Objective To evaluate the role of stone size in predicting urinary calculus composition using Hounsfield units on noncontrasted computed tomography (CT) scan. Methods A retrospective review was performed for all patients who underwent ureteroscopy or percutaneous nephrolithotomy during a 1-year period, had a stone analysis performed, and had CT imaging available for review. All CT scans were reviewed by a board-certified radiologist. Variables evaluated included age, sex, body mass index, stone size, stone location, Hounsfield units (HUs), and stone composition. Results We identified a total of 91 patients (41 men and 50 women) with CT imaging and stone analysis available for review. Stone analysis showed 41 calcium oxalate monohydrate (CaOxMH), 13 calcium oxalate dihydrate, 29 calcium phosphate, 5 uric acid, 2 struvite, and 1 cystine stone. Average age was 46 years, and average body mass index was 32 kg/m2. Measured HUs varied significantly with size for CaOxMH and calcium oxalate dihydrate stones (P values <.05), but not for calcium phosphate stones (P =.126). Using a CaOxMH identification value of 700-1000 HUs, 28 of 41 stone compositions (68%) would not have been correctly identified, including all 10 (100%) small (<5 mm) stones, 13 of 22 (59%) medium (5-10 mm) stones, and 5 of 9 large (>10 mm) stones (55%). Conclusion For calcium stones, the ability of CT HUs to predict stone composition was limited, likely due to the mixed stone composition. Within a cohort of CaOxMH stone formers, measured HUs varied linearly with stone size. All stones <5 mm were below thresholds for CaOxMH composition.