Specific language impairment (SLI) is a common developmental disorder in young children. To investigate the association between multilingual home environment and SLI, we conducted a case-control study in Hong Kong Chinese children over a 4-year period in the Duchess of Kent Children's Hospital. Consecutive medical records of all new referrals below 5 years of age were reviewed and children diagnosed with SLI (case) were compared with those referred with other developmental and behavioural problems who had been assessed as having normal language and overall development (control) using the Griffiths Mental Developmental Scale. SLI was defined as those with a language quotient more than one standard deviation below the mean and below the general developmental quotient in children with normal general developmental quotient, but without neurological or other organic diseases. We used binary and ordinal logistic regression to assess any association between SLI and multilingual exposure at home, adjusting for age and gender of subjects, parental age, education level and occupational status, number of siblings, family history of language delay and main caregiver at home. Multivariable linear regression was used to examine the effect of covariates on the language comprehension and expression standard scores assessed by the Reynell Developmental Language Scale. A total of 326 cases and 304 controls were included. The mean ages of cases and controls were 2.56 and 2.89 years respectively. Boys predominated in both groups (cases, 75.2%; controls, 60.2%). The children were exposed to between one and four languages at home, the major ones being Cantonese Chinese followed by English. The adjusted odds ratio (OR) of SLI was 2.94; [95% confidence interval (CI) 1.82, 4.74] for multilingual compared with monolingual exposure. A significant linear dose-response relationship was found (OR of SLI = 2.58 [1.72, 3.88] for each additional language to which the child was exposed). Male gender (OR = 1.88 [1.24, 2.87]), positive family history (OR = 2.01 [1.17, 3.47]), lower education levels of parents (P = 0.028 for father, P = 0.038 for mother) and lower occupational status of father (P = 0.005) were also risk factors for SLI. Multilingual exposure also significantly reduced the language quotient (P = 0.012) and language comprehension standard score (P = 0.016) of children with SLI, but not of normal children. We concluded that multilingual home environment is associated with SLI with a dose-response relationship. Exposure to multiple languages might adversely affect subsequent language ability.