The egg-recognition processes underlying egg rejection are assumed to be based on an imprinting-like process (a female learning the aspect of her own eggs during her first breeding attempt). The imprinting-like process and the misimprinting costs have been the objective of many theoretical models and frequently have a leading role in papers published on brood parasitism; however, an experiment has never been undertaken to test the existence of this imprinting-like process by manipulating egg appearance in first-time breeding females. Here, we present the first such experimental study using the house sparrow (Passer domesticus), which is a conspecific brood parasite and which has a good ability to reject conspecific eggs, as a model species. We found that contrary to what the hypothesis predicts first-time breeding females did not reject their own eggs in their second breeding attempt. This lack of response against unmanipulated eggs could indicate that females have an innate preference for their own eggs. However, in a second experimental group in which first-time breeding females were allowed to learn the aspect of their (unmanipulated) own eggs, none ejected manipulated eggs during their second clutch either—a finding that does not support the idea of recognition templates being inherited, but instead suggests that recognition templates could be acquired again at each new breeding attempt. Our results demonstrate that it is likely that egg discrimination is not influenced by egg appearance in the first breeding attempt.
Learning can be defined as an adaptive change in behaviour through the effect of acquired information resulting from experience, which is stored in the nervous system . Learned adjustments to the individual's environment are as important for survival and increasing fitness as physiological processes . Learning can be the consequence of imitation, copying, imprinting or teaching .
Imprinting is a special type of early learning that enables young birds subsequently to recognize members of their own species . This means that, if the ability to recognize conspecifics is an acquired trait, when a bird is raised by another species it will attempt to mate with birds of its foster species rather than with birds of its own species [3,4], and then its direct fitness will be zero. This possibility is known as the cost of misimprinting, and it has classically been considered an important theoretical problem for obligate interspecific avian brood parasites, in which females lay their eggs in the nests of other species (the hosts) leaving their offspring being cared for and raised by these host species . Thus, given that young brood parasites never encounter adult conspecifics, it was traditionally thought that, in brood parasites, recognition and selection of a conspecific mate was innate [6,7]. Today, we know that conspecific recognition in brood parasite species is not innate, at least according to studies in a few species in which the problem of whether sexual imprinting among brood parasites is learned or innate [8,9]. The cost of misimprinting is avoided, because the imprinting process is not as instantaneous and irreversible as classically believed , and the timing of the sensitive phase in young brood parasites can be delayed until the appropriate stimulus (i.e. the observation of a conspecific) is received [9,11].
Given that imprinting is just a special type of learning (see above), other cases of learning could be based on imprinting-like processes, which could be at risk of misimprinting costs as well. Probably the most important of these cases is the one that involves learning the appearance of their own eggs by host rejecters of brood parasite eggs (reviewed in ). Namely, given that brood parasitism triggers high fitness costs, this strong selective pressure has favoured the evolution of egg rejection as a host defence [5,12,13], and the recognition processes underlying egg rejection of parasitic eggs are assumed to be based on learning the aspect of the female's own eggs during her first breeding attempt (i.e. an imprinting-like process) and later rejection of foreign eggs differing in appearance from the learned aspect of the eggs [14–19].
This learning scenario in which a female imprints on her own eggs (which also applies to learning the appearance of their own chicks; ) has enormous theoretical implications in the study of coevolution between brood parasites and their hosts, because of the costs of misimprinting (i.e. to imprint on the parasite egg or chick; [5,20]). If a naive individual with recognition ability is parasitized during its first breeding attempt, the appearance of both the parasitic and the female's own eggs will be recognized as her own. Thus, in successive breeding attempts, the female will accept both egg types and thus will never reject a parasitic egg, although, if not parasitized, this misimprinted individual will be able to successfully raise its own offspring in the future, because it also imprinted on its own eggs. Thus, learning by an imprinting-like process is adaptive at the egg stage. However, given the implications of the misimprinting problem at the nestling stage, costs can be much more important, mainly in hosts of highly virulent brood parasites (those that evict or kill all host nestlings; [20,21]). In this case, a host individual that is parasitized at its first breeding attempt can imprint only on the parasite nestling, and so will reject its own nestlings in all subsequent unparasitized breeding attempts. This means that its lifetime reproductive success will be zero. Thus, this high cost of misimprinting at the nestling stage in theory selects against the evolution of chick discrimination in hosts of evictor brood parasites  (but see  and ).
