Despite decades of research and dozens of public health campaigns, stigma continues to negatively affect the well-being and life chances of people labeled with a mental illness. One of the most promising directions for reducing stigma lies in Allport’s (1954) theory of intergroup contact, suggesting that social interactions with people with mental illness invalidate negative stereotypes, decrease fear and anxiety, and enable perspective-taking and empathy. While the empirical literature is largely supportive of the contact hypothesis, social network theory indicates that the degree to which contact reduces stigma should depend on the nature, magnitude, and valence of exposure to people with mental illness. We address this question using data from the National Stigma Study – Replication II (NSS-R II), fielded on a special module of the 2018 General Social Survey (N=1,179). We find that simply knowing someone with mental illness, or even the number of people one knows, explains little about the public’s desire for social distance, endorsement of coercion, or perceptions of dangerousness. However, having stronger relationships with more people with mental illness, and having more friends and family (but not more peripheral ties) with mental illness, are associated with reductions in stigma. In contrast, exposure to more dangerousness or violence among people in the network with mental illness is associated with greater levels of stigma, while contact with more people who are in treatment for mental illness (as opposed to untreated) is linked to less stigma.