A puzzling question arising from imagery research is why no relationship has been found between self-reports of imagery and performance on spatial tests thought to require the use of imagery. To investigate this, spatial ability, measured by performance on two spatial tests, was compared with performance on the Vividness of Visual Imagery Questionnaire (VVIQ; Marks, 1973) and a newly constructed imagery questionnaire. The choice of items and ratings for the new questionnaire was based on Kosslyn's (1980, 1994) theories of the imagery system. Ratings on the new questionnaire consistently correlated signié cantly with performance on the spatial tests, whereas ratings from the VVIQ did not. The new ratings captured more of the imagery process than ratings of vividness alone, but the largest change depended upon the type of item imagined. Ratings of items of the same type as used on the spatial tests predicted performance on the spatial tests, whereas vividness ratings of items recalled or constructed from long-term memory did not. Participants can successfully introspect on several different properties of their images,and theirratings do predict performance on tasks thought to requireimagery. The largeeffect of itemtyperaises questions about the predictive value of existing models of the imagery system especially in relation to the role of phenomenological properties of our images. Explanations of the results of mental rotation experiments often suggest that mental imagery plays a functional role (e.g. Shepard & Metzler, 1971). The linear increase in response times with increasing angular disparity between the stimulus items that are being compared provides an objective measure that supports subjective reports that mental imagery is involved in spatial manipulations. Paivio (1971) and others have suggested that spatial tests based on this paradigm can be used as indirect measures of imagery ability, and Ernest (1977) refers to such tests as ' objective' measures of imagery.