The current study examined a new spatial integration (SI)task, based on figural rather than linguistic stimuli, tomeasure the construct of mental modeling ability.Previous tasks conflated linguistic ability with mentalmodeling ability by requiring sentence processing, whichmay have contributed to mixed findings with respect tothe relationship between mental model ability andworking memory capacity (WMC). The figural spatialintegration task produced the canonical continuity effect,such that discontinuous items had lower accuracy thancontinuous items. Furthermore, WMC and visuospatialability predicted SI task performance, and both werestronger predictors for the continuous condition. Theinteractions between predictors and task conditionssuggest reliance on heuristics and/or rehearsal duringperformance of the more difficult discontinuous items.