Oecologia (1998) 114:514±521 Ó Springer-Verlag 1998 Shin-Ichi Seki á Hajime Takano Caterpillar abundance in the territory affects the breeding performance of great tit Parus major minor Received: 10 June 1997 / Accepted: 29 December 1997 Abstract The e ects of caterpillar food supply on the breeding performance of a population of the Japanese great tit Parus major minor were investigated. Since more than 90% of the food items in our study site were caterpillars living on trees, we estimated the food availability using 20 frass traps per hectare. The sampling error of this method was about 10% on average, which was accurate enough to detect di erences between territories. Food abundance at laying in each territory a ected the timing of egg laying. However, food amount after hatching was correlated with clutch size. No relationship was found between edgling quality and food availability, probably because the e ects of local variation in food abundance could be canceled out by parental e ort such as extending the foraging area. There was a signi cant negative correlation between the length of the nestling period and food availability. We suggest that parent tits decide the timing of edging at the point where two factors, predation risk before edging and additional improvement of nestling quality, are balanced. Food availability just after edging a ected the length of post- edging parental care; it seems that edglings in ``poor'' territories would have had di culty in nding food and hence needed to depend on their parents longer than those in ``rich'' territories. S.-I. Seki 1 Laboratory of Forest Zoology, Faculty of Agriculture, The University of Tokyo, Bunkyo-ku, Tokyo 113-0032, Japan H. Takano Tama Forest Science Garden, Forestry and Forest Products Research Institute, 1833 Todori, Hachioji, Tokyo 193-0843, Japan Present address: 1 Wildlife Management Laboratory, Kyushu Research Center, Forestry and Forest Products Research Institute, 4-11-16 Kurokami, Kumamoto 860-0862, Japan Fax number: +81-96-3445054; e-mail: seki@ pri-kys.a rc.go.jp Key words Great tit á Caterpillar á Frass trap á Food availability á Breeding performance Introduction Breeding birds in the same population often di er widely in their breeding parameters, i.e., timing of breeding, clutch size, nestling weight, edgling survival and number of broods (Lack 1966; Nur 1986; van Noordwijk and van Balen 1988; McCleery and Perrins 1988; Gosler 1993; Cichon and Linden 1995). While these di erences have been reported in many species, the causes and the mechanisms are still controversial because it is often di cult to separate the contributions to the breeding performance made by parental quality from those by territory quality. Some recent studies have revealed that environmental conditions at the micro-habitat level, mainly food abundance, have marked e ects on several measures of breeding performance (HoÈ gstedt 1980; Goodburn 1991; Dhondt et al. 1992). These e ects are stronger than those of bird quality when habitats are heterogeneous at the scale of home ranges or territories. Dhondt et al. (1992) reported that female blue tits, Parus caeruleus, laid smaller clutches in poor quality sites than in good quality sites; they evaluated the quality of nest site by examining the breeding performances during the previous 5 years. Variance in clutch size of magpie, Pica pica, was mostly due to the e ect of territory quality when the habitat was heterogeneous in quality (HoÈ gstedt 1980; Goodburn 1991). In the above studies it is assumed that quality of a breeding site was a xed character, because of the di culty of measuring territory quality directly. However, food availability in a territory often changes through the season, and it is di cult for parent birds to predict precisely the future food availability (Perrins 1991). In this situation, adaptive change in clutch size is not possible. Parents can, however, enhance their breeding
