Evolutionary divergence of body size and wing and leg structure in relation to foraging mode in Darwin’s Galapagos finches

Abstract

The wings, legs, and tail in Darwin’s finches show many clear adaptations to different types of locomotion used during foraging. We use size scaling to analyse how various characters vary with body mass to clarify dimensional relationships. The selective advantage of a character is judged in terms of energy savings. The wing aspect ratio (4.6–4.9) is very low, so the energy costs for flight are high. Low body mass, low wing loading, and short arm wings in the warbler finch, small tree finch, and small ground finch promote agility and manoeuvrability among vegetation, along with short wings in the warbler finch. Evolution towards a shorter arm wing seems to be favoured in the smaller finch species. Long legs, long toes, and long curved claws are adaptations for climbing/clinging locomotion without tail support (woodpecker finch, small and large tree finches, cactus finch but having short legs). Selection for longer legs seems to act towards a lengthening of the tarsometatarsus. The climbing technique in the woodpecker finch is described. We discuss how the diversification in the beaks relates to the locomotion organs.

INTRODUCTION

The adaptive radiation of Darwin’s finches on the Galapagos Islands is a classic theme in textbooks on evolution because they show dramatic differences in beak size and form (e.g. Lack, 1947; Harris, 1974; Grant et al., 1985; Grant, 1999; Grant & Grant, 1989, 2002, 2008, 2014). Darwin’s finches derived from a common ancestor from the mainland about 2 Mya (Grant & Grant, 2002). Their diversification was relatively rapid and driven by ecological differences among locations (Lack, 1947; Bowman, 1961; Schluter & Grant, 1984; Grant, 1999). Spatial and temporal variations highly contribute to shape phenotypical traits in Darwin’s finches (Schluter & Grant, 1984; Grant, 1999; Kleindorfer et al., 2006, Grant & Grant, 2008, 2014; Carrión et al., 2022).

The size and form of the beak are strongly correlated with the kind of food they eat (Grant et al., 1985; Grant, 1999; Grant & Grant, 1989, 2008, 2014). The beak size increases disproportionately with increasing body mass, and we can still see the evidence of the differentiation, where hybridization seems to be an unusual but potent force in the adaptive radiation (Grant & Grant, 2014). Except for beak adaptations to different food choice for maximization of energy intake, a minimization of energy expenditure during foraging may be as important. Beak morphology has been the focus of many studies on Darwin’s finches, and fewer studies have concentrated on locomotor apparatus.

The exploitation of different feeding sites is usually associated with different structural adaptations of the wings, legs, and tail (e.g. Palmgren, 1932; Winkler & Bock, 1976; Norberg, 1979, 1981; Norberg, 1983, 1986, 2021; Norberg & Norberg, 1989). Because foraging flight and climbing are very energy demanding (e.g. Tucker, 1969; Pennycuick, 1975; Norberg, 1986), the finches can be expected to differ in the adaptive morphology of the locomotor organs for minimizing energy expenditure during foraging, particularly between those species which differ most in feeding behaviour. The purpose of this investigation is to explore how the body size, wings, legs, and tail of Darwin’s finches are adapted to their respective foraging mode and feeding posture.

We attempt to answer the following questions: (i) what are the quantitative, real, morphological differences in the locomotion apparatus among the nine species of Darwin’s finches resident on the Island of Santa Cruz? (ii) What are the differences in shape and proportions that are not due to scaling effects of different body sizes? (iii) What functional adaptations can be identified in the real and relative sizes and proportions of the wings, legs, and tail in relation to foraging locomotion and feeding postures? (iv) To what extent has the diversification in the locomotion organs proceeded in contrast to that of the beaks?

MATERIALS AND METHODS

The collection of morphological data was performed on living specimens on Santa Cruz on the Galapagos Islands in January and February 1981 in three different areas: in the arid coastal zone at Academy Bay, in the cultivated area near Bella Vista (transitional zone), and in the Scalesia zone near the craters Los Gemelos. With the use of mist-nets and song playback we captured eight to 38 specimens of each of the nine finch species resident on Santa Cruz. They were (with the keys used in the figures and tables) the small ground finch (Geospiza fuliginosa, Gful), the medium ground finch (Geospiza fortis, Gfor), the large ground finch (Geospiza magnirostris, Gm), the cactus finch (Geospiza scandens, Gs), the vegetarian finch (Platyspiza crassirostris, Pc), the small tree finch (Camarhynchus parvulus, Cpar), the large tree finch (Camarhynchus psittacula, Cpsi), the woodpecker finch (Camarhynchus pallidus, Cpal), and the warbler finch (Certhidea olivacea, Co). The small ground finch was captured in the coastal (N = 17) and transitional (N = 5) zones, the medium ground finch in the coastal (N = 25) and transitional (N = 13) zones, the large ground finch in the coastal zone, the cactus finch in the coastal (N = 4) and transitional (N = 18) zones, the vegetarian finch in the coastal (N = 19) and transitional (N = 1) zones, the small tree finch in the coastal (N = 11), transitional (N = 7), and Scalesia (N = 4) zones, the large tree finch in the transitional (N = 7) and Scalesia (N = 1) zones, the woodpecker finch in the transitional (N = 4) and Scalesia (N = 7) zones, and the warbler finch in the coastal (N = 1), transitional (N = 3), and Scalesia (N = 20) zones.

In December 1983 and January 1984, we filmed all finch species during foraging with a spring-driven Pathé 16 mm film camera run at 80 frames s^–1, to gain insight into their foraging behaviours.

Measurements

The following measurements were taken on all specimens: body mass, wingspan, wing area, hand-wing length, tail length, foot length, tarsometatarsus length, tibiotarsus length, ratios of hand-wing length to wingspan and of tibiotarsus length to tarsometatarsus length. The body mass, wingspan, and wing area were used to estimate the wing loading and aspect ratio. Body mass is given in g, all lengths in mm, wing area in cm², and wing loading in Nm^-2, whereas all ratios are non-dimensional.

