Testing the influence of crushing surface variation on seed-cracking performance among beak morphs of the African seedcracker Pyrenestes ostrinus

The African seedcracker Pyrenestes ostrinus (Estrildidae; Fig. 1) is distributed across West and Central Africa, an area with a tropical continental climate and little variability in the annual rainfall (Smith, 1987; Smith et al., 2001). It is well known as a classic polymorphic taxon, for which beak size and shape vary across three distinct yet interbreeding morphs produced by a single diallelic locus (Grant, 1986; Smith, 1987, 1993; Slabbekoorn and Smith, 2000; von Holdt et al., 2018). The three morphs vary primarily in mandible width, and are referred to as small (<14 mm), large (14–17.5 mm) and mega (>17.5 mm) (Fig. 1; Grant, 1986; Slabbekoorn and Smith, 2000; Smith et al., 2001). The coefficient of variation of the bill size is at least as large as in Darwin's finches, in which the variation is mainly referred to distinct species (Smith, 1987, 1993; Delaney et al., 2005). Unlike other polymorphic avian species, the Pyrenestes polymorphism is completely independent of geographic origin (Smith, 1987, 1993; Grant and Grant, 2002), sex (Smith et al., 2001), size (Grant and Grant, 2002) and age (Smith, 1997).

A recent genetic study (von Holdt et al., 2018) demonstrated that the differences in beak size between the small and large morphs can be explained by a simple genetic step, while the transition to the mega morph is likely driven by more complex controls. By contrast, in Darwin's finches, hundreds of loci are involved in maintaining their phenotypic diversity (Lawson and Petren, 2017).

The diet of P. ostrinus consists largely of three types of sedge seeds of varying hardness: soft (Scleria goossensii), hard (Scleria verrucosa) and very hard (Scleria racemosa) (Fig. 2; Clabaut et al., 2009). When the seeds are abundant, all morphs prefer the small, soft seeds, which are easiest to crack and therefore offer the most gain for least effort (Smith and Skúlason, 1996; Smith et al., 2001). During times of food scarcity, the large and mega morphs specialize on the hard and very hard seeds, respectively (Fig. 2B,C), whereas the small morphs adapt to a more generalist feeding habit including other grass seeds (Smith, 1997).

Pyrenestes’ polymorphism may be related to the different Scleria-based diets altering feeding performance (Skúlason and Smith, 1995; Smith and Skúlason, 1996; Slabbekoorn and Smith, 2000). The inferred ecological hypothesis is that the larger morphs are able to crack the harder seeds more efficiently than can the smaller morphs, which is driving the polymorphism.

Although previous analyses on food-processing structures often focused on mammalian groups, the basic principles can be generalized to any surface morphology used to crack stiff, resistant food items such as seeds (Benkman, 1987, 1988; van der Meij and Bout, 2000). The morphology of the surface used for food processing can have a strong effect on fracturing food (Evans and Sanson, 1998, 2003; Lucas, 2004; Ungar, 2004; Freeman and Lemen, 2006; Berthaume et al., 2010; Crofts and Summers, 2014; Kolmann et al., 2015). Several features characterize the surface morphology of the seedcracker beak, such as the size of the entire crushing area, as well as the width of the ridges and depth of depressions on the crushing surface. The constellation of ridges and depressions may also be important in fracture ability.

It is likely that the crushing surface morphology of the seedcracker beaks evolved specifically to resist fracture rather than for force application, the latter being the function of the jaw muscles (Clabaut et al., 2009). Therefore, the beak morphology is most likely optimized to simultaneously enhance force application to the seed while avoiding fracture in the beak by dissipating stress. Because the assumption that beak morphology predicts feeding ecology has rarely been quantified (Navalón et al., 2019), our aim was to assess whether the differences in width and morphology of the beaks' crushing surface seen in the three seedcracker morphs affect the efficiency of the beak for cracking the three different seed types. We therefore tested the direct interaction between the crushing surface, seed type and the work (energy) involved in seed cracking, not the bite force itself.

