Recognition is a critical underlying mechanism of any discriminatory behavioural pattern, from associating with kin or selecting suitable food types to habitat choice and recognition of enemies (Hauber & Sherman 2001). Not surprisingly, both direct (e.g. innate) and indirect (e.g. learned in a context-dependent manner) recognition are central to many questions in behavioural ecology (Sherman et al. 1997). One of the most thoroughly studied systems for recognition is brood parasitism (Davies 2000). Recognition and discrimination of parasitic eggs has received major scientific and popular attention (Davies & Brooke 1989; Lotem et al. 1995; for review, see Davies 2000), while discrimination of alien chicks has so far been rarely tested experimentally (Davies & Brooke 1989; Soler et al. 1995; Lichtenstein 2001; Payne et al. 2001; Schuetz 2005; for reviews, see Redondo 1993 and Grim 2006a). Several examples of well-developed cases of chick mimicry suggest that host discrimination between own and foreign chicks has arisen during the coevolution between a few hosts and parasites (Redondo 1993; Grim 2005, 2006a). In general, hosts can reject natural or experimental parasitic chicks by nest desertion (Grim et al. 2003; Langmore et al. 2003), refusal to feed them (Lichtenstein 2001; Payne et al. 2001) or by directly attacking, killing and/or ejecting them from the nest (Redondo 1993; Soler et al. 1995). However, the proximate mechanisms of host's differential responses towards foreign chicks under natural conditions are generally unknown. The only thoroughly studied case is the discrimination based on the structure of begging calls in parasitic bronze-cuckoos (Chrysococcyx spp.) by superb fairy-wrens (Malurus cyaneus) in Australia (Langmore et al. 2003).

Studies of brood parasitism in general focused on host responses to eggs, mainly on egg rejection—especially by ejection—based on recognition of own versus alien eggs (Davies 2000). The notion of a rarity of chick discrimination and mimicry (Davies 2000) might be attributed to the assumption that hosts would use similar behavioural (e.g. ejection) and cognitive (e.g. recognition) mechanisms when defending against parasitic chicks and eggs (Redondo 1993). But this need not be so because hosts facing parasitic eggs and chicks are confronted with very different constraints and trade-offs, as elaborated by Redondo (1993). Therefore, it is important to explore other behavioural and cognitive mechanisms when studying chick discrimination by hosts of brood parasitic birds. I have attempted this using a host–parasite system, where discrimination was hypothesized not to be based on recognition (Grim et al. 2003).

A theoretical model (Lotem 1993) predicted the absence of chick discrimination in hosts of evicting parasites. Surprisingly, parasitic chick discrimination has now been reported for one of the best-known evicting brood parasites, the common cuckoo (Cuculus canorus, hereafter, cuckoo). Approximately 15% of the cuckoo chicks were reported to be deserted by host reed warblers (Acrocephalus scirpaceus; Grim et al. 2003). The fact that reed warblers are victimized by an evicting parasite need not be in discrepancy with Lotem's (1993) model. While the model assumed learned chick recognition, Grim et al. (2003) predicted that adaptive host responses to parasitic chicks need not be based on learning or recognition using an internal template (Hauber & Sherman 2001). Accordingly, parent reed warblers may, in principle, discriminate against cuckoo chicks by refusing to provide care for longer periods and/or with increased investments than that needed for successfully fledging their own chicks (see also Holen et al. 2001). My previous study (Grim et al. 2003) provided preliminary, correlative evidence in favour of this ‘discrimination without recognition’ hypothesis. Therefore, I tested the idea experimentally in the same population where parasitic chick desertion had been originally observed. If warblers' responses to parasitic chicks were not based on learning or the recognition of the parasite, then they should work against any parasite-like brood, i.e. any individual(s) requiring longer and/or higher parental care than hosts' own chicks under normal conditions. I used a cross-fostering experiment to alter those normal rearing conditions.

