Harrington et al. (1999) cited two contrasting studies on the impact of climate change on the synchrony between winter moth larval emergence and bud-burst of its host plants, oak and Sitka spruce. Dewar and Watt (1992) concluded that an increase in temperature of 2–5 °C would lead to an increase in phenological asynchrony, whereas Buse and Good (1996) concluded that an increase in temperature of 3 °C would result in no change in the degree of phenological synchrony. The purpose of this paper is to reconsider these and similar studies because of their contrasting implications for the effect of climate change on phenological asynchrony and to consider potential weaknesses in some approaches to assessing the impact of climate change. Many temperate forest pests have spring-emerging larvae (Hunter, 1992) and therefore risk phenological asynchrony between larval emergence and bud-burst. The potential consequences of phenological asynchrony have been demonstrated in experiments on several important defoliators of forest trees, which have shown that stage of bud or shoot development can have a marked impact on the growth and survival of pests such as the pine beauty moth Panolis flammea (Watt, 1987), the eastern spruce budworm Choristoneura fumiferana (Thomas, 1987), and the green oak tortrix Tortrix viridana (Du Merle, 1983). Consequently, larvae emerging asynchronously with bud-burst will suffer high rates of mortality. The winter moth Operophtera brumata has received particular attention with respect to host plant phenology. It is a major defoliator of broadleaved trees in Europe and north America (e.g. Embree, 1965; Holliday, 1977) and a pest of Sitka spruce in Britain (Stoakley, 1985). Studies have shown that the growth and survival of winter moth larvae are affected by bud-burst phenology, particularly on oak (Feeny, 1970; Wint, 1983) but also, to some degree, on spruce (Watt & McFarlane, 1991). Although studies, such as those referred to above, have demonstrated that asynchrony has a negative impact on the individuals of many insect species, the effect on insect abundance has not been demonstrated clearly (Watt & Woiwod, 1999). This lack of a clear demonstration of the importance of phenological asynchrony on insect abundance is, however, arguably a result of the difficulty of demonstrating its impact on insect populations clearly in natural conditions. Nevertheless, phenological asynchrony was implicated as the key factor causing year-to-year fluctuations in abundance of the winter moth in one of the most notable long-term studies of the population dynamics of an insect, over 30 years ago (Varley & Gradwell, 1968). Given the potential importance of phenological synchrony in determining year-to-year fluctuations in abundance of the winter moth and other similar insects, many of them forest pests, Dewar and Watt (1992) and Buse and Good (1996) explored the possible consequences of climate warming on the abundance of winter moth. Dewar and Watt used models of insect and plant (Sitka spruce) phenology and concluded that phenological synchrony would be disrupted under climate warming. Buse and Good concluded that climate warming would have no impact on phenological synchrony between winter moth and oak, based on experiments carried out on winter moth on 4-year-old oak trees. There are several alternative explanations for the difference in results reported by these authors. Either both studies correctly reflect the probable consequences of climate change for the interaction between winter moth and Sitka spruce (Dewar & Watt, 1992) and oak (Buse & Good, 1996) or at least one of the studies has flawed conclusions. If both studies reflect the interactions between winter moth and two of its major host plants accurately, the contrasting conclusions imply that the mechanism for the emergence of oak buds in the spring differs from the mechanism for the emergence of spruce buds in the spring, and that the mechanism for emergence of winter moth larvae is similar to that of oak buds but not to that of spruce buds. Dewar and Watt (1992) assumed that there were different mechanisms for larval emergence and spruce bud-burst, the main difference being that they assumed that the timing of bud-burst is affected by winter chilling but that larval emergence is not. The lack of synchrony between winter moth larval emergence and Sitka spruce bud-burst under field conditions implies that the relationship between larval emergence and temperature differs from the relationship between Sitka spruce bud-burst and temperature. Between 1986 and 1991, 50% larval eclosion preceded (50%) bud-burst by 2 days (1986), 0 days (1987), 3 days (1988), 10 days (1989), 4 days (1990), and 6 days (1991) (A. D. Watt and A. M. McFarlane, unpublished). In contrast, Buse and Good (1996) suggested that for the synchrony that they observed at different temperatures to be achieved, the mechanism for emergence of winter moth larvae must be similar to the mechanism for the emergence of oak buds in the spring. An alternative explanation for these contradictory conclusions is that temperature has a different effect on the bud-burst of young pot-grown trees in solardomes (e.g. Buse & Good, 1996) and older field-grown trees (e.g. Dewar & Watt, 1992). Young pot-grown trees