Body mass, M:

We weighed the birds immediately after capture with a 30 g or 100 g Pesola spring balance that could be read to the nearest 0.1 g.

Wingspan, b:

Distance between the two wing tips when the wings are moderately extended laterally as described below under ‘total wing area’. We used a ruler that could be read to the nearest millimetre when measuring the wingspan, hand-wing length, and tail length.

Total wing area, S:

The low-pressure region on top of the two wings continues across the body and generates lift. Therefore, the total wing area is conventionally measured as the area of both wings plus the intercepted body area, as projected onto the wing-chord plane, tail area excluded. We drew the wings’ outline from photographs and measured the area with a polar compensating planimeter, taking the mean of three measurements in each case.

Wing loading, Mg/S:

Wing loading is the body weight in Newtons, calculated as body mass, M, multiplied by acceleration of gravity, g, and divided by wing area, S.

Aspect ratio, b²/S:

The wing aspect ratio measures the narrowness of the wing. It is the dimensionless ratio b/c between wingspan b and the mean wing chord c. To avoid having to find the mean wing chord, aspect ratio is estimated as b/c = bb/bc = b²/S.

Hand-wing length, l_hw:

Distance from wing bend (ulno-carpal joint) to wing tip. We also calculated the ratio between the length of the hand-wing to half wingspan, 2l_hw/b, to estimate what proportion of a half wingspan is the hand-wing. The length of the arm wing, l_aw (the length between the wrist and the humero-scapular joint), is difficult to measure in live birds but can be understood from the equation l_aw = (b-z)/2–l_hw, where z is the distance between the two humero-scapular joints (body width).

Tail length, l_tail:

Distance from the root of the two middle tail feathers, measured on the ventral side of the tail, to the tip of the longest tail feathers.

Tibiotarsus length, l_tt:

Distance from the top of the knee to the ventral side of the joint between the tibiotarsus and tarsometatarsus when the knee and the tibiotarsus-tarsometatarsus joint are set at 90°.

Tarsometatarsus length, l_tmt:

Distance from the posterior surface of the tibiotarsus at the tibiotarsus-tarsometatarsus joint to the outer edge of the most distal scale on the dorsal side of tarsometatarsus. We also calculated the ratio l_tt/l_tmt to find out which leg element contributes most to the leg length (femur and foot lengths excluded).

Foot length, l_foot:

Distance, measured on the dorsal side of the foot, from the edge of the skin at the claw-base of the middle forward toe to the edge of the skin at the claw-base of the hind toe when the toes are straightened and gently stretched forwards and backwards, respectively.

Using size scaling to compare shape and proportions between different-sized species

In this paper we use size scaling to analyse how various wing and leg measurements vary with body size among different-sized finches. Theory of scaling generally produces equations of the form y = αx^β, where x and y are dimensions or other quantitative properties of animals of different sizes, and α and β are constants (McMahon & Bonner, 1983; Alexander, 1985). This form of equation, widely used in functional analyses (e.g. Greenewalt, 1972; McMahon & Bonner, 1983; Schmidt-Nielsen, 1984; Pennycuick, 2008), gives straight lines when plotted on logarithmic coordinates. The scaling exponent β dictates how y changes with increasing x, whereas the value of the multiplicative factor α determines the elevation of the line, raising or lowering it in a log-log diagram (McMahon & Bonner, 1983: 30). The power function can also be written in logarithmic form as logy = logα + βlogx, where log means logarithm to the base of 10. It is the equation of a straight line, where logα is the y-intercept when logx = 0, which happens when x = 1 since log1 = 0, and β is the slope of the line in a log-log diagram.

In species with the same mass density, body mass M is proportional to body volume and is easy to measure. Therefore, we relate measurements of each of the nine finch species to their respective body mass M. Power functions of the form y = αM^β were fitted to empirical data by the least squares regression method. The sloping line shows how measurement y varies with body mass M across different-sized species. The effect of body mass of species i on the observed y-value can be specified by measuring the difference in elevation between the observed value y_obs of species i and the value y_trend that y would take if it were lying on the fitted regression line. The deviation of y_obs from the y-vs.-M trend-line can be quantified by a relative (body-size corrected) measure y_rel. It is defined as y_rel = y_obs/y_trend = y_obs/αM_obs^β, where M_obs is the observed mass of the species i inserted into the fitted power function to calculate the trend-line value y_trend for species i. The relative measure is used to trace deviations from the mean trend and to compare species of different sizes. The ratio takes the value 1 when y_obs lie on the regression line, and is larger than 1 when y_obs is larger than its trend-line value y_trend, and vice versa.

The percentage deviation from the trend line in a species is calculated from the expression (y_obs - y_trend)/(y_trend) × 100. Using the relative value, y_rel, we can detect if a particular character of a species deviates from the mean trend-line across all finches in the group, i.e. when the effect of body mass has been removed. It may reflect a difference in form and proportion due to a specific adaptation. Differences between the size-corrected values are tested with the Student’s t-test. Since many variables are tested among several species, there is a large risk of obtaining fortuitous significances. Therefore, we base conclusions mostly on differences with significance levels of P < 0.001.

We do not compensate for differences in the degree of phylogenetic relatedness but give all species equal weight when fitting the regression line. This is because the finch species present on Santa Cruz are closely related and likely to interfere more than distantly related ones and therefore should not have reduced weight in the regression.

Geometric and elastic similarities:

The most basic size-scaling rule requires that different-sized species of similar kind within a modest size range are geometric (isometric). The length of a structure, l, in geometrically similar animals then varies with body mass as l ∝ M^1/3 and wing area, S, as S ∝ M^2/3. For different-sized animals to maintain the same safety margin against failure due to maximal muscular forces, they need to be geometrically similar (McMahon, 1984; Norberg & Wetterholm Aldrin, 2010). An alternative to geometric similarity is elastic similarity. It requires that different-sized structures deform to the same bent shape under their own weight or under forces proportional to the weight (McMahon, 1984). To be elastically similar, different-sized structures must be thicker in relation to their length the larger they are. For elastically similar animals the length of a bone in the leg or wing scales with body mass as l ∝ M^1/4.

