Intraspecific variation in avian pectoral muscle mass: constraints on maintaining manoeuvrability with increasing body mass

Authors

  • MAURINE W. DIETZ,

    Corresponding author
    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750 AA Haren, The Netherlands;
    Search for more papers by this author
  • THEUNIS PIERSMA,

    1. Animal Ecology Group, Centre for Ecological and Evolutionary Studies, University of Groningen, PO Box 14, 9750 AA Haren, The Netherlands;
    2. Department of Marine Ecology and Evolution, Royal Netherlands Institute for Sea Research (NIOZ), PO Box 59, 1790 AB Den Burg, Texel, The Netherlands; and
    Search for more papers by this author
  • ANDERS HEDENSTRÖM,

    1. Theoretical Ecology, Lund University, Ecology Building, SE-22362 Lund, Sweden
    Search for more papers by this author
  • MAARTEN BRUGGE

    1. Department of Marine Ecology and Evolution, Royal Netherlands Institute for Sea Research (NIOZ), PO Box 59, 1790 AB Den Burg, Texel, The Netherlands; and
    Search for more papers by this author

†Author to whom correspondence should be addressed. E-mail: M.W.Dietz@rug.nl

Summary

  • 1Within a single year, long-distance migrants undergo a minimum of four cycles of fuel storage and depletion because their migrations have at least one stopover. Each cycle includes an almost twofold change in body mass (mb). Pervasive predation threats beg the question whether escape flight abilities keep up with such large changes in mb.
  • 2We derive aerodynamic predictions how pectoral muscle mass (mpm) should change with mb to maintain constant relative flight power.
  • 3We tested these predictions with data on red knot Calidris canutus, a long-distance migrating wader that breeds in arctic tundra and winters in temperate and tropical coastal areas. We focused on the subspecies C. c. islandica.
  • 4mpm varied with mb in a piecewise manner. In islandica knots with mb ≤ 148 g, the slope (1·06) was indistinguishable from the prediction (1·25). In heavy knots (mb > 148 g) the slope was significantly lower (0·63), yielding a mpm 0·81 times lower than predicted at pre-departure weights (210 g).
  • 5Manoeuvrability tests showed that above 160 g, knots were increasingly unable to make a 90° angle turn. This is consistent with mpm being increasingly smaller than predicted.
  • 6Relatively low mpm enables savings on mass and hence flight costs, and savings on overall energy expenditure. We predict that reduced escape flight ability at high mb will be compensated by behavioural strategies to minimize predation risk.

Introduction

Powered flight becomes more costly when the flying object gains mass (Pennycuick 1975). So, it is no surprise that birds and bats show weight-saving design features. Flying animals are predicted to adaptively and reversibly change the mass of most body parts, a phenomenon called phenotypic flexibility (Piersma & Lindström 1997; Piersma & Drent 2003). The largest avian organ is the pectoral muscle complex (on average 17·1% of body mass, and even up to 25% in hummingbirds; Greenewalt 1962) and we may assume that there is strong selection for these muscles to be as light as possible. The pectoral muscles power the movements of the wings during flight and may have an additional role as a nutrient (protein) store, especially during migration (Battley et al. 2000; Bauchinger & Biebach 2001; Schwilch et al. 2002). Long-distance migrants must store large amounts of nutrients to be able to migrate successfully, a necessity in conflict with the minimization of body mass (mb). The question is how this conflict affects pectoral muscle size and the related flight performance in migrating birds.

Red knots Calidris canutus, henceforward simply called knots, are long-distance migrating waders that twice a year cover distances of 5000–15 000 km between their high arctic tundra breeding grounds and their temperate or tropical coastal wintering areas. Depending on subspecies, these distances are covered in two or more flights (Piersma et al. 2005). Hence, in their annual cycle, knots go through at least four cycles of fuelling and depletion (Fig. 1), during which mb varies almost twofold. Knots adjust their organ sizes apparently adaptively in relation to the stopover/flight cycles (Piersma et al. 1999a; Piersma, Gudmundson & Lilliendahl 1999b). This is also the case for the pectoral muscles, which we know are structures of flexible size (Lindström et al. 2000); their size appears endogenously regulated on a seasonal basis (Dietz, Piersma & Dekinga 1999b). The large intraspecific and intraindividual variation in pectoral muscle mass (mpm) and mb in knots offers the possibility to quantitatively predict the change in mpm as a function of change in mb and compare them with empirical data. Using the aerodynamic theory we here derive a prediction of the intraindividual relationship between mpm and mb for birds that do not otherwise vary in size. This prediction is compared with dissection data of adult knots accumulated over 21 years. We focus on the subspecies C. c. islandica (hereafter called islandica) and compare the results with smaller data sets from four of the five other knot subspecies.