growing in solardomes and other types of glasshouse are likely to experience similar temperatures in their shoots, stems, and root systems. In contrast, older field-grown trees, particularly those growing in forests, are likely to experience more variable but overall higher spring temperatures in their shoots than in their stems and root systems. If, as also seems likely, the phenology of bud-burst is therefore dependent on some complex function of the temperature experienced throughout a tree, the phenology of shoot development of field-grown trees is likely to be different from that of small pot-grown trees. Unfortunately, there is little evidence of differences in the phenology of seedlings and mature trees, although Farnsworth et al. (1995) demonstrated that experimental soil warming affects the phenology of mature trees and shrubs. Thus, Buse and Good's (1996) study may be misleading, failing to provide an adequate prediction of the response of insects and their host plants to climate change. This is not to argue that all studies in controlled environments, such as the solardomes used by Buse and Good (1996) and colleagues (e.g. Docherty et al., 1996), and the ecotron (e.g. Naeem et al., 1995), are flawed; however the inability to grow anything other than young trees in these environments limits their value in comparing the relative effects of climate change on trees and the insects that feed on them. Although Dewar and Watt's (1992) study was concerned solely with winter moth, their conclusion, that climate warming will disrupt the phenological synchrony of winter moth and its host plants, may be applied generally to insects that feed on the spring foliage of trees. The complex nature of insect life cycles and tree morphology means that the winter and spring temperatures experienced by insects and their tree hosts are usually different. Some insects, like winter moth, overwinter in the egg stage, other tree-feeding insects overwinter as adults, producing eggs in spring, and some overwinter as larvae. In each case, they will experience a different temperature profile from their hosts. Similar responses to temperature would lead to phenological asynchrony so insects have evolved different responses to temperature from their hosts. Differences in response include differences in development thresholds, day-degree requirements, and the presence or absence of either a winter chilling requirement or a photoperiodic response. Dewar and Watt (1992) showed that, for example, larval emergence in an insect without a winter chilling requirement could be synchronised with bud-burst in a tree with a winter chilling requirement. As they demonstrated, however, synchrony may be disrupted under climate warming largely because of the absence of a winter chilling requirement in larval emergence. A similar breakdown in synchrony might be predicted between insects and plants where only one of them responds to photoperiod or where there are large differences in development thresholds. This prediction may be tested by modelling, the approach used by Dewar and Watt (1992). This approach was also one of two research strategies recommended by Harrington et al. (1999), who also advocated the analysis of long-term data-sets on phenology to predict the impact of climate warming on the interactions between different trophic levels. As they pointed out, however, the analysis of long-term data-sets is limited to past and present climates whereas modelling allows prediction of the consequences of future climates. There are various approaches to modelling, including empirical models such as CLIMEX (Sutherst et al., 1995). Although such models have value, they are also based on past and present climates, and therefore of most value when based on data collected from a wide range of climatic zones. A more powerful approach will be the development of models based on a knowledge of the phenological processes of larval emergence and bud-burst (e.g. Dewar & Watt, 1992). Improved predictions of the impact of climate change on trophic interactions require better understanding of these phenological processes. Controlled environment experiments provide the best method of quantifying the impact of temperature on the phenology of insects but, for the reasons discussed above, controlled environments are unsuitable for trees. Given the uncertainty that exists over the processes determining the phenological development of tree bud-burst (e.g. Nizinski & Saugier, 1988; Murray et al., 1989), the ability to predict the impact of climate change on the phenological synchrony between tree-feeding insects and their hosts is limited. For the reasons outlined here, it would be wrong to assume that Buse and Good's (1996) study is an accurate guide to the impact of climate warming on winter moth and insects with similar life histories. Further analysis of historical data, modelling, and monitoring of phenological development are clearly needed (Harrington et al., 1999). Although Harrington et al. reported many examples of phenological monitoring, these examples dealt with plants or animals, never both. Thus there is a need for more studies comparing the phenology of insects and their host plants (Watt & McFarlane, 1991; Hill & Hodkinson, 1995). Accepted 14 July 2001