Here, we performed a series of biomechanical fracture experiments on different seed varieties using tools that approximate the bony crushing surfaces of Pyrenestes beak morphs. We reasoned that the less energy it takes to crack the seeds, the less danger to the beaks' structural integrity. Furthermore, if beak surface morphology is related to feeding ability, we would expect morphological differences in crushing surfaces to affect the work (energy) required to fracture a seed. Specifically, we predicted that tools that resemble mega morph bill morphology should crush the mega seeds for a lower work expenditure than would tools that resemble the small and large morphs.

One skull of each of the three different morphs and accompanying computed tomography (CT) scan data of Pyrenestes ostrinus (Vieillot 1805) were available to scrutinize differences in beak shape and surface morphology (provided by Thomas B. Smith, UCLA, Los Angeles, CA, USA; small morph 4606, large morph 5617 and mega morph 5103). The mandibles were separated from the skulls by dissecting out the jaw muscles. The most obvious difference between the beaks is the size, which can be observed most clearly in total mandible width (Fig. 1, Fig. S1A, Table S1). The crushing surfaces of the mandibles differ in width and curvature in cross section. The widths of the crushing surface on the mandible were 2–3 mm in the small morph, 3–4 mm in the large morph and 4–5 mm in the mega morph (Table S1). The curvature variations were as follows: the small morph has a flat, horizontal crushing surface (Fig. 2A); the large morph has a slightly curved crushing surface, whereas the central part of the mandible is nearly flat and becomes more rounded towards the labial and lingual sides (Fig. 2B); in the mega morph, a strong curvature can be observed, whereas the most curved part is in the central part of the mandible forming a tip and becoming less strongly bent towards the labial and lingual sides (Fig. 2C). Measurements of different areas of the mandible were taken with a digital calliper with a resolution of 10 µm. A detailed description of the morphology is provided in Fig. S3 and its legend; measurements are shown in Table S1.

To consider whether it would still be possible for the jaw muscles to apply close to maximal bite force, the distance between the tip of the upper and the lower beak (gape) was measured with a digital calliper while placing each seed type, one at a time, between the beaks at the place where they are cracked. This was done after the dissection of the jaw muscles, but with the jaw joints in contact with each other. The seed, the lower beak and the upper beak were held with one hand, and the gape was measured with the other hand. This procedure visualized how wide each of the beaks have to be opened to hold each seed type.

Samples of S. goossensii, S. verrucosa (from Ndibi, Cameroon) and S. racemosa (from Wakwa, Cameroon) were collected by T. B. Smith ∼10–20 years ago. The three types of sedge seeds – S. goossensii (soft), S. verrucosa (hard) and S. racemosa (very hard) – are easy to distinguish by their size and morphology (Fig. 2). Scleria goossensii has a soft pericarp with a rough surface and tiny spikes. The mean diameter is 2.226 mm (N=12), and the cupule is only one-third of the seed size. Crushing the soft seed, the pericarp usually stayed in one piece and only burst open at one spot. The hardness is ∼13 N (Clabaut et al., 2009). Scleria verrucosa has a rough pericarp surface with remarkable thorns and spikes. The cupule is nearly as large as the whole seed. These hard seeds have a mean diameter of 2.878 mm (N=12), crushed in a brittle manner and have a hardness of ∼153 N (Clabaut et al., 2009). Scleria racemosa has a smooth, shiny pericarp with a cupule, which is nearly as large as the seed. The mean diameter is 4.850 mm (N=12). They also crushed in a brittle manner and have a hardness of ∼299 N (Clabaut et al., 2009).

