Plants and animals exhibit seasonal patterns in their activities because there is a clear seasonality in the suitability of their environment: there is often only a limited period in the year when conditions are favourable enough to successfully reproduce or grow. If reproduction or growth takes place outside this window of favourable conditions, there are often large fitness consequences. Ultimately, the activity that is the most demanding for an organism should take place at the time of optimal conditions. What we need to know to evaluate the observed shifts in phenology is how that period of optimal conditions shifts due to climate change. Thus, we should use the shift in the seasonal changes in the ecological conditions as a yardstick to assess whether the change in phenology observed is sufficient or not.

As these ecological conditions for a species are often set by organisms at other trophic levels it is important to realize that there is no a priori reason why the phenology of different trophic levels will shift at the same rate (Visser et al. 2004): plants are likely to have different mechanisms underlying their phenology than insects, which in turn will have different mechanisms than vertebrates. Although natural selection is expected to have shaped these mechanisms such that the response to temperatures is similar to that of, e.g. food species, this will only be true under the set of abiotic conditions under which these species have evolved, and not for the conditions altered by climate change (see Visser et al. (2004) for a more extended argument).

If the phenology of a species is shifting at a different rate from that of the species that make-up its ecological conditions, this will lead to mistiming of its seasonal activities (Visser et al. 2004) or, to use an alternative terminology, to a mismatch in phenology (Stenseth & Mysterud 2002). Such trophical decoupling of food web phenology may have severe consequences, including biodiversity loss (Visser et al. 2004).

This leaves us with the difficult tasks of, first, assessing the period of optimal conditions for a species and, second, quantifying how much this period has shifted due to climate change and thereby determining whether the response of a species is sufficient. In general, what is needed is identification of the variables that make-up the important selection pressures on the phenology of a species. For many species, the optimal period will be determined by the temporal distribution of food needed during reproduction or growth, as food abundance is often a major determinant of fitness (Martin 1987), and in most of this paper we will assume that the main selection pressure on phenology is food abundance (but see §8).

The phenology of caterpillar biomass has been monitored in a few study areas, including Wytham Wood in the UK (Perrins 1991; Cresswell & McCleery 2003) and the Hoge Veluwe in The Netherlands (Visser et al. 1998; Visser et al. in press b). For these populations, the shift in the date of the caterpillar biomass peak can be used as a yardstick for the shifts in timing of reproduction of the birds. The Wytham Wood great tit population has advanced its laying date over the past 39 years, but more so than the shift in caterpillar biomass phenology (Cresswell & McCleery 2003). In contrast, the Hoge Veluwe great tits have not advanced their laying date but the caterpillar peak date has advanced over the past two decades (Visser et al. 1998, in press b) (figure 1c). The Hoge Veluwe pied flycatchers have advanced their laying dates, but not enough to match the shift in caterpillar biomass peak date (Both & Visser 2001). In all cases, the responses of the birds seem to be different from their main food. Especially for the Wytham Wood great tits and the Hoge Veluwe pied flycatchers, where there has been an advance in laying date, it is essential to have a yardstick because otherwise it would be completely unclear whether the observed shifts in laying date have resulted in a mismatch in timing due to a too large (as the Wytham Wood great tits, but see §8) or too small (as the Hoge Veluwe pied flycatchers) response.

In species where the young search for their food themselves, such as golden plovers (Pluvialis apricaria), reproductive success also depends on the synchrony of hatching of their chicks with the phenology of the food, in this case adult tipulids (Tipulidae) (Pearce-Higgins & Yalden 2004). As for the forest caterpillars, tipulids are also abundant for only a 2–3-week period. Unfortunately, no long-term datasets are available to document shifts in the phenology of the birds and their prey. However, the birds' laying dates and tipulid emergence depend on temperatures in different date periods (Pearce-Higgins et al. 2005) and if temperatures in these different periods change at different rates the synchrony between the birds and their food may be distorted. The same may hold for the black grouse (Tetrao tetrix), whose precocious chicks depend critically on abundant insects for their survival during the first two weeks of life (Baines et al. 1996). In Finland, black grouse lekking has advanced over the past four decades, and as a consequence hatching date has shifted as well (G. X. Gilbert, unpublished data). Data on shifts in insect abundance are needed to provide a yardstick for the interpretation of this shift in avian phenology.

Bewick's swans (Cygnus columbianus bewickii) now depart earlier from the Netherlands to their artic breeding areas (J. van Gils, unpublished data), which may be advantageous as the phenology of their stop-over sites in the Baltic area is advancing. In North America, wood warblers (Parulidae) have not advanced their migration phenology, either at a stop-over location or at a location at the southern range of their breeding area (Strode 2003; Marra et al. 2005; Mills 2005). Temperatures north of the stop-over site, however, have significantly increased and as a consequence the phenology of their main prey species, the eastern spruce budworm (Choristoneura fumiferana), has advanced (as deduced from a simple model using temperature). Thus, the wood warblers arrive too late in their breeding grounds to fully profit from the peak in caterpillars, and the same holds for the conditions en route where they need to refuel. In contrast, temperatures south of the stopover site have decreased and, hence, caterpillars there appear later, perhaps hampering an earlier departure date, as birds need to fatten up before starting their migration.

