IMPORTANCE: Positive margins are associated with poor prognosis among patients with oral tongue squamous cell carcinoma (SCC). However, wide variation exists in the margin sampling technique. OBJECTIVE: To determine the effect of the margin sampling technique on local recurrence (LR) in patients with stage I or II oral tongue SCC. DESIGN, SETTING, AND PARTICIPANTS: A retrospective study was conducted from January 1, 1986, to December 31, 2012, in 5 tertiary care centers following tumor resection and elective neck dissection in 280 patients with pathologic (p)T1-2 pN0 oral tongue SCC. Analysis was conducted from June 1, 2013, to January 20, 2015. INTERVENTIONS: In group 1 (n = 119), tumor bed margins were not sampled. In group 2 (n = 61), margins were examined from the glossectomy specimen, found to be positive or suboptimal, and revised with additional tumor bed margins. In group 3 (n = 100), margins were primarily sampled from the tumor bed without preceding examination of the glossectomy specimen. The margin status (both as a binary [positive vs negative] and continuous [distance to the margin in millimeters] variable) and other clinicopathologic parameters were compared across the 3 groups and correlated with LR. MAIN OUTCOMES AND MEASURES: Local recurrence. RESULTS: Age, sex, pT stage, lymphovascular or perineural invasion, and adjuvant radiation treatment were similar across the 3 groups. The probability of LR-free survival at 3 years was 0.9 and 0.8 in groups 1 and 3, respectively (P = .03). The frequency of positive glossectomy margins was lowest in group 1 (9 of 117 [7.7%]) compared with groups 2 and 3 (28 of 61 [45.9%] and 23 of 95 [24.2%], respectively) (P < .001). Even after excluding cases with positive margins, the median distance to the closest margin was significantly narrower in group 3 (2 mm) compared with group 1 (3 mm) (P = .008). The status (positive vs negative) of margins obtained from the glossectomy specimen correlated with LR (P = .007), while the status of tumor bed margins did not. The status of the tumor bed margin was 24% sensitive (95% CI, 16%-34%) and 92% specific (95% CI, 85%-97%) for detecting a positive glossectomy margin. CONCLUSIONS AND RELEVANCE: The margin sampling technique affects local control in patients with oral tongue SCC. Reliance on margin sampling from the tumor bed is associated with worse local control, most likely owing to narrower margin clearance and greater incidence of positive margins. A resection specimen-based margin assessment is recommended.