The experiments were undertaken with a custom testing device, originally developed as a double-bladed guillotine (Anderson and LaBarbera, 2008; Anderson, 2009; Anderson and Rayfield, 2012; Fig. S1C) that combines aspects of a single-blade guillotine (Atkins and Mai, 1979; Veland and Torrissen, 1999) and the scissors tests (Atkins and Mai, 1979; Pereira et al., 1997; Lucas, 2004; Anderson, 2009). This device allows for testing materials to be compressed between two custom-designed tools aligned against each other. Although previous work used custom-designed blades for tools, we used tools designed according to beak morphology (described below). One tool is set on an upper moveable platform, which is attached to a large screw (20 mm diameter, 3 mm pitch; SKF, Göteborg, Sweden) anchored to the base box. The counterpart tool is mounted on a force transducer (LC703-100; Omegadyne, Sunbury, OH, USA) that is attached to the box, but can be moved forwards, backwards and in lateral directions to ensure a precise alignment between the upper and lower tools. The force transducer measures the force needed to crack the seeds (in eV). Displacement differences between the upper platform and the box were measured (in mm) with a linear displacement sensor (HS50 Linear Displacement Sensor LDS; Vishay Measurements Group UK Ltd, Basingstoke, UK), which is attached to the side of the machine. An AC induction motor (SD18M; Parvalux, Bournemouth, UK; including a gearbox) moves the upper platform up and down with a constant velocity (1.9 mm s⁻¹ by default) by rotating the screw. The part where the tools were mounted on the upper platform had to be stable and fixed to avoid tilting while cracking the seeds. Objects to be fractured are placed in the middle of the two tools (Anderson, 2009).

Two different sets of tools were made of high-quality tool steel (RS Components, Corby, UK) to represent two different aspects of the crushing surfaces of the mandible in Pyrenestes morphs; one set varying in width (2 mm, 3 mm, 4 mm and 5 mm) and another set varying in curvature (flat, weak curvature, strong curvature; Fig. 2). The specific curvature of this latter set was based on curvature measured from CT scan images provided by Annelies Genbrugge (Ghent University, Ghent, Belgium). Each of these tools was paired with an upper flat tool in the double guillotine. This is a reasonable simplification for the curvature tests, as the upper beaks are not curved in the same way as the mandibular surfaces (see Figs S1–S4 and Table S1). The material properties of steel tools differ from those of the beaks' keratinous sheaths; therefore, there will be differences in stiffness, i.e. the steel tools are expected to be more robust and rigid than the beaks.

In addition, we ran a series of experiments with a three-point bending tool, which has two triangular ridges with a V-shaped dent between them on the lower platform opposed by a single V-shaped ridge on the upper platform. The ridge and groove morphology of the beaks' crushing surfaces indicates that this might be applicable in cracking the seeds (see Fig. S1B). With this set-up, we tested whether tools resembling a pestle-and-mortar structure (Lucas, 2004) provide different results than the more simplified tools above.

Before every run, the force transducer and the displacement sensor had to be re-calibrated and zeroed, as the start height of the upper platform was different every time and to avoid biases from the force transducer. It was ensured that the linear displacement sensor touched the box. Seeds to be fractured were placed between the two tools, recording started, and the motor moved down the upper platform. The machine was stopped immediately after cracking the seed, and recording was stopped. Seed remains were removed, and everything was reset to the starting position. The signals from the force transducer and the displacement sensor were detected and amplified by a scanner (Vishay, 5000 series 5100B), and converted into force and displacement. The conversion is based on factory-set calibrations of the sensors (Anderson, 2009).

Each tool was tested on the three different types of seeds. As the seeds are a natural product, there is variability in their physical properties. Therefore, each tool–seed combination was tested 12 times to obtain a statistical mean. The peak force values (in N) of each run were also collected to compare the forces required to crack the seeds.

As the area under the curve is equal to the work done, the work value is the sum of the integral (I) values from the start to the end point (Fig. S2). The work values are important for comparisons of the effectiveness of the different set-ups, because the work represents how much energy it takes to break the seeds. This again is important, as it is in an organism's interest to maximize net energy gain (energy in food item minus energy cost of food acquisition). Hence, energy is used in comparisons between the different morphs.

We performed a two-way ANOVA, including a post hoc Tukey's honestly significant difference test, on the work values of each seed–tool combination using R (https://www.R-project.org/). The categorical variable ‘seed’ had five levels (soft, hard₁, hard, hard₂, very hard) and the categorical variable ‘tool’ had eight levels (2 mm, 3 mm, 4 mm, 5 mm, flat, weak curvature, strong curvature, three-point bending).

Regardless of the tool used, the work required to crack the three different types of seeds differed by about one order of magnitude (Table 1, Fig. 3). Differences in cracking behaviour were also observed during the experiments: soft seeds were compressed flat, whereas hard and very hard seeds generally broke into multiple pieces. A subset of the hard seeds, S. verrucosa, fractured in a manner similar to the soft seeds at intermediate forces between the harder S. verrucosa seeds and the softer S. goossensii seeds (see Tables S2–S4 for details). Thus, it was possible to identify two different types of hard seeds: hard₁ and hard₂.