Figure 1.

Life cycle stages of red knot Calidris canutus islandica. Abbreviations represent the following life phases: Fi, fuelling period i, PDi, pre-departure period i, FLi, migration flight i, and Ai, just arrived after migration period i. The nongrey-coloured phases, including wintering, are the phases of which dissection data of islandica are available. In subsequent graphs, the same colours are used to indicated these phases. Months are indicated (approximately) in the inner circle.

In addition to the conflict between being as light as possible and storing enough fuel to migrate successfully, knots have to deal with a pervasive threat of predation. Since birds of prey are the main predators at most fuelling areas, migrant waders depend strongly on their take-off ability, acceleration and manoeuvrability to escape from their avian predators (Piersma et al. 1993a; Ydenberg et al. 2004). Maintaining optimal escape flight performance is thus an essential part of their predator evasion strategy. This begs the question whether escape flight ability keeps up with the large changes in mb and mpm. To investigate this, we followed the change in manoeuvrability (i.e. the ability to make a 90° angle turn in an enclosed space) with a natural increase in mb in 18 captive knots during their spring fuelling period. During this period an increase in mb is shown even in captivity (Weber & Piersma 1996; Piersma 2002; Selman & Evans 2005). To determine if the change in manoeuvrability was related to mpm, pectoral muscle thickness was measured using ultrasonography when most birds were near peak mb.

Aerodynamical prediction

In flapping flight, the power required to fly at a particular speed increases with body size at a rate inline image(for an ‘ideal bird’sensu Pennycuick 1975). Hence, the pectoral muscles that provide power should increase with power requirements. However, this relationship in fact increases among a series of isometrically scaled species and cannot be applied to the within-individual bird for changes of power requirements due to changed mass. When deriving how the variability in mpm is required to sustain aerobic flight within individuals, we will assume that the added mass of fuel and pectoral muscle affects the body thickness, but not body length. The body frontal area (Sb) should therefore change in direct proportion to the added mass as Sb ∝m within an individual (cf. Hedenström 1992). It is also assumed that wing morphology (shape and wing span) remains unaffected by tissue accumulation. For an ideal bird, the power components considered are induced and parasite power only (hence disregarding profile power of the flapping wings), which are due to generating lift and overcoming body drag, respectively. For such an ideal bird the power required to fly (P) is (eqn 11 in Pennycuick 1975)

image( eqn 1)

where g is acceleration due to gravity, A is the equivalent flat-plate area, ρ is air density and Sd is the wing disk area. We assume that g and ρ are constant and hence do not affect the scaling of power in relation to added mass. Equivalent flat plate area is A = SbCDb, where CDb is the body drag coefficient. CDb depends on the Reynolds number Re (Re = Ul/ν, where U is flight speed, l is a characteristic length and ν is kinematic viscosity; Pennycuick 1989). Re and hence CDb varies among species but will remain relatively constant within a species, as the characteristic length (body length in the stream-wise direction) is a constant. Therefore, A will scale as Sb. The wing disc area Sd (= πb2/4) depends on the wing span (b), which is unaffected by added mass. Simplifying proportionality (eqn 1) according to our assumptions for within individual mb change, means that the power required to fly should vary as inline image