515 success if they change breeding behavior and hence other breeding parameters after egg laying (van Noordwijk and van Balen 1988; McCleery and Perrins 1988; Higuchi and Momose 1981). In this study we used frass traps to assess caterpillar abundance within the territories of great tits, Parus major minor. Frass traps have been used in many studies of birds for measuring the caterpillar availability at the habitat level, but, to our knowledge, have not been used at the scale of the territory (Lack 1966; Zandt 1994). The aims of the present paper are to discuss: (1) whether frass fall can be reliably estimated with reasonable e ort; and (2) how the uctuating food availability within territories a ects breeding performance of great tits. Materials and methods we visited the nests every day. From these visits we determined rst egg laying date (laying date), clutch size, hatching date, brood size and edging date. Breeding pairs were divided into two groups, `` rst brood'' pairs and ``late brood'' pairs; the latter include repeat and second broods. General breeding data are shown in Table 1. Nestlings were weighed four times: soon after hatching, and at 5, 10 and 15 days of age. We measured wing length and tarsus length 15 days after hatching. For calculating uctuating asymmetry [FA ˆj(length of right wing or tarsus) ) (length of left wing or tarsus)j], both right and left wings and tarsus were measured in 1995. All nestlings were ringed at 10 days. Feeding behavior of the parents was observed in 1995 using binoculars (8 ), telescope (20 ) and videocamera (14 ). We recorded the food items, the body length of food and the frequency of visits by each parent from these observations. The length of food was estimated using the parents' beak length as a scale. These observations were carried out in the mornings (6:00±11:00 a.m.) when the nestlings were 10±15 days old. We collected caterpillars in the surrounding forests and oven-dried them in the laboratory. There was a signi cant relationship between caterpillar body length (L, mm) and dry weight (W, mg) (logw ˆ 0.067L ) 0.53, Study site The study was carried out at Tama Forest Science Garden in Hachioji, Tokyo, Japan (35 39 N, 139 16 E, about 200 m altitude) from 1994 to 1995. This area is located at the northeast end of Tama Hill Forest region. The woods are composed of natural forests such as Quercus glauca and Abies rma, coniferous plantations of Cryptomeria japonica and Chamaecyparis obtusa, and deciduous plantations mainly of Prunus spp. The study site is quite heterogeneous since the sizes of forest patches are small, 0.05±3.07 ha each (0.52 ha on average) (Fig. 1). As a result, the caterpillar abundance varies even within the study site. Seventy nestboxes were set in 29 ha of the study area and 25±30 pairs used these nestboxes in each year. General breeding data Great tits were caught with mist nets mainly from October to early April, though some females were caught during the breeding season, in May. All the birds captured were marked with one metal ring and three color rings for individual identi cation. A total of 66% (1994) to 86% (1995) of breeding pairs were captured and marked. Song posts of males and observed points of all birds were mapped from late March to the end of July. Territory was estimated as the maximum area over which the male sang (song posts ³31). Home range of each pair was estimated as the maximum observed area of male and female (observed points ³57). Only those points obtained from ringed individuals or individuals tracked from particular nestboxes were used in these maps. Nestboxes were visited every 5 days from late March to July. Over 1 week around hatching and another week around edging, Fig. 1 Map showing vegetation of the study site Table 1 The general breeding parameters in the studied population of great tits. The values of laying date, clutchsize, and brood size are given as mean SE, and are followed by sample size in parentheses. Laying date was calculated taking 1 April as day 1 1994 1995 First brood Late brood First brood Late brood Laying date 12.3 5.0 (26) ± 21.1 4.4 (15) ± Clutch size 8.4 0.9 (21) 6.9 0.8 (13) 8.4 1.3 (11) 7.3 0.6 (6) Hatching success (%) 91 85 86 91 (except for predation and nest desertion) Brood Size 7.6 1.4 (21) 5.9 1.1 (9) 7.4 2.2 (11) 6.7 1.5 (3) Predation by Elaphe climacophora (%) 6 47 25 63 Fledging success (%) (except for predation) 100 100 100 100
516 r 2 ˆ 0.74, n ˆ 50), which enabled us to calculated the dry weight of caterpillars brought to the nest. We censused the study area three times a day for a few weeks after edging. When we found a family ock, we tracked the group and checked whether the edglings were fed by their parents. Because some families disappeared just after edging, we estimated the length of post- edging parental care using data obtained from those families that were observed after the o spring became independent. In this paper, we will not discuss reproductive success, since there is not enough data on the survival of edglings. Estimation of food availability In the study site 93% of the food items brought to the nest were caterpillars (12 nests, 655 