An alternative approach to the measurement of seasonal trends in bird breeding success: a case study of the bullfinch Pyrrhula pyrrhula


Dr I. Newton, Institute of Terrestrial Ecology, Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire PE17 2LS, UK. Tel: 01487 773381. Fax: 01487 773467. E-mail:


1. On the basis of juvenile:adult ratios among netted birds and studies of moult, inferences are drawn about seasonal and annual variation in the breeding output of a bullfinch population studied near Oxford, in southern England.

2. In most years during 1962–66, the juvenile:adult ratio among netted birds increased from July or August to October, and then declined. The ratio in October, taken as a measure of annual productivity, varied between 2·8 and 5·5 in different years.

3. In the Oxford area, egg-laying among bullfinches occurred between mid-April and late August, so that the last young of the year left the nest in late September/early October. However, not all pairs continued breeding this late in the year. Individual bullfinches started to moult when they finished breeding. The proportion of adults which started moult after 20 August (implying successful breeding from eggs laid after mid-July) varied between 7 and 68% in different years (or between 19 and 63% using a different method of estimation). Juveniles from these late nests, identified as those starting moult after 31 August, formed 24–61% of the total juvenile population in different years, the rest having been produced from clutches begun before mid-July.

4. In the 2 years when the greatest proportions of adults started to moult after 20 August and the greatest proportion of juveniles started after 31 August, the juvenile:adult ratios in October were highest. By implication, the two-fold variations in overall productivity between years were associated primarily with variations in the amount of late breeding.

5. This study, based on moults and age-ratios among netted birds, provided information on seasonal productivity that was practically impossible to obtain in this species by conventional nest finding. An implication was that an interaction between the seasonal pattern of breeding in particular years and a seasonal trend in predation on nest contents, had major effects on overall annual productivity.


The numbers of some seed-eating bird species in Britain and other parts of western Europe have declined markedly during the past 20 years (Marchant et al. 1990; Gibbons, Reid & Chapman 1993; Fuller et al. 1995; Baillie et al. 1998; Siriwardina et al. 1998). This finding has promoted interest, not only in the environmental changes causing the declines, but also in the demographic processes involved. The breeding rates of cardueline finches are especially difficult to study. In southern England, these birds can raise two or three (theoretically up to four) broods in a breeding season. However, pairs do not necessarily raise successive broods in the same location, tending to settle for each nest wherever suitable seeds are plentiful at the time. This itinerant behaviour makes it difficult to measure reliably the seasonal production of individual pairs (as opposed to individual nests) and, hence, to assess the annual reproductive rate of a population. An additional problem is that nests in late season (July–September), when cover has thickened, are extremely hard to find. The bullfinch Pyrrhula pyrrhula L. is perhaps less prone to move long distances between successive nests than are other cardueline finches, but because the pairs are quiet and secretive when breeding, they are more difficult to locate.

In this paper, I assess the annual productivity of bullfinches near Oxford in five successive years (1962–66), from the ratio of juveniles to adults netted in different months, using the state of the moult in birds of both age-groups to estimate the seasonal pattern in productivity. So far as I am aware, such a procedure, based on both moult and age ratios in captured samples, has not been used previously, but could be of wider application in species whose breeding rates are otherwise hard to measure. A major implication of the findings, of wider significance, is that an interaction between the seasonal pattern of breeding in different years and a seasonal trend in predation rates on nest contents, can underlie major variations in the annual production of young.

Background and methods

The work was done at Wytham, a wooded estate near Oxford, in south-central England (lat. 51°45′N, long. 1°20′W). Each year, I caught bullfinches in most months by placing mist nets at drinking pools or at feeding sites where favoured seeding herbaceous plants grew close to cover. Such sites included patches of sow-thistle Sonchus oleraceus L. (mainly July–August), meadowsweet Filipendula ulmaria (L.) (mainly August–October), nettle Urtica dioica L. (mainly September–December), bramble Rubus fruticosis L. (mainly September–February) and broad-leaved dock Rumex obtusifolius L. (mainly November–March). Where the birds fed on tree seeds, such as silver birch Betula pendula Roth or ash Fraxinus excelsior L., they were attracted to mist-net level by placing bunches of seeding dock stems on low bushes or on the ground nearby. In these ways, it was possible, in the 400 ha of woodland, to operate several net sites simultaneously, and to catch several hundred bullfinches over the period July–February each year and smaller numbers in March–June.