The argument of misimprinting costs linked to an imprinting-like process during the first breeding attempts, both in the egg and in the nestling stages, has frequently been evoked in the literature on brood parasitism [21,24–28], and has been the objective of many theoretical models [20,29–32]. However, the available evidence supporting the existence of this imprinting-like process is very scarce. For example, Rensch (see  for details of Rensch 1924, 1925) replaced the first three eggs of a garden warbler pair (Sylvia borin) with eggs of another passerine species; when the garden warbler female laid its fourth egg, she rejected it, which could be interpreted as a first-laying female having learned the characteristics of the added eggs as her own and later considered her fourth egg as foreign . Another case involves a grey catbird (Dumetella carolinensis) nest in which the entire clutch was replaced with cowbird eggs: the grey catbird parents removed one of their own eggs as a consequence of misimprinting on cowbird eggs  (see  for a similar result with village weavers (Ploceus cucullatus)). Furthermore, another case of misimprinting at the chick stage has recently been documented in which a pair of intraspecific brood parasite species (the American coot, Fulica americana) killed all of their own chicks after misimprinting on chicks of a neighbouring pair .
In addition, the existence of a misimprinting cost has recently been empirically supported in three host species of the brood parasitic brown-headed cowbird (Molothrus ater), in which it has been shown that, in areas where cowbirds are abundant, the rejection rate of foreign eggs is lower than in areas where the parasite is less common . However, an alternative possibility that was not discussed in this study is that the higher rejection rate found in areas where cowbirds were more abundant could be the consequence of brood parasites punishing hosts that reject the parasitic eggs (the mafia hypothesis; ), thereby forcing the hosts to accept the parasitic eggs in future breeding attempts. This nest-destruction behaviour in nests in which the parasitic egg was rejected has been experimentally demonstrated in brown-headed cowbirds  and, therefore, could explain the results found by Strausberger & Rothstein .
The obvious experiment to test the existence of this imprinting-like process using first-time breeding females has never been carried out because of the difficulties of ensuring which individuals are female first-time breeders under natural conditions. Here, using the house sparrow (Passer domesticus) as the model species and working with a captive breeding population, we were able to manipulate the appearance of the eggs laid by first-time breeding females as soon as the eggs were laid. If an imprinting-like process occurs during the first clutch of first-time breeders, it would be expected that, during their second breeding attempts, they would reject eggs differing in appearance from that learned during the first clutch.
2. Material and methods
(a) Study species, study population and general methods
The house sparrow is a very common passerine species in which intraspecific brood parasitism has been reported . It has been unambiguously demonstrated that the house sparrow has a fine capacity to discriminate conspecific eggs and to eject them while incurring very low rejection costs .
This study was performed in a captive population of house sparrows maintained in two aviaries in Hernán Valle, 60 km from Granada, Spain, during the 2011 breeding season. Each of these two aviaries consisted of four cages of ca 50 m3 each, interconnected by small holes through a central cage (of approx. 40 m3) in which food was provided. All sparrows were marked with coloured leg bands for individual identification. There were a total of 50 and 47 pairs in aviary 1 and aviary 2, respectively. In aviary 1, 23 pairs were formed by 1 year old females and 1 year old males, and 10 by 1 year old females and older males. In aviary 2, 18 pairs were formed by 1 year old females and 1 year old males, and seven by 1 year old females and older males. The rest of the pairs in both aviaries were formed by experienced females.
The birds were provided ad libitum access to commercial seed mix for canaries, nestling food for canaries with honey and small pieces of fruit added (egg food with fruits; Bogena), cracked grains of wheat and rice, fly maggots, apple, lettuce, sand, and vitamin-supplemented water. Each aviary was provided with more nest-boxes than pairs. Potential stress was minimized because, when we entered one of the four cages in each aviary, sparrows moved to the others, and they could not see that we were manipulating their nest-boxes. Various materials for nest construction (cotton, wood, mattress stuffing and vegetable material) were provided with ad libitum access during the breeding season. More detailed information on aviaries and sparrow care can be found in Soler et al. .
Each breeding season, all nestlings are ringed with a closed numbered ring for individual recognition, and later, in December, when all individuals are captured, the surviving juveniles are ringed with a unique colour combination. Thus, we can securely identify first-time breeders.
Pair members breeding in each nest-box were identified by observations or by video filming the nest entrance once the first egg was laid. Almost all breeding pairs used in this study remained together during the entire breeding season. There was only one case of divorce during the second clutch (n = 59 pairs) and consequently this pair was excluded from the analyses. From the beginning of the breeding season, nest-boxes were examined weekly, but when the construction of a nest was complete, the nest-box was checked daily just after sunrise, enabling us to detect the first egg immediately after it was laid. Each egg was numbered according to laying order.
