In a decision making task repeated encounters of a visual stimulus attribute to improved performance in identification and classification of that stimulus. This process is termed as repetition priming. Recent studies have proposed that response learning (RL), where previous decision gets associated with that item, accounts for a significant portion of performance gains observed in repetition priming. However, RL is unable to explain intact priming in amnesic patient population and residual priming in healthy population after the decision cue was switched. Furthermore, it is still unclear whether RL is dependent on the well structured knowledge and object representations. In this paper two-fold approach is taken to fully address the aforementioned issues. First, to test the boundaries of RL and its dependence on knowledge structures, a new repetition priming experiment using 2-D novel objects is presented in healthy population. Experimental results indicated that novel objects result in short-term priming, and presenting the object multiple times did not further facilitate reduction in response time. This pattern does not fit prevailing theories of repetition priming, including RL. A new perceptual memory-based association mechanism is thereby required. Second, to formalize and test such a mechanism, a computational model is developed. The model was further extended to formulate priming using known objects, thereby providing explanation for residual priming in case of cue-inversion in healthy population and intact priming in amnesic patient population. Altogether this paper presents the missing link, perceptual memory-based association mechanism, to fully capture all the memory processes involved in decision making.