Bullfinches were resident in the Oxford area year-round, inhabiting woods, tall overgrown hedgerows and shrubby gardens. I caught many individuals several times each year, over several years, with most successive captures in the same general area. Among the national ring recoveries, most bullfinches that were reported had moved less than 5 km from the ringing site (usually natal site) and no evidence was found for regular migration (Summers 1979).

In most years in the Oxford area, egg-laying in the population occurred from mid-April or early May, and extended into August, so that the last young left the nest in early September (see also Greig-Smith & Wilson 1984). In one year (1962), however, when food in late summer was especially plentiful, laying extended beyond mid-August, so that the last young left the nest in early October (Newton 1972). For each successful breeding attempt, egg-laying usually lasted 4–6 days (one egg per day), incubation 12–14 days, the nestling period 16–17 days and the post-fledging period (when young were fed by parents outside the nest) 17–18 days. Hence, the full breeding cycle from egg laying to independence of young lasted at least 50 days and the young became independent at about 35 days old (see also Nicolai 1956). However, there was sometimes overlap between successive broods, if the female started another nest in the post-fledging period, while the male fed the young (Doerbeck 1963; Newton 1972; Bijlsma 1982). One pair was found to raise three broods with intervals between the first eggs of successive broods of 33 and 35 days (Newton 1972). In theory, therefore, some pairs might have had time to raise up to three broods in a season when egg-laying occurred from mid-April to mid-August, and perhaps up to four in a season when egg-laying continued to mid-September. In practice, however, not all pairs extended laying into August (producing their last eggs in June or July), and many nesting attempts failed through predation (see later). After a loss of nest contents, bullfinches in Germany were found to start another clutch within a week (Ringleben 1960), and in studies in Germany and the Netherlands, pairs typically laid 3–5 clutches in a season, but raised only two broods (Nicolai 1956; Bijlsma 1982). These latter studies did not involve marked pairs, however, and allocations of nests to particular pairs were based on the locations and timings of nesting attempts.

In the bullfinch, as in many small birds, moult follows breeding (for full details, see Newton 1966). The adults start anytime from late July to early September (or early October in 1962), depending on when their breeding stops, and each individual takes 10–12 weeks to replace all its feathers. Individual adults may begin their moult while still feeding their last brood of young, the overlap being greatest in those individuals which breed latest in the year (Newton 1966). Juveniles replace only their body feathers, retaining the juvenile flight and tail feathers for another year. They moult over the same period of the year as adults, the juveniles from the earliest nests starting before those from the latest nests, with each individual taking 7–9 weeks (Newton 1966).

For purposes of study, adults were given a score from 0 to 45, depending on the state of moult in the primary feathers, the replacement of which spanned the entire moult period in most individuals. For most of the moult, this ‘primary score’ increased linearly with date, enabling the date when moult started to be estimated, usually to within 5 days, by backdating (for details, see Newton 1966, 1967). Thus, the onset date of moult for each adult was estimated by working back at 0·65 points per day from its score on the date of capture, except for those birds caught at the beginning of moult with only the first primary growing, when a rate of 0·2 points per day was assumed. These rates were calculated from birds caught more than once during moult (Newton 1966). For juveniles, which do not replace their primary feathers, body moult was divided into five stages of roughly equal duration (each of 9–12 days, for details, see Newton 1966). This procedure again allowed the starting dates for each individual caught in moult to be estimated by backdating, although less accurately than for adults. For purposes of analysis, starting dates of both age classes were grouped into 10-day periods. Most birds were caught in the early stages of moult and for no individuals should the estimates of starting date be in error by more than one 10-day period. The starting dates of moult were used as estimates of the dates that individual adults finishing breeding, or that individual juveniles left the nest, thus providing a retrospective measure of aspects of each season's breeding.

The validity of backdating as a measure of the timing of moult onset depended on moult rates being roughly constant from bird to bird, with no systematic variation associated with year or with onset date within a year. No such systematic variation was found (Newton & Rothery 1999), so for purposes of backdating, moulting rates were assumed to be constant between individual adults or between individual juveniles, regardless of when they moulted.

