The hypothesis that emotions influence our behavior via emotional action tendencies is at the core of many emotion theories. According to a strong version of this hypothesis, these emotional action tendencies are immediate, automatic (unintentional), stimulus-based and directly linked with specific muscle movements. Recent evidence, however, provides little empirical support for this strong version during routine behavior, especially when the task does not require the evaluation of the stimuli. The present study tested the prediction that surprise interrupts routine behavior and (...) triggers a threat avoidance response. In the presence of a threat-related stimulus, avoidance responses are relatively rapid, and approach responses impeded, even when the interrupted routine behavior is guided by a non-evaluative task goal. In contrast, approach and avoidance responses are predicted to be unaffected in the presence of a pleasant surprising stimulus. To test these predictions, in each trial the participants had to execute an approach or withdrawal movement depending on the location of a target stimulus. In the critical trial, either a picture of a pleasant or a threat-related animal was presented as target. Supporting the predictions, the initiation times for these movements were shorter in response to a threat-relevant than a pleasant surprising stimulus. Additionally, in the presence of a threat-related surprising stimulus, withdrawal movements were made faster than approach movements even though the participants performed a non-evaluative task. Implications and limitations of the present study are discussed. (shrink)