Two fundamental theories in social cognition are long debated about how people understand the mental states of others. The mentalizing account posits mind-reading abilities as reflective, theory-based inferential processes that develop gradually from childhood learning, whereas the embodiment account contends them as reflexive, simulation-based enactive processes that are genetically inherited. Recent research found new evidence that even infants and non-human primates can automatically represent what others see, intend, and believe. For instance, infants and chimpanzees can expect an agent to reach toward a location where he or she believes a desirable object to be hidden, even when the agent’s belief is false. This ‘implicit mentalizing’ is interpreted as the root form of explicit, language-based mind reading skills in adults, and is believed to build a bridge linking two theories, since it confirms the inferential component of mentalizing account and also favors the nativist view of embodied account. However, a new theory named ‘submentalizing’ was recently proposed to provide an alternative interpretation for implicit mentalizing. By re-examining methodological flaws of all relevant studies in children and adults, submentalizing theorists assert that implicit mentalizing depends on general purpose neurocognitive mechanisms, rather than mechanisms that are specialized for inferential process or embodied process. Specifically, they point out that most empirical evidence for implicit mentalizing to date could be simply caused by domain-general cognitive mechanisms (e.g. Involuntary attentional orienting, spatial memory, retroactive interference and distraction), because they all can simulate the behavioral effects of mentalizing. Therefore, the submentalizing theory concludes that humans may not need dedicated mentalizing/embodiment as much as previously thought for social cognition—domain-general processes alone can provide a fast and efficient substrate to understand other’s mind in everyday life, allowing people to navigate a wide range of social situations without really thinking about or simulating mental states. We suggest this ‘submentalizing’ theory provides a novel and promising approach to explore the underlying mechanisms of implicit mentalizing. It goes beyond the classical debate of ‘mentalizing vs embodiment’ in social cognition, and instead posits submentalizing processes can ultimately substitute both theories. Its theoretical framework not only challenges the core position of mental states ascription in mentalizing account, but also refutes the nativist view of embodied account. It motivates researchers to re-evaluate all existing evidence of implicit mentalizing and provides a reductionism perspective to understand the nature of social cognition (i.e. the relationship between social and non-social processes). In addition to its theoretical contributions, the submentalizing theory also gives constructive implications for methodological issues and sheds light on future research directions. For example, future studies in mind reading should start to utilize well-controlled procedures (e.g. self-informed belief induction) to manipulate subjects’ perceptual and belief states, and also need to include closely matched inanimate control conditions and subtle attentional measures to completely rule out the interpretations of domain-general processes. In conclusion, we suggest the submentalizing theory has taken an ambitious but solid first step to question current theories in social cognition. To fully validate its claims and implications (e.g. substitute mentalizing/embodiment theories), future studies are needed to be implemented in multiple disciplines (i.e. neuropsychology and primate research) and with novel experimental strategies and advanced technologies (e.g. eye-tracking, fMRI, fNIRS).