feeding visits), such as Geometridae and Notodontidae, and parent tits always fed at tree branches and never on the ground. Therefore food availability for the breeding tits was evaluated from the caterpillar abundance on trees. We used frass traps to estimate caterpillar density. Square traps with an opening of 50 50 cm were placed on a grid, about 20 traps/ha. The shape of the trap was designed not to squash the frass, using wire frames at the bottom. We collected frass with variable intervals, 4±7 days. We changed the collecting interval depending on the weather because the frass would be dissolved when it had rained. We sorted out the frass from litter, oven-dried (60 C, 48 h) and weighed it. The caterpillar density of each territory was measured as the average frass density (mg á m )2 á day )1 ) in that territory. Measurement of frass was started on 1 April in 1994 and 10 April in 1995, and ended 20 days after edging in each territory. In 1994, 100 frass traps were placed in a 10 10 grid over a 71 71 m (0.5 ha) study plot to estimate the sampling error. The size of this plot was within the range of a territory size at this study site. We collected frass every 7 days and weighed as stated above. The study plot was about in the centre of the study site. This research was carried out over two periods, 20±26 June and 26 June± 2 July 1994. We estimated the sampling error at low trap densities using the data from this plot. Figure 2 shows an example of a simulation for estimating the sampling error when 9 traps per 0.5 ha were used. First we divided the study plot into nine square blocks of the same size (A±I) and then chose four traps from each block, which were used as the source pool for the following simulation (A1, A2, A3, A4, B1,..., I4). The traps chosen were located near the centre of each block. Secondly, we selected one of the four traps in each block, and calculated the average of nine traps selected from all blocks. This procedure was repeated 100 times, and the j th average was designated as FDj. Thirdly, the absolute di erence between the average of 100 traps (FD) and FDj was calculated for each run (Ej ˆjFDj ) FDj). The average of all Ej was de ned as mean sampling error, because FD was assumed to be the true frass density. The simulation was carried out at four densities; 8, 18, 32, and 50 traps/ha. We sampled caterpillars in 12 known great tit nest sites to compare their biomass with frass fall at that time. We collected them by beating 100 branches per territory (60 cm each), ovendried (80 C, 24 h) and weighed them. The height of the trees in those 12 territories selected was about 8 m, which enabled us to collect caterpillars from various heights as well as various positions in the canopy. We collected caterpillars only once during the 7 days of frass collection, because intensive sampling might itself reduce caterpillar density. Results E ect of frass density on the timing of breeding and clutch size There was a signi cant positive correlation between the laying date of each pair and the date of maximum frass density in the territory in 1995 as well as 1994 (Fig. 3). Clutch size in rst broods decreased as the breeding season progressed; the negative relationship was signi- cant in both years (Fig. 4). However, when clutch size was plotted against frass density during egg-laying, no clear relationship was observed (r 2 ˆ 0.002 in both years). Clutch size was signi cantly correlated with frass fall during the nestling period in 1995 (r ˆ 0.66, n ˆ 12, P ˆ 0.02) while the relationship in 1994 was not signi cant (r ˆ 0.37, n ˆ 21, n.s.); there was also a signi cant relationship when data from the two years were combined by general linear model (ANCOVA) (F ˆ 5.04, df ˆ 1, P < 0.05). E ect of frass density on nestlings' quality Fig. 2 The design for choosing traps from the study plot (for 9 traps per 0.5 ha). One hundred frass traps were placed in a 10 10 grid over 71 71 m (0.5 ha) of study plot. The average of the 100 traps was assumed to be the true frass density (FD) inthisplot We used three measures of nestling quality: weight at 15 days (15-days weight), variance of 15-days weight in a brood, uctuating asymmetry (FA) of 15-days tarsus length and wing length (average within a brood). None of these was signi cantly correlated with frass density (15-days weight: r ˆ )0.40, n ˆ 21 in 1994; r ˆ 0.085, n ˆ 12 in 1995. The variance of 15-days weight: r ˆ 0.040, n ˆ 14 in 1994; r ˆ 0.17, n ˆ 9 in 1995. FA of wing length r ˆ )0.20, n ˆ 9; FA of tarsus length r ˆ )0.42, n ˆ 9).