Additional estimates of moult rate and timing were obtained using a model for avian moult proposed by Underhill & Zucchini (1988), in which all adults caught during the moulting season were placed in one of three classes: (a) moult not started, (b) moult in progress and (c) moult finished (their Type 1 method). Estimates of moult rates and start dates (mean and standard deviation) were obtained from changes in the proportions of these three classes of birds through the season (for details, see Newton & Rothery 1999). Because the method was sensitive to change in the composition of the underlying population through the season, it could be used only for adults. It could not be applied to juveniles, because freshly fledged (and unmoulted) juveniles were continually being produced during the first half of the moulting season, thereby altering the ratio of non-moulting to moulting birds in the population.


Juvenile:adult ratios

With at least some bullfinches starting egg-laying in late April/early May, the first fledglings of the season could have appeared by early June. In fact, no juveniles were seen or heard in Wytham Woods in any year until July, and until around mid-July net catches consisted entirely of adults (Appendix 1). The earliest date that an independent juvenile was caught in a mist net was 13 July. From then on, the proportion of juveniles in netted samples increased, usually reaching a peak in October, before declining (Appendix 1, Fig. 1). A decline in the juvenile:adult ratio after October was evident in 4 out of 5 years, but in 1964, the ratio in November–December was even higher than in October. This was associated with substantial local movement in that autumn and a net immigration of newly-moulted juveniles (‘first-winter birds’) into Wytham (as indicated by a sudden increase in the proportion of unringed among netted birds). As this event was evident only in one year, the ratio of juveniles to adults in October, after all breeding stopped, was taken as an index of overall productivity in each year. This ratio ranged from 2·8 to 5·5 in different years and implied that overall productivity varied by two-fold from year to year.

Figure 1.

Percentage of juveniles in monthly samples of bullfinches caught in mist nets at Wytham, near Oxford 1962–66, based on the pooled data from different years. Numbers of different birds caught each month are shown above the columns and the figures for individual years are given in Appendix 1. Almost all the birds for July were caught in the latter half of the month.

The juvenile:adult ratio in August, reflecting breeding output in May–July, also varied substantially from year to year (0·5–1·4), as did the ratio in September (2·1–5·1), reflecting output in May–August. However, a relatively high juvenile:adult ratio in August was not necessarily associated with a relatively high ratio in September or October. The only consistent finding from year to year was that the ratio increased from August to October (more than tripling overall). For the ratios quoted here, each bird was counted only once in each monthly sample, but some individuals were caught more than once per month. Inclusion of multiple captures altered the juvenile:adult ratios slightly, as shown in Appendix 1, but made no difference to the conclusions.

Dates of onset of moult in adults

In 1962 the start of moult in adult bullfinches was spread over the 11 weeks from late July to early October, and in the following 4 years over the 7 weeks from late July to early September (Fig. 2). Spread in the date of onset of moult resulted from individual variation in the dates that breeding activity finished. From late July onwards, pairs started to moult as their last young left the nest. The greater spread in the dates of onset of moult in 1962 was associated with the more prolonged breeding activity in that year.

Figure 2.

Percentage of bullfinches which started to moult at different dates in 1962–66. Data presented in 10-day periods. The records for 1962–64 are fewer than those presented in Newton (1966) for these years because of the exclusion from the present data of estimates for juveniles caught before they had started to moult. In addition, revised estimates were made for some individuals based on improved information on moult rates.

In each year, many adults caught in the early stages of moult still had dependent young. Some adults were seen with their young just before they were netted, while others on capture were found to have food in their throat pouches (an indication of breeding). The number of adults at different stages of moult which were known to have dependent young is shown in Table 1 based on the combined data for all 5 years. The totals provide minimum estimates for the proportions of adults caught that had parental duties at the time. Thus, at least 42% of all premoulting adults caught in August–October had dependent young, compared with 27% in the first stage of moult (scores 1–5), 17% in the second stage (scores 6–10), 12% in the third stage (scores 11–15) and none at any later stage. The findings were consistent with moult beginning at about the time the last young left the nest, even though the young had then to be fed for more than another 2 weeks. Similar overlap between the end of breeding and the start of moult has been noted in some other passerine species, being especially marked in late breeding individuals (Marler 1956; Eaton 1957; Dixon 1962; Evans 1966; Seel 1976; Bancroft & Woolfenden 1982).

