The use of visual cues by insect herbivores is likely to be an important component of plant-herbivore interactions in the wild, yet has until recently received little attention from researchers. In the last decade, however, interest in this topic has intensified following Hamilton & Brown's (2001) autumn colouration hypothesis, which proposes that the intensity of colouration of trees at autumn time is a signal of their defensive commitment to potential herbivores. This idea remains controversial and to date robust empirical data linking colouration with chemical defence and herbivory have been lacking. This thesis begins with a meta-analysis, in which I synthesize and analyse previously published data to determine the evidence for the use of host plant colouration by herbivores. I then move to explore the relationship between chemical defences and colouration in a classic plant-herbivore system: the wild cabbage (Brassica oleracea) and its herbivores the cabbage white butterfly (Pieris rapae) and the cabbage aphid (Brevicoryne brassicae). Both species have colour vision, and I use spectral sensitivity data to model the colour of the host ‘through the eyes' of the herbivores. First, I present data from a field study of wild cabbage populations showing significant relationships between herbivory, plant colouration and levels of glucosinolates defensive compounds. These results suggest that plant colouration could be used by herbivores to gain information about plant chemical defence. I then show colouration has a fixed genetic component in a common garden experiment; a necessary requirement for evolution of a colour signal. I explore the use of colouration in host choice by herbivores in more detail in a series of behavioural experiments. I show that cabbage aphids do not use leaf brightness as a cue when selecting among plants, but they do respond to different leaf colours. I also show that cabbage white butterflies do not choose hosts based on particular colour cues, even though this colouration potentially provides important information about host defence levels, which are shown to impact upon offspring fitness. Together, these results provide a clear demonstration of a link between plant chemistry and colouration in the wild cabbage system. However, the data presented in this thesis indicate that the use of colouration as a guide to host defence is limited, and I conclude by discussing possible reasons why this might be the case.

Plants come under attack by a variety of organisms, including insects and pathogenic microorganisms such as viruses. Plant viruses can interact indirectly with their vectors by inducing changes to plant chemistry which may alter its attractiveness as a host for herbivore vectors. Using red raspberry as a study system, this study aimed to investigate the host plant mediated interactions occurring between the large raspberry aphid, Amphorophora idaei, and two of the viruses that it transmits, Black raspberry necrosis virus (BRNV) and Raspberry leaf mottle virus (RLMV). In whole plant bioassays, BRNV and RLMV-infected plants were shown to be initially more attractive to A. idaei and aphids remained on the initially selected host plant for a period of approximately 30 minutes. In addition, A. idaei took three days longer to reach reproductive maturity compared with those feeding on non-infected plants, suggesting a virally-induced manipulation of aphid behaviour whereby a deceptive attraction of the vector to a host plant found to be nutritionally poor, presumably acts to promote virus transmission. Investigations of the underlying plant chemistry revealed that raspberry viruses may be capable of facilitating aphid feeding by reducing leaf phenolic concentration when aphids are feeding and that infection with BRNV and RLMV resulted in significantly elevated levels of carbon and free amino acids in the leaves. While increased concentrations of amino acids might be expected to promote aphid performance, the amino acid composition was dominated by glutamate (77% of total content of infected plants), a previously suggested indicator of reduced host-plant suitability for aphids. Volatile entrainments from virus-infected plants showed elevated levels of the green leaf volatile (Z)-3-hexenyl acetate. Bioassays subsequently revealed that this compound acted as an aphid attractant at a concentration of 50 ng ml-1 but that aphid behaviour was unaffected by lower concentrations. The combined utilisation of PCR diagnostics developed from newly sequenced viral genomes and the implementation of a non-invasive, targeted method of sampling plant headspace volatiles enabled this study to provide novel insights into the nature of host plant mediated interactions between aphids and the viral pathogens that they transmit.