517 E ect of frass density on territory size and home range size Territory size increased signi cantly as frass density during egg laying period decreased in 1995 (r ˆ )0.77, n ˆ 9, P < 0.02), while this relationship was not clear in 1994 (r ˆ )0.25, n ˆ 13, n.s.). Since the sample size for one year might not be large enough to detect the e ect of frass density, we used a general linear model (ANCOVA) to detect overall trends. Highly signi cant negative relationship was obtained between territory size and frass density during the egg-laying period (F ˆ 10.53, df ˆ 1, P < 0.004). When territory size was plotted against laying date, a signi cant negative relationship was obtained (general linear model, ANCOVA, F ˆ 6.71, df ˆ 1, P < 0.02), although the relationship in each year was not signi cant. There was a negative relationship between [(home range size)/(territory size)] and frass density (g á territory )1 á day )1 á nestling )1 ) (Fig. 5). This means that the extent to which parent birds extend their home range outside their territory was signi cantly correlated with food availability per nestling in their territories. Fig. 3 The correlations between the date of maximum frass fall and the laying date in a 1994 (r ˆ 0.663, n ˆ 13, P < 0.02) and b 1995 (r ˆ 0.784, n ˆ 9, P < 0.02). Each point shows a separate pair Fig. 4 Clutch size in relation to laying date in a 1994 (r ˆ )0.597, n ˆ 20, P < 0.001) and in b 1995 (r ˆ )0.591, n ˆ 15, P ˆ 0.02) Fig. 5 The parents' feeding area (home range size/territory size) in relation to food availability per nestling (frass fall á territory )1 á day )1 á nestling )1 )ina1994 (r ˆ )0.893, n ˆ 12, P < 0.001 in the 1st brood; r ˆ )0.877, n ˆ 8, P < 0.01 in the late brood) and b 1995 (r ˆ )0.901, n ˆ 9, P < 0.001 in the 1st brood; r ˆ )0.778, n ˆ 3, n.s. in the late brood). The 3 pairs in territories where food was extraordinarily abundant were excluded from the regression analysis because more than a certain level of food is thought to have no e ect on parental behavior
518 E ect of frass density on parental care Hourly dry weight of feeding items per nest increased with increasing frass density during the nestling period (r ˆ 0.63, n ˆ 9, P<0.05), but hourly dry weight of feeding items per nestling showed no correlation with frass density (r ˆ 0.17, n ˆ 9). Similar results were observed for feeding frequency (r ˆ 0.57, n.s. in feeding frequency per nest; r ˆ 0.16 in feeding frequency per nestling, n ˆ 9). When the length of the nestling period was plotted against the frass density, a negative relationship was found (Fig. 6). The correlation between them was signi cant in rst and late broods in 1994 as well as in the rst broods in 1995. There was a signi cant negative relationship between length of post- edging parental care and frass density during the 10 days after edging (Fig. 7). E ect of frass density on the number of broods and the interval between rst and second broods Neither the length of the nestling period nor the duration of post- edging parental care a ected the number of brood (t-test P > 0.1 in both years, see also Figs. 6 and 7). For pairs that bred twice, frass fall had some e ect on the interval between rst and second broods (general linear model, ANOVA; F ˆ 6.50, df ˆ 1, n ˆ 8, P ˆ 0.051) The reliability of frass trap method The relationship between frass trap density and the mean sampling error in frass is shown in Fig. 8. The mean error at the trap density of 18±32 was about 10% of the frass density estimated by 200 traps. The correlation between the frass density and the caterpillar biomass was highly signi cant (r ˆ 0.91, n ˆ 12, P<0.001), implying that the frass density in a great tit's territory strongly re ects the caterpillar density. Frass density during the nestling period or during the 10 days after edging did not a ect the number of broods (t-test P > 0.1 in both years, see also Figs. 6 and 7). Fig. 6 The relationships between frass fall during the nestling period and nestling duration a in 1994 (r ˆ )0.862, n ˆ 13, P < 0.001 in the 1st brood; r ˆ )0.877, n ˆ 8, P < 0.01 in the late brood) and b in 1995 (r ˆ )0.963, n ˆ 9, P < 0.001 in the 1st brood; r ˆ )0.812, n ˆ 3, n.s. in the late brood) Fig. 7 The relationships between frass fall during the 10 days after edging and the duration of post- edging parental care (r ˆ )0.905, n ˆ 11, P < 0.001 in the 1st brood, 1994; r ˆ )0.444, n ˆ 6, n.s. in the 1st brood, 1995). Data from rst brood only are shown in this gure because trees were thickly leaved by the time of second brood and detailed observation of edglings became impossible
