task. To account for the above
considered, as several studies suggest that proficiency differences between L1 and L2 processing, the S HALLOW
significantly affects processing (e.g., Hopp, 2006; S TRUCTURE H YPOTHESIS (SSH) has been proposed
Jackson, 2008). (Clahsen & Felser, 2006a), arguing that grammatical
This article is organized as follows. First, we processing in language learners is qualitatively different
summarize recent findings about L2 processing research from that of native speakers. The main idea of the
and review Good-Enough (GE) processing (Christianson, hypothesis is that the syntactic representations learners
Hollingworth, Halliwell & Ferreira, 2001; Ferreira, Bailey compute during online comprehension lack the full-
& Ferraro, 2002; Ferreira, Christianson & Hollingworth, fledged L2 grammar, and thus learners are dependent on
2001; Ferreira & Patson, 2007), which is the main lexical-semantic or pragmatic knowledge to compensate
theoretical focus in this study. Then, the rationale of using for incomplete syntactic structure.
a translation task is illustrated. The subsequent section On the other hand, a contrasting set of studies
describes research questions along with the predictions suggests that L2 speakers’ ability to use certain types
and methods of the current study, followed by our of morphosyntactic information during online processing
results, which are interpreted within current L2 processing is quantitatively rather than qualitatively different from
research. that of native speakers. Frenck-Mestre and Pynte
(1997) showed that English–French and French–English
bilinguals performed complete syntactic parses and
L2 sentence processing research
patterned in the same manner as monolingual counterparts
Some studies that have examined L2 speakers’ real- in reading sentences containing temporarily ambiguous
time comprehension provide evidence that L2 speakers prepositional phrases as in the English They accused the
lack the ability to use particular L2 structures during ambassador of espionage/Indonesia but nothing came
online processing. Papadopoulou and Clahsen (2003) of it, and He rejected the manuscript on purpose/on
found that using a self-paced reading paradigm, both horses because he hated its author. Non-syntactic, lexical
high-proficiency L2 learners (Spanish, German, and information also reliably influenced reading L2 sentences.
Russian L1) of Greek and a Greek native group preferred The authors claimed that non-native processing is not
low attachment in reading ambiguous relative clauses necessarily different from native processing, at least with
that contain a thematic preposition “with”. The results, respect to this particular structure.
however, included the finding that the L2 learners did Juffs (1998) also investigated whether L2 speakers are
not show any attachment preferences when a sentence able to process sentences that contain main verb/reduced
contains the preposition “of” as in “A man looked at the relative ambiguities in the same way as native speakers,
teacher of the pupil who was in the school yard”, whereas manipulating argument structures and lexical-semantic
native speakers constantly showed a high-attachment cues. The results of an acceptability judgment task and
preference. The authors concluded from these results a self-paced reading task suggested that the learners of
that L2 learners have difficulty integrating syntactic English as a second language (ESL) deployed complex
information during online comprehension, while native knowledge of English in online tasks and that the ESL
speakers are able to use both syntactic and lexical learners were sensitive to the interaction of various

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 520 Jung Hyun Lim and Kiel Christianson

information sources in much the same way as native in the L2 processing mechanism (Hopp, 2006; Jackson &
speakers. Another study, by Jackson and Dussias (2009), Bobb, 2009; Leeser et al., 2011; Williams, 2006)
also indicated that highly proficient L2 German speakers
were able to fully utilize case-marking information in wh-
Good-Enough processing
question sentences during online reading comprehension.
Furthermore, an L2 proficiency effect has been Interestingly, shallow processing does not appear to be
reported in several studies (Frenck-Mestre, 2002; Hopp, unique to L2 learners; the “shallow and less detailed”
2006; Jackson, 2008), where advanced speakers have language processing discussed in the L2 processing
been found to use morphosyntactic information more literature has also been observed in monolingual
automatically than less proficient speakers. Hopp (2006) processing studies. The G OOD -E NOUGH (GE) APPROACH
investigated L2 processing of subject–object ambiguities to language comprehension (Christianson et al., 2001;
in German with both advanced and near-native L2 learners Christianson & Luke, 2011; Christianson, Luke &
(L1 English and Dutch). German has a canonical subject– Ferreira, 2010; Christianson, Williams, Zacks & Ferreira,
object–verb (SOV) word order, but the object can precede 2006; Ferreira et al., 2002; Ferreira & Patson, 2007)
the subject in subordinate clauses. The question of interest proposes that comprehenders compute linguistic input
was whether advanced and near-native L2 speakers of using both syntactic and semantic or heuristic processing,
German would be able to use syntactic disambiguation but the outputs of these routes are not always reconciled
information (case-marking vs. verb agreement) in the if one output appears good enough for comprehension,
same way as native German speakers when processing thus yielding a final interpretation that is not completely
subject–object ambiguities. Results showed that near- faithful to the input. GE language comprehension has
native L2 speakers patterned like native speakers in online been supported by studies using active/passive structures
reading time, consistently showing local slowdowns in the (Christianson et al., 2010; Ferreira, 2003), and garden-
critical region of OS sentences in both types of syntactic path sentences (Christianson et al., 2001, 2006).
disambiguation (case vs. verbal agreement) conditions. Ferreira (2003) found that native English speakers
On the other hand, advanced L2 speakers’ reading times were more likely to misinterpret simple passive sentences
were generally slow in the OS word order only at the end when syntactic structure was inconsistent with world
of the sentence, regardless of syntactic disambiguation knowledge, as in The cat was chased by the mouse. In
types. These findings emphasized the role of proficiency particular, participants were less accurate in implausible
in L2 processing mechanisms, suggesting that sensitivity sentences (80%) than plausible sentences (90%) for
to syntactic information is not restricted in non-native passives, whereas for actives, they were as accurate
processing. with implausible sentences (95%) as with plausible
Jackson (2008) extended Hopp’s (2006) study, sentences (98%). The frequency of the syntactic structure
examining how L2 speakers process temporarily did not influence misinterpretations, as comprehenders
ambiguous wh-questions compared to native German responded to infrequent subject clefts as accurately
speakers. Accuracy results showed that both native and as to active sentences. These findings indicated that
non-native speakers had greater difficulty in reading native English speakers tend to employ heuristic
temporarily ambiguous OS word order than SO word processing mechanisms along with a syntactic processing
order. The online reading time results showed that the algorithm. These heuristics include semantic plausibility
use of case-marking information varied according to and Noun–Verb–Noun (NVN) word order template that
L2 proficiency level. Native speakers and advanced maps constraints of this order onto Agent–Verb–Patient
L2 speakers assigned grammatical roles incrementally, interpretations (Townsend & Bever, 2001; see also the
whereas intermediate L2 speakers did not show immediate Competition Model in connection to linear templates,
sensitivity to case-marking information. e.g., Bates & MacWhinney, 1987). The passive structure,
To summarize, the L2 studies conducted so far have which has an atypical Patient-before-Agent order, is
provided mixed results; some of the studies show L2 apparently harder to process than the active structure due
speaker’s ability to use syntactic knowledge during online to the conflict between the syntactic algorithm and the
comprehension in the same manner as native speakers semantic heuristic. The syntactic representations in the
(Frenck-Mestre & Pynte, 1997; Hopp, 2006; Jackson, comprehender’s mind are thus considered to be “fragile”,
2008; Jackson & Bobb, 2009; Williams, 2006), whereas and the heuristic parser to be “fast and frugal” at reaching
others suggest that L2 processing lacks abstract linguistic an interpretation of a sentence when processing linguistic
structure compared to native speakers (Felser et al., input.
2003; Felser & Roberts, 2007; Marinis et al., 2005; Christianson et al. (2010) also investigated the
Papadopoulou & Clahsen, 2003). Furthermore, there are interaction of semantic and syntactic routes in normal
studies to indicate that individual differences, such as comprehension, by combining a comprehension paradigm
proficiency, working memory capacity, and task play a role and a structural priming paradigm. Active/passive