Which scaling principle that applies to an animal species today depends on the past frequency of occurrence of structural failure of support elements due to excessively large loads. When maximal muscle forces have caused few fractures of support elements, elastic similarity has been selected for, but when there has been a high frequency of structural failure of support elements under maximal muscle forces, selection has favoured geometric similarity (Norberg & Wetterholm Aldrin, 2010). Geometric similarity seems to be the dominant scaling principle among animals of similar kind within a modest size range. In the following, we will examine which scaling principle best fits the size-scaling of Darwin’s finches. Student’s t-tests were used to test whether the slope β of the regression lines differs from those expected under geometric or elastic similarity.

Functional considerations and predictions on morphological adaptations

Feeding behaviour

A short summary of the feeding ecology of Darwin’s finches, described in Bowman (1961: 25–44), is given here for the nine finch species as a basis for our interpretations of the morphological characters for the different movements. Bowman (1961) also noted that several of the finch species overlap in habitat and food selection.

The ground finches (Figs 1, 2) feed mainly on different sizes of seeds of fruits and flowers, hopping around on the ground or while perching on branches among bushes and trees. The large ground finch (Fig. 2) is more often seen in denser brushy areas and low trees than on the ground feeding on larger seeds and sometimes arthropods. The cactus finch (Fig. 3A) takes seeds and fruits of the prickly pear both on fallen leaves on the ground and up in the cacti. It often hangs upside down when visiting a cactus flower (Fig. 3B). The small tree finch (Fig. 4) often forages head downwards, which also is what we observed.

The vegetarian finch (Fig. 5) is mainly herbivorous and feeds on fleshy fruits, seeds, leaves, and flowers, and rarely on insects. It inhabits dense brush and frequently higher trees, perching while eating. The warbler finch (Fig. 6A, B) forages for insects among bushes and at all foraging levels in trees with agile flight. It often hangs head downwards, perches within leaf clusters, flies around the foliage, and sometimes hovering (which we also observed).

The tree finches take insects in bushes and trees. The small tree finch searches leaves, bark, lichens, mosses, and crevices for insect food with agile movements like those of parid birds (Lack, 1945: 41). The large tree finch climbs on branches but rarely on vertical trunks of trees. It searches for food by moving forwards and backwards around a section of a branch and is often seen hanging upside down. It uses its bill to cut through the stem for larvae. The woodpecker finch (Fig. 7A, B) climbs vertical tree trunks upwards as well as downwards without bracing the tail against the trunk for support. It carries about a cactus spine or small twig, which it pokes into cracks, dropping the stick to seize any insect that emerges (Lack, 1953).

Wings and body size

Wing size and form vary among flying animals and are adapted to different flight modes, foraging behaviour, habitat selection, and food choice. Wing surface area in relation to the body weight and the form of the wing can be described by three morphological parameters: wing loading, aspect ratio, and wing pointedness (Pennycuick, 1975; Norberg, 1979, 1990). Because wing loading Mg/S ∝Μ/Μ^2/3 = M^1/3 for geometrically similar birds, a small bird has lower wing loading than a larger one. Low wing loading facilitates slow flight and hovering and reduces the minimum turning radius (Pennycuick, 1975, 2008), important for agile and manoeuvrable flight. Any characteristic flight speed V is proportional to the square root of the wing loading, V ∝ (Mg/S)^1/2 ∝ M^1/6 (Pennycuick, 2008).

The aspect ratio is a measure of the overall shape of the wing. Higher aspect ratios mean greater aerodynamic efficiency and lower energy losses in flight, particularly at low speeds. A ‘good’ wing should have a shape that allows it to obtain sufficient lift L without much drag D. Long, narrow, high aspect ratio wings with pointed wingtips (such as in albatrosses) have higher lift L to drag D ratios and require less flight power than short broad wings of low aspect ratio and rounded wingtips. But long wings are a hindrance when flying among vegetation and for manoeuvrability, while short wings facilitate flights among dense vegetation. A combination of different wing loadings and aspect ratios is certainly evolved for each species, depending on niche choice, and on commuting and migration requirements (Norberg, 1981, 1995; Norberg & Rayner, 1987). The selection pressures for various demands may also be conflicting, necessitating compromise solutions.

In hovering and slow flight, the inertial power (the power needed to flap the wings) should be low and to reduce this power and the inertial loads on the wing skeleton the bird should have a low mass of the wing, and/or have the mass located as proximally as possible (Weis-Fogh, 1972). The shorter the arm wing is in relation to the total length of the wing, the more proximally the main mass of the wing becomes, important to slow flight and hovering among vegetation (Norberg, 1979). Examples of birds with very short arm wings are the hummingbirds.

Predictions on wing morphology related to foraging

Short flights between foraging sites within vegetation:

short wings.

Slow agile flight among vegetation (small ground finch, small tree finch, warbler finch)

: low body mass, low wing loading, short wings, and short arm wings.

Tail

The tail can be used for longitudinal and directional control in level flight and for increasing lift when lowered and spread to act like a delta-wing during take-off and landing and during turning manoeuvres (Pennycuick, 1972; Thomas, 1983; Norberg, 1994; Taylor & Thomas, 2014). The tail may also help balancing the body during perching and landing on horizontal branches.

Predictions on tail size related to foraging

Given the short tails in Darwin’s finches, we would expect longer relative tail length in the heavier finches that often perch on branches among bushes (vegetarian finch, large ground finch) and in the more manoeuvrable small finches (warbler finch, small tree finch, small ground finch).

