Journal of Neurolinguistics 35 (2015) 96e108

Contents lists available at ScienceDirect

Journal of Neurolinguistics
journal homepage: www.elsevier.com/locate/
jneuroling

Hemispheric involvement in native and

non-native comprehension of conventional
metaphors
Nira Mashal a, b, *, Katy Borodkin c, 1, Omer Maliniak c,
Miriam Faust b, c
a

School of Education, Bar-Ilan University, Israel

The Leslie and Susan Gonda (Goldschmied) Multidisciplinary Brain Research Center,
Bar-Ilan University, Israel
c
Department of Psychology, Bar-Ilan University, Israel
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 30 November 2014
Received in revised form 7 April 2015
Accepted 8 April 2015
Available online

The present study examined hemispheric processing of conventional metaphors in native (L1) and non-native (L2) language using
the divided visual eld technique. Participants included 25 native
Hebrew speakers and 24 bilinguals who acquired English as L1 and
Hebrew as L2. In Experiment 1, the two groups performed a semantic judgment task on conventional metaphors and literal Hebrew word pairs, and in Experiment 2, the processing of the
expressions was compared between the two L1s. The results of the
two experiments demonstrated a left hemisphere advantage for
processing conventional metaphoric expressions in L1, but a right
hemisphere advantage for processing the same kind of stimuli in
L2. No such L1-L2 difference in hemispheric involvement was
observed for literal word pairs. These results support the FineCoarse Semantic Coding Theory and the Graded Salience Hypothesis and suggest that the metaphoric meanings of conventional
metaphors may appear less salient for a non-native speaker.
2015 Elsevier Ltd. All rights reserved.

Keywords:
Conventional metaphors
Salient meanings
Right hemisphere
Second language
Bilingualism

* Corresponding author. School of Education, Bar-Ilan University, Israel.

E-mail address: nmashal2@gmail.com (N. Mashal).
1
Katy Borodkin is currently at Department of Speech-Language-Hearing Sciences, Lehman College, City University of New
York, USA.
http://dx.doi.org/10.1016/j.jneuroling.2015.04.001
0911-6044/ 2015 Elsevier Ltd. All rights reserved.

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

97

1. Introduction
Figurative language and, especially, metaphoric language use is highly pervasive in everyday life. It
is necessary for social communication and used in different dialogue environments. Whereas native
language (L1) speakers use gurative language effortlessly, the difculty it exerts on second language
(L2) speakers is well-known (e.g., Kecskes, 2006). What makes it difcult to understand might be that a
gurative meaning of an utterance is grounded in the socio-cultural experience of the native speakers,
from which a particular highly accessible interpretation emerges. Due to the lack of (or limited)
experience with the language and the culture, what is accessible for L1 speakers will not necessarily be
accessible for L2 speakers (Kecskes, 2006). Consequently, the hemispheric processing of a gurative
expression by L2 speaker may differ from that made by L1 speaker. Whereas the neural basis for
processing gurative language by native language speakers was studied extensively, very little is
known about the processing of gurative language in second language speakers. The aim of the present
study is therefore to study the neural basis of bilingualism and to gain further insights into the processing of metaphoric expressions by L2 speakers.
According to the Fine-Coarse Semantic Coding Theory (FCSC Theory, Beeman, 1998; Jung-Beeman,
2005), the right hemisphere (RH) possesses a unique semantic coding characterized by high sensitivity
to distant semantic relations. Based on evidence from studies showing that semantic priming effects of
remotely related words are obtained in the RH but not in the left hemisphere (LH) (e.g., Chiarello,
2003), the FCSC Theory has suggested that semantic processing by the two cerebral hemispheres
differs qualitatively. According to this theory, immediately after encountering a word, the LH focuses on
a single dominant interpretation (ne semantic coding), whereas the RH loosely activates and maintains larger semantic elds containing more distant associates and more unconventional meanings
(coarse semantic coding). Since the metaphorical meaning of a word is usually more semantically
distant than its literal interpretation, the FCSC theory predicts RH semantic processes may be more apt
for metaphor comprehension. Although there is a growing body of research examining the predictions
of this theory for metaphoric language processing in native language (e.g., Faust & Mashal, 2007), it has
been rarely tested in non-native language. In particular, Faust, Ben-Artzi, and Vardi (2012) found
priming effect for weakly-related word pairs in Hebrew as L1 presented to the RH via the left visual
eld (LFV), using a relatively long time interval of 750 ms. However, no priming effect was observed for
weakly-related word pairs in English as L2, pointing to a weaker coarse semantic coding for a nonnative than native language.
Another psycholinguistic theory that addressed hemispheric lateralization in language processing is
the Graded Salience Hypothesis (GSH, Giora, 2002, 2003). According to the GSH, salient meanings of
words and utterances are the foremost meanings on our mind, i.e., they are coded in the mental lexicon
and can easily be accessed. What is required for a meaning to be most salient is its conventionality,
frequency, familiarity and/or prototypicality. Giora (1997, 2003) claims that the degree of meaning
salience, rather than the literal or metaphorical meaning of an utterance, determines the time-course
of meaning processing. According to the GSH, the order by which meanings are accessed and retrieved
from the mental lexicon depends on the degree of meaning salience of the linguistic expression such
that salient meanings are retrieved before less salient meanings regardless of context and their literality or non-literality. In other words, it is the salience-non-salience continuum rather than the literalmetaphoric distinction that determines how linguistic stimuli are processed. The GSH predicts a selective RH involvement in the processing of nonsalient meanings (such as novel metaphors in L1) and
traditional LH involvement in the processing of salient meanings (such as conventional metaphors in
L1).
The FCSC Theory and the GSH may have some important implications for studying the processing of
metaphors and other aspects of gurative utterances in L2. For native speakers, a conventional metaphor possesses a highly salient metaphoric meaning and a less salient literal interpretation since
native speakers have naturally encountered the expressions more often than L2 speakers. Thus, on
encountering a highly conventional metaphor, the LH engages in a ne coding and strong activation of
small semantic elds, related to the salient metaphoric meaning that will be immediately available for
retrieval for L1 speakers. For L2 speakers, on the other hand, the metaphoric meaning of the highly
conventional metaphors may be less salient, resulting in the engagement of coarse semantic coding of

