BackgroundEthnic minorities report poorer evaluations of primary health care compared to White British patients. Emerging evidence suggests that when a doctor and patient share ethnicity and/or language this is associated with more positive reports of patient experience. Whether this is true for adults in English general practices remains to be explored.MethodsWe analysed data from the 2010/2011 English General Practice Patient Survey, which were linked to data from the NHS Choices website to identify languages which were available at the practice. Our analysis was restricted to single-handed practices and included 190,582 patients across 1,068 practices. Including only single-handed practices enabled us to attribute, more accurately, reported patient experience to the languages that were listed as being available. We also carried out sensitivity analyses in multi-doctor practices.We created a composite score on a 0-100 scale from seven survey items assessing doctor-patient communication. Mixed-effect linear regression models were used to examine how differences in reported experience of doctor communication between patients of different self-reported ethnicities varied according to whether a South Asian language concordant with their ethnicity was available in their practice. Models were adjusted for patient characteristics and a random effect for practice.ResultsAvailability of a concordant language had the largest effect on communication ratings for Bangladeshis and the least for Indian respondents (p < 0.01). Bangladeshi, Pakistani and Indian respondents on average reported poorer communication than White British respondents [-2.9 (95%CI -4.2, -1.6), -1.9 (95%CI -2.6, -1.2) and -1.9 (95%CI -2.5, -1.4), respectively]. However, in practices where a concordant language was offered, the experience reported by Pakistani patients was not substantially worse than that reported by White British patients (-0.2, 95%CI -1.5,+1.0), and in the case of Bangladeshi patients was potentially much better (+4.5, 95%CI -1.0,+10.1). This contrasts with a worse experience reported among Bangladeshi (-3.3, 95%CI -4.6, -2.0) and Pakistani (-2.7, 95%CI -3.6, -1.9) respondents when a concordant language was not offered.ConclusionsSubstantial differences in reported patient experience exist between ethnic groups. Our results suggest that patient experience among Bangladeshis and Pakistanis is improved where the practice offers a language that is concordant with the patient’s ethnicity.Electronic supplementary materialThe online version of this article (doi:10.1186/s12875-015-0270-5) contains supplementary material, which is available to authorized users.