519 decide the timing of breeding using these cues, or they may be kept from laying eggs by nutritional constraints. E ect of food availability on nestlings' quality Fig. 8 The relationship between frass trap density and the mean sampling error ( SD) in frass density Discussion The reliability of the frass trap method The measurement of food availability using frass traps was successful at the scale of a great tit's territory. The trap density necessary for estimating caterpillar biomass should be examined in each habitat, because it would be a ected by many factors, such as habitat heterogeneity and life history characteristics of insects. The good estimates even at the low trap density may be due to the fairly homogeneous distribution of frass at this study site. Zandt (1994) pointed out some di culties in using frass traps: poor weather conditions, such as low temperature and rainfall, would a ect frass production by reducing the feeding activities of caterpillars. During this study these e ects of weather condition were not observed. E ect of food availability on egg laying Many studies reported that yearly average of egg laying date was signi cantly correlated with the timing of caterpillar appearance (Lack 1958; Perrins 1979; Perrins and McCleery 1989; van Noordwijk et al. 1995). Perrins (1991) concluded that female birds start breeding as soon as caterpillars, or some other foods, are available and, as a result, timed their breeding to have their nestlings when caterpillars were most abundant. In these studies the date of maximum frass or the half-fall date of fully grown larvae were used as a measure of the timing of caterpillar appearance. In the present study, the timing of caterpillar appearance varied by about a month between territories, and the females varied in their timing of laying in parallel. This result indicates that not only general factors, such as weather conditions, but also local factors differing between territories, such as vegetation or caterpillar abundance, a ect the timing of breeding. Females In the great tit, heavier nestlings have a better chance of surviving the period immediately after edging (Perrins 1965; Garnett 1981; Tinbergen and Boerlijst 1990) and a higher chance of acquiring territory in the following year (Drent 1984). Thus the nestling's quality is an important factor a ecting breeding success of their parents. In our study no signi cant relationship was observed between nestling quality and food availability. Why was no clear relationship observed? Nestlings do not feed themselves, but are fed by their parents, so the e ect of food availability in their territory could be greatly modi ed by changing parental e ort. For example, pairs with low food availability had large home ranges, tended to have fewer nestlings, and brought the same amount of food per nestling as those with high food availability. Time allocation of parents could also have changed, though we did not examine this. E ect of food availability on the timing of edging When the length of the nestling period was plotted against the food availability, there were negative relationships for both rst and late broods, but the slope of the regression line was greater in late broods. What are the causes of this negative correlation and the di erence in slopes? Obviously nestlings cannot directly assess the food availability. Also, nestlings could not indirectly assess the food availability, because there was no clear di erence in feeding frequency per nestling, feeding amount per nestling, nor the nestling's quality between territories that di ered in their food availability. Nevertheless the timing of edging was delayed when the food supply within the territory was scarce. Thus, it seems that the major factors governing the timing of edging are parental factors, not nestling factors or parent-nestling interaction. One potential bene t for the parents of early edging is that they have a better chance of second brood, but neither the length of the nestling period nor the duration of post- edging parental care a ected the number of broods in our study area and this possibility was excluded. The expected bene t (B) of those parents that let their chicks edge at date t is given by the following formula: B ˆ B t L t 1 where B(t) is the expected reproductive success of those parents that succeeded in having edged young at date t, and L(t) is the probability of nestling survival at date t. Assuming a constant daily mortality rate (d), we obtain L(t) ˆ (1 ) d) t. We also assume a minimum date of