I manipulated brood sizes to create single- and four-chick broods and used cross-fostering of different aged broods to force parents to care for nestlings for prolonged or shortened periods than normal with varying amounts of parental investment (for details on experimental procedure and definition of terms, see §2b below). Under the ‘discrimination without recognition’ scenario, reed warblers could use two types of information for the decision to desert a parasitic cuckoo chick (Grim et al. 2003). According to the ‘parental fatigue’ hypothesis, parents can respond to the amount of parental care elicited by the brood and desert a cuckoo chick when the investment into it is significantly higher than that required for raising a typical host brood. Physiological exhaustion could work as a proximate cue triggering the desertion. I manipulated parental investment by creating single- and four-chick brood nests to vary offspring need for parental care. Under the ‘time limit’ hypothesis, parents can respond to the parental period length irrespective of the level of parental investment delivered to a brood. The parental fatigue hypothesis predicts that desertion rate and chick mortality should be higher and nestling period should be shorter in prolonged four-chick broods than in all other treatment groups (i.e. shortened and control groups) owing to increased costs of parental care in prolonged four-chick broods. In contrast, the time limit hypothesis predicts increased mortality, desertion rate and shorter nestling periods in both four- and single-chick broods in the prolonged treatment group in comparison to all other groups, but no differences between four- and single-chick broods. The parental decisions to care for chicks for a relatively fixed period could also result in longer nestling periods at shortened parental care nests. Here, chicks would not be ‘under parental pressure’ to leave a nest. In shortened nests also, no starvation or desertion was expected.

The single-chick broods allow the testing of a third hypothesis. Under the ‘single chick’ hypothesis, parents could desert the cuckoo chick because it is alone in the nest (Langmore et al. 2003). According to this hypothesis, the mortality and desertion rate should be higher in single-chick broods, which should also fledge earlier.

Under all hypotheses, I expected that only a portion of tested host pairs would respond to prolonged or shortened nestling periods because not all host pairs desert cuckoo chicks in the study area (Grim et al. 2003).

All time variables (nestling periods, etc.) were measured in hours, but results are presented as days for the readers' convenience. When the data fitted normal distributions, I used parametric tests (the use of non-parametric tests gave qualitatively the same results). All the values shown are means±s.e.

First, I manipulated brood sizes for obtaining four- and single-chick broods (see also Grim & Honza 2001) to vary the level of parental expenditure to raise chicks (average brood size of warblers in the study area is 3.2; Grim & Honza 1997).

Second, for both four- and single-chick broods, I established four treatment groups (see also Nilsson & Svensson 1993; Johnsen et al. 1994): (i) unmanipulated control (nests were subjected to the same procedures as all other nests except for cross-fostering), (ii) manipulated control (brood exchanged for the same age brood from another nest at the age of 3 days posthatch), (iii) shortened brood/nest (an older brood that was moved to a nest originally containing a younger one), and (iv) prolonged brood/nest (a younger brood that was cross-fostered to a nest originally containing an older one). I found no differences between unmanipulated and manipulated four-chick control broods in the nestling period length (U_19,5=0.60, p=0.55), the length of parental care at the nest (U_19,5=0.96, p=0.34) and number of fledged chicks (U_19,5=0.54, p=0.59). Therefore, I pooled the data (hereafter control group).

Manipulated four-chick control nests did not differ in the ages of chicks (age difference in days=0.5±0.2, range=0.3–0.8, n=8 nests matched in 4 nest pairs). Matched pairs of prolonged–shortened broods differed 4–5 days on an average in their ages. Both four-chick (age difference=4.7±0.4, range=1.7–7.4, n=16 nest pairs) and single-chick broods (age difference=4.6±0.2, range=3.5–6.8, n=15 nest pairs) were subject to similar experimental changes in brood ages (t₃₁=0.14, p=0.88). I varied the age differences between matched pairs to test the prediction that the longer the prolongation of the care at the nest, the higher the probability of desertion or chick mortality would be.

Shortened broods were transferred to nests where parents were ‘programmed’ to care for chick(s) for a longer time than their new chicks needed, as their original chick(s) were younger. In contrast, prolonged broods were moved to nests where parents were programmed to care for chick(s) for a shorter time than their new chicks needed, as their original chick(s) were closer to fledging.

The interaction between brood size and the change in the length of care required by a particular nest allowed differetiation of the possible effects of: (i) the length of care and (ii) the amount of care which would be impossible to achieve with naturally parasitized nests only. Prolonged four- and single-chick broods greatly (approx. fourfold) differ in the amount of care (e.g. mass of food) needed, but not in the length of care needed. Prolonged four-chick broods need more care than a normal four-chick brood. However, the prolonged single-chick brood obviously needs a much lower amount of food (approx. less than half) than the average host brood under normal conditions. This simultaneous increase in the length of care and decrease in the amount of care provides a strong test of my alternative hypotheses.

