The hypothesis that protection of infants from exposure to infectious agents with delayed first exposure to one or more specific agents together contribute to the aetiology of childhood leukaemia, especially common acute lymphoblastic leukaemia (cALL), has substantial indirect support from descriptive epidemiology and case-control studies in developed Western countries. A case-control study of childhood leukaemia diagnosed at ages 2-14 years has now been conducted in Hong Kong. Cases (n=98) formed a consecutive series of Chinese children diagnosed with acute leukaemia; controls (n=228) were identified following a survey using random digit dialling and required to attend for medical examination by a paediatrician. Interviews with mothers were conducted in hospital by one trained interviewer using a structured questionnaire. Odds ratios (OR) and 95% confidence intervals (CI) are reported for exposure variables capable of serving as proxies for exposure to infection in two critical time periods: first year of life, year before reference date (diagnosis for cases, corresponding date for controls). Analyses used logistic regression with adjustment for appropriate confounders. Change of area of residence reduced risk if during the first time period (OR = 0.47 [95% CI 0.23, 0.98]) and increased risk if during the second (OR=3.92, [95% CI 1.47, 10.46]). Reported roseola and/or fever and rash in the first year of life reduced risk (OR=0.33 [95% CI 0.16, 0.68]) whereas tonsillitis in the period 3-12 months before reference date increased risk (OR=2.56 [95% CI 1.22, 5.38]). Some other proxies for exposure to infection at the critical times were associated with predicted patterns of risk but day-care attendance failed to show predicted associations. These results provide support for the delayed exposure hypothesis in an affluent geographical setting in which population exposure to infectious agents is quite distinct from the settings of previous case-control studies.