Table 1.  The numbers of adult bullfinches at different stages of moult that were known to have dependent young. The table does not include birds at later stages of moult when none had dependent young
Stage of moult
Stage 1
(score 1–5)
Stage 2
(score 6–10)
Stage 3
(score 11–15)
Stage 4
(score 16–20)
  • *

    Birds were known to have dependent young only if they were seen feeding them prior to capture or if they had food in the throat pouch. The figures, which are based on all adults at appropriate stages caught in August–October all years, are therefore minimum figures.

Number caught5388424222
Number (%) feeding young*22 (42)24 (27)7 (17)5 (12)0 (0)

Mid-July was the mid-point in a potential 6-month breeding season, mid-April–mid-October. From the starting dates of moult in the population (Fig. 2), the percentage of adults that began a successful nest after mid-July (thereby delaying the start of moult beyond 20 August) could be estimated by direct backdating from moult scores at 68, 14, 16, 7 and 28% in the 5 years (Table 2), or on the Underhill–Zucchini model at 63, 33, 33, 19 and 43% in the 5 years. The two methods would not be expected to give the same values because, while the first was based on backdating from individuals caught in moult, the second was a maximum likelihood estimate based on larger samples that included non-moulting birds (see Newton & Rothery 1999). None-the-less, estimates from the two methods were correlated with one another (r = 0·97, P < 0·01), and both methods gave the lowest percentages in 1965, and the highest in 1962 and 1966.

Table 2.  The proportions of late moulting adults and juveniles in different years in relation to the juvenile:adult ratio among netted bullfinches in October
YearNumber of adults
caught in moult*
Percentage of
adults that
started moult
after 20 August
Number of juveniles
caught in moult*
Percentage of
juveniles that
started moult
after 31 August
ratio in October
  • *

    Figures give the total number of birds caught in moult each year, throughout the moult season (July–November/ December).

  • †Reflecting the proportion of clutches started after mid-July. The first figure gives an estimate based on backdating from individual moult scores, and the second figure (in parentheses) gives an estimate based on the Underhill & Zucchini (1988) model for moult data (Type 1) which includes records of non-moulting birds (see Newton & Rothery 1999).

  • For full details, see Appendix 1.

  • Correlation between percentage of late adults and late juveniles in different years: r = 0·896, P < 0·02, (on Underhill-Zucchini estimate for adults, r = 0·857, P < 0·06).

  • Correlation between percentage of late adults and juvenile: adult ratio: r = 0·856, P < 0·05, (on Underhill-Zucchini estimate for adults, r = 0·825, P < 0·09).

  • Correlation between percentage of late juveniles and juvenile: adult ratio: r = 0·939, P < 0·01.

19622268 (63)82615·50
19639614 (33)171243·24
19648316 (33)125332·84
1965677 (19)118323·47
19668028 (43)127504·90

Dates of onset of moult in juveniles

Juveniles which began moult at the start of the moulting season, in late July, must have been out of the nest for 3 weeks or more (and, hence, independent of parental care), because all their feathers were fully grown. Juveniles which moulted later in the season sometimes started before they were 3 weeks out of the nest, and before their flight and tail feathers were fully grown. Some were seen to be fed by adults or heard to make begging sounds. The implication was that late season young started to moult at a later date, but at an earlier age, than early season ones.

In the population as a whole, young bullfinches could have left the nest anytime in a 20-week period from late May to late September/early October, but starting dates of moult were compressed into a shorter 15-week period from late July to early November (Fig. 2). Nonetheless, moulting dates were broadly correlated with fledging dates and, in general, juveniles which left the nest earliest in the season started moulting first (Fig. 3). In fact, the first juveniles to moult each year started before the latest had even hatched.

Figure 3.