At the time when hatching of a clutch or fledging was expected, the nest was checked three times per day. According to my previous experience (1994–2001), reed warbler chicks may sometimes fledge prematurely when handled or when the nest is touched at chick ages of 9 days or more. In contrast, chicks did not leave the nest when they were observed from a distance of 1m or more. This suggested that the cue for the premature nest leaving was mechanical (the touching of nest and the handling of chicks) rather than visual (seeing an observer close to the nest). Therefore, I avoided mechanical contact with chicks/nests when checking for fledging time and checked nests as quickly as possible and from a distance of at least 1m or more. Despite regular nest checks most nestlings hatched when an observer was not present. The exact hour of hatching (necessary to calculate the exact fledging age) for these individual chicks was calculated from a standard growth curve based on mass measurements. Chick mass was measured with a portable electronic balance to the nearest 0.01g. The curve was based on mass growth of nestlings that were directly observed during hatching and weighted immediately thereafter (1.25±0.02g; range, 1.18–1.31g, n=5 chicks from 5 different nests). I did not measure chicks until fledging (the last measurement was on day 8 to avoid premature fledging). Therefore, the logistic growth curve could not be fitted to these data as these most likely did not include asymptotic mass (cf Grim 2006b). Thus, the curve was calculated as the third order polynomial regression of chick mass in grams against chick age in hours which best fitted the data (R²=0.99, F_3,26=642.59, p<0.0001). The mass of some other nestlings that weighted 1.35–1.60g (n=12 chicks from 12 different nests) when first measured suggested that they were freshly hatched or only several hours old. The real measured masses of these nestlings during measurements at older ages were well predicted by the standard growth curve (measured mass vs. predicted mass at a particular age: Wilcoxon-paired tests, all p>0.25). This indicates that hatching time estimates are robust. Further, hatching times estimated by the growth curve were in concordance with data on hatching times obtained by direct nest checks. Growth curve estimates of hatching hour were done identically across all treatment-groups.

The fledging time was calculated as a midpoint hour between two nest checks when chick(s) left the nest. For some nests (n=10) the exact time of fledging was obtained from time-laps video recordings. Nestlings in four-chick broods were marked with a small colour dot on the nape with non-toxic colours for individual recognition when they were 8 days old. This enabled the checking of nests with minimal disturbance and without the need to handle chicks in the period shortly before fledging.

When I repeated all analyses presented in the results separately for the four- and single-chick brood datasets, I obtained qualitatively identical results. Therefore, I pooled the data when testing the effect of age differences in matched-pair broods on nestling period lengths and fledging success.

Sample sizes for fledging success (total n=105 nests) and nestling period lengths (total n=88 nests) differ between some groups as at some nests, the exact nestling period could not be reliably estimated.

Figure 1 Effects of experimental treatments on parental care periods at nests that successfully fledged. Mean total lengths of time (+s.e.) that the four-chick nests (filled bars) or single-chick nests (open bars) were attended by parents and fosterers. Inset (more ...)

As with parental care periods, the variation in nestling periods was best explained using the same predictors of both treatment and brood size and the interaction between treatments and brood sizes was not significant (F_2,85=1.21, p=0.30). Nestlings from shortened broods spent more time, whereas nestlings from prolonged broods spent less time, in the nest than those from control group (figure 2; F_2,85=30.81, p<0.0001). The nestling period was longer in four-chick broods than in single-chick broods across treatments (F_2,85=6.82, p=0.01).

Figure 2 Effect of experimental treatments on nestling periods in broods that successfully fledged. Mean total lengths of time (+s.e.) that the four-chick broods (filled bars) or single-chick broods (open bars) spent in their original and recipient nests. Sample (more ...)

Analyses of nestling periods at the level of individual chicks (with brood identity as a random effect) gave qualitatively similar results. Nestling periods significantly decreased from shortened through control to prolonged broods (general linear mixed model F_2,128.2=20.62, p<0.0001) and the brood size effect was also significant (F_1,121=11.17, p=0.001). The interaction between treatments and brood sizes was significant (F_2,128.2=6.12, p=0.003) because in shortened broods, there was no significant difference between single- and four-chick broods, whereas nestling periods of individual chicks in prolonged and control broods were longer in the single-chick broods than in the four-chick broods (Tukey–Kramer HSD, p<0.05). This test at the level of individual chicks is conservative as it artificially decreases the length of nestling periods in four-chick broods, but not in single-chick broods, in comparison with the total nestling period from hatching of the first to fledging of the last chick within a brood. Indeed, the average length (10.6±0.1 days) of nestling period in four chick broods was significantly shorter than the total length (11.4±0.1 days; paired t-test, t₄₉=−8.68, p<0.0001).

