When spatial ability was regressed on a measure of general intelligenc
e in a large national sample of 12th-graders, there was a distinct dow
nturn in the regression line at about -2 SD units. A bimodal distribut
ion of spatial ability was found among these low scoring students, sug
gesting 2 qualitatively different types of individuals. Relative to st
udents of expected spatial ability, those below expected in spatial ab
ility were high on verbal skills and low on performance skills. Studen
ts unexpectedly low in spatial ability performed relatively well on co
gnitive tests that required answering unambiguous questions by retriev
ing information directly from long term memory, whereas they performed
relatively poorly on tests requiring inference and perception of rela
tions among novel stimuli. Students unexpectedly low in spatial abilit
y performed especially poorly at tests involving visual-spatial percep
tion skills. Results were similar for both sexes. The relatively poor
performance of those unexpectedly low in spatial ability does not appe
ar to be related to health problems, personality, interest differences
, or biographical data. An organic impairment possibly related to late
ralization represents a reasonable causal hypothesis for these data.