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 521

structures were adapted from Ferreira (2003), and error recent existing studies pointing to the effects of task on L2
rates in answering comprehension questions rose when processing (Jackson & Bobb, 2009; Leeser et al., 2011;
structure conflicted with the plausibility of the sentence, Williams, 2006).
replicating Ferreira (2003). A more important finding Figure 1 demonstrates the mechanisms of the GE
was that both plausible-passive and implausible-active processing in comparison to the SSH, which are clearly
sentences tended to prime passive structures in production related views in language processing. Figure 1a shows the
data, suggesting that the outputs of both syntactic and SSH, proposing that lower, syntactic, route is restricted
semantic routes are not always seamlessly reconciled. for L2 speakers (see Clahsen & Felser, 2006b). In the
Depth of language processing has also been found SSH, the complete syntactic parsing is rather less detailed
to be influenced by reading goals and task demands. in L2 processing (at least at some levels of proficiency)
Swets, Desmet, Clifton, and Ferreira (2008) investigated than native speakers due to the inability to construct
whether frequency and the type of comprehension fully specified syntactic structures in real-time processing.
questions could affect depth of processing and final This results in “shallow parsing”, characterized by over-
interpretation in syntactic ambiguity resolution. They reliance on lexical-semantic information.
found that native English speakers were faster in reading On the other hand, the GE account predicts that, as
globally ambiguous sentences when asked superficial long as there is no deficit in syntactic knowledge (i.e., as
comprehension questions. Also, when task demand long as proficiency is sufficient), access to the syntactic
increased, that is, in a condition that asked specifically processing route is unimpaired. This is illustrated by the
about attachment of the globally ambiguous portion of solid line between the input and the full parsing route
the sentence, reading times were significantly longer in Figure 1b. Instead, what results in less-than-veridical
than in conditions with superficial questions. Lastly, final representations is incomplete (or absent) integration
reading times were slower in ambiguous sentences when of the outputs of the two processing routes, which is
answering detailed questions about the interpretation illustrated by the dotted, integrative line between good-
of the ambiguity than in the disambiguated sentence enough and full representations in Figure 1b. Adding this
conditions. Based on these findings, Swets et al. (2008) integrative line allows for integration or lack of it under
concluded that language processing can sometimes be certain circumstances. In other words, the GE account
INCOMPLETE , leaving ambiguous sentences unresolved proposes that even if the syntactic route is available, its
when there is no information to disambiguate them. Thus, output does not always get integrated with the output from
language processing can be strategic and goal-dependent. the heuristic route, which can happen in L1 processing as
Precisely how much of human language processing is well as L2 processing. The model shown in Figure 1b
“good enough” is not yet clear; however, this framework does not require the arrows tracing the derivation of the
leads to new ways of looking at human language full representation to be solid, i.e., full, for any given
processing mechanisms. Also, as the “good-enough” or speaker in any given situation. Rather, the point is that
“shallow” processing appears to exist in both native they CAN be, and the ultimate interpretation can STILL
speakers and L2 learners, and there is a growing body of be “good enough” if the outputs from the two processing
research that reports L2 speakers’ native-like processing routes are not integrated. Furthermore, whether the arrow
(Hopp, 2006; Jackson, 2008; Jackson & Bobb, 2009; from the L2 grammar into the “full parsing” box will be
Williams, 2006), the GE approach might provide the solid or dotted for any given L2 speaker is predicted to be
more integrated account of L1 and L2 processing. In dependent at least in part on proficiency and task.
the GE processing account, the language processing
system assumes two different routes, the grammar and
Translation task
semantic-based heuristics. The syntactic algorithm is
fully operational alongside semantic-based heuristics such In L2 processing research, comprehension questions
as lexical-semantic information, world knowledge, or have commonly been used as an offline measure
the NVN heuristic. But the syntactic representation to confirm readers’ comprehension. Comprehension
is “fragile” and quickly decaying (Sachs, 1967), so questions, however, might not provide much insight
that it is easily overridden by surface-level heuristics, into how L2 learners integrate morphosyntactic and
sometimes leading to interpretations that are unfaithful semantic information during comprehension. Question
to the original text. In context of L2 processing, it may form might in fact alter both response and processing
be that the syntactic output might be relatively more patterns (Christianson & Luke, 2011; Swets et al.,
fragile, depending on proficiency level, and thus be more 2008). With this concern in mind, the present study
susceptible to semantic/heuristic influences. In addition, employs a translation paradigm as an offline task.
within the GE approach the depth or veridicality of Translation is an important process that L2 speakers
language processing is hypothesized to be goal-dependent undertake when dealing with two languages, providing
and affected by task (Swets et al., 2008), not unlike some information about cognitive and linguistic development

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 522 Jung Hyun Lim and Kiel Christianson

Figure 1. Mechanisms of Good-Enough processing in comparison with the Shallow Structure Hypothesis. (a) Shallow
Structure Hypothesis for L2 processing, adapted from Clahsen & Felser (2006b). (b) Good-Enough processing, for both L1
and L2 processing. SRC-pl = subject relative clause, plausible; SRC-impl = subject relative clause, implausible; ORC-pl =
object relative clause, plausible; ORC-impl = object relative clause, implausible

(Campbell, 1998; Malakoff, 1992). It was also once In cognitive theories of the translation process, it
considered as a common practice to test learners’ is assumed that three major processes occur during
reading comprehension in language teaching situations translation, that is, to comprehend a source language,
(Buck, 1992; Cordero, 1984; Malakoff, 1992). As such, to switch between two languages, and to produce in a
translation may provide a window into L2 learners’ target language. One of the two views of how these
integration of morphosyntactic and semantic information, processes work during translation, the HORIZONTAL
because translation provides explicit evidence of how APPROACH , suggests that the translation process requires
a language learner has comprehended a given input additional attention and increase cognitive load compared
(Campbell, 1998). to normal reading, because one needs to establish

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 523

semantic and syntactic matches between two languages Because some of previous studies have pointed out task
during translation in a parallel manner (Danks & Griffin, differences influencing L2 processing (Jackson & Bobb,
1997; Gerver, 1976). Macizo and Bajo (2004, 2006) 2009; Leeser et al., 2011; Williams, 2006), the translation
found that both translators and L2 learners took longer task in this study not only provides offline data but also
to read sentences for translation than for repetition, broadens our understanding of how L2 processing can
indicating that translation includes reformulation process be modulated by reading goals/tasks. Task effects are
between the source language and target language during furthermore particularly informative to the question of
comprehension, which is consistent with the horizontal whether the GE processing account can account for L2
approach. processing as well as L1 processing. Because data from a
Since translation entails a complex cognitive process, translation task indicate how L2 speakers understand the
including comprehension, production, and additional input, it is useful for examining whether L2 processing can
working memory demands, compared to normal also be explained by the GE processing account, which
comprehension, one could ask whether translation can predicts effects of task or goal on language processing
serve as a valid measure of online processing. That and determines whether conflicting information sources
is, it could be argued that translation does not reflect are integrated or not (Ferreira & Patson, 2007).
online sentence comprehension, but rather measures the
“memory” for sentence content. There is no clear answer
The current study
for this concern, as few experimental cognitive studies
have been conducted on translation; however, if translation The first question addressed in the present study is how
is a “horizontal” process (Macizo & Bajo, 2004, 2006), it Korean learners of English integrate morphosyntactic and
should proceed in a relatively incremental fashion during semantic information during L2 processing using the self-
processing, and thus does in fact reflect processes that paced reading paradigm in combination with a translation
take place during online comprehension. Nevertheless, task. The term “semantic information” throughout the
even if translation is merely an indicator of the “memory” study is used to refer to general world knowledge, that
of a sentence, readers must still process that sentence, is, how entities referred to in a given sentence usually
including syntax, in real time during reading to produce behave in the real world. This study is in part motivated
their translation. Consider, for example, the output of by the study of Ferreira (2003), extending the basic
the L2-to-L1 translation process. It must derive from idea to L2 populations and examining subject relative
some combination of the input, the translator’s L2 and clauses (SRCs) and object relative clauses (ORCs),
L1 grammars, and the translator’s world knowledge. instead of active/passive structures. Sentences containing
Assuming the syntactic structure of the L2 input is SRCs like The man that bit the dog . . . and ORCs like
comprehensible to the translator and available in the L1, The man that the dog bit . . . have been of interest to
that structure should be maintained in the L1 translation. If psycholinguists, since these structures can assess the role
the L2 structure is not comprehensible (i.e., too difficult or of verbal working memory (Just & Carpenter, 1992;
unfamiliar), then we would expect an L1 translation based King & Just, 1991), and are useful in examining gaps
on L2 lexical semantics and the world knowledge of the and traces in syntax (Gibson, Hickok & Schütze, 1994;
translator, reflecting a default more frequent or simpler Nicol & Pickering, 1993). Previous studies in the L1
L2 structure that did not occur in the input. If the L2 processing literature show that ORCs generally cause
structure is comprehensible, but the semantic content of greater processing difficulty than SRCs. There are several
the input interferes with comprehension, we would expect accounts to explain this difference, including memory-
a translation that maintains the L2 input structure in the L1 based accounts, syntax-based accounts, and a perspective-
output, but that displays a semantic meaning that was not shifting account. Among them, in the syntax-based
in the L2 input. Thus, if a translation accurately reflects the accounts, comprehenders are required to treat the man as
syntax of the input, it seems quite clear that the syntax in a subject and an object at the same time. But the difficulty
the translation output is derived from the input, especially of assigning inconsistent thematic and syntactic roles to
if the syntax in the translation is difficult or non-canonical a single NP increases the difficulty of processing ORC
(e.g., an object relative clause or a passive structure). constructions. In addition, according to the active filler
In addition, translation is taken as a type of reading strategy (Clifton & Frazier, 1989), comprehenders may
task in this study, not just a method to elicit L2 experience misanalysis and require recovery processes
learner output. Thus, by comparing reading processes in to reach the correct syntactic analysis. Accordingly,
normal comprehension and translation, we can determine the ORC construction involves an initial misanalysis
whether there may be a possible effect of reading goals and a reanalysis process to assign correct syntactic
in L2 sentence comprehension, similar to recent L1 and thematic roles. Readers must “retrieve” correct
research reporting that reading goals and task demands syntactic information, and thus this process increases
can influence processing routines (Swets et al., 2008). reading times compared to the relatively easier SRC