98

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

the RH and weak activation of large and diffuse semantic elds (Faust & Mashal, 2007; Schmidt,
DeBuse, & Seger, 2007).
Both theories have gained considerable support from behavioral and neuroimaging studies in L1
speakers showing an enhanced role for the RH in the comprehension of distantly related or nonsalient
metaphoric word pairs presented to the LVF (e.g., Faust & Mashal, 2007; Laurent, Denhieres, Passerieux,
Iakimova, & Hardy-Bayle, 2006; Mashal, Faust, Hendler, & Jung-Beeman, 2005; Mashal & Faust, 2009;
Mashal, Faust, Hendler, & Jung-Beeman, 2007; Pobric, Mashal, Faust, & Lavidor, 2008; Schmidt et al.,
2007; Sotillo et al., 2005; but see Rapp, Mutschler, & Erb, 2012). In one study, the effect of meaning
salience on hemispheric processing was examined in a divided visual eld paradigm that used four
types of word pairs, conventional and novel metaphors, literal and unrelated word pairs (Faust &
Mashal, 2007). Conventional metaphors and literal word pairs represented the salient meanings,
whereas novel metaphors represented the nonsalient meanings. Participants were presented with the
stimuli and asked to decide whether the second word of each word pair (presented to either the right or
to the left visual eld) forms a meaningful expression with the rst word (presented centrally). The
results showed that words presented to the RH via the LVF (LVF/RH) produced faster and more accurate
responses than words presented to the LH via the right visual eld (RVF/LH) in performing semantic
judgment for novel (i.e., nonsalient) metaphoric expressions. The integration of the GSH (Giora, 2003)
and the FCSC Theory (Beeman, 1998; Jung-Beeman, 2005) thus provides a psycholinguistic framework
for semantic processing, in which the two cerebral hemispheres contribute to the on-line semantic
processes that subserve successful comprehension of literal as well as non-literal language.
Much is known about the comprehension of metaphoric meanings by the two cerebral hemispheres
in L1 on the one hand, and about bilingual language lateralization in an array of language components
on the other hand (for a meta-analysis of this line of research, see Hull & Vaid, 2007). To the best of our
knowledge, however, there are no studies of hemispheric involvement in non-native second language
metaphor processing. However, several studies investigated hemispheric processing of idiomatic expressions by native and nonnative language users (Cieslicka & Heredia, 2011; Cieslicka, 2013). Cieslicka
and Heredia (2011) tested the effects of salience and context on differential cerebral involvement
during the processing of a laterally presented targets related to the gurative or the literal interpretation of an idiom. Three Interstimulus Intervals (ISIs) were used: 0 ms, 300 ms, and 800 ms. The main
nding indicated that literal targets, namely the salient meaning of L2 idioms, showed robust priming
throughout all time delays in both visual elds, except at 300 ms ISI, where they were facilitated only in
the LH. These ndings highlight the special status that salient literal meanings of L2 idioms enjoy in the
course of gurative processing by L2 users. In contrast, the idiomatic meaning of the L2 idioms presented to the LVF/RH was primed only after 300 ms reecting the extra effort needed for the processing
of less salient and less automatized L2 idioms. Thus, the processing of idiomatic meaning is a more
demanding task for L2 users, probably because it requires the suppression of the plausible salient literal
meaning (see also Cieslicka, 2013).
The current study aims to test hemispheric processing of both L1 and L2 two-word expressions
using the divided visual eld technique. The expressions were either conventional metaphors or literal
word pairs in addition to meaningless word pairs, which were used in a semantic judgment task. Two
groups participated in the study: English-Hebrew bilinguals and native Hebrew speakers. The bilingual
group performed the task in English (L1) and Hebrew (L2) and the native Hebrew speakers e in Hebrew
only. Experiment 1 was aimed to test hemispheric involvement in processing metaphoric expressions
in Hebrew as L2 (in the bilingual group) and Hebrew as L1 (in the native Hebrew speakers group). In
line with the FCSC Theory (Beeman, 1998) and combined with the assumption that L1 metaphoric
meaning of a conventional metaphor is strongly coded in the lexicon and highly salient, we hypothesized that L1 conventional metaphors would show faster and/or more accurate responses for words
presented to the RVF/LH than LVF/RH. We predicted the opposite pattern of responses for L2 metaphoric processing (i.e., that L2 conventional metaphors would show faster and/or more accurate responses for words presented to the LVF/RH than RVF/LH), because L2 metaphoric meanings may rely
more on the coarse semantic coding of the RH specializing in processing distant semantic relations. In
addition, Experiment 2 was designed to compare the bilingual and the native Hebrew speakers groups
in the pattern of hemispheric involvement when processing conventional metaphors in L1 (English in
bilinguals and Hebrew in the other group). Based on the FCSC Theory and the GSH, we predicted that as

