This theoretical model of emotion is based on research using the startle-probe methodology. It explains inconsistencies in probe studies of attention and fear conditioning and provides a new approach to emotional perception, imagery, and memory. Emotions are organized biphasically, as appetitive or aversive (defensive). Reflexes with the same valence as an ongoing emotional state are augmented; mismatched reflexes are inhibited. Thus, the startle response (an aversive reflex) is enhanced during a fear state and is diminished in a pleasant emotional context. This affect-startle effect is not determined by general arousal, simple attention, or probe modality. The effect is found when affects are prompted by pictures or memory images, changes appropriately with aversive conditioning, and may be dependent on right-hemisphere processing. Implications for clinical, neurophysiological, and basic research in emotion are outlined.