The Corsi Block-Tapping (CB) task has been used as a measure of spatial memory since its development in 1971. However, a standard set of items has been developed for this task, and inconsistencies in performances within levels have been demonstrated in association with different path configuration. This study investigated item consistency by analyzing the performances of 94 young adult participants on a block-tapping task that involved five quasi-randomly determined sequences at each of nice levels of difficulty. In general, performance declined with increasing path length. Cochran Q-test comparisons were conducted on the items within each level, and differential performance were identified at levels 7 and 8 only. Pairwise comparisons determined the specific items for which performance was discrepant, and further analysis indicated that performance decrement were related to more complicated block-tapping paths. The findings suggest that this version of the CB task is relatively consistent overall, and the observed effect of path configuration indirectly corroborated the spatial nature of this task. Performance heterogeneity at higher levels reflected more complicated path configuration and, presumably, greater span capacity load. Differential intra-level item consistency should be considered in clinical applications of spatial (configural) memory tasks in order to avoid erroneous interpretations concerning sustained attention ability based on failures within levels.