Relationship between the dates of fledging (= nest-leaving) and the start of moult in individual juvenile bullfinches. The diagram is based on 10 birds ringed as nestlings and caught later in the same year in moult (filled circles), and on six other birds caught in moult before they had finished growing their juvenile flight and tail feathers (open circles). In the latter, fledging dates were estimated from feather length, assuming that young left the nest with feathers half grown and that feathers grew at about 2·5 mm per day, to reach their final length (outer primary) of about 58 mm (as found in captive bullfinches, Newton 1967). As the last birds moulted unusually early for their age, their effect was to reduce the slope of the relationship from that expected if other birds of their age could have been included (see text). Overall relationship: moulting date = 127 + 0·53 fledging date, r2 = 0·88, P < 0·001, day 1 = 1 April.

The precise relationship between the dates of fledging (= nest leaving) and moult was examined for two groups of juveniles. The first consisted of 10 birds which were ringed in the nest at about 10 days of age and then subsequently caught in moult, enabling the dates of both their fledging and moult onset to be estimated. The second group consisted of six late young caught with incompletely grown flight and tail feathers that had just begun their body moults. For these birds, fledging dates were estimated from the length of the outer primary feathers (for growth curves see Newton 1967), assuming that these birds left the nest when their feathers were about half grown. None of the birds in this second group had passed beyond stage 1 of moult at the time, indicating only slight overlap between the growth of juvenile flight feathers and the start of the post-juvenile body moult (most late young showed no such overlap). Juveniles from the two groups gave the pattern depicted in Fig. 3, in which different individuals moulted in roughly (but not exactly) the order in which they had left the nest. The effect of including the second group of birds, which moulted unusually early for their age, was to lower the slope of the calculated regression line (0·53) from the slope based on the first group alone (0·75). Other birds ringed as nestlings and caught before they had started moult, or after they had finished, fitted the same general relationship between fledging date and moulting date. Similar correlations between the age of the juveniles and the dates of their post-juvenile moults have been noted in some other passerines (Michener & Michener 1940; Selander 1958; Snow 1958).

Relationship between late breeding and overall annual productivity

From figures given earlier, bullfinch pairs which produced eggs after mid-July were expected to start their moults at about the time their last young left the nest, say from about 20 August onwards, while the young themselves were expected to start moult from early September onwards. For the present purposes, therefore, the amount of late breeding each year was assessed as the proportion of all adults examined that started moult after 20 August and as the proportion of juveniles that started after 31 August. These proportions varied greatly from year to year (as much as 7–68% in adults, 24–61% in juveniles) and in both age-groups were greatest in 1962, followed by 1966 (Fig. 2, Table 2). The proportions in the other 3 years were lower and varied inconsistently between age-groups. The 2 years with most late breeding were also the years with the greatest juvenile:adult ratios among birds netted in October. By implication, the amount of late breeding in different years had a big influence on overall annual productivity in those years. In 1962, when an estimated 63–68% of adults raised young from eggs laid after mid-July, twice as many young were produced overall than in 1964 when only 16–33% did so (the figures depending on the method of estimation).


Possible bias in juvenile:adult ratios

The value of the post-breeding juvenile:adult ratio as a measure of seasonal productivity depended on how closely the ratio in the trapped sample reflected the ratio in the population at large. Mist-netting is probably a less biased method of catching songbirds than is any other method available, but I had no way of assessing whether juveniles were more or less likely to be caught than adults. Whether biased or not, there was no reason to suppose that the ratios were not comparable between years, and the juvenile:adult ratios in October were consistent with those expected from known breeding success. To produce a ratio of 2·8–5·5 young per adult, each pair would have to raise to independence an average of 5·6–11·0 young per year, a total that could have been provided by an average of two or three successful broods per pair per year. Moreover, the highest ratio of 11·0 young per pair occurred in 1962, when breeding was most extended, time enough in total for some individuals to fit in an extra brood. The fact that overall productivity measured in October in different years was closely related to the timing of moult (reflecting the end of the breeding season), implied that the juvenile:adult ratio gave a realistic measure of the variation in productivity between years. Similar data were obtained from another area in Kent, where Greig-Smith & Wilson (1984) found among netted bullfinches a juvenile: adult ratio in September–October of around 3:1. This ratio was based on the combined totals for 2 years (1980 and 1981), but was within the range of values found in Wytham during 1962–66. These authors also noted a decline in the age-ratio after October, which they attributed to juveniles suffering greater mortality than adults. Other published data, giving a juvenile: adult ratio in bullfinches of 0·61 in 1995 and 1996 are not comparable with my data because they are based on the combined records for May–August, omitting the next 2 months when the ratio would be expected to peak (Balmer & Peach 1997). In the overall Wytham data for 1962–66, the juvenile: adult ratio increased more than three-fold from August to October. Clearly, if juvenile:adult ratios are to give a reliable measure of annual productivity, they are best taken at the end of the breeding season, when any within-season variations in nest success and brood numbers have taken effect.