(b) Experimental design, predictions and egg manipulation
In our sparrow population, most pairs lay more than one clutch . Thus, we designed an experiment involving first-time breeding females, considering first and second clutches and three experimental groups (table 1). The first two groups allowed us to test whether females reject eggs differing from those of their first clutches (predictions from groups 1 and 2 in table 1). If females confronted with manipulated eggs during their first clutches do not reject unmanipulated eggs during their second clutches, such a lack of response could be a consequence of recognition templates being at least partially inherited . However, our second experimental group allowed us to discard this possibility, because females allowed to learn the aspect of their own eggs during the first clutch should reject the manipulated eggs during the second clutch even if recognition templates are inherited (table 1). Finally, group 3, in case predictions 1 and 2 were supported, allowed us to test the effect of our painting manipulation. The acceptance of manipulated eggs in both the first and second clutches of the third group (table 1) would confirm the reliability of our experimental manipulation.
In manipulated clutches, in the three groups, eggs were manipulated as soon as they were laid, every day early in the morning (if one nest was visited before the egg was laid, it was visited again about 30 min later), thus minimizing the time a host would be exposed to its own eggs. Most eggs were laid early in the morning, during the first 2 h after sunrise (most of them, approx. 90%, between 07.00 and 08.30). Manipulation consisted of adding 20–25 extra spots of approximately 2–3 mm using a brown fibre pen (Faber-Castel OHP colour code 78, size 1525). Manipulated clutches clearly differed from normal house sparrow clutches (figure 1). We considered three possible sparrow responses to experimental egg manipulation: ejection, desertion, and acceptance of the eggs. We checked the nest every day after experimental manipulation, and considered acceptance when neither ejection nor desertion occurred during the 7 days after the manipulation of the last egg.
First-year breeding house sparrow females typically laid four eggs (first clutch, mean ± s.e. = 4.42 ± 0.14, n = 58), a clutch size very similar to that corresponding to old females (first clutch, 4.46 ± 0.18, n = 39; Mann–Witney U-test: U = 1003, p = 0.35). Moreover, the clutch size of first clutches did not differ among experimental groups (first clutches, Kruskal–Wallis ANOVA test: H (2, N = 59) =2.79, p = 0.25).
Expected frequencies of rejection rates in groups 1 and 2 were calculated based on the rejection rate determined from an egg-recognition test performed in these same breeding pairs and using a non-mimetic egg. That is, it is possible that not all individuals used in our study were genetically able to recognize and reject foreign eggs; moreover, it could also be possible that some individuals genetically able to recognize foreign eggs could decide to accept them (i.e. a plastic response; ). As non-mimetic models, real house sparrow's eggs painted red were used. Non-mimetic eggs of a red colour have frequently been used in egg-recognition experiments [40–42], including one performed in the house sparrow . Moreno-Rueda & Soler  found that the rejection rate of mimetic eggs (conspecific) and non-mimetic eggs (painted red) was very similar (23% and 30%, respectively; p = 0.73). Thus, we can assume that house sparrows respond similarly to the two types of experimental eggs used in this study.
We performed the egg-recognition test at the end of the second clutch and in subsequent ones. Briefly, one sparrow egg from previously failed clutches, which were kept in a refrigerator, was painted red and introduced into the nest. By using natural eggs, we excluded the possibility that unsuccessful attempts to puncture hard model eggs could increase the costs of rejection and/or provoke clutch desertion . We recorded the response to red eggs from a total of 50, 26 and four second, third and fourth clutches, respectively. Both the red eggs that were added as well as the female's own eggs after laying were manipulated out of sight of the breeding birds thanks to the characteristics of the aviaries (see above). We considered the red egg to have been accepted when it remained in the nest for at least 7 days. We considered the red egg to have been ejected when in one of our visits it had disappeared, and we considered a clutch deserted when the clutch remained cold for at least two consecutive days.
We used eight mini-cameras (2.5 × 2.5 × 1.8 cm; Euroma, KPC-S500) for filming the sparrows’ response to both experimental egg manipulation and addition of the red egg. For this purpose, nest-boxes had a transparent ceiling that allowed light into the box, with a small hole of the appropriate size for a camera to be installed on it after the introduction of the experimental egg. Each set of four cameras was connected to a recorder (Linux MPEG4) and a monitor that allowed us to view what was happening in the nest at any time during the video recording.
All statistical analyses were performed with Statistica v. 8.0 software (StatSoft 2008).