A second example where the advancement of arrival dates of a long distance migrant is insufficient, as deduced using food phenology as a yardstick, is the pied flycatcher (Ficedula hypoleuca) in The Netherlands. While there is a clear advancement in the phenology of their main prey, caterpillars, the birds are not arriving any earlier (Both & Visser 2001). Interestingly, despite this lack of advancement of arrival date, the birds have advanced their laying date. They do this by shortening the interval between arrival and breeding. Currently, this interval has shrunk to only a few days and further advancement of breeding is not possible without an earlier arrival. North American passerines have advanced their arrival date on the breeding grounds, probably as a result of faster spring migration at higher temperatures en route (Marra et al. 2005). However, this response to temperature is about one-third of the advance in flowering of at least one plant species in their breeding areas, and the birds' advance may thus be insufficient. A final example where there is a yardstick available to judge (the lack of) shifts in arrival dates is the honey-buzzard (Pernis apivorus), a bird of prey that is specialized on wasps (Vespula and Dolichovespula spp.). While there is a clear advancement in wasp phenology, there is again no change in Dutch honey-buzzard arrival dates (R. Bijlsma, unpublished data).

An example where climate change might not lead to mistiming is the orange tip butterfly (Anthocharis cardamines) because its mean date of first appearance has a very similar response to March temperatures as the flowering date of garlic mustard (Alliaria petiolata), one of its host plants (Sparks & Yates 1997; Harrington et al. 1999). However, this analysis is based on a period largely prior to climate change (1883–1993) and hence, except in the unlikely event that March temperatures are truly the causal mechanism, climate change may affect insect and plant phenology differently.

Another marine system where there has been a shift in phenology is the marine pelagic community: diatoms and dinoflagellates (primary producers), copepods (secondary producers), non-copepod holozooplankton and meroplankton (secondary and tertiary producers). Large differences in the extent of the shifts of the phenology of these components have been recorded over the past four decades (Edwards & Richardson 2004): of the producers, diatoms as a group have not shifted (as their phenology depends on photoperiod or light intensity, rather than temperature) but the dinoflagellates have by 23 days. The copepod and non-copepod zooplankton have advanced by 10 days while the largest advancement was observed in the meroplankton (27 days). As a result, there is currently a mismatch between the successive trophic levels and a change in the synchrony in the timing of primary, secondary and tertiary producers (Edwards & Richardson 2004).

At first sight it seems surprising that different parts of a food chain respond differently to variations in annual temperatures. There has always been temperature variation and hence selection on higher trophic levels to keep in line with the annual variation in the phenology of their food source. As a result, many life-history characters are phenotypically plastic: under different conditions the phenotypic expression of, for instance, flowering time differs. It is, however, important to realize that climate may not change uniformly over the season, and hence species which differ in the periods that causally affect their phenology may diverge in timing, with possibly detrimental effects for the higher trophic level and, perhaps, beneficial effects for the lower trophic level.

Species at different trophic levels often differ in the temporal and spatial scale at which decisions are made regarding timing. For at least some vertebrate species reproduction has to be initiated long before their young need provisioning. For instance, in great tits there is a four-week delay before the start of laying and the peak in food demands on the nestlings. If temperatures increase after the birds have started laying, caterpillar development is accelerated but the birds have very limited ability at this stage to advance the hatching of their eggs to match this (see van Noordwijk et al. 1995; Visser et al. 2004, for an elaborated argument). Long distance migrants not only face a temporal problem, but also the spatial problem of being thousands of kilometres away from their breeding site during winter. These birds may not use climatic variables to time their migration and as a consequence the birds do not advance their migration while the food sources at their breeding sites do advance (Coppack & Both 2002). This north–south geographical variation is similar to the variation in climate change at lower and higher elevations, which is responsible for mistiming in the American robin (Inouye et al. 2000).

Recently, a method has been developed to describe phenology in a more sophisticated way than just correlating annual phenology to mean temperatures over a fixed period (Gienapp et al. 2005). In this so-called proportional hazard model, developed for survival statistics, the probability of laying is calculated per day depending on the current and previous conditions. This model, albeit still a correlational model, avoids the problems with fixed periods and also allows for interactions, such as that between photoperiod and temperature, to be incorporated. The model has been used to predict great tit laying dates up to 2100 (Visser et al. in press b).

