The article presents the MAC/FAC model, a computational framework designed to explain three seemingly contradictory psychological phenomena related to similarity-based retrieval: (a) structural commonalities are more heavily weighted than surface commonalities in working memory, (b) superficial similarity is more important in retrieval, and (c) occasional purely structural (analogical) reminders occur. The model consists of two stages: MAC and FAC. The MAC stage uses a computationally efficient, non-structural matcher to filter candidate long-term memory items based on content vectors, which are redundant encodings of structured representations. The FAC stage employs the Structure-Mapping Engine (SME) to compute structural matches on the filtered items, providing a more accurate but computationally expensive process. The model's effectiveness is demonstrated through computational experiments that align with psychological data, sensitivity analyses, and comparisons with another model, ARCS. The article also discusses the limitations and potential extensions of the MAC/FAC model, placing it within the context of recent work on similarity-based retrieval.The article presents the MAC/FAC model, a computational framework designed to explain three seemingly contradictory psychological phenomena related to similarity-based retrieval: (a) structural commonalities are more heavily weighted than surface commonalities in working memory, (b) superficial similarity is more important in retrieval, and (c) occasional purely structural (analogical) reminders occur. The model consists of two stages: MAC and FAC. The MAC stage uses a computationally efficient, non-structural matcher to filter candidate long-term memory items based on content vectors, which are redundant encodings of structured representations. The FAC stage employs the Structure-Mapping Engine (SME) to compute structural matches on the filtered items, providing a more accurate but computationally expensive process. The model's effectiveness is demonstrated through computational experiments that align with psychological data, sensitivity analyses, and comparisons with another model, ARCS. The article also discusses the limitations and potential extensions of the MAC/FAC model, placing it within the context of recent work on similarity-based retrieval.