A computational model of spatial relational reasoning implemented in the ACT-R cognitive architecture allows for the simulation of a wide range of behavioral data in the context of both determinate and indeterminate deductive spatial reasoning tasks. In that respect the presented study bridges the gap between results of previous work that investigated determinacy conditions separately. ACT-R’s subsymbolic processing principles substantially contribute to the underlying theory of Preferred Mental Models as they add a powerful component making precise accuracy predictions possible. In addition, the data is informative about a possible strategy when the task is to judge if an externally presented spatial description matches a mental model that resulted from the current reasoning process.