Associative blocking in human conditioning was investigated using electrodermal and self-reported US expectancy measures. Previous null results using a design reported by Lovibond, Siddle, and Bond (1988) suggested that a clearly demarcated phase structure and visual cues with semantic content may have distracted attention from the experimental contingencies. Therefore the current experiment intermixed pre-training and compound training trials and masked the transition to the test phase to remove or reduce phase boundaries. Simple coloured squares were used as CSs to reduce semantic content. A significant blocking effect was observed on both the expectancy measure and on the electrodermal measure. Both results were due to improved transfer of conditioning to the target CS in the overshadowing control group in comparison to previous experiments. The results were interpreted as providing evidence that previous null results were due to failure of transfer of learning across clearly distinct phases. There was no evidence of a dissociation between the electrodermal and self-report measures. Theoretical and procedural implications for human Pavlovian conditioning are discussed.