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 524 Jung Hyun Lim and Kiel Christianson

structure. Thus, the pair of structures is asymmetrical in comprehension versus reading for translation. Predictions
canonicity and processing difficulty in English, analogous related to the task manipulation are largely informed
to active/passive structures (Ferreira, 2003; Holmes & by previous studies that indicate a possible influence
O’Regan, 1981; King & Just, 1991; Traxler, Morris & of task on language processing (Macizo & Bajo, 2004,
Seely, 2002). Like English, SRCs has also been found 2006; Swets et al., 2008). If L2 processing is affected
to be easier to comprehend than ORCs in Korean, despite by reading goals, then the availability of structural and
Korean being a left-branching language where the relative semantic information should differ depending on reading
clause precedes the modifying noun. This preference has sessions, but there could be two possible scenarios. First,
been demonstrated by both Korean L1 and L2 learners, since translation could motivate L2 learners to stay more
suggesting fundamental similarity across languages in the alert than during reading for normal comprehension, L2
processing of this particular syntactic representation (Cho, learners may be more involved in interpretive work,
1999; O’Grady, Lee & Choo, 2003).1 resulting in greater sensitivity to semantic information
By manipulating both syntactic and semantic in the translation session than in the comprehension
information, this study explores whether L2 speakers session. Alternatively, translation might increase L2
show similar processing patterns as native speakers with learners’ awareness of syntactic information compared to
regard to using structural and semantic knowledge. At normal reading for comprehension, as translation requires
the same time, the study aims to examine whether more attention to linguistic input. These differences in
the GE approach can account for L2 processing as either way would provide evidence that the depth of L2
well as L1 processing. If both native speakers and L2 processing is dependent on reading goals, consistent with
learners are able to use both syntactic and semantic the GE processing account. In addition, accuracy rates
routes in processing, but the syntactic representation in response to comprehension probes are expected to be
is sometimes overwhelmed by semantic information, higher in the translation session if translation requires
both groups should have more trouble interpreting ORC deeper interpretive work. Whether any possible different
structures with implausible semantic events than any effects of the two reading goals stem from additional
other condition, with more difficulty experienced by working memory load associated with the translation task
lower-proficiency L2 speakers. Also, L2 learners should or some other processes inherent in translation (such as
produce a number of translations that maintain the constraints imposed by the dual task of comprehension
correct morphosyntax, but not the thematic roles of the and production) remains to be explored, but this issue is
nouns from the input. Such a pattern would suggest that beyond the scope of this paper.
L2 processing is “good enough” in the same manner L2 proficiency is also predicted to influence both online
as native processing (Christianson et al., 2001, 2010; processing patterns and comprehension and translation
Ferreira, 2003), as it would provide evidence of both accuracy rates. Based on the previous studies that showed
syntactic and semantic/heuristic processing routes being proficiency effects in the use of syntactic information
active in parallel, and misinterpretation coming from the (Hopp, 2006; Jackson, 2008), L2 learners are expected to
lack of integration of the two routes as in Figure1b. use increasingly more syntactic information as proficiency
Furthermore, if L2 learners can retrieve structural increases, and accordingly accuracy rates would be lower
knowledge during online comprehension, reading times in the low-proficiency group. More precise predictions
should be different between the structures, specifically are difficult, as there have not been many studies done
yielding longer reading times in the ORC conditions than on task effects. Observations made in the present study
the SRC conditions. However, if language learners lack the will therefore broaden our understanding of L2 learner’s
ability to use morphosyntactic information during online reading strategies and how they are affected by proficiency
comprehension, then reading times should not differ and cognitive processes associated with different reading
across different structures, and participants’ translations tasks.
will not reflect the structures of the input sentences,
especially in the non-canonical, more difficult ORC
Method
conditions.
The second question is how reading goals and L2
Participants
proficiency affect L2 processing by comparing reading for
Participants were recruited from the University of Illinois
at the Urbana–Champaign community. Thirty-six native
1 Cho (1999) found that Korean subject relative clauses were easier Korean speakers who began learning English after puberty
both to comprehend and to produce than object relative clauses in
participated; 33 native English speakers served as a
Korean L1 child acquisition. In-depth morphosyntactic comparison
of Korean and English relative clause structures is outside the scope control group. Participants were paid $10 or received
of the present study, but interested readers are directed to O’Grady, course credit for participation. English proficiency of
Lee and Choo (2003) for a comparative analysis of these structures. the L2 speakers was measured using a cloze test (Giuli

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 525

Dussias, Pennsylvania State University, with permission), Table 1. The mean and standard deviation of the
which had a maximum possible score of 40. norming survey (standard deviations in
parentheses).

Materials Mean rating from

There were 48 experimental items and 96 filler sentences. 1 (implausible) to
The experimental sentences can be found in the Appendix. Sentence type 7 (plausible)
The fillers consisted of a variety of structures, none Biased reversible, plausible 6.51
of which included asymmetry in plausibility. Forty- (the doctor treated the patient) (0.70)
eight experimental sentences were constructed as in (1)
Biased reversible, implausible 2.20
below. Types of structure (SRC, ORC) and plausibility
(the patient treated the doctor) (0.87)
(plausible, implausible) were manipulated, resulting in
four conditions. Plausibility was manipulated only in the
relative clauses. The sentences below are examples of the
four conditions: (1a) illustrates a subject relative clause only one version of any given sentence. Participants were
with plausible event; (1b), a subject relative clause with directed to read each sentence and rate each item on a
implausible event; (1c), an object relative clause scale from 1 (“implausible”) to 7 (“extremely likely”).
with plausible event; and (1d), an object relative clause The results of the norming study are presented in Table 1.
with implausible event. Twenty-nine sentences were chosen among the 40 items
based on the grand mean in each version. As shown in
(1) a. The cat that chased the mouse was fast. Table 1, the grand mean of the 40 items was 6.51 for
(SRC-plausible) the plausible version and 2.20 for the implausible version.
b. The mouse that chased the cat was fast. Twenty-nine sentences were selected with means over 6.51
(SRC-implausible) in the plausible version and under 2.20 in the implausible
version. The grand mean of the selected 29 items was 6.74
c. The mouse that the cat chased was fast. in plausible version and 1.92 in the implausible version. A
(ORC-plausible) paired-sample t-test showed that the implausible sentences
d. The cat that the mouse chased was fast. were less plausible than the plausible sentences (t = 26.29,
(ORC-implausible) p < .01).

The experimental sentences were distributed across

four different lists in Latin Square design, such that each Procedure
participant read only one version of each experimental The main experiment was run on E-prime professional
sentence. Both experimental and filler sentences were software version 2.0. Native speakers read each sentence
randomly ordered during presentation. in a self-paced, word-by-word, non-cumulative reading
In the relative clause, 19 of 48 stimulus items paradigm. They were then asked to identify whether or
contained the same semantic relations (plausibility) as the not a word appeared in the sentence by pressing the
items adapted from Ferreira (2003). These 19 semantic “yes” or “no” button, in order to verify that participants
properties such as The cat chased the mouse/The mouse read each sentence completely, i.e., they did not skip any
chased the cat were normed by English native speakers in words during reading. After the brief word identification
Ferreira’s (2003) study. Plausibility for the relative clauses task, a paraphrase verification sentence was presented to
in the remaining 29 items was determined via a norming make a judgment on whether the verification sentence
study that was separately conducted before the current represented the same meaning of the sentence that they
main experiment. The norming study was conducted on read just before (Kim, 2008). An example of a paraphrase
46 native speakers of English who did not participate verification question is in (2) for the stimulus item The cat
in the main experiment. The purpose of the norming that the mouse chased was fast. A brief practice session
study was to select the remaining 29 sentences for the was provided before the main experiment for participants
main experiment. The norming survey contained a total to get used to the procedure.
of 70 sentences, consisting of 40 items and 30 fillers,
the structure of which was a simple active such as The (2) a. The cat chased the mouse. The cat was fast.
mom bathed the baby. The 40 biased, reversible items b. The mouse chased the cat. The cat was fast.
were presented both in plausible and implausible versions
across the two lists. The remaining 30 sentences contained For L2 learners, the procedure was largely similar;
symmetrical events as fillers. The target sentences were however, the experiment for the L2 group consisted of two
distributed over two lists, such that each participant saw sessions: (i) reading for comprehension and (ii) reading

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 526 Jung Hyun Lim and Kiel Christianson

Table 2. Grand mean reading times in milliseconds (standard deviations in

parentheses), native speakers.