(a) Estimation of the expected rejection rates
One real house sparrow egg painted red was experimentally added to a total of 80 clutches. The red egg was ejected in 11 out of 50 second clutches, in four out of 26 third clutches and in zero out of four fourth clutches. No cases of desertion or other possible types of response were found. Notably, none of the four house sparrow pairs that ejected the red experimental egg during the third clutch ejected it during the second clutch. Thus, this egg-recognition experiment shows that at least 15 out of 50 (30%) house sparrow pairs were able to recognize and reject the experimental non-mimetic red egg. Therefore, we assume a 30 per cent rejection rate for these breeding pairs and, thus, if the hypothesis concerning the existence of an imprinting-like process is correct, we would expect that, during their second clutches, 30 per cent of females in groups 1 and 2 should reject their own eggs or manipulated eggs, respectively (expected frequencies in table 2).
In house sparrows, both males and females are able to discriminate foreign eggs . As we installed a camera in each nest-box (see §2), we were able to film 13 ejections, of which six were perpetrated by males (three by 1 year old males and three by older males) and seven by females.
(b) Experimental test of the imprinting-like process
Altogether, this experiment was successfully conducted in 51 first clutches and 44 second clutches of first-year breeding house sparrow females. All house sparrow breeding pairs (except one case of ejection in an unpainted clutch in group 3) showed the same response: acceptance of their second clutches (table 2). That is, contrary to the hypothetical occurrence of an imprinting-like process during first clutches, no cases of ejection or desertion in second clutches were found in those groups in which the egg coloration of second clutches differed from that experienced by the breeding pairs during first clutches (i.e. groups 1 and 2; table 2). Consequently, the expected frequencies of rejection/acceptance for these groups considering the hypothesis as correct differed significantly from observed frequencies (table 2), with regard to groups 1 and 2 both separately (predictions of groups 1 and 2) and jointly (general prediction; table 1).
Most rejecter host species recognize a foreign egg even in the absence of their own eggs; that is, egg discrimination is usually based on true egg recognition [15,46,47]. While recognition by discordancy means that hosts eject an egg which differs from the rest (i.e. based on direct comparison), true egg recognition implies that rejecters know the appearance of their own eggs and reject those that differ from their memorized templates [15,16,24,48–50]. In this memory-based mechanism, recognition templates are assumed to be most probably learned by an imprinting-like process, which can be highly costly if a rejecter individual imprints on the parasite egg or chick (see references discussed earlier).
Our experiment has shown that for first-time breeders, when the egg appearance of second clutches differed from that of first clutches, the second clutches were not ejected or abandoned, as predicted by the hypothesis concerning the existence of an imprinting-like process during first clutches (table 1). In fact, the frequencies of rejection/acceptance predicted by this hypothesis differed significantly from observed frequencies (table 2).
A result similar to that found in our experimental group 1 has been reported in other experimental studies where individuals of unknown age, after first accepting non-mimetic eggs, later also accepted their own eggs [19,35,51]. Of special note is the study by Lotem et al. , in which first clutches of mid-season breeding (likely to be first-time breeders) great reed warbler (Acrocephalus arundinaceus) females, which had been exposed only to brown eggs, were artificially depredated in order to provoke a re-nesting attempt. However, they did not reject their own eggs in the re-nesting clutches.
Results reported by Lotem et al. , as the authors themselves stated, do not necessarily contradict the hypothesis about the occurrence of an imprinting-like process during first clutches for the following three reasons. First, their experimental procedure under natural conditions perhaps could not prevent great reed warbler females from learning their own eggs before experimental replacement, because females could have enough time to imprint on them . Second, another methodological weakness could be that mid-season breeding females were not first-time breeders, but were experienced females of low physical condition. And third, the lack of egg discrimination could have resulted from females having an innate recognition of their own eggs (i.e. recognition templates may be inherited [52–54]), which would favour the acceptance of their own eggs .
Our results in group 1 are very similar to those found by Lotem et al. . However, the first two reasons stated above, which did not allow Lotem et al.  to conclude that their results contradicted the hypothetical existence of an imprinting-like process during first clutches, are not likely to apply to our results. This is because, in the first case, captivity conditions allowed us to manipulate the coloration of the sparrows' eggs very soon after the female laid them (see §2). Certainly, although eggs were manipulated as soon as they were laid (see §2), females still had several minutes to observe and learn egg appearance before egg manipulation, and this time lag could be long enough to imprint on their eggs. However, this potential weakness in our experimental design is unlikely for two reasons: first, it could be applied to females but not to males, and in house sparrows males as well as females are able to discriminate foreign eggs (see §3); and second, females could delay learning until the clutch is completed [18,19]. When intra-clutch variation in egg morphology is high (as in the case of the house sparrow; figure 1), such a delay will be adaptive, given that it would allow females to learn the appearance of the last egg laid, which often is very different from the previous ones (figure 1). In the second case, it is clear that our naive females were without a doubt first-time breeding females.