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

99

both groups were examined in their respective L1s in this experiment, they would demonstrate the
same pattern of hemispheric involvement in metaphor processing.
2. Experiment 1: metaphoric comprehension in Hebrew as native and non-native language
2.1. Participants
All participants were undergraduate students at Bar-Ilan University, who received academic credit
or were paid for participation. The bilingual sample included 24 participants (14 women), aged 20e28
(M 22.88, SD 1.92). They were born in the US and lived in Israel at the time of testing. Self-reported
language proles were assessed using the Language Experience and Prociency Questionnaire (Marian,
Blumenfeld, & Kaushanskaya, 2007), as summarized in Table 1. The bilinguals were exposed to English
from birth and began speaking and reading in Hebrew at childhood. They have been living in Israel for
more than 10 years on average and rated themselves as highly procient in reading, listening, and
speaking in Hebrew. The bilinguals preferred reading in English to reading in Hebrew, but were more
balanced in their preference of language for speaking.
The native Hebrew speakers sample consisted of 25 individuals (15 women), aged 21e29
(M 24.47, SD 2.17). They were all born in Israel and learned English as a foreign language at school.
As Table 1 indicates, they were exposed relatively early to English (around the age of 8), but had
virtually no immersion experience and rated their current exposure to English as only minimal. They
had a strong preference to read and speak in Hebrew rather than in English, suggesting their language
competence in English was low; however, their self-rated English prociency was relatively high and
comparable to Hebrew prociency in the bilingual sample. At debrieng, it became clear that the native
Hebrew speakers rated their English prociency in relation to other English as a foreign language
learners in Israel, whereas the bilinguals rated their Hebrew prociency relative to the Hebrewspeaking population in Israel. Thus, the self-rated L2 prociency was perhaps not as reliable in
native Hebrew speakers as it was in bilinguals as an index of language competence. Supporting this
argument, the correlations between L2 prociency ratings and preference to communicate in L2 were
strong and signicant in the bilingual sample (r .60, p .004 for speaking and r .61, p .004 for

Table 1
Means (standard deviations) of L2 history and prociency measures (Hebrew for the bilinguals and English for the native Hebrew
speakers).
English-Hebrew bilinguals
L2 age milestones (years)
Started learning
Started reading
Time (in years) spent in an L2-speaking
Country
Family
School/workplace
Current exposure (% of time)a
L1
L2
Self-reported prociency in L2b
Speaking
Listening
Reading
Preference to speak (% of time)a
L1
L2
Preference to read (% of time)a
L1
L2
a
b

Native Hebrew speakers

6.68 (6.33)
6.95 (3.26)

8.12 (1.27)
9.24 (1.82)

11.48 (7.98)
6.35 (9.90)
10.09 (7.76)

.17 (.51)
.06 (.24)
.11 (.47)

57 (20)
43 (20)

87 (10)
12 (7)

7.95 (1.56)
8.59 (1.30)
7.64 (1.87)

6.76 (1.44)
7.65 (1.32)
6.94 (1.56)

60 (29)
40 (29)

91 (12)
8 (11)

73 (29)
26 (29)

91 (14)
7 (13)

Percentages not always sum up to 100% due to missing data.

Prociency was rated on a scale ranging from 0 none to 10 perfect.

100

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

reading), whereas these correlations were not signicant in the native Hebrew speakers sample
(ps > .05). Given these characteristics, the native Hebrew speakers were not considered sufciently
uent in English and thus were not tested on English metaphor comprehension. All participants were
right handed according to self-report with normal or corrected to normal vision.
2.2. Hebrew word decoding task
Because performance on the semantic judgment task used in this study required good Hebrew
reading skills, participants completed a word decoding task, which consisted of a list of 217 single
words of three to seven letters (Shatil, 1995). Participants were given 1 min to read aloud the words
as quickly and accurately as possible. Reading speed was assessed by the total number of items
read.
2.3. Semantic judgment experiment in Hebrew
2.3.1. Stimuli
The stimulus pool consisted of 120 word pairs, all in Hebrew. The two words formed three types of
semantic relations: literal (loyal friend), conventional metaphoric (heated debate) or unrelated (angle
laundry). The unrelated word pairs were used as llers in a semantic judgment task. Both the prime and
the target word in each word pair consisted of two to six letters and word length was counterbalanced
across the three types of word pairs.
Several pretests were performed to determine the type of semantic relationship between the two
words in each pair and the word frequency. The aim of the rst pretest was to determine the type of the
semantic relationship of each word pair (metaphoric, literal, or unrelated word pairs). In order to do so,
46 judges, who did not participate in the experiment, were presented with a list of two-word expressions and asked to decide whether they are literally plausible, metaphorically plausible or not
plausible. Expressions that were rated by at least 72% of the judges as either metaphorically/literally
plausible or not plausible were selected as expressions with either a metaphoric or a literal meaning or
as unrelated word pairs, respectively. The familiarity level of the conventional metaphors was rated by
another group of 36 judges, who also did not participate in the experiment. Judges were presented
with a list of only the plausible metaphors from the rst pretest. Subjects were asked to rate the expressions on a 7-point familiarity scale ranging from 1 (highly nonfamiliar) to 7 (highly familiar).
Metaphoric expressions scoring more than 4 on this scale were selected as conventional metaphors
(rating average 6.17).
Since in Hebrew there is no extensive database for word frequency, the third pretest tested the word
frequency. Forty-ve additional judges, who participated in neither the former pretests nor the
experiment, were presented with the list of all the words and asked to rate them on a 5-point frequency scale ranging from 1 (highly nonfrequent) to 5 (highly frequent). The average rates on the
frequency scale for the target words were 3.43, 3.56, and 3.50 for the unrelated word pairs, literal
expressions, and conventional metaphors, respectively. The average ratings on the frequency scale for
the priming words were 3.90, 3.81, and 3.80 for the unrelated word pairs, literal expressions, and
conventional metaphors, respectively. No signicant difference was found for the target and the
priming words between the three conditions (Fs < 1).
Finally, 10 additional judges were asked to rate the concreteness of the primes and the targets on a
scale ranging from 1 (highly abstract) to 5 (highly concrete). Words with an average of less than 3 (on
the 1e5 scale) were considered as abstract words whereas words with an average of more than 3
were considered as concrete words. For the priming words, 71%, 74% and 74% of the words were
judged as concrete for the conventional metaphors, literal expressions, and unrelated word pairs,
respectively. For the target words, 63%, 64% and 64%, of the words were judged as abstract for the
conventional metaphors, literal expressions, and unrelated word pairs, respectively.
Stimuli of each type of semantic relation were divided into two equal-sized sets of 20 word pairs to
be presented to either the left or right visual eld (LVF/RH and RVF/LH, respectively). The mean length,
frequency, and concreteness values of the target words in a pair were matched between the three
conditions (all Fs(3, 57) < 1).