SRC-plausible SRC-implausible ORC-plausible ORC-implausible

RC region 1222.97 1215.55 1280.56 1320.98

(319.71) (395.81) (382.93) (466.11)
Matrix verb 357.77 464.58 456.61 577.38
(94.40) (227.44) (164.36) (333.75)
Post-verb region 1231.17 1318.11 1258.57 1450.93
(376.85) (416.03) (362.09) (426.13)

RC = relative clause; SRC = subject relative clause; ORC = object relative clause

for translation, resulting in L2 participants reading 24 (RC) region, which consisted of the words within the
experimental sentences and 48 fillers in each session. relative clause (reading times for each word in the RC
The 48 experimental items and 96 fillers were randomly were summed to create this single dependent variable),
divided in half between the comprehension and translation the MATRIX VERB region, and the POST- VERB region,
session. The order of the comprehension and translation consisting of the rest of the sentence. Reading times for
session was counterbalanced across participants to control the rest of the words in the sentence were summed in the
for possible block order effects, and a brief practice same way as in the RC to create the single dependent
session was performed at the beginning of each session. variable. An example of the critical regions is presented
In the comprehension session, the participants read in (3).
each sentence according to the previously described
procedures. In the translation session, the procedure was (3) The cat that chased the mouse was fast.
the same as the comprehension session, except that the RC matrix verb post-verb
participants translated the English (L2) sentence they had
read into Korean (L1) verbally after the word identification Reading time data were analyzed in a linear mixed-
task. Translations were recorded for analysis. Then, they effects model (Baayen, Davidson & Bates, 2008). Each
completed the paraphrase verification task by making the model was fitted using a stepwise selection procedure
yes/no judgment. After the whole experiment, participants where factors failing to reach at least p < .10 significance
completed the cloze test and language survey to provide were removed from the final model (Luke & Christianson,
language background information. 2011). Therefore, the effect of structure, plausibility, and
the interaction between structure and plausibility were
included as predictors in the final, best-fitted model.
Table 2 presents the mean reading times for the three
Results
critical regions in the four conditions, and Table 3 presents
the result of the linear mixed-effects model for native
Reading time data
speakers.
Native speakers Participants read the RC region significantly faster in
Data-trimming was conducted on the comprehension the SRC conditions regardless of plausibility (t = 2.37,
accuracy of the paraphrase verification task for the p = .01). The ORC-implausible condition yielded the
filler items only, because experimental items were longest reading times numerically, but neither main effect
predicted to have some amount of incorrect answers of plausibility nor interaction between the two factors
(i.e., in implausible versions). Thirty-two participants approached significance. At the matrix verb, both main
answered correctly over 90% of the time, but one had effect of structure (t = 4.31, p < .01) and plausibility were
just 73% accuracy, and was therefore excluded from reliable (t = 4.70, p < .01). The SRC conditions were read
further analyses. Reading times greater than two standard faster than the ORC conditions, and implausible sentences
deviations from the mean were replaced with the mean took longer to read than plausible sentences. There was
for the particular sentence in the same condition. In no significant interaction. Even after the matrix verb, a
addition, reading time analyses were conducted over all similar pattern was observed, yielding main effects of
trials except the filler items, regardless of correctness of both structure (t = 2.03, p = .04) and plausibility (t =
the trials, because incorrect answers were expected in 3.67, p < .01) with no interaction.
some conditions, especially implausible versions. Three In sum, the SRC-plausible condition was the easiest
critical regions of interest were the RELATIVE CLAUSE condition to read, and the ORC-implausible condition was

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 527

Table 3. Fixed effects in the linear mixed effects model of reading times, native speakers.

Predictor Estimate SE t-value p

RC region (R2 = 0.29) Intercept 1260.05 68.06 18.52 <.01

Structure 39.41 16.60 2.37 .01∗
Plausibility 9.47 16.60 0.57 .56
Structure × Plausibility 11.64 16.60 0.70 .48
Verb region (R2 = 0.16) Intercept 464.13 32.98 14.08 <.01
Structure 52.34 12.13 4.31 <.01∗
Plausibility 57.01 12.13 4.70 <.01∗
Structure × Plausibility 2.92 12.13 0.24 .80
Post-verb region (R2 = 0.38) Intercept 1314.83 87.78 14.98 <.01
Structure 39.35 19.33 2.03 .04∗
Plausibility 70.99 19.33 3.67 <.01∗
Structure × Plausibility 26.53 19.33 1.37 .17

∗
p <.05; † p <.10; RC = relative clause; SE = standard error

the most difficult, yielding the longest reading times at all regions of interest in both reading sessions. Table 5 shows
three regions, although no interactions were found at any the results of the final model for three regions of interests
regions. Participants were sensitive to both structure and in each reading session.
plausibility from the verb region through the rest of the In comprehension, L2 speakers showed a reliable
sentence, but the effect of structure was seen earlier in the interaction between structure and plausibility (t =
RC region than the effect of plausibility. –3.42, p < .01) in the RC region, indicating that only in
the plausible sentences was there a structural difference,
L2 speakers and only in the SRC sentences was there a plausibility
Comprehension accuracy performance on the filler items effect. An effect of plausibility (t = 2.01, p = .04) was
was overall 92% correct, and data were trimmed in the observed as well to indicate that implausible sentences
same way as for the native speakers. The experimental yielded longer reading times than plausible sentences.
items for L2 participants were the same as the ones Proficiency did not significantly contribute to the model.
used for the native speakers. The two reading sessions At the verb, there were main effects of both structure
(comprehension vs. translation) for the L2 group resulted (t = 3.94, p < .01) and plausibility (t = 6.33, p <
in two separate analyses to examine whether reading .01), but no interaction was found. Reading times were
goals influence the way that participants process syntactic inflated when the sentences were implausible, and the
and semantic information. L2 speakers’ reading times ORC conditions were read more slowly than the SRC con-
greater than two standard deviations from the mean were ditions. In addition, proficiency marginally interacted with
replaced with the mean of each L2 learners’ mean reading plausibility (t = 1.75, p = .08), indicating that only higher-
times for the particular sentence in the specific condition. proficiency learners displayed reduced reading times in
Reading times for L2 speakers in each reading session the plausible conditions than the implausible conditions.
were analyzed in a mixed-effect model in the same way After the matrix verb, plausibility exhibited a reliable main
as the native speaker data, treating proficiency scores effect (t = 4.64, p < .01), but there was no main effect of
as a continuous variable to examine proficiency effects. structure. Also, there was a significant interaction between
As a result of the stepwise selection procedure (Luke proficiency and plausibility (t = 2.00, p = .04).
& Christianson, 2011), effects of structure, plausibility, In the translation session, there was a main effect of
proficiency, and the interactions between each factor were plausibility at the RC, (t = 6.12, p < .01), indicating that
included as predictors in the final, best-fitted models, when reading times for implausible sentences were significantly
proficiency contributed to the model. When proficiency longer than for plausible sentences. Neither an effect
did not reach significance (p < .05), only effects of of structure nor an interaction between structure and
structure, plausibility, and the interaction of the two plausibility was found at the RC. The interaction of
factors were included in the final model. Subject and item structure and proficiency reached marginal significance
were also included as crossed random effects. Table 4 (t = –1.82, p = .06), showing that the reading times of the
presents the mean reading times of L2 speakers at three lower-proficiency learners increased in ORCs more than

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 528 Jung Hyun Lim and Kiel Christianson

Table 4. Grand mean reading times in milliseconds (standard deviations in parentheses), L2

speakers.

(a) Grand mean reading times in milliseconds, comprehension session (standard deviations in parentheses).

SRC-plausible SRC-implausible ORC-plausible ORC-implausible

RC region 2033.47 2452.09 2361.82 2284

(659.75) (1226.82) (1089.98) (917.57)
Matrix verb 546.71 819.89 706.71 976.14
(235.60) (457.59) (382.86) (600.19)
Post-verb region 1537.99 1868.33 1567.33 1935.40
(502.82) (858.95) (489.23) (718.02)
(b) Grand mean reading times in milliseconds, translation session (standard deviations in parentheses).

RC region 2698.22 3154.42 2661.69 3336.81

(1430.74) (1682.74) (1418.05) (1717.57)
Matrix verb 605.69 978.16 817.36 1038.42
(319.27) (554.87) (405.36) (635.46)
Post-verb region 2681.10 3448.68 3079.91 3523.99
(1221.04) (1378.41) (1349.13) (1662.42)

RC = relative clause; SRC = subject relative clause; ORC = object relative clause

in SRCs, whereas the reading times were similar between addition, the translation task appeared to encourage L2
the two conditions for higher-proficiency learners. At the learners to be attentive to structural information to a
verb region, the structure (t = 2.57, p = .01) and the greater extent even in the post-verb region, in the same
plausibility (t = 5.80, p < .01) effects were significant, way as native speakers.
showing that reading behavior at this region in the
translation task was similar to the comprehension session.
Accuracy data
No interaction between structure and plausibility was
observed. In the remainder of the sentence, plausibility Native speakers
yielded a significant effect (t = 4.48, p < .01), and Accuracy data were analyzed using logit mixed models,
structure yielded marginal significance (t = 1.80, p = which handle binary distributions, such as accuracy data,
.07). Proficiency had no relation to either structure or better than ANOVA (Jaeger, 2008). Models were fitted
plausibility in the post-verb region. using a stepwise selection procedure in which factors
In summary, L2 speakers’ reading patterns were quite failing to reach at least p < .10 significance were removed
similar to the native speakers as both groups were sensitive from the model (Luke & Christianson, 2011). Effects
to structure and plausibility at the verb region in both of structure, plausibility and the interaction of structure
reading sessions. This finding suggests that L2 learners and plausibility were included as predictors in the final,
were capable of processing structural information along best fitted model. Subject and item were also included
with semantic information during online comprehension as crossed random effects. Figure 2 illustrates the mean
similar to native speakers. However, while native speakers proportions of accuracy in native speakers, and Table 6
were affected by structural differences at the early region shows the results of the final model. (Although logit mixed
of the sentences, i.e., the RC region, L2 learners were model analyses are performed over raw data, Figure 2 plots
influenced by plausibility information in both reading means for ease of exposition.)
sessions. This difference between native and L2 speakers As illustrated in Figure 2, native speakers were
would appear to be a qualitative one. Proficiency and significantly less accurate in the ORC condition than
task were also related to the use of information in the SRC condition (z = –5.26, p < .01). The effect of
L2 processing. More specifically, the low-proficiency plausibility was also significant (z = –5.29, p < .01),
learners were more likely than high-proficiency learners showing that the implausible conditions yielded more
to focus on structural information when the reading goal inaccurate answers than the plausible conditions. Both
was to translate, yielding a significant difference in using effects were qualified by a reliable interaction (z = –2.41,
syntactic information depending on reading goals. In p = .02), such that the native speakers had the most

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 529

Table 5. Fixed effects in the linear mixed effects model of reading times, L2 speakers.