In the bullfinch, high juvenile to adult ratios in October were followed by relatively high ratios for the whole of the next year, although the ratios declined as the year progressed (Newton 1999). In other words, the effect of the amount of late-breeding on the age composition of the population was apparent for at least the next 12 months, as long as the two age-groups could be distinguished.

Possible bias in moulting dates

The amount of late breeding in 1962, as judged from the moult dates of both adults and juveniles, may have been over-estimated. This was because, in 1962, I did not start catching bullfinches until 12 September. Hence, in this year, adults that started in late July had only 3–5 weeks (juveniles less than 3 weeks) in which to be caught in moult, while adults that started after 12 September had the full 10–12 weeks of moult (juveniles 7–9 weeks) in which to be caught. This bias, in favour of late moulters, should not have affected the extended spread of starting dates observed in 1962, but could have led to over-estimation of the proportions of birds moulting late in that year. In addition, the sample of moulting adults (n = 22) was small in 1962. In the remaining years, I started netting activities well before the earliest bullfinches began to moult in late July and continued (as in 1962) until well beyond the latest ones had finished. Hence, in 1963–66, birds starting at different dates in the season should have had similar opportunities to be caught, giving a more representative picture of the moulting season as a whole.

Use of moult dates to assess the seasonal pattern of productivity

The fact that many juveniles started moulting in the latter half of the moulting season implied that many had been produced in the latter half of the breeding season, even though most adults had stopped breeding by then (except in 1962). This finding was consistent with the scarcity of observations of juveniles before July and with their absence from netted samples before the middle of that month. The few observations of dependent and independent juveniles before July were in farmland and gardens, and not in woodland. The finding was also consistent with known seasonal trends in nest success in the Oxford area. In gardens and farmland, some 44% of nests started in May produced young old enough to ring (at 10 days), compared with 67% of nests started in June and 84% in July–August. Almost all the failures were attributable to predation of eggs and young. The comparable figures for woodland, where predation rates were higher, were 15, 50 and 70% (Newton 1972). Moreover, these figures gave maximum estimates of nest success, because some of the broods concerned might have been taken by predators after they had been ringed. Similar seasonal trends in nest success were found in an area of the Netherlands, with figures for the proportion of eggs (rather than nests) that produced young at 39% for April, 46% for May, 53% for June, 60% for July and 67% for August (Bijlsma 1982). Hence, the fact that most young caught in moult seem to have been produced in the latter half of the breeding season could be attributed partly or entirely to the greater success of later nests. Bullfinches nest in bushes and the seasonal trend in success was, in turn, consistent with the general thickening of shrubby vegetation and growth of herbaceous vegetation in late summer, which could have made nests increasingly difficult for predators to find.

Patterns of breeding and survival

Late breeding in bullfinches could have arisen in two different ways: (1) by some birds extending the length of their breeding season, and thereby raising an additional brood in certain years; or (2) by individuals having a similar-length season, but some starting and finishing much later than others in certain years. Because nest success was so much greater late in the season than earlier, either of these processes could produce the results observed, with a higher proportion of late moulting birds and a higher October juvenile to adult ratio, in some years than in others. In practice, it would have made no difference to the final outcome whether birds started early in the season and failed (having successful nests later) or whether some delayed starting until later in the season when successful breeding was more likely. The key point was that, the greater the proportion of pairs that raised broods late in the season, the greater the ratio of juveniles to adults in October, after all breeding had finished.