520 edging t 0, because nestlings are unable to edge earlier than a certain date due to physical constraints. Equation 1 can be rewritten as follows: B ˆ B 0 DB t 1 Š 1 d t 1 t 0 t 1 ˆ t t 0 2 where B 0 is a bene t value of those parents that have succeeded in raising young until the date t 0, and DB(t 1 ) shows additional bene t value that parents would have if they lengthen the nestling period for t 1 days. Since the observed di erence in caterpillar availability would a ect the feeding e ciency and, hence, the reproductive success, we can assume that B(t) is a function of territory quality (q). Then Eq. 2 becomes B ˆ B 0 q DB q;t 1 Š 1 d t 1 t 0 3 Fledglings need much more energy than nestlings for ying (Royama 1966; Moreno 1984) and it takes some time before they can feed by themselves. Hence the period soon after edging is the most costly for the parents, and edglings may have a higher risk of starvation. Because edglings in rich territories would be able to get food more easily than those in poor territories, db 0 (q)/ dq>0. Since DB(q, t 1 ) would increase either by increasing the nestling's fat reserves (Perrins 1965; Garnett 1981; Tinbergen and Boerlijst 1990) or by waiting for the growth of wing feathers (Gosler 1993), the bene t of higher fat reserves and mobility would be small in rich territories. That is, ddb(q,t 1 )/dq<0. Furthermore, ddb(q,t 1 )/dt 1 >0 and d 2 DB(q,t 1 )/dt1 2 < 0, because increases in fat reserves or wing growth would last only a few days after t 0. Under these conditions, we assume two simple functions for B 0 (q) and DB(q,t 1 ): B 0 q ˆf gq DB(q,t 1 ) ˆ h[1)exp()it 1 )]/q where f, g, h, i are constants and i >0. Substituting these equations into Eq. 3, we obtain B ˆff gq h 1 exp it 1 Š=qg 1 d t 1 t 0 4 Because 18 days was the average length of the nestling period, d is expressed as (1 ) d) ˆ 18p L where L * shows the survivorship from hatching to edging. The commonest predator at the study site is the blue-green snake Elaphe climacophora which takes whole broods. The predation by the snake was 6% in the rst brood of 1994, but rose above 63% in the late brood of 1995 (see Table 1). That is, d values di er greatly between breeding groups. The t 0 was assumed to be 15 days because 15 days was the shortest nestling duration observed in this research. The best timing of edging for parents is the point where the B value is maximized (db/dt 1 ˆ 0). We estimated the above mentioned four constants by the least-squares method, using the q, d, andt 1 data of 33 pairs in 1994 and 1995 (software STATISTICA for Macintosh ver4.1j, non-linear estimation). The results were f ˆ )1.71, g ˆ 0.0748, h ˆ 10.4, and i ˆ 0.721. Then we substituted these constants into Eq. 4 and calculated the t 1 that maximizes B at each q and d. The relationship between observed and predicted dates of edging is shown in Fig. 9. The observed date tted very well with the predicted dates and the regression coe cient between them was not signi cantly di erent from 1.0. Accordingly, we suggest that parent tits decide the timing of edging at the point where the two factors, predation risk before edging and additional improvement of nestling quality, are balanced. The larger increase of nestling quality expected in poor territories causes the negative relationships between nestling period and food availability. Two problems are left unsolved: the rst problem is how the parents control the timing of edging, and the second is how they know the predation risk in each season. E ect of food availability on the timing of independence Food availability soon after edging a ected the length of post- edging parental care; in ``poor'' territories with low food availability the edglings depended on their parents longer than those in ``rich'' territories. Some other studies have also reported the variation of the length of post- edging parental care (Kluyver 1951; Higuchi and Momose 1981). They have related the prolonged post- edging parental care to the low food availability. Davies (1976, 1978) noted that parental ``meanness'' reduced the pro t of the ``begging food'' strategy and o spring switched from ``begging food'' to ``feeding themselves'' when the self-feeding strategy became more pro table in terms of energy intake. At our study site, where food availability di ers between territories, edglings might di er in their choice of feeding strategy; edglings would have di culty in nding prey in poor territories and self-feeding would not be so Fig. 9 Predicted versus observed date of edging (r 2 ˆ 0.65, n ˆ 33). The regression coe cient between them was not signi cantly di erent from 1.0 (0.97 0.27; regression coe cient 95% con dence interval)
