The semantic integration between the daily source domain and the scientific target domain involved in scientific metaphor processing provides a perspective for exploring the similarities and differences between processing mechanisms of native language and non-native language (Jankowiak et al., 2017). However, previous studies focused on conventional metaphors and devoted little attention to scientific metaphors. Such metaphors are frequently deployed by scientists who use concrete and intuitive objects and knowledge to describe abstract concepts or objects that are impossible to observe directly. The mechanism of scientific metaphors reflects the core idea of conceptual metaphor theory, the metaphorical understanding of abstract concepts through more specific knowledge structures (Lakoff and Johnson, 1980). The continuous innovation of natural science requires the reorganization and construction of thinking and expression methods in systems related to new concept expression (Tang et al., 2017a,b). Compared with conventional metaphors, the processing of scientific metaphors involves a unique mechanism involving a complicated contextual structure and knowledge inference, which might provide insight in observing the between-language differences of the brain correlates modulated by nativeness, thus deepening the comparative study of bilingual processing mechanisms. Moreover, the comparative study of two different types of metaphors (scientific metaphors and conventional metaphors) might shed new light on the discussion of the processing mechanism of figurative language.

The processing of scientific metaphors differs from that of conventional metaphors mainly in two aspects. First, the contextual structure of scientific metaphors is more complex than that of conventional metaphors, which has been established in previous N400 experiments (Tang et al., 2017a). For example, the source domain (code) of the scientific metaphor (A mitochondrion is a code) is more frequently used in everyday contexts, while the target domain (mitochondria) is more frequently used in scientific contexts. Second, the late processing period of scientific metaphors probably involves a reasoning process from the daily concrete concept to the scientific abstract concept, so as to better comprehend the embedded scientific knowledge. The late components of ERP [the late positive component (LPC) and the late negativity] are very sensitive to this process.

Event-related potentials (ERPs) with extremely high time resolution and rich component dimensions are increasingly used in the study of metaphorical cognitive neural mechanisms. The N400 is considered to be very sensitive to semantic violations (Kutas and Hillyard, 1980), and is an important index to measure the difficulty in retrieving information stored in the mind. Several monolingual studies on figurative meaning processing reported a modulation of N400 amplitudes by the degree of metaphor conventionality (Pynte et al., 1996; Tartter et al., 2002; Coulson and Van Petten, 2007; Lai et al., 2009). Such results seem to be in line with the Career of Metaphor Model (Bowdle and Gentner, 2005), which postulates that novel metaphors are understood through comparison, while conventional metaphors are preferentially analyzed by categorization. Namely, the processing of novel metaphors involves structure alignment between a metaphoric target and a literal base. In contrast, when processing conventional metaphors, the metaphoric target concept is understood as a member of a superordinate category specified by the literal base term (Jankowiak et al., 2017). Thus, understanding novel metaphors requires engagement in sense creation, while understanding conventional metaphors involves easier sense retrieval, as revealed by enhanced N400 amplitudes for novel metaphors (Arzouan et al., 2007a; Lai et al., 2009). In addition to novelty, N400 is also sensitive to concreteness. Forgács et al. (2014, 2015) proposed that most metaphors describe abstract concepts, which, in their study, elicited larger N400s compared with abstract literal expressions when stimulus novelty was controlled across types. In this study, the target words of scientific metaphors were scientific terms and were relatively more abstract than the daily targets of conventional metaphors, which might contribute to the enhanced N400 amplitudes.

The late positive component (LPC) is considered to reflect integration or reprocessing at the sentence level (Kaan et al., 2000). Some monolingual studies found that novel metaphors elicited greater LPCs than conventional metaphors (De Grauwe et al., 2010; Weiland et al., 2014), which indicates that the semantic processes of metaphorical sentences require additional retrieval of information from semantic memory. Other monolingual studies found that novel metaphors elicited smaller LPCs than conventional metaphors (Arzouan et al., 2007a,b; Zhao et al., 2011; Ma et al., 2016), which might be caused by a late negativity overlapping in the LPC time window, reflecting the secondary processing of further semantic integration. Such late negativity is also considered as a continuation of the N400, indicating a sustained difficulty in fusing two concepts from the source domain and the target domain (Goldstein et al., 2012; Rutter et al., 2012).

So far, few studies focused on the processing mechanism of scientific metaphors for bilingual speakers. Thus, expanding a monolingual electrophysiological study on scientific metaphor processing to include a bilingual component might be insightful to enunciate conceptual integration mechanisms when processing semantically complex meanings. Some bilingual ERP studies have reported a lower N400 amplitude elicited by L2 relative to L1 stimuli (Proverbio et al., 2002; Moreno et al., 2008; Midgley et al., 2009; Newman et al., 2012; Heidlmayr et al., 2015), probably indicating weaker interconnectivity for L2 words within the semantic network compared with L1 words (Midgley et al., 2009). Consequently, when processing non-native language, the activity in long-term memory decreases due to weaker interconnectivity. The functional role of the N400 effect is linked to memory operations involved in information retrieval (Kutas and Federmeier, 2000; Kotz et al., 2012). The current study seeks to build on existing research as to semantic processing in bilinguals by using materials with different semantic complexity ranging from highly complex (scientific metaphoric) and relatively complex (conventional metaphoric) to relatively simple (literal) utterances. To minimize the potential influence of any between-group individual differences, a within-subject design was adopted. Accordingly, contrastive analyses of cognitive mechanisms were carried out to investigate native and non-native figurative language comprehension, which can further reveal how semantic complexity and language dominance interact with each other when processing bilingual languages.