Comprehension session

Predictor Estimate SE t-value p

RC region (R2 = 0.42) Intercept 2276.47 163.251 13.94 <.01

Structure 48.33 39.32 1.23 .22
Plausibility 78.82 39.19 2.01 .04∗
Structure × Plausibility –127.70 37.40 –3.42 <.01∗
Verb region (R2 = 0.32) Intercept 775.38 70.03 11.08 <.01
Structure 92.65 23.52 3.94 <.01∗
Plausibility 148.26 23.43 6.33 <.01∗
Proficiency 16.91 20.09 0.84 .40
Plausibility × Proficiency 13.09 7.50 1.75 .08†
Post-verb region (R2 = 0.29) Intercept 1734.30 122.58 14.15 <.01
Structure 23.60 39.54 0.60 .55
Plausibility 182.78 39.42 4.64 <.01∗
Proficiency –3.09 28.60 –0.11 .92
Plausibility × Proficiency 25.03 12.54 2.00 .04∗
Translation session

RC region (R2 = 0.40) Intercept 2972.91 273.78 10.86 <.01

Structure 29.27 47.25 0.62 .54
Plausibility 288.97 47.25 6.12 <.01∗
Proficiency 38.36 82.63 0.46 .64
Structure × Proficiency –27.29 15.51 –1.82 .06†
Verb region (R2 = 0.28) Intercept 861.37 73.11 11.78 <.01
Structure 67.46 26.23 2.57 .01∗
Plausibility 152.23 26.23 5.80 <.01∗
Proficiency 5.53 20.69 0.23 .78
Plausibility × Proficiency 14.58 8.59 1.70 .09†
Post-verb region (R2 = 0.34) Intercept 3183.42 230.00 13.84 <.01
Structure 118.53 67.66 1.80 .07†
Plausibility 302.91 67.66 4.48 <.01∗
Structure × Plausibility –80.88 67.66 –1.20 .23

∗
p < .05; † p < .10; RC = relative clause; SE = standard error

Table 6. Fixed effects in the logit mixed model of difficulty in the ORC-implausible condition. This result
question accuracy in native speakers. replicates Ferreira (2003), but with RC constructions
rather than active/passive constructions.
Question accuracy (C = 4.33)

Predictor Coefficient SE Wald Z p L2 speakers

As with the native speakers, accuracy data for L2 speakers
Intercept 1.66 0.16 10.03 <.00 were analyzed using a logit mixed model. Identical to
Structure −0.37 0.07 −5.26 <.01∗ the native speakers, effects of structure, plausibility and
Plausibility −0.37 0.07 −5.29 <.01∗ the interaction of structure and plausibility were included
Structure × Plausibility −0.17 0.07 −2.41 .02∗ as predictors in a final, best-fitted model. Subject and
item were also included as crossed random effects. In
∗
p < .05; † p < .10; SE = standard error both reading sessions, proficiency was also added as

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 530 Jung Hyun Lim and Kiel Christianson

Table 7. Fixed effects in the logit mixed model of

question accuracy in L2 learners for both reading
sessions.

Comprehension session

Question accuracy (C = 4.00)

Predictors Coefficient SE Wald Z p

Intercept 1.41 0.14 10.29 <.00

Structure –0.17 0.09 –1.83 .06†
Plausibility –0.71 0.09 –7.69 <.01∗
Proficiency 0.09 0.04 2.18 .03∗
Structure × Plausibility –0.01 0.08 –0.01 .95
Translation session
Figure 2. Mean proportion of comprehension accuracy in Question accuracy (C = 3.90)
native speakers. SRC-pl = subject relative clause, plausible;
SRC-impl = subject relative clause, implausible; ORC-pl = Predictors Coefficient SE Wald Z p
object relative clause, plausible; ORC-impl = object
relative clause, implausible Intercept 1.46 0.20 7.18 <.00
Structure –0.36 0.10 –3.51 <.01∗
Plausibility –0.97 0.10 –9.29 <.01∗
Proficiency 0.13 0.06 2.34 .02∗
a continuous variable in the model (see Table 7). The Plausibility × Proficiency –0.06 0.03 –1.78 .07†
mean proportions of accuracy in L2 learners for both
∗
reading sessions are illustrated in Figure 3 and the p < .05; † p < .10; SE = standard error
results of the final, best-fitting models are presented in
Table 7.
As can be seen in Figure 3, there was no numerical
difference between the reading sessions in terms of comprehension; z = 2.34, p = .02 for translation) in
accuracy. In both reading sessions, the main effects of both reading sessions, indicating that the higher the
both structure and plausibility were reliable with no proficiency was, the more accurate the answers. In the
interactions (see Table 7), indicating that L2 learners had translation session, proficiency marginally interacted with
more difficulty in understanding ORCs and implausible plausibility (z = –1.78, p = .07), indicating that as
sentences, independent of other factors. Also, proficiency proficiency increased, accuracy in the plausible conditions
yielded significant effects (z = 2.18, p = .03 for increased compared to the implausible conditions,

Figure 3. Mean proportions of comprehension accuracy in L2 learners for the two reading sessions.

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 531

Table 8. The numbers of translation in each condition.

Numbers of translation

Condition SRC_correct SRC_incorrect ORC_correct ORC_incorrect

SRC-plausible 196 1 6 0
SRC-implausible 103 21 4 27
ORC-plausibhle 7 4 176 3
ORC-implausible 10 75 74 33

SRC = subject relative clause; ORC = object relative clause

whereas lower-proficiency learners did not show this incorrect translations in the ORC-implausible condition,
significant difference depending on plausibility. provides strong evidence that L2 learners exhibited
command of syntactic knowledge (providing an accurate
morphosyntactic Korean translation of a non-canonical,
Translation data more difficult English structure), but that the semantic
Translations were another dependent variable for the L2 information overrode the fragile syntactic representation,
group. Participants’ translations were transcribed by the resulting in misinterpretation. There was also a significant
experimenter and coded as SRC-correct, SRC-incorrect, interaction between structure and plausibility (z =
ORC-correct, and ORC-incorrect (where ‘correct’ means –7.10, p < .01), indicating the probability of more
thematic roles were correctly assigned and ‘incorrect’ incorrect translations for ORC-implausible sentences than
means thematic roles were translated in a reversed order for any other sentence type. Finally, low-proficiency
from the original). There were inappropriate translations learners produced more incorrect translations in the ORC-
where participants did not produce any SRC or ORC implausible condition than high-proficiency learners did
structures, or they did not comprehend the input. (an average of 72% errors vs. an average of 43% errors,
These outliers (n = 124) were about 14% of the total respectively). This result is expected within an account
864 translations and were removed from the further in which syntactic structure becomes especially fragile
analyses. To be more specific, 6% in the SRC-plausible when it conflicts with semantic information, as in the GE
condition, 28% in SRC-implausible condition, 12% in processing account.
ORC-plausible condition, and 11% in ORC-implausible
condition were removed. A Poisson Regression model
Discussion
with repeated measures (using a Generalized Estimating
Equations (GEE) model in SAS software; see Agresti, In this study, we investigated L2 learner syntactic and
2007) was used to examine the effects of morphosyntax semantic processing compared to native speaker syntactic
and semantic knowledge on translation performance and semantic processing using both a self-paced reading
as the translation data were multi-categorical (Agresti, task and a translation task. Subject relative clause and
2007). object relative clause structures were utilized with a
Table 8 presents a contingency table for the numbers of manipulation of plausibility in order to examine the use of
translations produced in each condition. As can be seen hypothesized syntactic and semantic/heuristic processing
in the contingency table, Korean translations generally routes during online comprehension. Another question
reflected the structure of the to-be-translated sentences asked was how reading goals, or task, affect L2 reading
and implausible sentences resulted in an increased patterns, by asking Korean learners of English to read
probability of incorrect translations (with thematic roles sentences either for comprehension or for translation in
reversed). the other session. L2 proficiency was also explored as a
A Poisson loglinear regression showed that both possible factor. It was predicted that L2 learners’ reading
structure (z = 39.99, p < .01) and plausibility (z = times would be modulated by structure type, if they are
6.37, p < .01) yielded main effects in L2 learners’ able to access to syntactic information online during
translations. To be specific, expected odds of the reading. Also, comprehension accuracy was predicted to
correct translation when the structure was SRC were be lower when non-canonical structure was combined
2.75 times larger than when the structure was ORC. with implausible events, if the interpretation derived
The plausibility effect was 1.61 times larger when from the semantic route overrides the interpretation
the event was plausible than implausible. In particular, derived from the syntactic route (cf. Ferreira, 2003).
the cell (n = 33), where participants produced ORC- Higher L2 proficiency learners were predicted to show

