It has been shown previously that manipulation of the M demand (amount of information needed for processing) of chemistry problems affects student performance, which suggests that manipulation of logical structure of chemistry problems could also lead to significant changes in performance. The objective of this study is to investigate the following: Given the opportunity for training, what is the effect of increasing (manipulation) the complexity of logical structure of chemistry problems on student performance, and to what extent can cognitive variables explain changes in performance. Results obtained show that (a) even a small increase in the logical structure of a problem can change the role of cognitive variables (mental capacity and formal reasoning) to the extent that increase in logical complexity outweighs the advantage students may have gained through training on a similar problem; (b) the use of algorithms and training on particular types of chemistry problems could lead to a situation in which formal reasoning is the only cognitive variable that explains variance in performance significantly; and (c) after having solved very similar problems on two different occasions with improving performance, the improvement is not retained if the logical structure of a third problem increases considerably. It is concluded that when dealing with significant changes in logical complexity of chemistry problems, developmental level of students is the most consistent predictor of success. A model for the qualitative analysis of logical complexity of chemistry problems is presented.