This model neglects the profile power required to overcome the drag of flapping wings and inertial power that is required to accelerate the wings at each stroke. Profile power is low at slow speed and relatively constant at cruising speeds (Pennycuick 1975), and it should not affect the power required to fly in relation to body mass. Inertial power is usually ignored at cruising speeds, while it only contributes a small fraction of the total power at slow speeds (Norberg 1990). Inertial power depends on the moment of inertia of the wings, which is unaffected by mass changes due to fuel deposition. Therefore, we assume that inertial power will not affect scaling relationships. When changing body mass, a bird should adjust its characteristic flight speed (Hedenström & Alerstam 1995), which will affect the Re and potentially CDb. However, changes of Re due to changes in flight speed due to within-individual mass change will be relatively small (Re = 60 000 at U = 15 ms−1 vs. Re = 75 000 at U = 19 ms−1 for a bird with dimensions of a knot) compared with the Re range over which CDb varies in birds (CDb = 0·4 at Re = 50 000, CDb = 0·25 at Re = 200 000; Pennycuick 1989). Therefore our assumption of a constant Re and hence CDb due to changes in mass, is justified. The aerodynamic model chosen for this analysis is deliberately simple because we believe it captures the essential aerodynamic costs. A recent aerodynamic study of bird wakes, carried out in a wind tunnel, show that the momentum shed in the wake can be accurately modelled across a wide range of speeds, even though the model ignored the fact that the birds are flapping their wings (Hedenström, Rosén & Spedding 2006).

In order to express the available power from the pectoral muscles of a given mass, we followed that of Pennycuick & Rezende (1984) who found that the mass-specific power output of the myofibrils is Pm =21·2f W kg−1. We used this figure as a measure of the muscle mass required to sustain aerobic flight as a function of flapping frequency (f). Previous analyses have been mainly concerned with interspecific variation of flapping frequency, and Pennycuick (1996) derived a dimension consistent formula for such a relationship.

image( eqn 2)

Where S is the wing area, I is the wing moment of inertia and the other variables as previously defined. A reduced version of this formula is in good agreement with data (Pennycuick et al. 1996; Pennycuick 2001). Within an individual bird, b, S and I are unchanged and so we might expect that wing beat frequency will vary according to inline image Very few measurements of within-individual variation of wing beat frequency are available, but for two species flying in a wind tunnel the wing beat frequency changed in relation to body mass in agreement with our assumption (Pennycuick et al. 1996). Consequently, equating power delivered by the beating pectoral muscles and power required to fly gives

image( eqn 3)

Hence, on a double logarithmic plot of mpm against mb the slope of a regression is predicted to be 5/4, i.e. 1·25. This is our aerodynamics-based prediction for the allometric slope, a benchmark against which data from carcass analyses can be compared.

Materials and methods

body composition analysis

Body composition data of adult knots were accumulated between 1981 and 2002. Most of the dissected specimens were victims as a result of flying into lighthouses, poachers, or died during capture (Battley & Piersma 2005), but some were purposefully collected (Iceland, Piersma et al. 1999b; Delaware Bay, Baker et al. 2004; Arctic Canada, Morrison, Davidson & Piersma 2005). Islandica knots (ntotal = 155) were available from most phases of their annual cycle (Fig. 1). Adults of the other subspecies were also available but only for a few of the annual cycle phases (C. c. canutus: n = 35; C. c. piersmai: n = 3; C. c. rogersi: n = 10; and C. c. rufa: n = 98). The birds were dissected following the procedures of Piersma et al. (1999b) and Battley & Piersma (2005). In brief, the birds were plucked before the skin was removed and the two pectoral muscles, Musculus pectoralis and Musculus supracoracoideus, were excised from both sides of the keel and weighed to the nearest 0·01 g. The water content was determined by drying to constant mass at 60 °C. Fat content was determined following drying via extraction with petroleum ether. However, the water and fat content were not always determined.

mpm was calculated as the total fresh mass of M. pectoralis and M. supracoracoideus from both sides of the body. mb (± 0·1 g) was sometimes determined immediately following killing or finding the sample, but always just before dissection. Hence, this latter mb was used in the analyses. mb includes the small intestinal content, which averaged approximately 2·56 g (± 0·11 SEM; n = 88). The birds were not necessarily dissected immediately after death, sometimes only following a considerable period in the freezer. Therefore, we corrected mpm to a water percentage of 70%.

To control for size differences, we calculated the standard muscle volume for both pectoral muscles (SMV, cm3) using four sternum measurements (Piersma, Davidson & Evans 1984, eqn 8). Within islandica, SMV did not differ between annual phases (anova, n = 149, P = 0·233). However, SMV varied significantly between subspecies (anova, n = 293, P < 0·001), with piersmai being the smallest (17·125 ± 0·318 cm3) and islandica (19·5716 ± 0·145 cm3) the largest. mpm and mb of the smaller subspecies were corrected to islandica-size by multiplying with the ratio (mean SMV islandica)/(mean SMV subspecies).