Legs and feet

The lengths of the different leg elements are related to the bird’s feeding behaviour (e.g. Palmgren, 1932; Winkler & Bock, 1976; Norberg, 1979; Norberg, 1983, 1986; Leisler et al., 1989; Norberg & Norberg, 1989; Zeffer & Lindhe Norberg, 2003). For example, a bird which often hangs under branches and/or climbs should have a short tarsometatarsus to minimize the muscle force to keep the legs flexed (Palmgren, 1932). On the other hand, the step length during climbing increases with the length of the legs, primarily the tarsometatarsus (Norberg, 1979; Norberg, 1986).

The ratio between the tibiotarsus and the tarsometatarsus, l_tt/l_tmt, shows which of these leg elements contributes most to the overall leg length (femur and foot lengths excluded). Norberg (1979) showed in a biomechanical analysis that, in climbing species, the tarsometatarsus affects stride length the most, the femur less, and the tibiotarsus least. A bird with long legs usually has both long tibiotarsi and tarsometatarsi and can make long steps but is not well adapted to hang under branches. But birds with short legs, on the other hand, can do this with less muscle force, especially with a short tarsometatarsus. A reduction of the tarsometatarsus under the evolutionary pathway towards lower costs for a hanging and clinging behaviour may be visible by the l_tt/l_tmt ratio. The shorter the tarsometatarsus is in relation to the total leg length the larger the ratio becomes.

Woodpeckers, nuthatches, and treecreepers are trunk climbers, and adaptations for vertical climbing locomotion are long curved claws and a long tail (e.g. Winkler & Bock, 1976). Curved claws increase the ability to get hold of the bark. However, while woodpeckers and treecreepers climb vertically upwards with tail support, nuthatches and the woodpecker finch climb both head up and head down on the trunk without tail support (Lack, 1945: 41).

Predictions on the legs and feet related to foraging

Near-vertical climbing without tail support (large tree finch, woodpecker finch):

long legs, long toes, and long curved claws in relation to body size, long relative length of tarsometatarsus and low l_tt/l_tmt ratio.

Clinging on vegetation (small tree finch, warbler finch, cactus finch):

long curved claws.

Hanging upside down under branches (small tree finch, large tree finch, cactus finch, warbler finch):

short relative length of tarsometatarsus and higher l_tt/l_tmt ratio than in the other finches.

RESULTS

Morphology

The power functions are plotted in log-log diagrams, but for some ratios where it is difficult to separate the species because of the logarithmic vertical scale, they are shown in linear regression diagrams (Figs 11, 14B, 18). The measurements of wings, legs, and tail are given in Tables 1–3, where also the percentage deviations from the regression lines are included for each species. The relationships between the various dimensions and body mass, expressed as power functions, are listed in Supporting Information, Table S1, including comparisons with geometric and elastic similarities.

The comparisons among the birds are based on the size-corrected (relative) measurements in the form y_rel = y_obs/αM_obs^β (Supporting Information, Tables S2–S3). The Supporting Information (Table S4) shows the significance levels for relative measurements among the finches. Since each species is tested against eight other species for each character, we state significant differences only when P < 0.001 to reduce fortuitous significances. When one species is said to be adapted in a certain way it is regarded to be so in relation to one or more of the other species treated here, and in relation to their body masses. We discuss mainly the main deviations from the trend lines, those we find important for differences in their foraging locomotion.

Body size (Table 1):

There is a body-size-dependent gradation in the foraging mode among Darwin’s finches, from the warbler finch (10 g) and the small tree finch (13 g) to the large ground finch (33 g) and the vegetarian finch (35 g).

Wing form (Fig. 8):

The wings of Darwin’s finches are short and broad, and the wing shape is nearly rectangular with rounded wing tips. There are very small differences in wing form among the finches, as exemplified in Figure 8.

Wingspan (Table 1; Fig. 9; Supporting Information, Tables S1, S2, S4):

All the Galapagos finches have short wings compared to most other birds (Norberg, 1990: fig. 10: 2). The regression coefficient for wingspan vs. body mass (β = 0.352) is slightly larger than expected under geometric similarity (β_g = 1/3; P < 0.001), so the span increases faster with increasing body mass than it would under geometric similarity. The small and large tree finches have the longest wings of the species, +3.6% and +3.3%, respectively, longer than their trend-line values, and they both differ significantly from most of the other finches (Supporting Information, Table S4). The warbler finch has the significantly shortest wingspan of all species, -5% shorter than the trend-line value. The large ground finch has the second shortest wingspan (-2.1%), and differs significantly from all but the cactus finch, the vegetarian finch, and the woodpecker finch.

Hand-wing length (Table 1; Fig. 10; Supporting Information, S1, S2, S4):

The hand-wing length is a part of the wingspan, and the two regression diagrams are like each other. The regression coefficient for the hand-wing length vs. body mass (β = 0.320) is like that for geometric similarity (β_g = 1/3). The small and large tree finches have the longest relative length of the hand wing of the finches (+3.5% and +4%, respectively), and differs significantly from all the other finches but the ground finches and the vegetarian finch. The warbler finch has the shortest hand wing (-4%) and differs significantly from all but the large ground finch and the cactus finch.

The ratio hand-wing length to half wingspan vs. body mass (Table 1; Fig. 11; Supporting Information, Tables S1, S2) should be the same for geometrically similar species. However, the trait-mass slope for 2lhw/b (β = -0.030) does not follow that for geometric similarity (β_g = 0; P < 0.001), but all the species investigated differ less than about ± 1% from the mean trend. Figure 11 shows the linear regression for the relationship between 2lhw/b and body mass, with a more expanded y-axis. The observed values of this non-dimensional ratio varies from 0.639 in the cactus finch to 0.668 in the warbler finch, indicating that the hand wing makes up a larger part of the wing length, and thus the arm wing a smaller part, in the warbler finch than in the cactus finch and the other finches with low values. The small and large tree finches also have rather large values, 0.654 and 0.655, respectively, thus having slightly shorter arm wings in relation to the length of the wingspan than the larger species.

