In a population-based cohort of 6613 women with invasive breast cancer, who had breast-conserving surgery between 1981 and 1990, 391 recurrences in the operated breast were identified. The main aim of this study was to examine the prognosis and prognostic factors in different subgroups of local recurrences, characterised by the time to recurrence, location of recurrence and previously given radiotherapy. The median follow-up for women who had a local recurrence was 7.9 years. The life-table estimates for breast cancer-specific survival in women with local recurrences were 84.5% (standard error (S.E.) 1.8) at 5 years and 70.9% (S.E. 2.7) at 10 years. The risk of breast cancer death was highest among women who had an early (<or=2 years) recurrence in the same quadrant as the primary tumour, with a breast cancer-specific survival of 67.9% (S.E. 4.8) at 5 years and 56.0% (S.E. 5.9) at 10 years. There was a statistically significant difference in the probability of breast cancer-specific survival, as measured from the recurrence, between women who initially did or did not receive radiotherapy (P=0.0123). However, when measured from primary treatment, there was no significant difference, indicating that the difference in prognosis could be due to a lead-time bias. Independent prognostic factors for breast cancer-specific survival in women with local recurrences were time to local recurrence and the Nottingham Prognostic Index (NPI).