Research investigating top-down attentional capture has demonstrated a tight coupling of working memory content with attention and eye movements. By capitalizing on this relationship, we have developed a novel methodology called the Memory Activation Capture (MAC) procedure for measuring the dynamics of working memory content supporting complex cognitive tasks (e.g., decision making, problem solving). By observing which items are preferentially fixated in task irrelevant arrays containing task relevant information, we gain a measure of working memory content as the task evolves through time. The efficacy of the MAC procedure is demonstrated in a hypothesis generation task. Results suggest a two-stage process following hypothesis retrieval whereby it undergoes a brief period of heightened activation before entering a lower activation state while being maintained for output. The present effects are of additional general interest as they represent the first demonstrations of top-down attentional capture driven by participant-established WM content retrieved from long-term memory.