Wing area (Table 2; Fig. 12; Supporting Information, Tables S1, S2, S4):

The slope of the regression line for wing area (β = 0.672) is as for geometric similarity (β_g = 2/3). The small tree finch has the largest size-correlated wing area of the finches (+7.3% above the trend-line value), followed by the small ground finch (+3.6). The warbler finch has a s0 ignificantly smaller area in relation to the body size (-9.8%) than all the other finches apart from the large ground finch (-3.1%).

Wing loading (Table 2; Fig. 13; Supporting Information, Table S1):

The slope of the regression line for wing loading vs. body mass (β = 0.323) is as for geometric similarity (β_g = 1/3). We find the lowest wing loadings in the small tree finch (15.3 Nm^-2), the small ground finch (16.3 Nm^-2), and the warbler finch (16.5 Nm^-2). The vegetarian finch and the large ground finch have the highest wing loadings of the finches (22.5 Nm^-2 and 23 Nm^-2, respectively), which is a scaling effect of their large sizes.

Aspect ratio (Table 2; Fig. 14A, B; Supporting Information, Table S1):

The aspect ratio should be constant among geometrically similar species (β_g = 0) but increases with body mass among the species. The slope, β = 0.032, differs significantly from geometric similarity. This results from the slight increase of the wingspan with body mass than expected for geometric similarity, whereas wing area follows geometric similarity. All finches have very low aspect ratios, ranging from 4.6 to 4.9. The regression line for aspect ratio vs. body mass in the Darwin’s finches lies below those for most other flying birds and bats (Norberg, 1990: fig. 10.3). Figure 14A includes the regression lines for Greenewalt’s (1975) passeriform birds (p), shorebirds (s), and hummingbirds (h). Figure 14B presents a linear regression for the finches with expanded y-axis, showing that the Camarhynchus species and the cactus finch have somewhat higher ratios than the other species. The very short wings in the warbler finch contribute to its low value.

Tail length (Table 2; Fig. 15; Supporting Information, S1, S2, S4):

The slope of the regression line for tail length vs. body mass (β = 0.276) lies between those for geometric similarity (β_g = 1/3) and elastic similarity (β_e = 1/4). The tail is very short and weak, but rather wide in Darwin’s finches (Bowman, 1961). The vegetarian finch and has the longest tail of the finches (+8.2% longer than its trend-line value) and differs significantly from all the Geospiza species and the woodpecker finch. The warbler finch has the second longest tail (+2.6%) and differs significantly only from the cactus finch, which has the shortest tail of the finches (-3.9%), followed by the other Geospiza species.

Legs (Table 3, Figs 16-19; Supporting Information, S1, S3, S4):

Considering resistance to bending under the body load, the slopes of the regression lines for the leg bones would be expected to fit elastic similarity (β_e = 1/4) rather than geometric similarity (β_g = 1/3). In fact, the exponents are more like 1/4 than 1/3 although the exponent for the tibiotarsus length l_tt (β = 0.273) is slightly higher and the exponent for the tarsometatarsus length l_tmt (β = 0.236) is slightly lower than expected for elastic similarity.

The vegetarian finch and the warbler finch have the longest tibiotarsi in relation to body size of the finches (+6.5% and +3.9%, respectively, above the trend line), and the longest tarsometatarsi (+7.5% and +5.9%, respectively), and they differ significantly from all the other finches but the large tree finch (and, for the warbler finch, not from the woodpecker finch in tarsometatarsus length; Supporting Information, Table S4). The woodpecker finch also have long tarsometatarsi, +3.4% longer than its trend-line value. The small ground finch has the shortest tibiotarsus (-4.2%) and tarsometatarsus (-7.2%) of the species and differs significantly from all the other finches but the large ground finch and large tree finch (and not from the woodpecker finch in tibiotarsus length).

Tibiotarsus length/tarsometatarsus length, ltt/ltmt (Table 3; Fig. 18; Supporting Information, S1, S3, S4):

The ratio l_tt/l_tmt varies slightly among the finches and has a higher exponent (β = 0.038) than expected under geometric as well as elastic similarities (for which β_g =0 and β_e = 0). The ratio varies from 1.245 in the warbler finch (-2% below trend line) to 1.334 in the small ground finch (+3.7% above trend line). The small ground finch differs significantly from all the others but the medium ground finch, and the warbler finch differs significantly only from The Small and medium ground finches and the cactus finch.

Foot length (Table 3; Fig. 19; Supporting Information, S1, S3, S4):

The slope of the regression line for the foot length vs. body mass is significantly higher (β =0.295) than expected for elastic similarity (β_e = 1/4) and lower than expected for geometric similarity (β_g = 1/3). The woodpecker finch and the large tree finch have the longest feet of the species investigated with +5.5% and +4.0%, respectively, longer feet than their respective regression-line values. The woodpecker finch differs significantly from all but the cactus finch and the large tree finch, and the large tree finch differs significantly from all but the cactus finch and the woodpecker finch. The cactus finch also has long feet (+2.5%), whereas the other Geospiza species have short feet. The foot lengths in the vegetarian finch and the warbler finch are much shorter than their trend-line values (-3.9% and -4.5%, respectively). The warbler finch differs significantly from all the other finches.

Kleindorfer et al. (2006) found that the small ground finch in the lowlands of Santa Cruz has significantly longer feet and claws, and a shorter beak than in the highland forest. We found no such differences between our specimens from the coastal and the transitional zones, but the sample size for the transitional zone is small. Foot length for the small ground finch from the coastal zone is 23.60 mm (N = 15, SD = 1.07) and from the transitional zone 24.50 mm (N = 5, SD = 0.79). Our measures of the beak length for the same species are 12.34 mm (N = 15, SD = 0.44) from the coastal zone and 12.40 mm (N = 5, SD = 0.38) from the transitional zone.