Fat percentage of the pectoral muscles was very low, on average 3·6 ± 0·1% for all subspecies (n = 287). The fat percentage did differ between subspecies (anova, n = 287, P < 0·01), but this was solely due to rufa (3·1 ± 0·2%) that differed significantly only from islandica (3·5 ± 0·2%, post hoc Bonferroni test). Within islandica, fat percentage did not differ between most phases (anova, n = 75, P = 0·099), only in wintering birds was it slightly lower (2·8%; anova, n = 142, P < 0·001, post hoc Bonferroni test). Therefore, we did not correct for fat percentage.

We assigned the knots to the various annual phases on the basis of collection location, time of the year, mb and wing moult (Piersma & Davidson 1992; Battley & Piersma 2005). Winter starved birds were excluded from the analysis.

allometric regression

In this study, the allometric relationship is best described by a continuous piecewise regression. Such a continuous piecewise regression or broken-stick relationship can (for mpm and mb) be described by

image

Where a is the intercept, b1 is the slope of the first (left) part, b2 is the slope of the second (right) part of the regression, c is the estimated breakpoint between the two phases, and r is a smoothness parameter set at 0·001 (Koops & Grossman 1993; Kwakkel, Ducro & Koops 1993). All allometric regressions were fitted using the nonlinear regression algorithm procedures from the nonlin 2·5 package (shareware program, P.H. Sherrod, based on the nonlinear least-squares algorithm described in Dennis, Gay & Welsch 1981). To test whether a piecewise regression was preferred over a linear regression, we used F-test (Kwakkel et al. 1993).

To be able to compare the empirical data with the theoretical aerodynamically based predictions, it is needed to estimate the model intercept, as well as the slope. We estimated the intercept by using the theoretical slope (1·25) and intersecting this model at the breakpoint of the piecewise regression (x = 2·17, i.e. at mb = 148 g, and y = 1·47, i.e. at mpm = 29·7 g).

manoeuvrability test

In aircraft, the power required to turn at a prescribed rate and load factor, increases with increasing weight (Vinh 1993). Although turning flight mechanics differ between the flapping flight of birds and fixed wing aircraft, it is reasonable to assume, at least as a first approximation, that power available from flight muscles will be related to turning performance, also in birds. We therefore used a simple turning flight assay to test this possibility.

The change in manoeuvrability with increasing mb was followed in 18 captive knots during their natural spring fuelling period in 2005 (31 March−23 June). The knots were housed under natural conditions in two outdoor aviaries at the Royal Netherlands Institute for Sea Research (NIOZ, Texel, The Netherlands). In each aviary (l × w × h: 3 × 2 × 2 m) the knots had access to a small, barren artificial mudflat to practice their probing activity. Food (trout pellets; Trouvit, Produits Trouw, Vervins, France) and fresh water were available ad libitum. As part of the care-taking routine, condition, plumage, moult and mb (± 1 g) were determined weekly on a table halfway along the corridor bordering the aviaries. After these procedures, the knots returned to their aviary on their own accord. This involved a voluntary take-off from the hand of the researcher and flying a short distance (7 m) at low speed through the corridor (width: 1·4 m) to the entrance door of their aviary (width: 0·8 m). When reaching the door, the flying birds had to make a 90° turn to enter the aviary.

We scored whether the knots made the 90° turn and if successful, the time from take-off to entering the aviary (±0·1 s, flight time). birds that failed tended to land at the end of the corridor and walk back to the aviary. In some weeks, some birds took part in metabolic trials and their manoeuvrability was not tested. Two of the 18 knots never made the turn at low mb (ranges 109–137 g and 112–173 g), and were therefore excluded from the analyses. We also excluded data obtained after peak mb was reached as we did not collect data for all birds during this period. The number of observations per individual ranged from 6 to 11.