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 532 Jung Hyun Lim and Kiel Christianson

greater sensitivity to structural information than lower- nor in the translation session by L2 speakers did we
proficiency learners. In addition, the translation task observe an interaction between the two factors, unlike
was expected to yield longer reading times than when previous studies. We speculate that this is because
answering comprehension questions, and the translation of proficiency, task, or possibly other uncontrolled
output was also predicted to yield explicit evidence factors, since task apparently affects the timing of the
of how L2 learners comprehended the linguistic input, interaction.
specifically, evidence of how, or if, syntactic and semantic Furthermore, these observations have several im-
information had been integrated. plications. First, the results demonstrate that reading
Results from the native speakers in the present goals affected the sorts of information L2 learners
study were in line with previous L1 studies examining computed during online reading from the beginning of
relatively easier processing of SRCs than ORCs (Holmes the sentence. Second, these results hint at the cognitive
& O’Regan, 1981; King & Just, 1991), in that native processes involved in translation. The fact that the lower-
participants showed longer reading times in the ORC proficiency participants had longer reading times in the
conditions than the SRC conditions early in the RC ORC sentences in the translation session suggests that
region. At both the verb and the post-verb regions, translation increases low-proficiency learners’ attention
native speakers were influenced by both structural to the form due to having to reformulate linguistic
and plausibility information, independently. The result components of the source language to those of the
that implausible conditions made processing harder target language during online reading (HORIZONTAL
independently of structure is consistent with the claim PERSPECTIVE OF TRANSLATION ; Danks & Griffins, 1997;
that language comprehension proceeds along two routes: Gerver, 1976). Lastly, L2 learners seem to be dependent on
a morphosyntactic and a semantic route (Christianson semantic/plausibility information to a larger extent early
et al., 2010; Kuperberg, 2007). The comprehension in sentences compared to native speakers, given that the
accuracy results further strengthen the argument that native speakers showed structure effects first at the RC
both processing routes are indeed at work. For the region. This pattern of results is quite consistent with the
ORC conditions, native speakers’ accuracy was lower previous studies that found L2 learners’ large reliance on
overall (76%), and lower for implausible sentences semantic information (Felser et al. 2003; Frenck-Mestre
than plausible sentences (65% vs. 83%). This result & Pynte, 1997; Juffs, 1998; Papadopoulou & Clahsen,
fits with findings from previous studies proposing the 2003; Williams, Mobius & Kim, 2001). The reliance on
Good-Enough processing account (Christianson et al., semantic knowledge is also very intuitive, in that the
2001, 2010). The data here are also noteworthy as L2 learners are in the process of learning the structure
they replicate the findings of Ferreira (2003) using of English, but have already acquired knowledge about
subject/object relative clauses, instead of active/passive how the world works. Pinker (1984) hypothesized that
structures. infants use their world knowledge to recognize syntactic
Unlike the early sensitivity to structural differences in relationships within sentences before they acquire all
native speakers, L2 speakers did not exhibit the same the syntax, which is called SEMANTIC BOOTSTRAPPING.
sensitivity to structural information in the RC region; Similarly, in the current study, world knowledge appears to
instead they showed a plausibility effect in both reading be a good way for L2 learners to bootstrap the syntax of the
sessions. In addition, the degree of sensitivity differed target language (Fisher, Hall, Rakowitz & Gleitman, 1994;
depending on reading goal and proficiency. Specifically, Pinker, 1984). This large reliance on semantic information
in the comprehension session, L2 speakers, regardless by L2 speakers in the first place can also be found in
of proficiency, showed an interaction between structure the translation data where L2 participants yielded SRC-
and plausibility, indicating a significant slowdown for incorrect sentences (n = 75) in the ORC-implausible
implausible sentences only in the SRC conditions. In the condition. These responses suggest that L2 speakers may
translation session, on the other hand, lower-proficiency rely on world knowledge rather than computing full
learners were more attentive to structural information in syntactic representations of the target sentences before
the RC region, yielding longer reading times in the ORC they integrate syntactic outputs with semantic/heuristic
than the SRC conditions. The interaction of structural outputs.
and semantic information in the comprehension session At the verb region, however, L2 speakers were
is in line with the previous L1 and L2 studies that show influenced by both structure and plausibility information
an immediate interaction, which has been interpreted in the two reading sessions, which was identical to the
as plausibility information influencing syntactic analysis native speakers. This suggests that L2 speakers were able
or postponing syntactic assignment (Jackson & Roberts, to compute both syntactic and semantic/world knowledge
2010; Traxler et al., 2002). However, the current study during reading in the same manner as native speakers,
found the interaction in the L2 comprehension session which reflects the concept of parallel routes towards
at the RC region only; neither in native speakers comprehension. The fact that the L2 learners showed

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 533

different reading times according to structure types is SSH and the GE account, as the SSH suggests that
in line with studies showing that L2 learners are able L2 learners depend necessarily on lexical-semantic or
to utilize morphosyntactic information during processing world knowledge during sentence processing, and the
(Hopp, 2006; Jackson, 2008; Jackson & Dussias, 2009), GE account predicts that semantic representations will
and it does not seem that the syntactic representation they sometimes override syntactic representations, presumably
compute during reading is less detailed than that of native to a greater extent in L2 populations.
speakers. In addition, the fact that L2 participants were Another facet of this observation is that L2 learners
not less detailed in the verb region, unlike the RC region, displayed a similar pattern of accuracy to native speakers
suggests that L2 speakers are just slower in retrieving in the other three conditions. As with native speakers, L2
syntactic information than native speakers are, but that learners attempted to comprehend the sentences through
they are eventually able to access the syntax similarly both a morphosyntactic and a heuristic route. Consistent
to native speakers. In the remainder of the sentence, with the GE processing, when the two outputs from each
L2 learners’ reading patterns were different from native route conflict, especially in the cases where implausible
speakers in the comprehension session. Plausibility events were presented in non-canonical structures – the
information lingered in L2 learners’ minds longer than ORC-implausible condition in this experiment – the
structural information, which is again consistent with output from the syntactic route seemed to be overridden
previous findings in the L2 processing literature that by the output from the semantic/heuristic route, ultimately
language learners are more sensitive to lexical-semantic or increasing misinterpretations.
world knowledge than native speakers (Felser et al. 2003; The argument that L2 learners processed the input
Frenck-Mestre & Pynte, 1997; Juffs, 1998; Papadopoulou through two routes but the lack of integration yielded the
& Clahsen, 2003; Williams et al., 2001). However, in misinterpretations is augmented by the translation data
the translation session, L2 learners showed both structure in this study. Translation data from L2 speakers provide
(although marginal) and plausibility effects in the post- explicit evidence to show that, at least some of the time, L2
verb region, similar to native speakers. This suggests that participants actually computed the full syntactic analysis,
the additional translation task keyed L2 learners’ access to but the syntactic representation was overridden when
syntactic knowledge during online reading more so than the sentence meaning was implausible. The occurrence
just reading for comprehension. It is possible that previous of incorrect translations where L2 learners produced
studies demonstrating that L2 learners are insensitive to accurate structural translations with reversed thematic
morphosyntactic input during online simply did not use a roles in the SRC-implausible condition (n = 21) and the
task that would have motivated L2 learners to process in ORC-implausible condition (n = 33) are the strongest
a more native-like manner. evidence that the syntactic analysis was detailed and
With regard to the offline comprehension questions, complete, but the implausible propositions overrode the
L2 learners’ accuracy results were slightly different from syntactic parser’s analysis, resulting in misinterpretations.
those of native speakers in that the L2 participants did This result agrees with the comprehension behavior of
not show an interaction of structure and plausibility native speakers reported in previous L1 processing studies
on the paraphrase verification task. In other words, (Christianson et al., 2010; Ferreira, 2003). In addition,
ORC sentences produced less accurate answers than most of the outliers in the SRC-implausible condition
SRC sentences regardless of plausibility, and implausible included ‘passive’ structure with reversible thematic roles
sentences yielded less accurate answers than plausible to make the sentence more sensible; for example, they
ones independent of structure. The reason for the absence usually translated the original input “The mouse that
of an interaction appears to lie in the fact that L2 chased the cat was fast” into “The mouse was chased by
learners failed to reach the accuracy level of the native the cat”. Although this observation cannot be statistically
speakers in the SRC-implausible condition. To be specific, tested, it is speculated that this behavior represents L2
accuracy proportions in the other three conditions (SRC- speakers’ both syntactic and semantic route at work
plausible, ORC-plausible, ORC-implausible) were similar during comprehension, resulting in passive structure
between native speakers and L2 learners (see Figures used to produce a translation that accords with world
2 and 3 above). However, the difference was large knowledge.
between the two groups in the SRC-implausible condition Most of the studies in L2 processing literature to
(83% for native speakers vs. 69% for L2 learners), date have derived their conclusions exclusively from
which resulted in no interaction in the L2 group. This online reading time data (Felser et al., 2003; Hopp, 2006;
discrepancy is interpreted as showing that L2 learners Jackson, 2008; Jackson & Dussias, 2009; Papadopoulou
rely on world knowledge in sentence processing more & Clahsen, 2003). As pointed out in Christianson, Luke,
than native speakers, such that L2 speakers interpreted and Wochna (2012), online reading times do not always
the implausible sentences based on how the world usually reflect how readers interpret sentences. As such, the fact
works. The data can be explained well by both the that L2 learners show different reading times from native