It is interesting that the scaling exponents for the lengths of the tibiotarsus and tarsometatarsus vs. body mass are almost as expected under elastic similarity and therefore built to resist bending under their own weight, whereas wing measurements scale more according to geometric similarity and therefore are more frequently exposed to fractures under maximal muscle forces. Conversely, the feet do not follow either elastic or geometric similarities. Paws, claws, and hands do not obey any strict allometric rules, but can be adapted to other things than for avoiding bending forces and fractures (McMahon & Bonner, 1983). The short and flexible toe bones (tarsals) in Darwin’s finches seem not to be subjected to large forces, and the feet may instead in a freer way meet the requirements for better climbing and grasping. Another, or additional, explanation is that an overall regression line can be obtained as a ‘statistical artifact’, where data for different subgroups (for example, birds with long or short feet) fall on lines with different slopes (explained in Schmidt-Nielsen, 1984).

DISCUSSION

Overall comments on wing and leg structure and function

The low aspect ratios in Darwin’s finches cannot be adapted for high-performance aerodynamic function but must meet very different demands. The birds, however, are not bad flyers, but they need to flap their wings at higher frequencies, with higher costs for flight. We were able to measure the mass of the pectoralis and supracoracoideus muscles in four specimens of the medium ground finch and one specimen of the small ground finch. All had about the same body masses as the live finches measured. This spot check indicates that the flight muscle mass (of both muscles) makes up only 12–13% of the total body mass in both species. This is much lower than the average percentage for passerines (18%), hummingbirds (27%), and most other flying birds (Greenewalt, 1975), indicating less potential flying power. Climbing and clinging are also energy consuming (Norberg, 1986) and require considerable adaptations for minimization of energy costs.

The ecology and habitat choice of the finches (e.g. Lack, 1947; Bowman, 1961; Grant, 1999) most probably require that they can move easily by hops on the ground, take-off from the ground, and move with hops and short flights between branches and twigs amidst dense vegetation. High aspect ratio wings are a hindrance for movements among vegetation. During short flights the speed is in general below the maximum-range speed and the minimum-power speed (Pennycuick, 1975, 2008). The aerodynamic lift force varies in direct proportion to the wing surface area and with the square of the relative air velocity (Pennycuick, 1975), so low flight speeds must be compensated for by a large wing area. Therefore, the short wings of the finches must have a wide wing chord to make the wing surface area large enough to generate the lift required in slow flight and for tight manoeuvres among vegetation. The result is low-aspect ratio wings. The wings have evolved a wide wing chord from wing base and almost to the wingtip, which increases the area of the outer part of the wing and helps to increase total wing area. In this way the aerodynamic centre of pressure is moved farther out on the wing. The distal position of the aerodynamic centre of pressure improves roll acceleration and enhances manoeuvrability but makes the wingtip broad and rounded, which increases the wingtip trailing vortex, aerodynamic drag, and flight energy costs.

For a given flight mode, small flying species have lower flight costs than large species (Tucker, 1969) and are therefore more likely to use high-efficiency, high-cost search and capture methods and to exploit patches that require agility and energy-demanding locomotion modes, and they can also benefit from high-efficiency, energy-expensive foraging methods at lower food densities than larger species can do (Norberg, 2021). To compensate for increasing locomotion costs with increasing body size, the foraging efficiency must be higher the larger the animal is.

High flight costs may also be a hindrance for movements between islands unless strong winds would facilitate their transportation. This may have implications for finch dispersal among islands and thus facilitates divergencies within groups on an island.

The congeneric Geospiza species usually show more similarities with each other than with the other finches, both in real values and in relation to their body sizes. All four species have shorter legs than the other finches relative to body size. Both the small warbler finch and the large vegetarian finch have the longest legs and tail of the finches relative to body size. However, the large difference in body size makes a big difference in their foraging modes.

Manoeuvrable and agile flight

Agility during flight among vegetation is promoted by low body mass, low wing loading, and, if they often fly very slowly or hover during food snapping, a short arm wing (Norberg, 1979). Only the smaller species in this investigation have low wing loading for agile and manoeuvrable flight.

The small tree finch (Fig. 4), the warbler finch (Fig. 6), and the small ground finch (Fig. 1) have the lowest wing loadings of the finches. They are very agile, and we often observed them foraging among twigs in the periphery of the tree canopy. Although the wingspan and hand-wing length are long in the small tree finch and short in the warbler finch in relation to their body masses, both (and the small ground finch) have larger values of the hand-wing ratio 2lhw/b than the other Geospiza species and the vegetarian finch, which means that their arm wings are making up a smaller part of the total wing length. This may indicate that the evolution towards a shorter arm wing is favoured in the smaller finch species, improving agility and energy-saving flights by minimizing the inertial loads on the wing skeleton in slow flight.

Flight manoeuvres within vegetation may also be favoured by a large tail area to produce additional lift. The vegetarian finch has the significantly longest tail of the finches, 8.2% longer than its corresponding value on the trend-line across all the finches, followed by the warbler finch (+2.6%) Any increase of the tail length would add to the tail area, and more so when spread, and give more aerodynamic forces for manoeuvring. However, the large body size in the vegetarian finch rules out high manoeuvrability among vegetation, and its longer tail may be an adaptation mainly for longitudinal and directional control in level flight and for increasing lift during take-off and landing (Thomas, 1983; Taylor & Thomas, 2014).

Fulfilments of the predictions

Slow agile flight among vegetation:

As predicted, the small ground finch, the small tree finch, and the warbler finch have low body mass, low wing loading, and shorter arm wings in relation to the length of the wingspan than have the larger species. The warbler finch is the only one of these species that has the predicted short wings and quite a long tail.

Short flights between foraging sites within vegetation:

Among those which forage within vegetation the warbler finch and the large ground finch are the only ones with short wings in relation to their body sizes. Contrary to the prediction for flight within vegetation, the tree finches have longer relative wing lengths than any of the other finches.

