This thesis is conducted in two parts. The first part investigates expert and lay representations of GM food risk using the Carnegie Mellon University approach (Granger Morgan at al, 2002) and social representations theory (Moscovici,1984). Elite interviews were followed by focus group discussions with lay people to explore how GM food risks are understood by both parties. Expert and lay perceptions differ on a wide range of issues. In contrast to experts, lay participants are particularly concerned with various non-scientific risk dimensions, such as trust and values. These moral and ethical concerns cannot be addressed through the provision of scientific information alone as the Carnegie Mellon University approach suggests. Social representations theory offers an alternative to this top-down approach by showing that, rather than erroneous knowledge, lay participants' non-scientific concerns express ways of understanding GM food. The second part of the thesis focuses on one particular non-scientific dimension of lay GM food representations, namely the concept 'messing with nature', and examines the social construction of nature and naturalness in relation to food. 188 respondents completed an internet-administered free associations task to establish in depth what lay people mean when they judge GM food to be 'unnatural'. The findings show that the concept of 'messing with nature ' can be firmly established as a major risk dimension in relation to GM food. Moreover, rather than being solely material, both food and nature have significant symbolic and affective dimensions. This indicates that GM food risks are social in part- they are representations that have socio-cultural and value dimensions and with all this they are disputed. In turn, this has a range of implications for risk communication.