The length of the period that nests were prolonged (+days) or shortened (−days) significantly and negatively correlated with the nestling period length (r=−0.65, n=88, p<0.0001). However, when only prolonged nests were analysed, the relationship disappeared (r_s=−0.12, n=21, p=0.60). This means that not all individuals in the population follow the same decision rules when forced to care for nest for prolonged periods. This suggests that there is a high inter-individual variability in host responses to ‘parasitic’ chicks in the study population corresponding to similarly high intrapopulation variability in responses to parasitic eggs in hosts of parasitic birds in general (Davies 2000).

Contrary to the parental fatigue hypothesis, the desertion rate was not higher in prolonged four-chick broods (3 out of 13) than single-chick broods (3 out of 14; one-tailed Fisher's exact test, p=0.71). Deserted single- and four-chick broods did not significantly differ in any of their breeding parameters (table 1).

Table 1 Breeding parameters of deserted prolonged single- and four-chick broods. (Mean±s.e. and p-values for Mann–Whitney tests are shown.)

Successfully fledged prolonged broods were similar to deserted prolonged broods in most of their breeding characteristics (table 2). The total length of parental care for deserted nests was lower than that for prolonged, but successful, nests (table 2). Desertions occurred significantly later (12.1±0.3 days) than was the normal parental care period of 11.2±0.1 days at control nests (U_6,37=2.47, p=0.01). The probability of brood desertion significantly increased with increasing prolongation of parental care when all data were included in the analyses (nominal logistic regression: χ²=16.50, n=105, p<0.0001).

Table 2 Breeding parameters of prolonged broods compared between successfully fledged and deserted broods. (Single- and four-chick broods pooled in both datasets; mean ±s.e. and p-values for Mann–Whitney tests are shown.)

The desertion of some broods and the shorter nestling periods in prolonged nests are unlikely to be explained by visitation rates of particular nests. In general, observer visitation rates could have little effect on both desertion rates and nestling period lengths as: (i) reed warblers are tolerant to the presence of humans in our study area (for supporting data see Honza et al. 2004), (ii) despite much more frequent visits to nests in earlier studies, no adverse effect on chicks' growth and no desertions were detected in those studies (Grim & Honza 1997, 2001; Grim 2006b), and (iii) length of the nestling period did not differ between nests checked personally (11.4±0.1) and nests checked by long-term monitoring with time-laps video (11.7±0.2), which were not visited for 2 days before fledging (U_56,10=0.98, p=0.33). Specifically, the number of visits throughout the nestling period only slightly differed between the three groups of nests (ANOVA: R²=0.14, F_2,91=7.51, p=0.001). In addition, shortened broods were visited more frequently (7.5±0.3) than both prolonged (6.4±0.3) and control broods (6.1±0.2) (Tukey–Kramer HSD, p=0.03), while the last two groups did not differ. This was because shortened broods fledged at higher nestling ages and thus these nests had to be checked more times than prolonged broods that fledged earlier. Most importantly, deserted prolonged broods were visited significantly fewer times (4.7±0.6) than successful prolonged broods (6.9±0.3; U_6,21=2.86, p=0.0043).

Furthermore, the single-chick brood treatment had a significant effect on fledging success (ANOVA: R²=0.17, F_2,48=4.75, p=0.01). The shortened and control single-chick groups did not differ from each other (1.0±0.0 versus 1.0±0.0 fledglings per brood, Tukey–Kramer HSD, p>0.05), but prolonged broods showed significantly lower fledging success (0.8±0.1) than both control and shortened groups (Tukey–Kramer HSD, p=0.04). I did not repeat the test excluding deserted nests because there can be no partial mortality in single-chick broods and thus the fledging success across treatments, in principle, cannot vary.

Owing to desertions and partial brood mortality, the fledging success (percentage of hatched chicks that fledged, predation excluded) was significantly lower in prolonged broods in comparison with both control and shortened broods in the pooled four- and single-chick broods dataset (figure 3; ANOVA, R²=0.22, F_2,102=14.74, p<0.0001; Tukey–Kramer HSD, p<0.0001). The control and shortened broods did not differ from each other (Tukey–Kramer HSD, p>0.05). Excluding deserted nests did not change the results qualitatively (ANOVA, R²=0.08, F_2,96=4.07, p=0.02; prolonged broods versus control and shortened broods, Tukey–Kramer HSD, p=0.046). Fledging success significantly decreased with increasing prolongation of the care at the nest (r_s=−0.40, n=105, p<0.0001). The effect was significant even when excluding deserted nests (r_s=−0.23, n=99, p=0.02).