So far we have considered just a single life-history stage, but species have to perform different tasks at different times, all of which may be affected by climate related ecological factors. This is most pronounced in the annual cycle of long distant migrants, where the birds have to decide when to leave their wintering site, how fast to travel, how much energy in the form of body reserves should be brought to the breeding site, and when to start breeding after arrival. Especially for these species it is important to take the entire life cycle into account when predicting future shift in phenology. One way to model this is to use stochastic dynamic programming (McNamara & Houston 1996). This has been applied to the migration of pink-footed geese (Anser brachyrhynchus) from The Netherlands to the Artic by Bauer et al. (submitted), who showed that the onset of spring on the stop-over sites affects the optimal departure times, stressing the need to take the whole migratory journey into account.

In this review, we have considered a very simple form of yardstick: in most of our examples we have assumed that the selection acting on the phenology of a species comes from just a single selection agent: their food source. Furthermore, we have assumed that there is just a single activity (i.e. breeding, spawning) a year for which timing is important. Although we admit that this ‘single critical activity–single selection agent’ scenario is highly unlikely to be the case, we fear that it is the best possible yardstick at this moment. And even so, most examples we reviewed come from simple ecological situations, because more complex trophic interactions are more difficult to study, and long-time series on all relevant species within a food web are rarer than these simple food chains.

When we go beyond measuring the shifts in phenology of a single selection agent it becomes more difficult to define a yardstick, as we then need to integrate these different selection pressures. The one example where there are data on both food and predator phenology (the Macoma example, Philippart et al. 2003) there was no change in food phenology but a strong shift in predator risk phenology, and it is unclear how these two selection pressures should be combined into a single yardstick. Another example where the phenology of predation may be an equally important selection pressure as the phenology of food is the frog (Rana temporaria), and its predators (newts, Triturus sp.). The newts have advanced their entry in ponds, whereas frogs have not substantially altered their reproductive phenology (Beebee 1995). Therefore, embryos and larvae of early breeding frogs are now exposed to higher levels of newt predation (Walther et al. 2002).

Ultimately, the way forwards is to measure selection on phenology and to assess whether there is increased directional selection (Visser et al. in press a). For this however, long-term studies where individual fitness can be calculated are needed rather than just population means. The very few studies that have calculated whether there is increased directional selection are all on birds (Visser et al. 1998; Both & Visser 2001; Cresswell & McCleery 2003) as only for this group such long-term studies are available. Clearly, individual fitness measures will not be available for the Macoma system. But also for such systems, the ultimate way to assess whether their shift in phenology is sufficient is to measure the reproductive success of early and late spawning individuals. Even if we are able to measure the selection acting on the phenology, a yardstick is still essential as the changes in these fitness-based measures may also be caused by other changes in the environment. A full understanding of whether species are reacting sufficiently to the climate change-induced advance of its food requires therefore both these fitness estimates and the yardstick. In the case of great tits, the change in selection differential over the years is consistent with the improved (Cresswell & McCleery 2003) or the deteriorated synchrony (Visser et al. 1998) between food and reproduction.

The second critical assumption of our review is that there is just a single critical activity per year that is under selection. In reality, this will not be the case and a life cycle approach is more appropriate: the entire life cycle has to be fitted into the seasonal changes in suitability of the habitat. This is even more important as different life history stages may each be affected by climate change, as we have already discussed for migrant birds (Winkler et al. 2002; Both & Visser 2005). Such changes of multiple life-history traits have also been reported for resident bird species. The Hoge Veluwe great tits, which have not shifted their laying dates, have responded to climate change by no longer producing second clutches (Visser et al. 2003). In the case of the Wytham Wood great tits, they have prolonged the time between the laying of their last egg and the hatching of their chicks. As a result the interval between hatch date and caterpillar peak has not changed over the years and hatching asynchrony has been reduced (Cresswell & McCleery 2003). It is, therefore, very well possible that the shift in laying date in this population has been sufficient despite the fact that the shift in the phenology of their food, the peak caterpillar biomass date, was weaker than the shift in laying date, and hence we have classified this study as ‘sufficient’ rather than to ‘too much’ in table 1. As for the ‘single selection agent’ assumption, also the ‘single critical activity’ assumption can only be lifted for a handful of studies, again mainly birds.

Finally, there are a few more complications with the assumption that shifts in food phenology is a useful yardstick. It may also be that more generalist species are less affected by climate change, because they can more easily switch to alternative prey if they are out of synchrony with one of their prey species. Another complication is that so far we have just considered changes in the timing of the optimal period, e.g. food peak, but also the width of the optimal window may change as climate changes. These changes have been only rarely mentioned in the literature (Buse et al. 1999) and will make the use of a yardstick more difficult. However, changes in the width of the optimal period may be even more important than the actual date of the peak, because if the window becomes too narrow reproduction may become impossible, whatever the change in phenology.