Models of attention and emotion assign a special status to the processing of threat. While evidence for threat-related attentional bias in highly anxious individuals is robust, effects in the normal population are mixed. An important explanation for the absence of threat-related attentional bias in nonanxious individuals may relate to the spatial frequency components of stimuli. Here we report behavioral data from two experiments examining the relationship between spatial frequency components of emotional and neutral faces and fast saccadic orienting behavior. In Experiment 1 participants had to saccade toward a single face, filtered to include mostly low, high or broad spatial frequencies (LSF, HSF or BSF), posing a fearful, happy or neutral expression presented for 20 ms in the periphery. At BSF a general emotional effect was found whereby saccadic responses were faster for fearful and happy faces relative to neutral, with no significant differences between fearful and happy faces. At LSF both fearful and happy faces had shorter saccadic latencies in comparison to neutral, demonstrating an emotional bias consistent with the BSF data. However, at LSF fearful faces resulted in significantly faster saccades than happy faces indicating that this bias was stronger for threat-related faces. There was no difference in saccadic responses between any emotions at HSF. Experiment 2 showed that the emotional bias diminished for inverted stimuli suggesting that the results were not attributable to low-level image properties. The findings suggest an overall advantage in the oculomotor system for orientation to emotional stimuli and at LSF in particular, a significantly faster localization of threat conveyed by the face stimuli in all individuals.