The third reason could be applied to our results for group 1, but not for group 2 (table 1), in which the female's eggs of the first clutch remained unmanipulated, whereas those of the second clutch were manipulated immediately after laying (table 1). If the explanation for the results found by Lotem et al.  and by us in group 1 was that females have an innate preference for their own eggs, in group 2, in which the female was allowed to learn the aspect of its (unmanipulated) own eggs, manipulated eggs during the second clutch should have been ejected even if recognition templates are inherited, but they were not. Our experimental design does not provide a fuller explanation of the mechanism by which females learn to recognize their own eggs, but our results in group 2 demonstrate that recognition templates are not inherited, and suggest that egg learning is not restricted to the first breeding attempt, signifying that new recognition templates would be acquired at each subsequent breeding attempt.
The fact that rejecters could refine their recognition template through some form of template-updating mechanism has been suggested in several studies [19,24,26,48], but we have provided evidence showing for the first time the possibility that a new learning process may occur during the second clutch. Thus, recognition of their own eggs by house sparrow females is not based on an imprinting-like process and, thus, the misimprinting costs frequently cited in the brood parasitism literature (see §1) do not exist, at least in this species. Our results reinforce the findings of several previous empirical studies that failed to find any evidence of experience-dependent egg-discrimination ability ([48,49,55–58]; but see  for evidence of an age effect on chick discrimination). Our experimental study using first-year breeding females shows that learning or imprinting on her own eggs has no consequences for egg discrimination in later clutches even during the same breeding season.
Our study has two potential problems. First, that our egg manipulation (addition of spots) may not have been strong enough to provoke egg rejection by house sparrows; however, this is very unlikely, because sparrows are able to discriminate against very similar conspecific eggs  and sparrows’ response to foreign eggs (i.e. rejection rate) is higher after spot addition . Second, the fact that both males and females are able to recognize and reject odd eggs (as found in other host species ) could be viewed as a potential problem for the interpretation of egg-recognition experiments trying to determine the mechanism by which rejecters learn to recognize their own eggs. This is not true because by installing cameras in nest-boxes, which is recommended in egg-recognition experiments in general , it is easy to determine which sex is responsible for the egg ejection, and the discrimination mechanism can be studied by comparing the two sexes.
In fact, the possibility of our results suggesting that females could learn the appearance of their own eggs in each new breeding attempt is the only mechanism that could be considered adaptive for males, given that one male could divorce or lose his female and then will pair with another female whose eggs, in species with high inter-clutch egg variation, will have a different appearance and then, if the male has egg-rejection ability, will reject the eggs laid by the new female . As can be seen in figure 1, inter-clutch egg variation is high in house sparrows, and thus this could be the explanation for the lack of an imprinting-like process in house sparrow males. However, the following question arises: how can the lack of this process in females be explained? Furthermore, we might ask: what then is the mechanism responsible for egg recognition in both house sparrow males and females? Clearly, further experimental studies are needed to answer these questions.
In conclusion, our results demonstrate that first-time breeding house sparrow females, in which egg coloration of second clutches differed from that learned during first clutches, did not eject or abandon their second clutches as the hypothesis about the existence of an imprinting-like process predicts. This result supports neither the assumption of the existence of high misimprinting costs nor the possibility of recognition templates being inherited, but rather opens up the possibility of the existence of a memory-based mechanism in which recognition templates are learned in each new breeding attempt (see the electronic supplementary material).
We thank Mariola Sánchez, Ismael Lamas, José María Villaescusa, Cristina Curado and Alejandro for their help in the care of the sparrows; Silvia Alsina for helping us with videotape analysis; and David Nesbitt for improving the English. We especially thank Juan J. Soler for useful comments on the manuscript and statistical advice. Financial support was provided by Junta de Andalucía (to the RNM 339 research group) and by the Spanish Ministerio de Educación y Ciencia/FEDER (research project no. CGL2011-25634). Research has been conducted according to relevant national (Real Decreto 1201/2005, de 10 de Octubre) and regional (permissions provided yearly by la Consejería de Medio Ambiente de la Junta de Andalucía) guidelines.
- Received October 23, 2012.
- Accepted November 21, 2012.
- © 2012 The Author(s) Published by the Royal Society. All rights reserved.