Over a period of six years, the presence or absence of 23 species of arboreal insects (13 Homoptera, 1 Heteroptera, 9 ants (Hymenoptera: Formicidae)) was recorded on each of 866 young Amazon and Amazon hybrid cocoa trees at the Cocoa Research Institute, Tafo, Ghana. The general form of the time-series, derived from the proportion of trees occupied by each species, is described. Analysis of 18 of these series, taken either singly or jointly, was carried out using analysis of variance and principal components analysis and the degree of spatial association between 16 of the species was also investigated. For the latter, the number of joint occurrences of pairs of species was compared with the expected number by means of a chi squared test. An index of association was derived from the proportion of occasions on which the difference was significant. For all but four species, there were highly significant differences between annual means in the analysis of variance and, in many, the overall annual trend was downward with time. For those species where significant differences were not shown between years, trends were often apparent on individual plots but were negated because of inconsistencies between plots in the direction of the trend. In the joint analysis of all 18 species by principal components, the first component accounted for approximately 33% of the total variance and could be identified as representing to a large extent the long-term upward or downward trends shown up by the analysis of variance. The second principal component, accounting for a further 13% of total variance, was interpreted as being a further manifestation of the long-term trend, representing a tendency for deviation from the trend line shown up by the first component to become less with time and to approach the trend line in a damped oscillation. This is thought to be partly caused by a drought at the start of the experiment which affected certain species adversely. The third principal component accounted for 11% of total variance and the scores followed an irregular annual cycle which was related to relative humidity. The scores on this axis were high during dry periods of the year and low during wet periods. In the analysis of variance it was found that 13 species showed significant differences in spatial distribution between seasons and, for most, there was evidence for an annual cycle. There was also some evidence of correlation between the seasonal cycles of some insects with seasonal rainfall level or morning or afternoon relative humidity. These same species are important in contributing to the third principal component scores. Spatial association analysis showed eight species to be strongly associated. These fell into three groups with positive associations between members of the same group and negative association between members of other groups. Each group featured an ant-homopteran association, Crematogaster spp. with Planococcoides njalensis (Laing) in group 1, Oecophylla longinoda (Latreille) with Stictococcus spp. in group 2, and Camponotus spp. and Polyrhachis laboriosa F. Smith with Planococcus citri (Risso) and Toxoptera aurantii Boyer de Fonscolombe in group 3. Three other coccids were loosely linked to these groups through particular members, Waxiella sp. nr zonatus (Newstead) to group 1 through Planococcoides njalensis and to group 2 through Camponotus spp. and Phenacoccus hargreavesi (Laing) and Steatococcus spp. to group 2 through Planococcus citri. Changes taking place with time in the relative abundance of members of these groups could be discerned in the trajectories of the first two principal components. The spatial re-arrangements taking place with time of the territories of the major ant species are followed in some detail, and the effect of these changes on Homoptera is discussed.