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 534 Jung Hyun Lim and Kiel Christianson

speakers does not necessarily imply that the syntactic morphosyntactic representations in some tasks, but do in
information they compute is shallower or less detailed. others.
The translation data in this study are unique in this respect Taken all together, the data presented here are quite
because they represent L2 speakers’ final interpretations, consistent with the GE processing account (Figure 1b
demonstrating both syntactic and semantic knowledge above), although some of the results are also certainly
that is computed. It is also worth noting that these data consistent with the SSH (Figure 1a). In the GE framework,
are the first to demonstrate a misinterpretation effect of the critical difference between L1 and L2 processing
unambiguous RC structures. lies not in the availability of the syntactic processing
We now turn to the second question probed in this (assuming underlying syntactic knowledge is sufficient),
study, namely, the roles of task and proficiency in L2 but rather in the ability of L2 speakers to integrate the
processing. Based on L2 learners’ reading times, both outputs of syntactic and semantic/heuristic routes to arrive
L2 proficiency and reading goals affected the depth at accurate comprehension. The results reported here
of processing during online reading. It was predicted clearly support this view: The interaction between syntax
that the translation task might increase awareness of and semantics observed in the translation data patterned
either semantic or syntactic information to a greater with native speakers, and the completeness of this
extent than normal comprehension due to the possible integration, as did the online reading times. Furthermore,
task demands associated with translation. Reading these patterns were modulated by proficiency and,
times revealed that translation generally increased L2 importantly, task.
speakers’ focus on structural information from the In conclusion, the present study investigated L2
beginning of the sentence to the latter part of the learners’ reading behaviors compared to native speakers,
sentence. Especially for lower-proficiency speakers, along with the possible influence of reading goals
their reading times were inflated in ORCs compared and L2 proficiency on reading behaviors of Korean
to SRCs during the translation session compared to speakers of English. Although L2 learners had a greater
the comprehension session, whereas there was no tendency to rely on semantic or world knowledge than
significant difference between structures in higher- native speakers, both online and offline data showed
proficiency learners. This difference suggests that L2 that the L2 learners were able to use both syntactic
learners’ reading strategies vary with both proficiency knowledge and plausibility information during real-time
and, importantly here, task. This variability is expected comprehension. Misinterpretations occurred when the
under the the GE processing view. The difference outputs from morphosyntactic and semantic processing
is consistent with data from an eye-tracking study routes were inconsistent, as previously observed in native-
examining reading for comprehension vs. translation speaker processing. Results also showed that translation
(Lim & Christianson, 2012. These results deserve further leads L2 learners, especially low-proficiency learners, to
investigation, especially in connection with cognitive allocate more attention to structural sources, suggesting
theories of translation. reading goals and L2 proficiency play important roles
As for the accuracy results, it was surprising that in L2 processing. All together, these results suggest that
accuracy rates were almost identical between the two L2 processing is quantitatively rather than qualitatively
reading sessions in L2 group. Accuracy rates were different from L1 processing, i.e., sometimes “good
expected to be lower in the comprehension session. A enough”, just as with native speakers. Furthermore, the
possible explanation for this result would be that people translation paradigm used in this study, which provided
comprehended sentences essentially the same way while explicit output from language learners, appears to be
reading for translation as reading for comprehension. a useful window into understanding language learners’
Indeed, this similar accuracy pattern between the two final interpretations and integration of morphosyntactic
reading sessions can even be taken to validate translation and semantic information in L2 processing research.
as a method to examine online processing, as the increased The comparisons of processes involved in reading for
cognitive load did not alter comprehension levels in either comprehension and translation are also informative about
direction. the nature of translation itself. Translation appears to
Accuracy in both reading sessions was significantly involve deeper analysis of linguistic input with more
modulated by proficiency level, with accuracy positively attention to structure of the input, especially for lower-
associated with proficiency. This finding suggests that proficiency learners.
L2 proficiency plays a crucial role in L2 processing
mechanisms. All together, these results indicate the
Appendix. Stimuli
task-dependent nature of processing, which is explicitly
predicted within the GE processing view (Ferreira & The sentences below are only presented in subject relative
Patson, 2007; Swets et al., 2008). Indeed, it seems difficult clause and plausible version. The implausible version
to claim that L2 speakers do not have access to detailed is created with reversing the nouns. The object relative

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 535

clause is created as follows: e.g., The dog that bit the man 28. The volunteer that helped the blind man was very
was in the yard → The man that the dog bit was in the handsome.
yard.
29. The homeowner that paid the gardener loved the
garden.
1. The dog that bit the man was in the yard.
30. The wrecker that towed the car drove at a high speed.
2. The chef that ruined the food was in the kitchen.
31. The boy that petted the puppy had cute eyes.
3. The bird that ate the worm was small.
32. The bird that protected the egg was in the big tree.
4. The lawyer that sued the doctor was smart.
33. The secretary that assisted the President was in a
5. The soldier that protected the villager was brave.
conference room.
6. The cat that chased the mouse was fast.
34. The tutor that taught the student solved the math
7. The teacher that quizzed the student was in the problem.
classroom.
35. The boss that fired the worker was unhappy.
8. The cop that pursued the thief was driving a car.
36. The parent that raised the twins lived in New York.
9. The waiter that served the customer was tall.
37. The terrorist that held the hostage was located in the
10. The owner that fed the cat was sitting on a sofa. building.
11. The detective that investigated the suspect was very 38. The criminal that kidnapped the girl was on CNN
tired. news.
12. The doctor that treated the patient was female. 39. The conductor that led the orchestra was pleased with
the performance.
13. The politician that deceived the voter was Korean.
40. The consultant that advised the client was very clever.
14. The golfer that hit the ball was in the shade.
41. The lawyer that defended the client was worried about
15. The hunter that shot the deer was in the Rocky
the result.
Mountains.
42. The parents that punished the child went to church.
16. The frog that ate the fly was green.
43. The fan that cheered for the baseball player wore a
17. The ghost that scared the boy was hiding behind a
red shirt.
curtain.
44. The nuns that took care of the orphans live in a small
18. The police officer that arrested the citizen was
village.
handsome.
45. The dean that awarded the student a prize was very
19. The fisherman that caught the fish was in the middle
famous in the school.
of the ocean.
46. The dog that herded the sheep was very furry.
20. The man that walked the dog was in the park.
47. The fan that admired the Hollywood star loved to wear
21. The coach that scolded the player won the
big sunglasses.
championship twice.
48. The guard that locked up the prisoner regretted past
22. The mother that bathed the child smelled nice.
decisions.
23. The father that scolded the teenager was in the living
room. References
24. The guide that led the tourist liked Europe a lot. Agresti, A. (2007). An introduction to categorical data analysis
(2nd edn.). Hoboken, NJ: John Wiley & Sons.
25. The reporter that interviewed the actress was at the
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-
coffee shop.
effects modeling with crossed random effects for subjects
26. The kids that obeyed the teacher enjoyed the summer and items. Journal of Memory and Langage, 59, 390–412.
break. Bates, E., & MacWhinney, B. (1987). Competition, variation,
and language learning. In B. MacWhinney (ed.),
27. The grandmother that dressed the child had a beautiful Mechanisms of language acquisition, pp. 157–193.
smile. Hillsdale, NJ: Lawrence Erlbaum.