The main objective of this study was to observe brain responses to scientific metaphors, conventional metaphors, and literal expressions in Chinese (L1) and English (L2). First, we aimed to observe whether scientific metaphors would evoke higher N400 amplitudes compared with conventional metaphors in both L1 and L2. That is to say, our first focus was the modulation of conventionality for the N400 effect in two languages. A linear N400 effect has been reported by two previous monolingual studies (Tang et al., 2017a,b), with more pronounced N400 amplitudes for scientific metaphors relative to conventional metaphors. This increased processing difficulty was evident due to more demanding mappings required for scientific metaphor comprehension when complicated context and scientific reasoning are engaged. Second, we aimed to observe whether English scientific metaphors would elicit attenuated N400 effects in comparison to Chinese ones, as has been found in previous bilingual research on semantic processing (Weber-Fox and Neville, 1996; Proverbio et al., 2002; Phillips et al., 2004; Moreno and Kutas, 2005; Midgley et al., 2009; Braunstein et al., 2012; Newman et al., 2012; Heidlmayr et al., 2015). Lower N400s elicited by English materials compared with Chinese ones would further imply weaker semantic interconnectivity for L2 when compared with L1 stimuli (Midgley et al., 2009). Third, we aimed to examine whether language nativeness would modulate cognitive mechanisms involved in semantic reintegration at the later stage of processing, as indexed by the LPC response. Several monolingual studies have reported higher LPC amplitudes as the index of secondary semantic integration (van Herten et al., 2005; De Grauwe et al., 2010; Brouwer et al., 2012; Rataj, 2014). It was expected that English scientific metaphors would elicit decreased LPC amplitudes in comparison to Chinese ones. Lower LPCs elicited by English materials relative to Chinese ones would point to weaker semantic processing for L2 compared with L1 sentences (Newman et al., 2012). Furthermore, if the LPC effect is modulated by the conventionality of metaphors (Arzouan et al., 2007a), scientific metaphors should evoke more pronounced LPC amplitudes than conventional metaphors, indicating the continuation of information retrieval or access to the non-literal route. Finally, we expected that main effects between LPCs across items could be observed in both L1 and L2, indicating a similar sensitivity to different levels of conventionality of metaphors.

The first aim of the current study was to observe brain responses to scientific metaphoric, conventional metaphoric, and literal sentences in Chinese (L1) and English (L2). In line with what was hypothesized, in both the Chinese and English experiments, we observed general between-condition differences, with scientific metaphors eliciting higher N400s than conventional metaphors. However, in the English experiment, more significant effects were found in larger regions over frontal, central, and parietal regions, with a slight right hemisphere bias. Importantly, this finding accords with the behavioral results drawn from the Chinese experiment, which showed lower accuracy rates and longer reaction time for scientific metaphors than with conventional metaphors. This supports the graded salience hypothesis for metaphorical expression. The unique complexity and abstraction of scientific language reduce the explicitness of language expression and have an impact on semantic integration in the later stage of processing. The longer distance between the scientific target domain and the daily source domain for scientific metaphor comprehension might increase cognitive effort in conceptual integration (Tang et al., 2017a,b). According to the Career of Metaphor Model (Bowdle and Gentner, 2005), the processing mechanism of metaphors is regulated by their familiarity. The understanding of metaphors with low familiarity requires the construction of metaphorical meaning through lexical semantic processes, while the understanding of metaphors with high familiarity is mainly carried out through semantic retrieval (Coulson and Van Petten, 2002; Lai and Curran, 2013). In addition, compared with literal expressions, conventional metaphors elicited a larger N400 response, which might indicate that despite being frequently used with high familiarity, conventional metaphors require more resource-intensive mappings between concepts than literal expressions.

In both Chinese and English experiments, LPC amplitudes evoked by scientific metaphors were smaller than those elicited by conventional metaphors and literal sentences. Lower LPC amplitudes for scientific metaphors might be caused by a sustained negativity, indexing the integration process of conceptually taxing meanings. Another explanation could be due to reprocessing operations after an initial failure in meaning interpretation (Ruchkin et al., 1988; Rataj, 2014) or alternatively extra working memory load for complex semantic processing (Ruchkin et al., 1988; Anderson et al., 1996; Oberauer et al., 2001; Jiang et al., 2009). According to some monolingual research on metaphors, novel metaphors elicited larger late negativity amplitudes than conventional metaphors, which was interpreted as the continuation of information retrieval or access to the non-literal route when understanding novel metaphors (Arzouan et al., 2007a), and as the ongoing difficulty of meaning integration indexed as the continuation of the N400 effect (Rutter et al., 2012). The larger late negativity of scientific metaphors might further indicate more difficult reintegration of the two domains, especially when the knowledge inference is involved at the later stage of scientific metaphor processing.