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

101

2.3.2. Procedure
Each participant was presented with all 120 word pairs. Participants were instructed to silently read
the rst word, and then when the second word appears, to indicate as rapidly and accurately as possible
whether the word pair formed a meaningful expression. The rst word of each pair was presented
centrally, followed by the target stimulus (second word) displayed 2.8 (center of lateralized word to
center of xation) to the right or to the left of xation. Participants were instructed to maintain central
xation. The participants placed their right index nger between two keys on the computer keyboard
and waited for a central xation cue (2500 ms duration), which appeared in the center of the screen and
indicated the onset of a trial. Then the xation disappeared, and the rst word appeared for 300 ms, at
the center of the screen, followed by xation for 100 ms and then the target word for additional 180 ms.
The xation remained on the screen during the target stimulus presentation, to ensure full xation cue,
and then for additional 2500 ms when response was allowed. The session began with a practice list,
consisting of six word pairs not used in the experimental lists. Note that the current stimulus onset
asynchrony (SOA) of 400 ms was previously used in a similar task (Faust & Mashal, 2007).
2.4. Results
Reaction times for correct responses and the percentage of correct responses in the semantic
judgment experiment were calculated, after omitting the outliers (responses 2 SD from the mean),
for each participant in all experimental conditions of interest. Responses for unrelated word pairs that
served as llers were not analyzed. Performance on the Hebrew word decoding task was negatively
correlated with reaction times (rs ranging from .63 to .72) and positively correlated with the percentage of correct responses (rs ranging from .53 to .69) in the experiment. Thus, word decoding speed
was entered as a covariate in all subsequent analyses.
A 2 (Visual Field: left, right)  2 (Semantic Relations: literal, metaphoric)  2 (Hebrew Language
Status: L1, L2) mixed ANCOVA was conducted on mean reaction times, with Visual Field and Semantic
Relations as within-subjects factors, Hebrew Language Status as a between-subjects factor, and word
decoding speed as a covariate. This analysis yielded a signicant three-way interaction, F(1,46) 6.39,
p .02, partial h2 .12 (for descriptive statistics, see Fig. 1 and Table 2). Simple effects analysis
demonstrated that in native Hebrew speakers, reaction times were faster for conventional metaphors
in the RVF/LH compared to conventional metaphors presented in the LVF/RH, F(1,46) 6.75, p .01. In
contrast, in bilingual speakers of Hebrew as L2, reaction times were slower in RVF/LH than in LVF/RH
for these expressions, F(1,46) 4.80, p .03. Although RVF/LH tended for faster response times than
the LVF/RH for literal expressions, the difference was not signicant in either group (p > .05). The
remaining effects of the analysis, including the main effect of Visual Field, F(1,46) < 1, Semantic Relations, F(1, 46) < 1, and Hebrew Language Status, F(1,46) 2.29, p .14, partial h2 .05, were not
signicant. The two-way interactions of Visual Field  Semantic Relations, F(1,46) 2.08, p .16,
partial h2 .04, Visual Field  Hebrew Language Status, F(1,46) 2.65, p .11, partial h2 .05, and
Semantic Relations  Hebrew Language Status, F(1,46) < 1, were also not signicant.
A three-way 2 (Visual Field: left, right)  2 (Semantic Relations: literal, metaphoric)  2 (Hebrew
Language Status: L1, L2) mixed ANCOVA was conducted on the percentage of correct responses. This
analysis resulted in a signicant effect of Hebrew Language Status, such that the overall performance in
the semantic judgment task was more accurate in native Hebrew speakers (M 93.21, SD 5.35)
compared to L2 speakers (M 76.52, SD 12.26), F(1,46) 8.90, p .005, partial h2 .16. There were no
other signicant main effects (F(1,46) < 1 for Visual Field and F(1,46) 3.31, p .08, partial h2 .07 for
Semantic Relations) or interactions (all Fs < 1). Means and standard deviations are presented in Table 2.
Although our a priori hypotheses concerned L1-L2 differences in hemispheric involvement, which were
expected for metaphoric but not literal expressions, inspection of the adjusted RT means also suggested that
in L2 speakers the LH but not the RH becomes slower when processing metaphoric expressions compared
to literal expressions. Conrming this post hoc observation, re-analysis of L2 speakers data indicated
signicantly slower responses to metaphoric compared to literal expressions in the RVF/LH, F(1,46) 12.91,
p .001, whereas RTs were comparable across expression types in the LVF/RH, F(1,46) < 1, p > .05.
The correlations between accuracy and reaction times were all negative (rs ranging from .33 to
.72, ps < .05), indicating that no accuracy-reaction times trade-off was present in the performance.