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 536 Jung Hyun Lim and Kiel Christianson

Buck, G. (1992). Translation as a language testing procedure: Ferreira, F., Christianson, K., & Hollingworth, A. (2001).
Does it work? Language Testing, 9 (2), 123–148. Misinterpretations of garden-path sentences: Implications
Campbell, S. (1998). Translation into the second language. for models of sentence processing and reanalysis. Journal
New York: Longman. of Psycholinguistic Research, 30 (1), 3–20.
Cho, S. (1999). The acquisition of relative clauses: Experimental Ferreira, F., & Patson, N. D. (2007). The ‘good-enough’ approach
studies on Korean. Ph.D. dissertation, University of Hawaii to language comprehension. Language and Linguistics
at Manoa. Compass, 1 (1–2), 71–83.
Christianson, K., Hollingworth, A., Halliwell, J. F., & Ferreira, Fisher, C., Hall, D., Rakowitz, S., & Gleitman, L. (1994). When
F. (2001). Thematic roles ssigned along the garden path it is better to receive than to give: Syntactic and conceptual
linger. Cognitive Psychology, 42, 368–407. constraints on vocabulary growth. Lingua, 92, 333–375.
Christianson, K., & Luke, S. G. (2011). Context strengthens Frenck-Mestre, C. (2002). An on-line look at sentence
initial misinterpretations of text. Scientific Studies of processing in the second language. In R. Heredia & J.
Reading, 15, 136–166. Altarriba (eds.), Bilingual sentence processing, pp. 218–
Christianson, K., Luke, S. G., & Ferreira, F. (2010). Effects of 236. Amsterdam: Elsevier.
plausibility on structural priming. Journal of Experimental Frenck-Mestre, C., & Pynte, J. (1997). Syntactic ambiguity
Psychology, 36 (2), 538–544. resolution while reading in second and native languages.
Christianson, K., Luke, S. G., & Wochna, K. (2012). Garden- Quarterly Journal of Experimental Psychology, 50A, 119–
path vs. local coherence structures: Similarities and 148.
differences in parsing and interpretation. Ms., University Gerver, D. (1976). Empirical studies of simultaneous
of Illinois at Urbana–Champaign. interpretation: A review and a model. In R. W. Brislin (ed.),
Christianson, K., Williams, C., Zacks, R., & Ferreira, F. (2006). Translation: Application and research, pp. 165–207. New
Younger and older adults’ good enough interpretations York: Gardiner.
of garden-path sentences. Discourse Processes, 42, 205– Gibson, E., Hickok, G., & Schütze, C. T. (1994). Processing
238. empty categories: A parallel approach. Journal of
Clahsen, H., & Felser, C. (2006a). Grammatical processing in Psycholinguistic Research, 23, 381–405.
language learners. Applied Psycholinguistics, 27, 3–42. Gregg, K. (1996). The logical and developmental problems of
Clahsen, H., & Felser, C. (2006b). Continuity and second language acquisition. In C. R. William & T. K.
shallow structures in language processing. Applied Bhatia (eds.), Handbook of second language acquisition,
Psycholinguistics, 27, 107–126. pp. 50–84. New York: Academic Press.
Clifton, C. Jr., & Frazier, L. (1989). Comprehending sentences Holmes, V. M., & O’Regan, J. K. (1981). Eye fixation patterns
with long distance dependencies. In G. Carlson & during the reading of relative clause sentences. Journal of
M. Tanenhaus (eds.), Linguistic structure in language Verbal Learning and Verbal Behavior, 20, 417–430.
processing, pp. 273–317. Dordrecht: Kluwer. Hopp, H. (2006). Syntactic features and reanalysis in near-native
Conway, A. R. A., Kane, M. J., Bunting, M. F., Hambrick, D. Z., processing. Second Language Research, 22, 369–397.
Wilhelm, O., & Engle, R. W. (2005). Working memory Jackson, C. (2008). Proficiency level and the interaction
span tasks: A methodological review and user’s guide. of lexical and morphosyntactic information during L2
Psychonomic Bulletin & Review, 12, 769–786. sentence processing. Language Learning, 58 (4), 875–
Cordero, A. D. (1984). The role of translation in second language 909.
acquisition. The French Review, 57 (3), 350–355. Jackson, C., & Bobb, S. C. (2009). The processing and
Danks, J. H., & Griffin, J. (1997). Reading and translation. In comprehension of wh-questions among second language
J. H. Danks, G. M. Shreve, S. B. Fountain & M. K. McBeath speakers of German. Applied Psycholinguistics, 30, 603–
(eds.), Cognitive processes in translation and interpreting, 636.
pp. 161–175. Thousand Oaks, CA: Sage. Jackson, C., & Dussias, P. E. (2009). Cross-linguistic differences
Dussias, P. E. (2001). Sentence parsing in fluent Spanish–English and their impact on L2 sentence processing. Bilingualism:
bilinguals. In J. Nicol (ed.), One mind, two languages: Language and Cognition, 12 (1), 65–82.
Bilingual language processing, pp. 159–176. Malden, MA: Jackson, C., & Roberts, L. (2010). Animacy affects
Blackwell. the processing of subject–object ambiguities in the
Felser, C., & Roberts, L. (2007). Processing wh-dependencies in second language: Evidence from self-paced reading with
a second language: A cross-modal priming study. Second German and second language learners of Dutch. Applied
Language Research, 23 (1), 9–36. Psycholinguistics, 31, 671–691.
Felser, C., Roberts, L., Gross, R., & Marinis, T. (2003). The Jaeger, T. F. (2008). Categorical data analysis: Away from
processing of ambiguous sentences by first and second ANOVAs (transformation or not) and towards logit mixed
language learners of English. Applied Psycholinguistics, models. Journal of Memory and Language, 59, 434–446.
24, 453–489. Juffs, A. (1998). Main verb versus reduced relative clause
Ferreira, F. (2003). The misinterpretation of noncanonical ambiguity resolution in second language sentence
sentences. Cognitive Psychology, 47, 164–203. processing. Language Learning, 48, 107–147.
Ferreira, F., Bailey, K., & Ferraro, V. (2002). Good enough Just, M. A., & Carpenter, P. A. (1992). A capacity theory of
representations in language comprehension. Current comprehension: Individual differences in working memory
Directions in Psychological Science, 11, 11–15. capacity. Psychological Review, 99, 122–149.

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100

 L2 processing in comprehension and translation 537

King, J., & Just, M. A. (1991). Individual differences in syntactic Nicol, J., & Pickering, M. J. (1993). Processing syntac-
processing: The role of working memory. Journal of tically ambiguous sentences: Evidence from semantic
Memory and Language, 30, 580–602. priming. Journal of Psycholinguistic Research, 22, 207–
Kim, J.-H. (2008). Working memory effects on bilingual 237.
sentence processing. Ph.D. dissertation, University of O’Grady, W., Lee, M., & Choo, M. (2003). A subject–
Illinois at Urbana–Champaign. object asymmetry in the acquisition of relative clauses in
Kuperberg, G. R. (2007). Neural mechanisms of language Korean as a second language. Studies in Second Language
comprehension: Challenges to syntax. Brain Research Acquisition, 25, 433–448.
(Special Issue), 1146, 23–49. Papadopoulou, D., & Clahsen, H. (2003). Parsing strategies
Leeser, M. J., Brandl, A., & Whiteglass, C. (2011). Task in L1 and L2 sentence processing: A study of relative
effects in second language processing research. In P. clause attachment in Greek. Studies in Second Language
Tromovich & K. McDonough (eds). Applied priming Acquisition, 24, 501–528.
methods to L2 learning, teaching and research: Insights Pinker, S. (1984). Language learnability and language
from psycholinguistics, pp. 179–198. Amsterdam: John development. Cambridge, MA: Harvard University
Benjamins. Press.
Lim, J.-H., & Christianson, K. (2012). L2 morphological Sachs, J. S. (1967). Recognition memory for syntactic and
sensitivity in comprehension and translation: Evidence semantic aspects of connected discourse. Perception and
from eye-movements. Ms., University of Illinois at Urbana– Psychophysics, 2, 437–442.
Champaign. Swets, B., Desmet, T., Clifton, C., & Ferreira, F. (2008).
Luke, S. G., & Christianson, K. (2011). Stem and whole-word Undersepcification of syntactic ambiguities: Evidence from
frequency effects in the processing of inflected verbs in self-paced reading. Memory & Cognition, 36 (1), 201–
and out of a sentence context. Language and Cognitive 216.
Processes, 26, 1173–1192. Townsend, D. J., & Bever, T. G. (2001). Sentence
Macizo, P., & Bajo, M. T. (2004). When translation makes the comprehension: The integration of habits and rules.
difference: Sentence processing in reading and translation. Cambridge, MA: MIT Press.
Psicologica, 25, 181–205. Traxler, M. J., Morris, R. K., & Seely, R. E. (2002). Processing
Macizo, P., & Bajo, M. T. (2006). Reading for repetition and subject and object relative clauses: Evidence from eye
reading for translation: Do they involve the same process? movements. Journal of Memory and Language, 47, 69–
Cognition, 99, 1–34. 90.
Malakoff, M. E. (1992). Translation ability: A natural bilingual Williams, J. N. (2006). Incremental interpretation in second
and metalinguistic skill. In R. J. Harris (ed.), Cognitive language sentence processing. Bilingualism: Language and
processing in bilinguals (vol. 83). pp. 515–529. New York: Cognition, 9, 71–81.
Elsevier. Williams, J. N., Mobius, P., & Kim, C.-K. (2001). Native
Marinis, T., Roberts, L., Felser, C., & Clahsen, H. (2005). Gaps and non-native processing of English wh-questions:
in second language sentence processing. Studies in Second Parsing strategies and plausibility constraints. Applied
Language Acquisition, 27, 53–78. Psycholinguistics, 22 (4), 509–540.

http://journals.cambridge.org Downloaded: 31 Oct 2013 IP address: 130.126.255.100