As not only mb but also wing surface, determines the manoeuvrability of a bird (Pennycuick 1992), mb must be corrected for interindividual variations in wing surface. The surface of the right wing was determined following the method described by Pennycuick (1989). Total wing surface (cm2) was calculated assuming that the right and left wing areas were equal, except for one bird that differed in the length and number of broken feathers tips in its left and right wings. For this bird, we additionally determined the left wing surface. We corrected mb for total wing surface as follows

image

where the mean total wing surface was 249·7 cm2 (± 3·5 SEM, n = 16).

When most knots approached peak mb (day 63 of the experiment; 2 June) the pectoral muscle thickness (±0·01 cm) was determined by MD by ultrasonography (10 MHz linear probe, Aquila, Esaote Pie Medical, Maastricht, the Netherlands) following the method described in Dietz et al. (1999a, 1999b). The following calibration curve (of MD) was used to calculate total fresh pectoral muscle mass

mpm = pectoral muscle thickness × 20·52 + 1·47(g).

Results

allometry of pectoral muscle mass

mpm increased with increasing mb (Fig. 2a). Fitting a linear regression through the log-transformed data yielded a slope of 0·77 (± 0·05 SEM) that was significantly lower than the theoretical slope of 1·25 (Student t153 = −9·60, P < 0·001). However, there were clearly two phases: the slope was lower in fuelling and pre-departure knots with high body masses than in knots of lower body mass (Fig. 2a). The piecewise regression model gives a better fit for the data compared with a linear regression (F2,153 = 3·955, P < 0·025). The breakpoint of the piecewise regression was at 2·17 (± 0·03 SEM), i.e. at a mb of c. 148 g. This marked the threshold value (a mb of 150 g) which is rarely exceeded in nonmigratory contexts (pers. obs.).

Figure 2.

(a) The allometric relationship between mpm and mb for C. c. islandica. The black line represents the continuous piecewise allometric regression through the data (log(mpm) =−0·83 + 1·06 log(mb) −[0·001 − (1·06 − 0·63)lninline image]; r2 = 0·655; P < 0·001); the grey line is the theoretical allometric regression log(mpm) = −1·24 + 1·25 × log(mb); and the dotted line marks where mb is 160 g, the point from which red knots encounter manoeuvrability problems (see Fig. 3d). Life phases are indicated with the same colours as in Fig. 1: breeding, red; F2 and F4 blue, A3, yellow; PD4, purple; and wintering, white. (b) Mean (± SEM) mpm for four other subspecies of red knot. The black line represents the general continuous piecewise allometric regression through the data of all subspecies together (including islandica) (log(mpm) =−0·69 + 1·00 log(mb) − [0·001 − (1·00 − 0·66)lninline image]; r2 = 0·663; P < 0·001); the grey line is the theoretical allometric regression and the dotted line marks where mb is 160 g, the point from which red knots encounter manoeuvrability problems. Symbols: square, C. c. canutus; circle, C. c. rufa; triangle up, C. c. piersmai; triangle down, C. c. rogersi. Numbers per point are given in figure. Colours and lines as in (a).

The slope of the first section of the piecewise regression was indistinguishable from the theoretical prediction (1·06 ± 0·13 SEM; Student t153 = −1·462, P > 0·05), but the slope of the second section was significantly lower than predicted (0·63 ± 0·08 SEM; Student t153 = −7·750, P < 0·001). This led to considerable differences between theoretical and actual mpm in pre-migration knots: mpm was 0·81 times lower than predicted in a typical pre-departure knot of 210 g.

To investigate if the results for islandica were representative for knots worldwide, we fitted a linear and a piecewise regression to the data from the other subspecies. Again, the piecewise regression gave a better fit to the data (F2,143 = 4·121, P < 0·025). Next, we compared the piecewise regression of islandica with that of the other subspecies. These did not differ significantly (F4,296 = 2·118, P > 0·05). The general model differed slightly from that of islandica (Fig. 2b). The breakpoint was slightly lower, at 2·15 ± 0·03 SEM, i.e. at 141·3 g. The first slope was also a slightly lower (1·00 ± 0·10 SEM), differing significantly from the aerodynamical slope of 1·25 (Student t298 = −2·500, P < 0·02), while the second slope was slightly higher, but still significantly lower, than the aerodynamical slope (0·66 ± 0·06 SEM, Student t153 = −9·667, P < 0·001). These small differences did not change the effects on mpm in pre-migration knots: mpm was still 0·81 times lower than predicted in a typical pre-departure knot of 210 g.