102

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

Fig. 1. Mean reaction times (adjusted for reading speed in Hebrew) for metaphoric and literal expressions as a function of Hebrew
language status and visual eld. Bars represent standard errors. L1 native language; L2 second language; LVF/RH left visual
eld/right hemisphere; RVF/LH right visual eld/left hemisphere.

ANCOVAs on reaction times and accuracy data on ller trials yielded no signicant differences between
groups and visual elds (all ps > .05), suggesting there was no response bias.
The results of Experiment 1 indicate that there is an RVF/LH advantage in native Hebrew speakers
when processing conventional metaphors, but an LVF/RH advantage in non-native Hebrew speakers
when processing the same expressions. Experiment 2 was designed as a control experiment, which
aimed to examine whether the groups demonstrate the same pattern of hemispheric involvement in
their respective L1s.
3. Experiment 2: metaphoric comprehension in native language (English vs. Hebrew)
3.1. Participants
The bilingual participants who took part in Experiment 1 also participated in Experiment 2.
3.2. Stimuli
The stimuli pool consisted of 120 word pairs in English. The word pairs formed three types of semantic relations: familiar metaphoric (sweet revenge), literal word pairs (hunger strike), and unrelated
word pairs (coral exam). The unrelated word pairs were used as llers in a semantic judgment task.

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

103

Table 2
Mean percent correct responses and reaction times (standard deviations) by Language, Visual Field, and Semantic Relations.

RVF/LH

LVF/RH

Metaphoric
Reaction time
Accuracy
Literal
Reaction time
Accuracy
Unrelated
Reaction time
Accuracy
Metaphoric
Reaction time
Accuracy
Literal
Reaction time
Accuracy
Unrelated
Reaction time
Accuracy

Hebrew as L1

Hebrew as L2

English as L1

749 (73.18)
92.42 (7.91)

967 (191.78)
71.71 (16.56)

847 (151.35)
89.27 (10.00)

705 (84.31)
97.40 (3.26)

906 (195.12)
82.50 (11.42)

763 (91.94)
93.13 (17.31)

980 (201.34)
91.40 (11.95)

1088 (223.39)
81.04 (17.69)

1064 (163.46)
78.54 (14.92)

785 (83.99)
88.00 (14.07)

939 (210.51)
69.38 (16.70)

945 (131.22)
75.21 (12.11)

728 (69.10)
95.00 (4.79)

929 (226.49)
82.50 (14.59)

901 (124.35)
83.13 (18.58)

1038 (243.06)
87.00 (11.90)

1119 (236.53)
76.04 (23.36)

1129 (173.21)
84.36 (14.39)

Ten judges who did not participate in the study were asked to decide whether each expression is
literally plausible, metaphorically plausible, or unrelated. Word pairs that were rated by at least 75% of
the judges as metaphorically/literally plausible or unrelated were selected for use in the corresponding
conditions. Another group of six judges received a list of only the metaphors rated as plausible in the
rst pretest, and were asked to rate their degree of familiarity on a 5-point familiarity scale ranging
from 1 (highly unfamiliar) to 5 (highly familiar). Metaphoric expressions scoring above 3 on this scale
were selected as conventional metaphors (rating average 4.32).
Each type of semantic relation contained 40 word pairs that were further divided into two equalsized sets of 20 word pairs to be presented to either the left or the right visual eld. Stimuli
belonging to the three experimental conditions were balanced between conditions according to word
length (mean 5.82 letters), frequency (mean 111 per million, based on the validated Kucera-Frances
frequency norms on the MRC database; Coltheart, 1981), and concreteness (mean 455, based on
the MRC database; Coltheart, 1981).
3.3. Procedure
Experiment 2 was conducted on bilingual speakers only, using identical to Experiment 1 procedure.
Both experiments were completed on the same day. The order of the experiments was counterbalanced across participants.
3.4. Results
The performance of the two groups in their respective L1s (Hebrew for native Hebrew speakers and
English for English-Hebrew bilinguals) in a semantic judgment task was compared. Data that was
collected in Experiment 1 from the Hebrew native speakers were re-analyzed for this purpose. Reaction times for correct responses and the percentage of correct responses in the semantic judgment
experiment were calculated, after omitting the outliers (responses 2 SD from the mean), for each
participant in all experimental conditions of interest. Responses for unrelated word pairs, which were
llers, were not analyzed.
A 2(Visual Field: left, right)  2(Semantic Relations: metaphoric, literal)  2(L1 Language: Hebrew,
English) ANOVA on reaction times was conducted. This analysis yielded a signicant effect of Visual
Field, F(1,47) 56.10, p < .001, partial h2 .54, Semantic Relations, F(1,47) 47.74, p < .001, partial
h2 .50, L1 Language, F(1,47) 23.35, p < .001, partial h2 .33, and Visual Field  L1 Language

