Like human face, human body is a reliable tool to convey emotional information in social interaction. Previous studies on human body have two important findings: (1) Threat-related body expressions gain more attention and processing than other body expressions. (2) The inversion effect for faces has been found for bodies, indicated that both faces and bodies are processed configurally. The configural processing of facial expression has also been confirmed. However, we don’t know whether body expressions are also processed configurally, given that only one problematic study combined threat-related body expressions and body-inversion effect. In addition, the time course of body expression processing is still unclear. Given all that, we employed the event-related potential (ERP) technique to investigate the time-course of perceiving threat-related body expressions (fear and anger). We also try to find out whether body expressions can be processed configurally. In the current study, neutral, fear and anger body expression pictures chose from The Bodily Expressive Action Stimulus Test (BEAST) (de Gelder & Van den Stock, 2011) were uprightly or inversely presented to 50 participants aged around 19 to 25. They were instructed to make an emotional categorization (neutral, fear or anger) to these pictures as rapid and accurate as possible. In the meantime, the ERP data was recorded from 64 scalp sites using Ag/AgCl electrodes mounted in an elastic cap. In the ERP data analysis, components such as P1, N1, N170, P3 and LPP were mainly concerned. Behavioral results confirmed the body-inversion effect, meaning that body expressions are recognized configurally. Moreover, the result showed that threat-related body expressions elicited a larger inversion effect, which proved our hypothesis. An interesting phenomenon was that even the recognition of emotion was impaired by inversion, the performance was still more accurate than chance level, demonstrated that featural processing can approximately convey emotional information. ERP results verified the configural processing of body expressions and revealed the distinct time course of threat-related body expressions processing from three cognitive process stages: (1) The P1 amplitude was increased to threat-related body expressions instead of neutral body expression, indicated that threat-related expression was rapidly processed and distinguished from neutral expression in the first stage. Moreover, the N1 was enhanced and delayed to inverted bodies, indicated that body-inversion effect existed in early process stage (N1) instead of later processing stage (N170). (2) The N170 in inverted condition was larger and more delayed than that in upright condition, demonstrated that body expressions are configurally processed. We can also see that the behavior result (a larger inversion effect to threat-related body expressions) was confirmed by the ERP result (a larger and more delayed N170 to threat-related body expressions). (3) In the third stage, threat-related information from the body gained more attention in upright condition than that in inverted condition, reflected by more pronounced P3 and LPP effects. Taken together, the current finding suggested a distinct time-course processing of threat-related body expressions. We also provide behavioral and electrophysiological evidence that the recognition of human body expressions is mainly based on the configural information.