Incoming chemistry students at tertiary institutions have a variety of academic backgrounds that influence their prospects of success at first-year level. The proficiencies of incoming students are currently changing due to the introduction of outcomes-based education and new syllabi for physical science in secondary schools. In order to ensure a smooth transition from secondary to tertiary education, university lecturers should be well informed about the content knowledge, conceptual understanding and skills development of prospective first-year students. This study evaluates the proficiencies of Grade 12 learners in physical science in terms of a number of clearly identified problem areas: conceptual understanding, logical scientific reasoning, basic mathematical ability, knowledge of subject content and scientific process skills. A test instrument was developed that consisted mainly of conceptual questions rather than recall or algorithmic items. Paired questions (two-tier methodology) and pictorial representations were used extensively. A follow-up question about certainty of response was included for all fixed-response items in order to evaluate the influence of guessing on response distributions. The test instrument was administered at the end of the third term to Grade 12 learners taking physical science at three schools in privileged environments (1 English and 2 Afrikaans medium) and four township schools, and to all University of Pretoria Foundation Year (UPFY) students. Analysis of results highlighted the generally poor performance of students from township schools and the significant improvement in performance after one year of intensive instruction of UPFY students, who generally came from similar or more impoverished backgrounds. The poor performance for all cohorts on basic concepts, such as the mole concept, stoichiometry and the limiting reagent, as well as on several special topics, indicates that students lack a sound basis for tertiary chemistry. Of real concern is the evidence of over-confidence obtained from the certainty of response analyses. This result indicates that respondents failed to judge the complexity and level of difficulty of questions accurately. Learners/students from all groups displayed weak understanding of events at molecular level. In order to address this situation, lecturers at tertiary level will have to actively promote conceptual understanding of all basic concepts in chemistry and resist the temptation to teach and assess mainly procedural fluency. Analysis of certainty of response data showed that the guess factor was less serious a complication than anticipated.