The construction of a high quality mental model from a complex visual
display relies on the capacity of learners to extract appropriate info
rmation from that display. Beginning students of meteorology complied
written records of generalisations extracted from animated weather map
sequences in order to prepare themselves for a subsequent prediction
task. Analysis of these records revealed that much of the information
extracted was perceptually salient rather than thematically relevant.
This perceptual dominance effect was found for both visuospatial and t
emporal aspects of the display. The statements produced were deficient
with regard to the causal explanations that would be necessary to bui
ld a satisfactory mental model of the depicted situation. These defici
encies involved both the proportion of causal material recorded and th
e attribution of causality on an everyday rather than a domain-appropr
iate basis.