Climbing

The woodpecker finch and the small and large tree finches climb on trunks and large branches during foraging, and we have observed that the small tree finch climbs both upwards and downwards. Because all of Darwin’s finches have short and weak tails, none of them has evolved tail-supporting climbing. Therefore, the climbing technique of the woodpecker finch is radically different from that of the woodpeckers. Adaptations for near-vertical climbing locomotion without tail support are long legs, long toes (long foot length), and long curved claws (Winkler & Bock, 1976; Norberg, 1986). The largest force the climber must resist is the horizontal force to keep the body close to the trunk (Norberg, 1986). Long legs and toes increase the distance between the points of attachment of the upper and lower foot, thus reducing the horizontal component of the pulling force between the claws of the upper foot and the trunk and the equally large horizontal pushing force that the lower foot must exert against the trunk (shown in Norberg, 1986).

The lengths of tibiotarsus and tarsometatarsus are longer than the trend-line values in the large tree finch, slightly longer in the woodpecker finch, and like the trend-line values in the small tree finch. All three have longer feet and claws than the other finches in relation to body size. We measured the claw curvature (according to the method in Feduccia, 1993) and claw length from photos of the third claw in a few specimens of each species, which may give a hint to their claw adaptations. Our estimates of the claw curvature for the tree finches, the woodpecker finch, and the cactus finch are 135–150°, which fall within the range for Feduccia’s trunk-climbing species (130–162°). The approximate length of the third claw is 20–22% of the foot lengths in these species, compared to 16–17% in the other finches.

Climbing technique in the woodpecker finch:

The climbing technique in the woodpecker finch, based on films, notes, and sketches, is here described. The woodpecker finch climbs vertical tree trunks upwards as well as downwards (Fig. 7A, B), just like the nuthatches (Sitta; Norberg, 1986). It places one foot above the other, usually more or less on the same vertical line along the trunk and seems to hitch upwards as well as downwards with the same ease. To facilitate this foot placement during climbing the finch usually orients the body obliquely sideways about 45° head-up to the vertical or 135° head-down depending on whether it moves upwards or downwards, in a similar manner to nuthatches. The body may also be kept vertical in a head-up or head-down attitude, particularly when the finch clings to the trunk and extracts insect larvae from the wood. This is an entirely different tree-climbing technique than that used by woodpeckers and treecreepers, which climb vertical trunks mainly in the upwards direction and always prop the tail against the trunk for support (Norberg, 1986).

The upper foot, and in particular the claws of the uppermost toes, pull on the trunk, whereas the lower foot pushes against it. The larger the distance is between the upper and lower points of foot attachment to the bark, the smaller the forces between the feet and tree trunk and the smaller the force that the finch needs to work against when climbing (described in detail in the treecreeper in Norberg, 1986). The rather long legs and long toes in the woodpecker finch enable it to keep a long distance between the points of attachment of the feet to the trunk. Judging by the way the legs are flexed when the woodpecker finch clings on to a trunk, the tarsometatarsus adds the most to lengthening the baseline between the feet during climbing, just as predicted, which also is indicated by its rather low value of the ratio between tibiotarsus and tarsometatarsus lengths (l_tt/l_tmt = 1.29, cf. below). The body needs to be kept close to the trunk so that the moment arm of the body weight is minimized to reduce the pull and push of the body that the feet must apply with respect to the tree trunk. With the climbing technique used by the woodpecker finch the legs need not move below the body and pass between the body and tree trunk during a stride, so the body can be kept close to the trunk by spreading the legs to either side of the body as required, without being short to reduce the distance to the trunk as in woodpeckers. With the woodpecker mode of climbing, the legs and feet must pass between the body and trunk both in the power and the recovery strokes during a vertical leap, which constrains the lengths of different leg elements (Palmgren, 1932; Winkler & Bock, 1976; Norberg, 1979; Norberg, 1983, 1986; Norberg & Norberg, 1989). The same reasoning applies to the tree finches and the warbler finch during clinging. This can be seen in photographs showing birds with one foot placed above the other (as in Figs 2, 5), just like the woodpecker finch does (Fig. 7A), and for the same mechanical and energetic reasons.

Indeed, the woodpecker finch is not particularly woodpecker-like. As already emphasized, it climbs without tail support, unlike that of woodpeckers. Second, while its beak approaches a woodpeckers in shape, it has not evolved a long tongue with which a woodpecker probes insects from crannies. Instead, it carries about a cactus spine or small twig, which it pokes into cracks, dropping the stick to seize any insect that emerges (Lack, 1953). We often noted that the woodpecker finch got itself a stick and only then went about searching for suitable holes to probe, rather than the other way around. It also picks with its open beak on the trunk. Bowman (1961: 39) suggested that since the woodpecker finch lacks long stiff tail feathers for support on the tree trunks it should not be named ‘woodpecker finch’. Because its climbing behaviour is more like that of a nuthatch, the woodpecker finch should perhaps be renamed the ‘nuthatch finch!’?

Fulfilments of the predictions for climbing

As predicted, the large tree finch and the woodpecker finch have rather long legs in relation to body size, especially long tarsometatarsi, and relatively low l_tt/l_tmt ratios. They also have long feet and long curved claws. The small tree finch has long curved claws and a low l_tt/l_tmt ratio, but both the tibiotarsus and tarsometatarsus are only slightly longer than the corresponding values on the trend line for the finches. The low body mass may facilitate its climbing behaviour. The cactus finch also has long feet and long, curved claws, but very short legs and a high l_tt/l_tmt ratio, so its feet may be more adapted for clinging and grasping than for climbing.