Second, the current study aimed to observe whether language nativeness modulates meaning integration mechanisms of scientific metaphors, to which the N400 response is sensitive. More specifically speaking, we were interested in whether L2 scientific metaphors would elicit similar N400s as L1 equivalents for late unbalanced bilingual speakers.

Between-language effects were found with English scientific metaphors eliciting more negative N400s than Chinese ones at the parietal region. Such enhanced N400 response to L2 relative to L1 stimuli is contrary to the results of some bilingual research (Proverbio et al., 2002; Moreno et al., 2008; Midgley et al., 2009; Newman et al., 2012; Heidlmayr et al., 2015). Within the memory system, weaker semantic interconnectivity for L2 compared with L1 words might explain the increased N400 amplitude (Midgley et al., 2009). From the opposite perspective, within the semantic network, larger interconnectivity for L1 words might be linked to the N400 L1/L2 effect (increased N400 amplitudes for L1 relative to L2 words). However, for scientific words, the picture might be different. Compared with conventional words, the semantic connectivity of scientific words might be quite weak in both native and non-native languages. This conjecture would seem to be supported by the behavioral results of the present English experiment. Such weak interconnectivity for scientific words might evoke similar activity in long-term memory in both languages for information retrieval. Unlike the daily words used in conventional metaphors, the scientific terms used in scientific metaphors are even less frequently used, which might result in the much lower familiarity for L2 relative to L1 scientific words. Therefore, the semantic processing of second language scientific terms is more challenging than that of Chinese ones. Moreover, for ordinary L2 speakers (Chang and Wang, 2016), the processing of L2 vocabulary often requires a greater degree of suppression of native vocabulary (Heidlmayr et al., 2015; Wu and Thierry, 2017), which further leads to difficulty in processing L2 words, especially scientific terms.

In addition, at the frontal, central, and parietal regions of the right hemisphere, English scientific metaphors elicited higher N400s than English conventional metaphors, indicating the special role of the right hemisphere in second language processing (Van Der Meij et al., 2011), which supports the Fine-Coarse Semantic Coding Theory (Mashal et al., 2015). Meanwhile, at the left parietal region, there was a marginally significant difference between the two metaphorical sentences, probably showing that the left hemisphere is also involved in understanding L2 scientific metaphors (Kim et al., 2017; Segal and Gollan, 2018).

Third, the current study aimed to observe the modulation of language nativeness for cognitive mechanisms involved at the later stage of meaning reintegration, which the LPC is sensitive to. In line with our predictions, within the LPC time frame, at the parietal region, we observed between-language differences with a slight right hemisphere bias, with smaller LPC responses evoked by the three English conditions than their Chinese counterparts. Reduced LPC amplitudes for scientific and conventional metaphors in the non-native language suggest more demanding cognitive effort to integrate the meaning of both novel and familiar metaphors in an L2 context. This result might indicate that late speakers who master two languages asymmetrically are less sensitive to the levels of conventionality of metaphoric meanings at the later stage of metaphoric language processing.

Inconsistent with the processing of Chinese scientific metaphors, English scientific metaphors only elicited lower LPCs than English conventional metaphors at the parietal region. That is to say, the LPC distribution of L1 scientific metaphors covered a larger area than that of L2 scientific metaphors (Jankowiak et al., 2017; Segal and Gollan, 2018), probably because it might be quite difficult for late L2 learners to reach a similar processing depth as displayed by the native speakers (Newman et al., 2012). In addition, both the left and right hemispheres are involved in the processing of L2 scientific metaphors (Chen et al., 2013), and the right parietal region might play a particularly important role.

Through a comparative analysis of the ERP components elicited by scientific metaphors in English and Chinese, this study examined brain responses to scientific metaphors in L2. It was found that nativeness modulates the cognitive cost for semantic integration at an early stage and for semantic reintegration and knowledge inference at a later period, supporting the Career of Metaphor Model and the Graded Salience Hypothesis.

In addition, the scalp distributions of the N400s and the late component elicited by scientific metaphors in L2 reinforce the essential role of the parietal region (especially the right parietal region) in processing L2, supporting the Fine-Coarse Semantic Coding Theory. More abundant types of stimuli could be added to subsequent follow-up experiments for more comparative analysis so as to continuously verify and improve existing studies.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics Committee in Shaanxi Normal University. The patients/participants provided their written informed consent to participate in this study.

XT and WR contributed to conception and design of the study. XT, MH, YH, and SH performed the data collection. LS, MH, and PY performed the analysis. XT wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.