manoeuvrability test

As expected, mean mb of the captive knots increased with time during their natural spring fuelling period (Fig. 3a). Peak mb (180 ± 5 g SEM) varied between 143 and 211 g, while peak mbcor (180 ± 3 g SEM) had a smaller range, 156–211 g. The fraction of knots that did not make the 90° turn (Ffailed) increased with time (Fig. 3b). Flight time (from hand to the aviary) was on average 2·9 s (± 0·1 SEM, n = 108, successful flights only), and did not vary with time nor with mb (P = 0·688 and P = 0·437, respectively).

Figure 3.

(a) The increase in mean mb (± SEM) with time during fuelling in 16 captive red knots. Numbers per measuring day are the same as in panel b. Day 0, the start of the experiment, is 31 March 2005. (b) The variation in the fraction of birds that did not make the 90° angle turn with time. Numbers are indicated above the bars. (c) The variation in the fraction of birds that did not make the 90° turn with mbcor class (mb corrected for wing surface). Numbers per mbcor class are indicated above the bars. The solid line represents the logistic regression fitted trough the individual data points (see text; inline image). (d) The relation between Ffailed (average per mbcor class) and the difference between theoretically predicted mpm and estimated actual mpm (calculated using mean mbcor for each mbcor class).

From the 10 g class of 165 g, i.e. a mb of c. 160 g, knots had increasing problems in making the 90° turn (Fig. 3c). Five of the 16 knots never failed to make the turn; three of them had high peak mbcor's (between 180 and 184 g). Of the remaining 11 knots, three failed only once (peak mbcor between 176 and 182 g), leaving eight knots that failed repeatably to make the turn (peak mbcor between 167 and 211 g). We analysed the variation in Ffailed with mbcor with a logistic regression model, combining a binomial distribution with the logit link and using the iterative generalized least squares algorithm (MLwiN 2·0; Rabasch et al. 2000). Individual, added as a random effect to the model, was not significant, while the other parameters were significant. Hence only mbcor explained the variation in Ffailed. mpm of captive knots varied with mb similarly as in wild islandica knots of equal mb (mb range 143–211 g; n = 72 and n = 16 for wild and captive knots, respectively; univariate analysis of variance, P > 0·05 for slope and intercept comparison). Therefore, we assumed that mpm of the captive knots could be estimated from the piecewise regression of wild knots. For the mean of each mbcor class, we calculated the difference between theoretical and estimated actual mpm and plotted this difference against mean Ffailed per mbcor class (Fig. 3d). This figure shows that the onset of increase in failed turns occurred at a mb of c. 160 g. Above this mass, the difference between actual and theoretical predicted mpm becomes substantial (c. 7% at 160 g, Fig. 2a); actual mpm being smaller than expected.

Discussion

The pectoral muscles of knots fuelled up for long-distance flight were considerably lighter than theoretically predicted for birds maintaining a constant aerodynamic performance. This cannot be a result of methodological problems (e.g. delayed dissection) because mpm was corrected to a standard water percentage. Note that the coefficients found here (1·06 and 0·63 for normal weight and heavy pre-migration knots, respectively) were close to earlier reports of coefficients for islandica knots in similar states (0·91 in just arrived and 0·64 in pre-departure knots, in northern Norway; Davidson & Evans 1986).

The empirical allometric line was based on an interindividual comparison while the theoretical allometric expectation was based on an intraindividual comparison. Lindström et al. (2002) showed that for 10 knots (12–35 data points per individual) the slopes of the individual linear regressions of mpm on mb were similar, but that the intercepts differed. Reanalyses of the data in Dietz et al. (1999b) showed that for the eight knots in the study (four data points per individual), the individual linear allometric regressions did not differ in slope nor intercept (univariate analysis of variance, n = 32, both P > 0·4). On this basis we conclude the difference between the empirical data and the theoretical expectation was not caused by the difference between interindividual and intraindividual comparisons.