104

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

interaction, F(1,47) 20.50, p < .001, partial h2 .30. Both groups responded faster for word pairs in the
RVF/LH compared to word pairs in the LVF/RH, but this difference was more pronounced in English.
Neither the Visual Field  Semantic Relations nor the Semantic Relations  L1 Language interaction
was signicant (both Fs < 1). More importantly and consistent with our hypotheses, the three-way
interaction was not signicant, F(1,47) 2.11, p > .05, partial h2 .04, indicating that there was an
RVF/LH advantage in both groups performing the task in their respective L1s (Hebrew for native Hebrew speakers and English for bilinguals) for processing conventional as well as literal expressions (see
Fig. 2 and Table 2 for descriptive statistics).
We also conducted a 2(Visual Field: left, right)  2(Semantic Relations: metaphoric, literal)  2(L1
Language: Hebrew, English) ANOVA on the percentage of correct responses, which resulted in a signicant effect of Visual Field, F(1,47) 72.72, p < .001, partial h2 .61, Semantic Relations,
F(1,47) 8.04, p .01, partial h2 .15, L1 Language, F(1,47) 10.74, p .002, partial h2 .19, and Visual
Field  L1 Language interaction, F(1,47) 22.66, p < .001, partial h2 .33. Both groups were more
accurate responding to word pairs in the RVF/LH compared to words in the LVF/RH, but this difference
was greater for English as L1 speakers. The remaining two-way interactions were not signicant
(F(1,47) 2.99, p .09, partial h2 .06 for Visual Field  Semantic Relations, and F(1,47) < 1 for Semantic Relations  L1 Language). Similar to the reaction times analysis, there was no signicant threeway interaction, F(1,47) < 1, indicating that performance in L1 (both Hebrew and English) was more

Fig. 2. Mean reaction times for metaphoric and literal expressions as a function of native language and visual eld. Bars represent
standard errors. L1 native language; LVF/RH left visual eld/right hemisphere; RVF/LH right visual eld/left hemisphere.

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

105

accurate when stimuli were presented in the RVF/LH compared to the LVF/RH, regardless of semantic
relations (for means and standard deviations, see Table 2).
To control for reading uency, we re-analyzed the data of Experiment 2 with Hebrew reading speed
as a covariate. The results of the reaction times did not change. For accuracy, the results were also very
similar with and without the covariate. The only difference was observed in the Semantic Relations
effect, which was now only marginally signicant, F(1,46) 3.30, p .076. These results suggest that
the different outcomes of Experiment 1 and 2 are not merely an artifact of the differences between the
analyses.
As in Experiment 1, we calculated the correlations between accuracy and reaction times, which
were all negative (rs ranging from .28 to .56, ps < .05). These results indicate that no accuracyreaction time trade-off was present in the performance. To test for response biases, ANCOVAs were
carried out on reaction times and percentage of correct responses on ller trials. Responses were faster
for unrelated word pairs presented in the RVF/LH compared to the LVF/RH, F(1,47) 11.93, p .001.
This nding indicates that both groups process the llers similarly. There were no other signicant
differences between visual elds or groups on ller trials.
4. General discussion
The results of the present study give a glimpse into the brain mechanism responsible for metaphor
processing by native and non-native language speakers. This study demonstrated that whereas literal
word pair processing is not modulated by language, there are differences in how native and non-native
speakers process metaphoric expressions. When conventional metaphors were processed in L1 (either
Hebrew or English), responses were faster to stimuli presented in the RVF/LH compared to stimuli
presented in the LVF/RH. In contrast, the same bilingual speakers who exhibited an RVF/LH advantage
for L1 (English) metaphoric expressions showed an LVF/RH advantage when processing this kind of
word pairs in their L2 (Hebrew). Notably, the L1-L2 differences reported here were observed in highly
procient L2 speakers, as evident from their self-ratings and the fact that they all were undergraduate
students in a Hebrew speaking Israeli university, who were exposed to their L2 from young age and
used it extensively in an immersion context.
The GSH (Giora, 1997, 2003) postulates that less salient meanings, be they metaphoric or literal, are
mainly processed by the RH. Consistent with this view, the LVF advantage observed in processing L2
conventional metaphors may indicate that the metaphoric meanings of conventional metaphors are
probably less salient for a non-native as compared to a native speaker. Whereas salient meanings are
stored in the mental lexicon and are automatically accessed, less salient meanings are accessed only
after the dominant meaning has been retrieved. In other words, meanings are gradually retrieved from
the mental lexicon, with the less salient meanings receiving secondary priority. Kecskes (2006) proposed that the salient metaphoric meanings of metaphoric expressions are more salient in L1 than in
L2 because of prior experience and prior encounters with the utterances in relatively similar and
typical contexts. Thus, although L2 speakers know the meanings of the conventional metaphors, their
metaphorical meanings do not receive priority in the mental lexicon and therefore, as predicted by the
GSH, these metaphoric expressions are processed by their non-dominant language hemisphere. Future
research should test directly the salience of the metaphoric expressions using participants' subjective
ratings to conrm that the LVF/RH advantage obtained in the present study reects the effects of L2
meaning salience on hemispheric involvement.
Our results support both the GSH (Giora, 2003) and the FCSC Theory (Beeman, 1998; Jung-Beeman,
2005). The RVF/LH advantage in the native languages for processing conventional metaphors corroborates with the FCSC Theory that postulates that the LH specializes in ne semantic coding, i.e., in
the core semantic features of words. On the other hand, the FCSC Theory argues that the more distantly
related and subordinate meaning of words or utterances are activated and maintained by the RH coarse
semantic mechanisms. More specically, the RH is involved in processing metaphors because they
involve more distant semantic relationships than literal language. Consistent with this view, and with
Cieslicka and Heredia (2011) who found more priming for nonsalient L2 gurative meanings of idiomatic expressions in the LVF/RH than in the RVF/LH, our results show LVF/RH advantage in the processing of L2 conventional metaphors. Thus, because these two approaches are to some degree