521 pro table as in rich territories. As a result, edglings in poor territories with low food availability might depend on their parents longer than those in rich territories. While the changes in edgling behavio alone are enough to explain the variation in post- edging parental care, parental behavio might also change depending on food availability; edglings in a poor territory would have di culty in nding food and be more dependent on their parents, and parents in a poor territory would decrease feeding frequency more slowly than in rich territories, because this would enhance edgling survival and, as a result, reproductive success of their parents. Acknowledgements We thank Tadashi Miyashita for helpful comments on an earlier draft of the manuscript, Masatoshi Yasuda for his advice on data analysis and model construction, and Hiroyoshi Higuchi, C. M. Perrins, Andre A. Dhondt, Erik Matthysen, Takasi Kagaya, Teruaki Hino and Shigeru Matsuoka for helpful discussion. We also thank Kazumasa Katagiri, Kimito Furuta and sta of Tama Forest Science Garden for encouragement and their support during the course of this study. References Cichon M, Linde n M (1995) The timing of breeding and o spring size in great tits Parus major. Ibis 137:364±370 Davies NB (1976) Parental care and the transition to independent feeding in the young spotted ycatcher (Muscicapa striata). Behaviour 59:280±295 Davies NB (1978) Parental meanness and o spring independence: an experiment with hand-reared great tits Parus major. Ibis 120:509±514 Dhondt AA, Kempenaers B, Adriaensen F (1992) Density-dependent clutch size caused by habitat heterogeneity. J Anim Ecol 61:643±648 Drent PJ (1984) Mortality and dispersal in summer and its consequences for the density of great tits, Parus major, at the onset of autumn. Ardea 72:127±162 Garnett MC (1981) Body size, its heritability and in uence on juvenile survival among great tits, Parus major. Ibis 123:31±41 Goodburn SF (1991) Territory quality or bird quality? Factors determining breeding success in magpie Pica pica. Ibis 133:85± 90 Gosler A (1993) The great tit. Hamlyn, London Higuchi H, Momose H (1981) Deferred independence and prolonged infantile behaviour in young varied tits, Parus varius, of an island population. Anim Behav 29:523±528 HoÈ gstedt G (1980) Evolution of clutch size in birds: adaptive variation in relation to territory quality. Science 210:1148±50 Kluyver HN (1951) The population ecology of the great tit, Parus major L. Ardea 39:1±135 Lack D (1958) A quantitative breeding study of British tits. Ardea 46:91±124 Lack D (1966) Population studies of birds. Oxford University Press, Oxford McCleery RH, Perrins CM (1988) Life time reproductive success of the great tit, Parus major. In: Clutton TH (ed), Reproductive success. The University of Chicago Press, Chicago, pp 136±153 Moreno J (1984) Parental care of the edged young, division of labor, and the development of foraging techniques in the northern wheatear (Oenanthe oenanthe L.). Auk 101:741±752 Noordwijk AJ van, Balen JH van (1988) The great tit, Parus major. In: Clutton TH (ed), Reproductive success. The University of Chicago Press, Chicago, pp 119±135 Noordwijk AJ van, McCleery RH, Perrins CM (1995) Selection for the timing of great tit breeding in relation to caterpillar growth and temperature. J Anim Ecol 64:451±458 Nur N (1986) Is clutch size variation in blue tit (Parus caeruleus) adaptive? An experimental study. J Anim Ecol 55:983±999 Perrins CM (1965) Population uctuations and clutch size in the great tit Parus major L. J Anim Ecol 34:601±647 Perrins CM (1979) British tits. Collins, London Perrins CM (1991) Tits and their caterpillar supply. Ibis 133:49±54 Perrins CM, McCleery RH (1989) Laying dates and clutch-size in the great tit. Wilson Bull 101:236±253 Royama T (1966) Factors governing feeding rate, food requirement and brood size of nestling great tit Parus major. Ibis 108:313± 347 Tinbergen JM, Boerlijst MC (1990) Nestling weight and survival in individual great tit (Parus major). J Anim Ecol 59:1113±27 Zandt HS (1994) A comparison of three sampling techniques to estimate the population size of caterpillars in trees. Oecologia 97:399±406