Characterization of invasive peritoneal implants from patients with noninvasive serous ovarian tumors has important prognostic and treatment implications, but the criteria for distinguishing invasive and noninvasive implants vary among investigators and can be difficult to apply. The authors studied 148 implants from 60 patients, 33 with primary atypical proliferative serous tumor, and 27 with primary noninvasive micropapillary serous carcinoma, with a mean follow-up of 62 months (median follow-up, 52 months). Previously reported and newly proposed histologic features for implant classification were evaluated and correlated with clinical outcome. Three criteria were applied for the diagnosis of "invasive" implants: invasion of underlying normal tissue, micropapillary architecture, and solid epithelial nests surrounded by clefts. Implants displaying any one of these three features were classified as "invasive," whereas those lacking all three features were classified as "noninvasive." Sixty-six implants were invasive and 82 were noninvasive. Of the 31 patients with invasive implants, six were dead of disease (DOD), 13 were alive with progressive disease (AWPD), and 12 were alive with no evidence of disease (NED). Of the 29 patients with noninvasive implants, two were DOD, one was dead of uncertain causes, one was AWPD, and 25 were alive with NED. Eighty-nine percent of invasive implants had a micropapillary architecture and 83% had solid epithelial nests surrounded by clefts. A minority of invasive implants (14% of those with underlying normal tissue) demonstrated invasion of normal underlying tissue. Nuclear atypia, mitoses, calcification, necrosis, and identification of individual cells "infiltrating" the stroma did not correlate with implant type. The proposed criteria permitted recognition of implants that correlated strongly with adverse outcome. Sixty-one percent of patients with implants displaying any one of the three features used to diagnose invasive implants were AWPD or DOD compared with 10% of patients whose implants lacked these features (p = 0.00001). Because implants associated with an adverse outcome can be identified before they invade underlying normal tissue, the term invasive implant to describe them is inaccurate and misleading. These implants resemble patterns of growth in micropapillary serous carcinoma of the ovary and the recurrent tumor that is obvious carcinoma. Accordingly, we propose that these extraovarian lesions be designated "well-differentiated serous carcinoma."