106

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

overlapping our ndings support both: The second experiment supports the RVF/LH advantage in L1s
by showing faster response times for both conventional metaphors and literal word pairs in the RVF/LH
than the LVF/RH, and the ndings of the rst experiment indicate that less salient and subordinate
metaphoric meanings show unique RH involvement in L2. Thus, within 400 ms the coarse semantic
mechanism of the RH seems to contribute to the processing of the subordinate metaphoric meanings in
L2. It is unclear however, whether longer SOA would lead to the same pattern of results, as no unique
RH involvement was observed in the processing of weak semantic relations in L2 using SOA of 750 ms
(Faust et al., 2012).
The process of lexicalizing the gurative meanings of conventional metaphors seems to affect the
pattern of hemispheric involvement. Because L1 speakers have more experience with conventional
metaphors, the expressions lose their metaphoricity over time and become more and more familiar
and lexicalized (Bowdle & Gentner, 2005). When metaphors are highly conventionalized, close associations have been formed between the words and their processing is based on the ne semantic
coding mechanism of the LH. It is possible that in L2, conventional metaphor processing continues to
rely on the coarse coding of the RH, much like the processing of novel metaphors in L1 (Faust & Mashal,
2007), and only after many years of language exposure and use the L2 associative network matures to
the extent that allows ne coding of these expressions in the LH. Recent research shows that the
acquisition of the creative aspects of language (e.g. metaphors, humor, ambiguity) is completed only in
early adulthood, indicating that it is a long-term process even for native language speakers (Patael,
Borodkin, & Faust, in preparation). Thus, it could be that because L2 speakers are less exposed to
this process of metaphor conventionalization, they show a different pattern of hemispheric involvement. A possible prediction stemming from the current study is that L2 speakers who experience
repeated exposure to conventional metaphors, i.e., a process that strengthens the connections between
the words, will show a similar (or close to similar) pattern of hemispheric lateralization as L1 speakers.
This prediction is based on a recent divided visual eld study that has shown that novel non-salient
metaphors change the pattern of hemispheric involvement from right lateralization at rst exposure
to equal involvement of both hemispheres at second exposure (Mashal & Faust, 2009). Note that only
two exposures to novel metaphors in L1 alter hemisphere processing but bilinguals who are very
procient in Hebrew, and who probably encountered conventional metaphors several times, still show
LVF/RH advantage to conventional metaphors. Our claim is that for bilinguals the gurative meanings
of conventional metaphors are non-salient, and future studies should test the type of conventionalization process required for bilingual speakers in order for them to process L2 conventional metaphors
similarly to L1 speakers.
Another nding that emerged from the present study is the slower response times to conventional
metaphors relative to the literal word pairs in L2 presented to the RVF/LH. Such a difference was not
observed in the LVF/RH. A possible explanation may be the increased cognitive load exerted on the LH
when the competing literal interpretation of a conventional metaphor has to be suppressed and the
appropriate gurative meaning has to be selected. Such cognitive load is manifested only in L2. Because
the literal interpretation of an L2 conventional metaphor is salient and highly accessible and enjoys
processing priority over the gurative interpretation (Cieslicka, 2006; Cieslicka & Heredia, 2011), more
cognitive control is required for its suppression. Thus, more cognitive control is required when bilinguals process L2 conventional metaphors than computing the literal meaning of literal word pairs. In
contrast, for native speakers, the literal interpretation of an L1 conventional metaphor is less salient,
and therefore does not compete with the strongly coded gurative meaning. In addition to this
unpredicted nding, we also observed that L1 English stimuli were processed slower than the L1
Hebrew stimuli regardless of stimulus type. This is probably because the target words were not entirely
comparable in psycholinguistic parameters as the two stimuli pools were drawn from different languages. Alternatively, the fact that English speakers participated in two experiments may somehow
contributed to this difference. Furthermore, the two-way interaction of VF  L1 Language in Experiment 2 points to a larger RVF/LH advantage for L1 English than for L1 Hebrew in both reaction times
and accuracy. Thus, in addition to the overall slowdown of L1 English words as compared to L1 Hebrew
words, the RH seems to contribute to the larger gap between the two L1 languages in favor of larger
RVF/LH advantage in L1 English. This advantage seems to stem from the slower and less accurate RH