Clinging and hanging

Except for climbing, the small tree finch (Fig. 4) and the large tree finch often cling to the vegetation during foraging, sometimes upside down (Bowman, 1961). We have also observed that the cactus finch (Fig. 3B) and the warbler finch sometimes hang under branches when picking food items from the vegetation. Like in climbing, the characteristics for clinging are long toes and curved claws. In relation to its body size the cactus finch has very long toes and claws, whereas both short in the warbler finch. Still, the small size in the warbler finch facilitates a clinging behaviour.

Hanging would be favoured by a short tarsometatarsus to reduce muscle forces needed for keeping the legs flexed to be able to reach a food particle (Palmgren, 1932). However, there is no clear adaptation in this bone for hanging in these species. All four species have different leg lengths in relation to their body masses, but all the finches have longer tibiotarsi than tarsometatarsi, and in different proportions (Table 3). Like the other Geospiza species, the cactus finch has very short legs, and they have the highest ratios between the tibiotarsus and tarsometatarsus (l_tt/l_tmt = 1.32–1.33). The warbler finch has very long legs, but the lowest l_tt/l_tmt ratio (1.25), which means that the tarsometatarsus makes up a larger proportion of the leg length in the warbler finch than in the Geospiza species. There is a gradient in this ratio in the clinging and hanging species, from 1.25 in the warbler finch, 1.27 in the small tree finch, 1.28 in the large tree finch, to 1.32 in the cactus finch, indicating that the tarsometatarsus adds less and less to the leg length the larger the ratio becomes. A shortening of the tarsometatarsus (with maintained leg length) under the evolutionary pathway towards lower costs for hanging therefore does not seem to be important in these species. Longer legs for climbing, clinging, take-off from and landing on branches, and increase of step length seem to be more important for the species foraging within vegetation. Figure 6B demonstrates the use of the long legs in the warbler finch during taking off from a small branch within vegetation.

Fulfilments of the predictions

Clinging (warbler finch, small and large tree finches, cactus finch):

The tree finches and the cactus finch have long curved claws, but not the warbler finch.

Hanging upside down under branches (small tree finch, warbler finch, cactus finch):

The only one of these species that has shorter tarsometatarsi relative to its body size and a higher l_tt/l_tmt ratio, as predicted, is the cactus finch.

Diversification in beaks vs. locomotion organs

The observed combination of morphological and ecological traits is usually assumed to represent a near-optimal solution maximizing the fitness of the individual. Optimal models are often built to maximize net energy intake and/or to minimize foraging time and energy expenditure, which are optimization criteria for fitness (Pyke et al., 1977). If maximization of net energy intake has been more important during the evolution of Darwin’s finches than minimization of energy expenditure during foraging, then we may expect larger differences among the finches in beak morphology than in wing and leg structures.

Darwin’s finches show a large variation in beak size and shape, enabling them to exploit different food and sizes of food (Grant et al., 1985). Grant (1999) showed that all populations of coexisting ground finches differ by at least 15% in at least one beak dimension, and the greater the difference between two species, the greater their dietary difference.

The differences in wing and leg structures are usually smaller than in beak morphology. In relation to body size, wingspan differs at most by 9% but wing area (which is related to the wingspan) as much as 17% between the smallest and largest species, which has large implications for flight. Tail length differs by 12% between the vegetarian finch and the cactus finch. The size-corrected tarsometatarsus length differs in length by 15% between the small ground finch and the vegetarian finch, and foot length by 10% between the warbler finch and the woodpecker finch. In relation to body size, all four Geospiza species have legs and toes with rather flattened claws (UM Norberg & RÅ Norberg, pers. obs.), except for the cactus finch which uses a different foraging strategy. The variation among the ground finches is at most 5% in tarsometatarsus length and 4% in wingspan, and least in foot length (0.6%).

The selection pressures towards different adaptations in beak structure seem to have been more important than adaptations in wing and leg structures during the evolution of Darwin’s finches. Crucial factors affecting the evolution of the finches may have been the relationship between energetic demands, such as energy intake vs. energy expenditure during foraging, and also how easily different structures can change.

‘Overall, the warbler finch has evolved much closer to a warbler than the woodpecker finch has to a woodpecker’ (Lack, 1953).

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article on the publisher's website.

Table S1. Power functions of wing, tail, and leg characters (y) vs. body mass (M) of nine species of Darwin’s finches on Santa Cruz. The slopes of the regression lines are compared to those for geometric and elastic similarities. r is the correlation coefficient.

Table S2. The size-corrected (relative) values y_rel = y_obs/y_trend for wing and tail measurements in nine species of Darwin’s finches on Santa Cruz. N is number of specimens in sample, X is mean value, SD is standard deviation.

Table S3. The size-corrected (relative) values y_rel = y_obs/y_trend for leg measurements in nine species of Darwin’s finches on Santa Cruz. N is number of specimens in sample, X is mean value, SD is standard deviation.

Table S4. Statistical tests of the differences between the size-corrected (relative) values, y_rel= y_obs/y_trend, for some wing, tail, and leg characters for males and females combined. Each species is tested against each of the other finches with a Student’s t-test. Significance levels: +++, P < 0.001; ++, P < 0.01; +, P < 0.05; ns, not significant.

ACKNOWLEDGEMENTS

We are grateful to the Charles Darwin Research Station for permission to do research on Santa Cruz and for lodging. We are thankful to Robert Bowman for lending us tape recordings of the finch songs and for information on the finches and their best locations on Santa Cruz. We are indebted to two anonymous referees for valuable suggestions for improvements of the manuscript. We further thank Peter and Rosemary Grant for inspiring discussions, and Peter Grant for valuable comments on a first draft of this manuscript. We also thank our son Björn Norberg for help with the drawings. The research was funded by the Swedish Natural Science Research Council (Nos B-BU 4455 to U.M.L.N. and B-BU 4450 to R.Å.N.). We have no conflicts of interest to declare.

DATA AVAILABILITY

Tables with separate data on wings, tail, and legs for males and ‘brown’ birds (females and adult young) are available from the corresponding author on request.

REFERENCES