Background: Type 2 diabetes has been found to be associated with cognitive impairments in planning, problem solving, organization, and working memory and also with an increased risk of dementia. Neurocognitive impairment may impact self-care and other health behaviours increasing the risk of poor health outcomes in this patient population. Detection of neurocognitive impairment in low and middle-income settings is challenging; there is a lack of validated screening tools suitable for local use in primary care and outpatient settings and access to formal neuropsychological testing services is limited. The inability to easily identify people with type 2 diabetes with neurocognitive impairments is constraining the development of context appropriate interventions to improve the care and outcomes in this sub-group of patients. Aim: The aim of the current analysis is to explore associations between neurocognitive function and measures of diabetes control (HbA1c, disease duration, type of blood glucose lowering treatment) at baseline in a population of people with type 2 diabetes participating in a clinical trial of treatment adherence support using SMS-text messages. Materials and Methods: Sms text Adherence suppoRt for type 2 Diabetes (StAR2D) is a randomised clinical trial testing if a system of SMS-text messages to support treatment adherence is more effective than usual care for controlling blood sugar among people with type 2 diabetes in sub-Sahara Africa (ISRCTN70768808). We have embedded neurocognitive assessment sub-studies into the Cape Town trial site. At baseline participants in the StAR2D trial complete a novel mobile-device based cognitive assessment, NeuroScreen, assisted by a field research assistant. The assessment contains 9 variants of tests found in the gold-standard neuropsychological test battery that have been adapted and normed for use in South Africa. It is available in English or isiXhosa. The assessment takes between 20 to 40 minutes depending on participant error rate. This cross-sectional analysis of baseline data uses linear and logistic regression models to explore associations between neurocognitive function and measures of diabetes control. Results: Six hundred participants eligible for enrolment in the StAR2D trial were recruited from the Cape Town trial site; 499 participants completed the baseline neurocognitive screening assessment (20 to 40 minutes to complete); 101 participants did not complete the assessment (commonly due to eyesight, hearing or motor difficulties e.g. hemiplegia due to previous stroke or technical difficulties.) We found differences in the scores in some but not all the neuropsychological tests. Using cut points suggested by an earlier validation study of NeuroScreen tool more than half of study participants would be scored as having at least mild neurocognitive impairment. HbA1c, duration of disease, type of blood glucose lowering treatment were not significantly associated with individual or overall neuropsychological test scores or odds of neurocognitive impairment. Conclusions: The prevalence of neurocognitive impairment may be substantial in this patient population. A novel tablet based neurocognitive screening tool was broadly feasible and acceptable to lay researchers and trial participants. There was no evidence that HbA1c, duration of disease, or type of blood glucose lowering treatment (oral agents alone or insulin containing regimens) was significantly associated with individual or overall neuropsychological test scores or odds of neurocognitive impairment. Validating this tool for this patient population and optimising its role in routine clinical care need further study.