The hypotheses of this investigation were based on attachment theory and Bowlby's conception of "internal working models", supposed to consist of one mainly emotional (model-of-self) and one more conscious cognitive structure (model-of-others), which are assumed to operate at different temporal stages of information processing. Facial muscle reactions in individuals with positive versus negative internal working models were compared at different stages of information processing. The Relationship Scale Questionnaire (RSQ) was used to categorize subjects into positive or negative model-of-self and model-of-others and the State-Trait Anxiety Inventory was used to measure trait anxiety (STAI-T). Pictures of happy and angry faces followed by backward masking stimuli were exposed to 61 subjects at three different exposure times (17 ms, 56 ms, 2,350 ms) in order to elicit reactions first at an automatic level and then consecutively at more cognitively elaborated levels. Facial muscle reactions were recorded by electromyography (EMG), a higher corrugator activity representing more negative emotions and a higher zygomaticus activity more positive emotions. In line with the hypothesis, subjects with a negative model-of-self scored significantly higher on STAI-T than subjects with a positive model-of-self. They also showed an overall stronger corrugator than zygomatic activity, giving further evidence of a negative tonic affective state. At the longest exposure time (2,350 ms), representing emotionally regulated responses, negative model-of-self subjects showed a significantly stronger corrugator response and reported more negative feelings than subjects with a positive model-of-self. These results supported the hypothesis that subjects with a negative model-of-self would show difficulties in self-regulation of negative affect. In line with expectations, model-of-others, assumed to represent mainly knowledge structures, did not interact with the physiological emotional measures employed, facial muscle reactions or tonic affective state.