Figure 3 Fledging success (percentage of fledged chicks) in the four-chick broods (filled bars) or single-chick broods (open bars) (excluding predated nests) in relation to experimental treatments. Mean+s.e. shown for four- and single-chick shortened (n=16 and (more ...)

Reed warblers made no rejection errors as they did not desert any of shortened (n=30) or control (n=48) nests. The total desertion rate of prolonged warbler broods (22.2%, n=27) did not differ significantly from that of naturally parasitized broods with single cuckoo chicks in the same study area (15.8%, n=57; χ²=0.50, p=0.48). However, warbler broods were deserted at younger ages (12.1±0.3 days) than cuckoo chicks (14.6±0.3 days; U_6,9=−3.04, p=0.002).

I experimentally tested the predictions of three hypotheses that were specifically designed to untangle the proximate mechanisms of chick desertion behaviour in a common cuckoo host, reed warbler. I observed that prolonged broods suffered from higher rates of breeding failure than control and shortened broods. Latencies from cross-fostering to desertion at deserted prolonged nests were 3–6 days, implying that the desertions occurred not as a response to a sudden change of chick appearance. Moreover, virtually the same level of change in chick appearance at shortened nests did not elicit any desertions (see also Davies & Brooke 1989). Chicks in prolonged broods also fledged at an earlier age. Experimentally induced changes in fledging success and nestling periods did not differ between four- and single-chick broods. These results provide strong support for the time limit hypothesis. The temporal pattern of desertion of cuckoo chicks also supports the time limit hypothesis (see §4a). In contrast, I found no support for the two alternative hypotheses. Specifically, the slightly longer care observed in four-chick nests than single-chick nests is contrary to what was predicted by the parental fatigue hypothesis and is in agreement with the time limit hypothesis. Zero desertion rates in control single-chick broods are also contrary to the single chick hypothesis and imply that brood size of one is not the cue triggering desertion of cuckoo chicks (see also Grim & Honza 2001).

Surprisingly, warbler chicks were deserted ca 2 days earlier than cuckoo chicks. As cuckoo chicks provide their warbler hosts with a supernormal stimulus (Grim & Honza 2001), they could perhaps delay the desertion response of their fosterers as predicted by the parental manipulation hypothesis (Redondo 1993). This seems to be a suitable explanation as other cuckoo species chicks can even save themselves from host physical aggression with intense begging (Redondo 1993; Grim 2006a). Alternatively or additionally, around the time of desertion cuckoo chicks are much bigger than warbler chicks (approx. 50g versus less than approx. 10g). Thus, they have more resources and better thermoregulation abilities (Hund & Prinzinger 1980) than small warbler chicks. This might help them to survive for longer after being deserted.

The longer latencies until desertion at nests parasitized by cuckoos than at nests experimentally prolonged through conspecific ‘parasitism’ also support the time limit hypothesis, but not the parental fatigue or single chick brood hypotheses. If foster parents used the amount of parental expenditure as a cue for desertion (the parental fatigue hypothesis), then they would have to desert cuckoo chicks much earlier than they actually did. In fact, they should desert parasitic chicks earlier than they deserted prolonged warbler broods, as a cuckoo chick overgrows average host brood before the fledging period of warblers, at the age of 8–9 days (Grim 2006b). Already, Grim et al. (2003) provided evidence against parental fatigue hypothesis (‘cuckoo chicks … require more food than an average sized host brood when 8 days old or older’, p. 74), although they did not interpret the data in that way. The parental fatigue hypothesis cannot be reconciled with the long delay between the 8-day size threshold and the desertion of the cuckoo chick 6–7 days later. This is because during this period, the cuckoo chick receives approximately the same amount of food as the average host brood from hatching to fledging (Grim & Honza 2001).

The proximate mechanism underlying the temporary restrictive parental care in reed warblers in the study population could be the hormonal control of the duration of parental care delivered to parasitized versus non-parasitized broods. Silverin & Goldsmith (1984) provided experimental evidence for the partly endogenous programmed period of high-prolactin plasma concentrations in pied flycatchers (Ficedula hypoleuca). Prolactin remained high for a fixed period (16–17 days) after the onset of incubation, but fell sharply in females incubating eggs in experimentally prolonged nests before the eggs hatched (Silverin & Goldsmith 1984).

