Emerging evidence demonstrates that postgraduate researchers have high rates of mental health problems. These problems are distressing, affect PhD studies, and have longer-term potential effects beyond the duration of the PhD. Yet large-scale studies of multiple risk and protective factors are rare.

We aimed to test the predictive validity of a comprehensive set of potential determinants of mental health symptoms (depression, anxiety and suicidality) among postgraduate researchers in the UK, including personal, study-related, and supervision characteristics.

We used regression models applied to data obtained from a national online survey of UK postgraduate researchers (Understanding DOCtoral researcher mental health; U-DOC, 2018–2019) to test predictors of mental health symptoms.

These models show that postgraduate researchers' mental health symptoms are predicted by demographic, occupational, psychological, social and supervisory relationship factors. Greater perfectionism, more impostor thoughts and reduced supervisory communion most strongly and consistently predict mental health symptoms.

Institutions training postgraduate researchers should focus interventions intended to improve depression, anxiety, suicidality, on self-beliefs and social connectedness. Moreover, supervisors should be provided with training that improves the degree of agency, and especially communion, in the relationships they form with postgraduate researchers.