To investigate breast-feeding patterns and factors encouraging early weaning, a survey was conducted in Tonalá and Tlaquepaque, two suburbs within Guadalajara's metropolitan area, in 1991. For this purpose a multiphase probabilistic sample of infants born in these areas from May 1990 through April 1991 was selected. This was done by choosing at random primary health care units in the study areas, health posts associated with these units, and all infants meeting the above criteria at each selected post. A total of 166 homes was visited and interviews were conducted with 141 mothers (91% of the 155 predicted) in June and July 1991. These interviews made use of a 33-item questionnaire developed for the purpose; the interviewers were social workers previously trained in such activities. Logistic regression models were used to calculate the relative risk (RR) and probability of early weaning being associated with certain variables. To help ensure the validity of the results, several regression models were constructed for the purpose of selecting the one best fitting the data. In addition, the attributable population risk (APR) was calculated. The results indicate that failure to breast-feed and early weaning were prevalent in the study population, 34.8% of the study infants being breast-fed less than 1 month. Three risk factors were associated statistically (P < 0.05) with early weaning, these being maternal age < 20 years (RR = 3.75; 95% CI = 1.53-9.19), maternal marital status single (RR = 2.88; 95% CI = 1.08-7.69), and social status of the main family provider other than "worker"--i.e., employee, professional (RR = 2.72; 95% CI = 1.17-6.28). The likelihood that a study infant would have been breast-fed less than a month was 0.84 if the infant was exposed to all three of these identified risk factors and 0.15 if he or she was exposed to none of them. The high percentages of study mothers less then 20 years old and with a social status other than "worker" were reflected in high attributable population found for these variables. In general, the findings point up a need to reduce the influence of these risk factors and to prolong maternal breast-feeding in the study population.