The manoeuvrability tests were based on voluntary choices with respect to the increase in mb and the flight performance, in knots that had spent several years in captivity. The limited flight possibilities in captivity may have had a negative effect on their flight abilities such as manoeuvrability. However, previous experiments with flying long-term captive knots in wind tunnels indicate that endurance flight capacities are not affected by long-term captivity (Lindström et al. 2000; Kvist et al. 2001; and see Dietz et al. 1999b). Consequently, one can assume that manoeuvrability is also little affected.

Above c.160 g, knots were increasingly unable to make a 90° turn. At this point, measured mpm is 0·949 times the theoretical mpm. This difference may seem small, but 0·949 is the critical value below which manoeuvrability decreases. mb of almost all dissected wintering, breeding and just arrived knots, was below 160 g (Fig. 2a), suggesting that knots only encounter manoeuvrability reductions in the days and weeks before take-off on long-distance flights.

constraints and trade-offs in performance

Pre-departure knots were on average 1·5 times heavier than wintering knots (209·9 vs. 138·9 g) and their pectoral muscles were 1·4 times larger (37·8 vs. 26·2 g). In a previous study, freshly arrived and pre-departure knots had similar mb's (141 and 205 g, respectively, at the beginning and end of a 2-week spring stopover in Norway; Evans et al. 1992), but the difference in mpm (6%) was significantly lower than the differences between newly arrived and pre-departure knots in our study (24·5% of mpm of freshly arrived knots). We cannot explain these differences, but both studies came to the same conclusion, namely that mpm of pre-departure knots is much smaller than theoretically predicted to match power requirements.

Why do fuelling and pre-departure knots not increase their pectoral muscles more in order to keep escape flight abilities constant? There are at least four possible, not mutually exclusive, hypotheses to account for this. The first explanation is that there is a physiological constraint on maximal muscle size. Muscle size increases via increases in muscle fibre size (hypertrophy) or in muscle fibre number (hyperplasia). Hyperplasia occurs as a result of chronic stretch of a muscle (Kelley 1996); during fuelling only hypertrophy is expected to occur. In birds, pectoral muscles require a high O2 flux rate and nutrient delivery rate due to their very high energy demand (Mathieu-Costello, Suarez & Hochachka 1992; Mathieu-Costello & Hepple 2002). This can only be achieved via small fibre size because then the diffusion distance within a fibre is short, while the fibre surface to volume ratio is large, enabling a high capillary-to-fibre surface ratio (Lundgren & Kiessling 1988; Mathieu-Costello et al. 1992; Guglielmo et al. 2002; Mathieu-Costello & Hepple 2002). Indeed, fibre area of bird pectoral muscles is small compared with rat leg muscles (Mathieu-Costello & Hepple 2002). In passerines, fibre area was smaller in long-distance migrants than in short-distance, partial or nonmigrants, and did not differ between migrants and breeders (Lundgren & Kiessling 1988). However, fibre area was larger in fuelling than in wintering sanderlings Calidris alba and dunlins Calidris alpina (1·2 and 1·6 times, respectively; Evans et al. 1992). Nevertheless, fibre area of fuelling sanderling and dunlin was smaller than that of long-distance migrating passerines, which migrate with shorter flights (Lundgren & Kiessling 1988). Hence in knot's pectoral muscles, fibre size and thus mpm, may increase during fuelling, but this increase may be limited due to the muscle's high energy demands.

An alternative hypothesis to account for a lack of increase in mpm is that muscle efficiency increases with increasing mb. The wind tunnel data of Kvist et al. (2001) suggest that heavy knots use less power than expected during sustained flight, indicating that muscle efficiency increases with increasing mb. Aerobic capacity, percentage mitochondria or myofibrils of pectoral muscles may increase during fuelling (Lundgren & Kiessling 1985; Lundgren & Kiessling 1986; Evans et al. 1992; Bauchinger & Biebach 2001) or during migration (Bauchinger & Biebach 2001; Guglielmo et al. 2002). Yet, in knots there is no evidence for a change in mitochondria, sarcoplasma, myofibril content (Evans et al. 1992) or cytochrome c oxidase activity (Weber & Piersma 1996), while succinate dehydrogenase activity decreased during (early) fuelling (Selman & Evans 2005). However, there are, indications that muscle physiology changes only after endurance exercise and thus can only be found in actively flying birds (Guglielmo et al. 2002). This may explain the lack of evidence for knots. Nevertheless, the manoeuvrability experiment suggests strongly that any increase in muscle efficiency is insufficient to maintain manoeuvrability.

