The way in which abstract concepts are represented has been discussed more intensely because of the emergence of the embodied cognition. According to the embodied theories, cognition is fully grounded in the body, and human concepts are “rooted” in perception, action and emotion systems. Accumulating evidence suggests that concrete concepts are closely linked to sensorimotor system, whereas researches on the role of the sensorimotor system for abstract words have not come to an agreement. The present study addresses the issue of whether the abstract concepts, specifically abstract verbs, are embodied or grounded in sensorimotor system. In most of the previous studies, abstract concepts have been considered as an unitary whole, which neglect the fact that different abstract words are linked to sensorimotor experiences differently. We thus hypothesize that it is the various proportion of the sensorimotor features in the abstract word stimuli that may explain the inconsistent results. Here we performed two behavioral experiments to test our hypothesis. The Action-sentence Compatibility Effect, a typical and powerful evidence of embodied theory, is adopted to indicate the involvement of sensorimotor system in semantic processing. In both experiments, participants were required to judge the word a verb or not. Experiment 1 investigated the ACE of the concrete and abstract verbs (e.g., “投掷” / “抛弃”), with the aim of examining whether individual two-syllable Chinese action verb meaning was processed through mental simulation. The repeated measurement design of 2 (action verb types: concrete/abstract verbs) × 2 (action congruency: the direction of the verb and the required movement is congruent/incongruent) was adopted. Forty-five participants volunteered in this study. Experiment 2 further classified the abstract verbs into two types according to the proportion ratings of the motor feature, and investigated the ACE of high and low motion abstract verbs (e.g., “承受” / “流露”), in order to elaborate the influence of motor features in the verb on the sensorimotor representation of abstract concept. The repeated measurement design of 2 (abstract verb types: high/low motion abstract verbs) × 2 (action congruency: the direction of the verb and the required movement is congruent/incongruent) was adopted. Forty-nine participants volunteered in this study. Results showed that ACE was only observed in the concrete verb processing of experiment 1(p<.05), and in the high motion abstract verb processing of experiment 2(p<.05), but not in the low motion abstract verbs(p>.05). That is, when the direction of the verbs and the required movement was congruent (i.e., both were toward or away from the participant), the reaction time of semantic judgement was significantly faster. Furthermore, in the experiment 2 the correlation analysis indicated the proportion ratings of the motor feature was significantly correlated to the magnitude of the ACE (i.e., difference between reaction times for incongruent and congruent trials) (r=.461, p<.001). In conclusion, the findings demonstrate that a role of motor system in semantic processing is not restricted to concrete words but extends to at least the abstract verbs in which the proportion of the motor feature are high. Specifically, the results suggest that 1) ACE is present in the two-syllable Chinese action verbs, so verbs are processed through sensorimotor simulation. and 2) motor system plays a role in the abstract verb processing, yet the proportion of the motor feature in the verbs could affect the involvement of motor system. Overall, the present study supports the embodied representation views indicating that sensorimotor system is involved in the abstract concept processing and sensorimotor features in the abstract words do matter a lot.