Because juveniles from early broods were older by the time of moult than juveniles from late broods, the early juveniles could also have suffered greater mortality before moult, reducing their proportion (relative to younger juveniles) in the population in late summer. In addition, any variation between years in the extent of differential mortality between juveniles and adults could also have affected the age ratios prevailing at the end of breeding. Neither possibility could be excluded, but food for bullfinches was extremely plentiful in summer, and their main predator, the sparrowhawk Accipiter nisus (L.), was absent from the Oxford area in the years of study. It would therefore be surprising if much mortality occurred among independent young at that time. However, whether it did or not, the age ratio in October presumably depended on all the factors that affected productivity and survival over the preceding months, among which the amount of late breeding was paramount.


Studies of moult among netted bullfinches during August–October seemed to give a reliable measure of the seasonal pattern of breeding within years and the juvenile to adult ratios in October of the year-to-year variation in overall productivity. The data emphasize the great contribution to overall productivity of late clutches, begun after mid-July and after many adults had finished breeding that year. Annual variations in the amount of late breeding largely accounted for annual variations in overall productivity, although the young from late nests may not have survived the winter as well as young from early nests (Newton 1966). These data were collected in the 1960s, when bullfinch numbers in southern England were probably higher than at any other time this century and well before their recent decline. They should therefore provide a sound comparative basis for contemporary and future measures of breeding success obtained in the same way.

A more general implication of the findings is that a marked seasonal pattern in predation of nest contents could impose strong selection pressures on the timing of breeding, over and above any influence of other potentially important factors, such as food-supply. Most studies on the timing of breeding in birds have involved species that nest in tree-cavities (= nest-boxes) or other protected sites, in which predation is low or non-existent. They have generally shown that, within years, early nests are most productive of both fledglings and of future recruits to the breeding population (e.g. Perrins 1970; Daan et al. 1989; Dijkstra, Daan & Baker 1990; Verhulst, van Balen & Tinbergen 1995; Verboven & Visser 1998). Species whose nest-contents are subject to a downward seasonal trend in predation rate, such as many of those that nest in accessible sites (e.g. Snow 1955; Newton 1964), might be expected to show different seasonal patterns of fledgling production and recruitment, as well as more marked annual variation in productivity associated with annual variation in the timing and duration of breeding.


I am grateful to Drs Rhys Green, Jeremy Greenwood, Shelley Hinsley and an anonymous referee for helpful comments on an earlier manuscript.

Received 15 May 1998;revisionreceived 16 September 1998


Appendix 1

Table 3. Numbers of adult and juvenile bullfinches caught in different periods 1962–66. The main figures show the numbers of different individuals caught in each period and the figures in parentheses show the total number of captures (including recaptures) in those periods. A, adults; J, juveniles; R, number of juveniles per adult; –, no data.  
196210 (10)51 (56)5·10 (5·60)12 (12)66 (70)5·50 (5·83)31 (34)109 (115)3·52 (3·38)
196323 (27)0 (0)29 (29)13 (13)0·45 (0·45)82 (97)94 (104)1·15 (1·07)38 (38)117 (119)3·08 (3·13)21 (21)68 (70)3·24 (3·33)36 (42)87 (95)2·42 (2·26)
196445 (47)0 (0)8 (8)1 (1)0·13 (0·13)55 (61)78 (89)1·42 (1·48)28 (28)78 (89)2·72 (3·10)43 (44)122 (143)2·84 (3·25)75 (80)277 (358)3·69 (4·48)
196541 (41)0 (0)11 (12)22 (23)2·00 (1·92)49 (54)83 (99)1·69 (1·83)30 (31)62 (72)2·07 (2·32)19 (20)66 (73)3·47 (3·65)50 (51)82 (98)1·64 (1·92)
19664 (4)3 (3)0·75 (0·75)63 (69)34 (38)0·54 (0·55)37 (38)94 (119)2·54 (3·13)20 (20)98 (98)4·90 (4·90)26 (26)65(65)2·50 (2·50)
Overall109 (115)0 (0)52 (53)39 (40)0·75 (0·75)249 (281)289 (330)1·16 (1·17)143 (145)402 (455)2·81 (3·14)115 (117)420 (454)3·65 (3·88)218 (233)620 (731)2·84 (3·14)