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

107

performance in processing the salient meanings of both L1 English literal and conventional metaphors
relative to the LH.
There are several limitations to the current study. First, L2 English speakers were not tested.
Optimally, both groups of speakers would participate in both the Hebrew and the English tasks. This
optimal design would have claried whether the differential bilingual performance on L1 and L2 tasks
was evidence of differential metaphor processing in L1 versus L2 or simply a by-product of differences
between the Hebrew and the English stimuli. Secondly, a more ecologically valid design that tests
conventional metaphor processing embedded within a sentence is recommended for future study.
Third, except for the ndings from the visual eld experiments, there is no independent verication
that the bilingual speakers process L2 conventional metaphors differently.
5. Conclusion
Given the difculties that non-native speakers experience with gurative language comprehension
and the paucity of studies investigating the qualitative differences in non-literal representations in the
brain between native and non-native speakers, our study may have signicant implications for the
research of both linguistic and social aspects of communication in L2. To gain further insights in this
important and surprisingly under-researched area, future studies may expand the scope to include
other forms of gurative language such as irony and humor to be explored using both behavioral and
neuroimaging techniques.
References
Beeman, M. (1998). Coarse semantic coding and discourse comprehension. In M. Beeman, & C. Chiarello (Eds.), Right hemisphere
language comprehension: Perspectives from cognitive neuroscience (pp. 255e284). Mahwah, NJ: Erlbaum.
Bowdle, B. F., & Gentner, D. (2005). The career of metaphor. Psychological Review, 112, 193e216.
Chiarello, C. (2003). Parallel systems for processing language: hemispheric complementarity in the normal brain. In M. T.
Banich, & M. Mack (Eds.), Mind, brain and language (pp. 229e247). Mahwah, NJ: Erlbaum.
Cieslicka, A. (2006). Literal salience in on-line processing of idiomatic expressions by second language learners. Second Language
Research, 22(2), 115e144.
Cieslicka, A. (2013). Do nonnative language speakers chew the fat and spill the beans with different brain hemispheres?
Investigating idiom ecomposability with the divided visual eld paradigm. Journal of Psycholinguist Research, 42, 475e503.
Cieslicka, A. B., & Heredia, R. R. (2011). Hemispheric asymmetries in processing L1 and L2 idioms: effects of salience and context.
Brain and Language, 116, 136e150.
Coltheart, M. (1981). The MRC psycholinguistic database. Quarterly Journal of Experimental Psychology, 33A, 497e505.
Faust, M., Ben-Artzi, E., & Vardi, N. (2012). Semantic processing in native and second language: evidence from hemispheric
differences in ne and coarse semantic coding. Brain and Language, 123, 228e233.
Faust, M., & Mashal, N. (2007). Right hemisphere advantage in processing novel metaphoric expressions: behavioral data.
Neuropsychologia, 45, 860e870.
Giora, R. (1997). Understanding gurative and literal language: the graded salience hypothesis. Cognitive Linguistics, 7, 183e206.
Giora, R. (2002). Literal vs. gurative language: different or equal? Journal of Pragmatics, 34, 487e506.
Giora, R. (2003). On our mind: Salience, context and gurative language. New York: Oxford University Press.
Hull, R., & Vaid, J. (2007). Bilingual language lateralization: a meta-analytic tale of two hemispheres. Neuropsychologia, 45,
1987e2008.
Jung-Beeman, M. (2005). Bilateral brain processes for comprehending natural language. Trends in Cognitive Science, 9, 512e518.
Kecskes, I. (2006). On my mind: thoughts about salience, context and gurative language from a second language perspective.
Second Language Research, 22, 219e237.
Laurent, J.-P., Denhieres, G., Passerieux, C., Iakimova, G., & Hardy-Bayle, M.-C. (2006). On understanding idiomatic language: the
salience hypothesis assessed by ERPs. Brain Research, 1068, 151e160.
Marian, V., Blumenfeld, H. K., & Kaushanskaya, M. (2007). The language experience and prociency questionnaire (LEAP-Q):
assessing language proles in bilinguals and multilinguals. Journal of Speech, Language and Hearing Research, 50, 940e967.
Mashal, N., & Faust, M. (2009). Conventionalization of novel metaphors: a shift in hemispheric asymmetry. Laterality, 14,
573e589.
Mashal, N., Faust, M., & Hendler, T. (2005). The role of the right hemisphere in processing nonsalient metaphorical meanings:
application of principal components analysis to fMRI data. Neuropsychologia, 43, 2084e2100.
Mashal, N., Faust, M., Hendler, T., & Jung-Beeman, M. (2007). An fMRI investigation of the neural correlates underlying the
processing of novel metaphoric expressions. Brain and Language, 100, 115e126.
Patael, S., Borodkin, K., & Faust, M. (In preparation). Lateralization and Language Creativity: Developmental Transition from
Adolescence to Young Adulthood.
Pobric, G., Mashal, N., Faust, M., & Lavidor, M. (2008). The causal role of the right cerebral hemisphere in processing novel
metaphoric expressions taken from poetry: a TMS study. Journal of Cognitive Neuroscience, 20, 170e181.
Rapp, A., Mutschler, D. E., & Erb, M. (2012). Where in the brain is nonliteral language? A coordinate-based meta-analysis of
functional magnetic resonance imaging studies. NeuroImage, 63, 600e610.

108

N. Mashal et al. / Journal of Neurolinguistics 35 (2015) 96e108

Schmidt, G. L., DeBuse, C. J., & Seger, C. A. (2007). Right hemisphere metaphor processing? Characterizing the lateralization of
semantic processes. Brain and Language, 100, 127e141.
Shatil, E. (1995). One-minute test for words. Unpublished test, the support center for students with learning disabilities. Haifa, Israel:
University of Haifa.
, L., Hinojosa, J. A., Manuel, M., Mercado, F., Lo
 pez-Martn, S., et al. (2005). Neural activity associated with
Sotillo, M., Carretie
metaphor comprehension: spatial analysis. Neuroscience Letters, 373, 5e9.