Introduction: The diagnostic process is wrought with potential sources of error. Psychologists seek to coach physicians to refine their cognition. Researchers try to create cognitive scaffolds to guide decision-making. Physicians however, are caught in middle between their own daily cognitive processes and these external theories that might influence their behaviour. Few attempts have been made to understand how experienced clinicians integrate guidelines or clinical decision rules (CDRs) into their decision-making. We sought to explore experienced clinicians decision-making via a simulated exercise, to develop a model of how physicians integrate CDRs into their diagnostic thinking. Methods: From July 2015-March 2016, 16 practicing emergency physicians (EPs) were interviewed via a think aloud protocol study. Six cases were constructed and video recorded as prompts to spur the clinicians to think aloud and describe their approach to the cases. Cases were designed to be slightly suggestive for pulmonary embolism or deep vein thrombosis, since these conditions are associated with CDRs. Using a constructivist grounded theory analysis, three investigators independently reviewed the transcripts from the interviews, meeting regularly to discuss emergent themes and subthemes until sufficiency was reached. Disagreements about themes were resolved by discussion and consensus. Results: Our analysis suggests that physicians engage in an iterative process when they are faced with undifferentiated chest pain and leg pain cases. After generating an original differential diagnosis, EPs engage in an iterative diagnostic process. They flip between hypothesis-driven data collection (e.g. history, physical exam, tests) and analysis of this data, and use this process to weigh probabilities of various diagnoses. EPs only apply CDRs once they are sufficiently suspicious of a diagnosis requiring guidance from a CDR and when they experience diagnostic uncertainty or wish to bolster their decision with evidence. Conclusion: EP cognition around diagnosis is a dynamic and iterative process, and may only peripherally integrate relevant CDRs if a threshold level of suspicion is met. Our findings may be useful for improving knowledge translation of CDRs and prevent diagnostic error.