During fuelling, the digestive organs must process considerable amounts of food. To be able to do this, size and metabolic activity per gram tissue increases (Piersma et al. 1999b; Battley & Piersma 2005; Selman & Evans 2005). Thus, a third hypothesis is that mpm is kept relatively small to save energy to counter the high mass/space and metabolic demands of the digestive organs.

The fourth and last hypothesis to explain the relatively small pectoral muscles at peak mass is that organ mass limitation serves as a weight-saving mechanism in order to reduce flight costs. Such adaptive weight-savings seem to occur in several organs in waders preparing for long-distance migration (Piersma, Koolhaas & Dekinga 1993b; Piersma 1998; Piersma & Gill 1998; Piersma et al. 1999b). Limiting the increase in mpm reduces flight costs not only during migration, but also during fuelling.

In conclusion, pectoral muscle hypertrophy may be limited by physiological constraints on fibre size (hypothesis 1). As hyperplasia does not occur, this constraint results in a relative small mpm, but possibly extra efficient muscles in heavy knots (hypothesis 2). However, this size limitation also has some benefits, as it either accommodates energetically and/or spatially other organs (hypothesis 3) or has overall weight-saving benefits (hypothesis 4).

towards a functional ecology of pectoral muscle mass

According to the aerodynamic prediction, heavy pre-migratory knots should have increased mpm much more than they did to avoid decreasing escape flight abilities with increasing mb. Increased mb impairing escape flight abilities of fuelling migrants was previously found in other waders (least sandpipers Calidris minutilla; western sandpipers Calidris mauri; Burns & Ydenberg 2002) and passerines (e.g. blackcap Sylvia atricapilla, Kullberg, Fransson & Jakobsson 1996; European robin Erithacus rubecula, Lind et al. 1999; sedge warbler Acrocephalus schoenobaenus, Kullberg, Jakobsson & Fransson 2000). So far, it has been assumed that this impaired escape ability is caused by increased wing loading. Our results show, however, that the problem associated with increased wing loading is due to an insufficient increase of mpm.

These findings lead to the conclusion that maintaining mpm at optimal flight performance size, when mb increases considerably, must be costly. To avoid these costs, birds are prepared to sacrifice part of their escape flight ability. As this probably increases their predation risk, birds seem at first to have chosen a strange compromise. However, the probability of surviving predator attacks is not only determined by the bird's escape flight ability, but also strongly by its antipredator behaviour, such as flocking (Creswell 1994a; Whitfield 2003), choice of feeding or stopover location (Piersma et al. 1993a; Creswell 1994b; Whitfield 2003; Nebel & Ydenberg 2005), vigilance (Lind 2004) and timing of behaviour (e.g. foraging; Lank et al. 2003; Burton & Armitage 2005). The antipredator behaviour of a bird may even affect its predation risk more than its escape flight ability (Lind 2004). In certain circumstances body condition (mb/wing length) has little impact on the chance of being captured by a raptor (Whitfield et al. 1999). In addition, magnitude of mb variation and possible antipredator behaviour may vary between sexes, ages and dominance levels (e.g. Creswell 1994b; Krams 2002; Nebel & Ydenberg 2005). The final effect of reduced escape flight ability on the predation risk of a bird depends thus on its personal characteristics (such as sex, age and dominance) and the environmental conditions that determine the possible antipredation strategies. All of these considerations suggest that a full evaluation of the functional ecology of flight muscle mass should not only incorporate detailed phenotypic assessments, but also measurements of vigilance and behaviour in an ecological context.

Acknowledgements

We thank Anne Dekinga, Phil Battley, Georg Nehls and many others who helped with the dissections. Bernard Spaans and Yvonne Verkuil assisted with the manoeuvrability tests. Martijn van der Pol did the logistic regression model analysis for us. We thank Judy Shamoun-Baranes for her constructive comments. Jan van de Kam took the photographs of the knots used in Fig. 1. The flight manoeuvrability experiment complied with Dutch law regarding animal experiments.

Ancillary