The shortening of nestling periods in the majority of prolonged nests raises a question of how cuckoo chicks can survive in such a population. Yet, there was a high variation in host responses to prolonged care at the nest: from the ability to care for cross-fostered chicks according to their demands, through forcing them to fledge prematurely, to deserting them if they did not fledge in time (the strategy was independent of the length of prolongation period). This variability in host responses to own chicks may explain the relatively low desertion rate of alien cuckoo chicks. Some cuckoo chicks ‘hit upon’ hosts that vary in their tolerance to prolonged care and fledge at varying fledging ages (17–21 days; Grim 2006b), while other chicks find themselves in nests of deserters and are not able to fledge at all (Grim et al. 2003). Additionally, my study population has probably been parasitized only for a historically short time which is in line with relatively low egg rejection rates (Honza et al. 2004), low parasitic chick rejection rates (Grim et al. 2003; this study) and low adult enemy recognition ability shown by reed warblers in my study area (Honza et al. 2004). Finally, chick desertion is more costly to fitness than egg desertion both with respect to time and energy, which may slow down the spread of chick discrimination through evolutionary time (Grim 2006a).

Observations of longer nestling periods in shortened rather than in control nests (figure 2) indicate that chicks fledge earlier during normal, compared with less-restrictive, experimental conditions. The absence of evidence for similarly restrictive parental behaviour in other reed warbler populations may be owing to methodological reasons because other authors rarely studied older cuckoo chicks (see discussions in Redondo 1993; Grim et al. 2003; Grim 2006a) and, to my knowledge, there are no published experimental cross-fostering studies of determination of nestling period length in open nesting passerines. In addition, high predation rates in the study population (M. Honza, B. Matysioková, T. Grim, unpublished data) may increase the benefits of forcing the brood to fledge as soon as possible. The extent of experimental manipulation (approx. 5 days difference in ages of matched-pair broods) was much larger than observed differences in nestling periods between prolonged and shortened warbler broods (approx. 1 day). This suggests that there are limits to parental manipulation resulting probably from ontogenetic constraints on chick growth (Starck & Ricklefs 1998).

The resulting spread of this adaptation in the long term may render such a host an unsuitable fosterer for cuckoos. This option has not previously been considered in debates on host selection by parasites. Only suitable diet composition, large population sizes, short nestling periods (Soler et al. 1999) and presence or absence of host brood reduction strategy (Soler 2002) are thought to be the main factors facilitating parasitism by cuckoos. Thus, the results of the present study have implications not only for coevolution, signalling and causal mechanisms in evolutionary perspective, but also for host selection by brood parasites. The above presented scenario may also prove fruitful for theoretical models of host–parasite coevolution (e.g. Planqué et al. 2002; Britton et al. in press).

The results of this study highlight an important point made earlier by Redondo (1993, p. 282): ‘Let us assume that birds can only recognize parasitic eggs and chicks by the same mechanism and we shall conclude that chick-recognition will never evolve but in a few rare cases’. Indeed, general study focus on egg-related adaptations in hosts and brood parasites and extrapolations of our knowledge of egg-related adaptations to potential chick-related adaptations may have obscured our understanding of what happens after the parasite egg hatches in a host nest (Redondo 1993; Grim 2005, 2006a). Hopefully, the present study will induce more research on the poorly explored host responses to parasite chicks.

The findings of the present study strikingly differ from those predicted by other studies on host–parasite discrimination, in that hosts do not need to have an internal representation or recognition template of the parasite's appearance to afford discrimination (Hauber & Sherman 2001). In turn, physiological or temporal decision rules, evolved perhaps even in the absence of parasitism, regarding when to terminate the feeding of a brood, whether conspecific or parasitic, alone are sufficient to implement effective anti-parasite responses.

Suggestions by two anonymous referees and the editor substantially improved the paper. This study could not have been carried out without the help of co-workers who assisted with the collection of data: A. Dvorská, M. Honza, B. Matysioková, P. Procházka, P. Samaš and Z. Skoumalová. I am grateful for comments by N. B. Davies, M. E. Hauber, O. Kleven, M. Krist, V. Remeš and E. Tkadlec, and the hospitality of the School of Biological Sciences, University of Auckland. I was supported by grants MSM6198959212 and GACR 206/03/D234.