This study investigates the relationship between the number of recovered lymph nodes and lymph node metastases in colorectal resection specimens. All of the slides from 2427 pT3 colorectal resection specimens from patients operated on at William Beaumont Hospital during the 45 years from 1955 through 2000 were reviewed. Lymph node metastases were present in 333 of 1499 (22.2%) specimens with fewer than 15 recovered lymph nodes, compared with 789 of 928 (85.0%) specimens with 15 or greater recovered lymph nodes (p <0.01). The proportion of lymph node metastases increased as a function of the number of recovered lymph nodes (p <0.01). Similarly, in patients without lymph node metastases, survival increased as a function of the number of recovered lymph nodes. Among these patients, the 5-year overall survival rate was 62.2% among patients with seven or fewer recovered lymph nodes and 75.8% among patients with 18 or more recovered lymph nodes (p = 0.018). Statistical analysis found the predictive probability of identifying the single lymph node metastasis in a theoretical specimen with a single lymph node metastasis is 0.25 if 12 lymph nodes are recovered and 0.46 if 18 lymph nodes are recovered. The predictive probability increased as the number of recovered lymph nodes increased, suggesting there is no minimum number that reliably or accurately stages all patients. Thus, all palpable lymph nodes should be recovered, including those that are 1 or 2 mm.