Inconsistent risk estimates for dietary heterocyclic amine (HCA) exposure and cancers may be due to differences in exposure assessment methods and the associated measurement error. We evaluated repeatability and comparability of intake estimates of the HCA 2-amino-1-methyl-6-phenylimidazo[4,5b]pyridine (PhIP) among two food frequency questionnaire (FFQ) collections, three diary collections, and three measurements of urinary PhIP and its metabolites in 36 non-smokers in Baltimore, Maryland, during 2004-2005. Collections spanned ∼9 months. Method repeatability was characterised with intraclass correlation coefficients (ICCs). Comparability among methods was assessed with Spearman correlation coefficients. Within-subject variability in PhIP intake was comparably high across all methods (ICCs of 0.20, 0.30, and 0.15 for FFQ, diary, and creatinine-adjusted urinary PhIP, respectively). Mean diary-based PhIP intake and mean urinary PhIP concentration were strongly correlated when restricting the analysis to participants with at least one non-zero diary-based estimate of PhIP intake (n = 15, r = 0.75, p = 0.001), but not in the full study population (n = 36, r = 0.18, p = 0.28). Mean PhIP intake from the FFQ was not associated with that either based on the diary or urinary PhIP separately, but was modestly correlated with a metric that combined the diary- and biomarker-based approaches (r = 0.30, p = 0.08). The high within-subject variability will result in significantly attenuated associations if a single measure is used to estimate exposure within an epidemiologic study. Improved HCA assessment tools, such as a combination of methods or validated biomarkers that capture long term exposure, are needed.