Abstract: Embodied language cognition highlights that language processing not only involves internal representation of purely abstract symbols but also body and environment which play an important role. The majority of the evidences are from the field of the first language (L1), and there is still debate on whether this effect exists in the second language (L2) or not. We proposed two hypotheses, “automatic activation” and “native language mediation”, after summarizing the evidence for L2 embodiment effect from behavioral and neurophysiological studies. “Automatic activation” means that embodiment effect occurs synchronously between L1 and L2. However, the “native language mediation” hypothesis expresses that the embodiment effect in the L2 may parasite on the L1, so it will not occur synchronously between L1 and L2. L1 embodiment effect occurs earlier than the L2. Several factors, such as proficiency, age of acquisition (AoA), and acquiring methods of L2, could account for the difference between L1 and L2 in the timing of the embodiment effect. For L2 learners with earlier AoA and high proficiency, automatic activation of the embodied cognition may be more evident, while late acquisition and low proficiency may mean that L2 learners need to rely on L1 mediation. Furthermore, if an L2 is learned in a classroom or a relatively artificial environment, linguistic symbols lack direct perceptual connection with the objects, events, or actions they describe, so learners are more likely to carry out psychological operations inside their brains, preventing them from forming a complete mental representation. In this case, the embodiment effect of L2 is more likely the consequence of conceptual or mental representation simulation based on linguistic symbol description.
Although some valuable conclusions have been drawn from existing studies, there are still many controversies. Firstly, because there are significant differences between L1 and L2 in the way of learning, learning duration, and proficiency, researchers argued that the L2 was represented less grain-fined than L1 and the processing of L2 is “disembodied”. However, we argue that the embodiment effect is not “all or none”, but a “continuum”. The embodiment effect of L2 may be at the beginning of the “continuum”, and so its intensity and mode of expression are different from that of L1. Secondly, although both hypotheses are plausible, it is not clear which explanation is more universal, though not completely contradictory. With the increase of L2 proficiency, the embodiment effect gradually transforms from “native language mediation” to “automatic activation”.
Still, there are some unresolved questions. For example, at what proficiency level does L2 achieve the transition from “native language mediation” to “automatic activation”? Is the transition precipitous or gradual? In addition to L2 proficiency, what are the roles do AoA, language similarities, styles of language acquisition, and task types play in the transition between them? Probably, studies on the L2 embodiment effect among bilinguals with different reading and writing habits may help to resolve the above debate. Finally, if “native language mediation” is suitable for L2 learners with low proficiency, is the mediation realized at the lexical level or semantic conceptual level? Although research on the L2 embodiment effect is still controversial, studies on this issue are helpful to enrich the theory of language embodied cognition and further clarify the mechanism of bilingual representation, and effectively improve L2 teaching.

Key words: embodied cognition, automatic activation, native language mediation, L2 processing