Fidelity requirements of flight simulators are addressed by proposing a classification based on informational, perceptual, experiential, and action implications. Fidelity assesment is argued to begin with cue inventory definition and empirical research. Cue evaluation is complex due to the presense of multiple cues for a single referent. A dominance hierarchy may emerge allowing little incidental learning to weak cues. An altitude holding experiment is reported to illustrate cue dominance effects. Other studies are discussed which also indicate possible non-optimun effects of using scenes with multiply redundant cues. Since certain flight situations may delete dominant cues, pilots need to be experienced using the weaker cues and this implies using simulators in reduced cue modes. Hence, maximun fidelity is not always desirable.