Each seed–tool combination was tested 12 times; because of the bimodal split of the hard seeds, sample sizes for the hard₁ and hard₂ seeds were smaller than 12 in most cases, according to our observations and the bimodal distribution of the values. The mean work value for S. goossensii was ∼0.0030 J. Work values ranged from 0.0105 J to 0.0296 J in the hard₁ seeds and from 0.0451 J to 0.0893 J in the hard₂ seeds. For S. racemosa, the work values were between 0.1727 J and 0.2754 J (Table 1; Table S3). The highest measured peak force value was 445.89 N (S. racemosa). The mean peak force values for S. goossensii were between 7.0 N and 10.0 N. For S. verrucosa (hard₁), they ranged from 23.91 N to 34.07 N; for S. verrucosa (hard₂), they ranged from 91.51 N to 128.76 N; and for S. racemosa, they were between 218.91 N and 313.69 N (Fig. 3; Table S2).

The two-way ANOVA revealed statistically significant difference in work by seed type (F_4,28=485.50; P≪0.001) and less significant difference in work by tool type (F_7,28=2.39; P=0.02). The interaction between seed and tool type was not significant. The degrees of freedom of the residuals is 330. The post hoc Tukey test showed a significant pairwise difference between almost all seed–seed combinations (P≪0.05); for the hard₁–hard and hard–hard₂ combinations, the amount of variation was significant, but with larger P-values (P=0.001), and for the soft–hard₁ combination, the amount of variation was not significant (P=0.06). The pairwise difference of all tool–tool combinations was not significant (P>0.05), except for the 4 mm–three-point (P=0.02) and the flat–three-point (P=0.01) combinations. The complete results of the ANOVA including P-values of the interaction between seed and tool types are provided in Dataset 1.

Thus, regardless of the seed type tested, there were no remarkable differences in work to fracture using tools based on the different crushing surface morphologies. Neither tools mimicking differences in surface width nor curvature had significant impact on the work values calculated (Table 1). Thus, the tools representing different crushing surface morphologies had no differentiated effect on seed-cracking efficiency.

The three-point bending tool had the lowest energy for the very hard seeds and the hard₂ seeds but was less economical for the soft and hard₁ seeds (Table 1), which might be caused by the smaller seeds fitting between the triangular ridges and, thus, having a lateral restriction rather than a three-point bending set-up.

For the small morph, the gape was negligible with the soft seed, 4 mm with the hard seed and 9–10 mm with the very hard seed. The large morph showed a negligible gape with the soft seed; a gape of 3.5 mm with the hard seeds, which also fit between the upper and lower beak without wobbling; and a gape of 9.5 mm with the very hard seed, which also fit between the beaks (without the cupule). For the mega morph, there was a negligible gape with the soft seed, a gape of 2–3 mm with the hard seed and a gape of 8.5 mm with the very hard seed. The very hard seed did not fit as well between the upper and lower beaks of the mega morph as the hard seed did between the upper and lower beaks of the large morph.

Scleria goossensii smashed easily without much resistance. For S. verrucosa, we observed that the hard seeds could be divided into two types of seeds. One type (hard₂) cracked with an audible noise and resistance; the other type (hard₁) smashed rather than cracked, showing very little resistance. Differences in the mean work value between the two types are significant for all tool pairings (ANOVA, P≪0.05). The reason for this observation is not evident; it could not be predicted from external morphology whether the seed will crack or smash. Both types of hard seeds always had intact fruits; hence, degeneration of the seeds is not the reason (Fig. S4). Scleria racemosa showed either cracking or crushing behaviour accompanied by a loud cracking noise and strong resistance (Fig. S4). Nonetheless, in some of the very hard seeds, very little resistance was observed. These seeds contained dust inside instead of a proper fruit and were white on the outside instead of brown; hence, it was predicable whether a seed would show resistance. It is very likely that the white, very hard seeds were poorly preserved, and thus these seeds were excluded from the analyses. Even though poorly preserved seeds could be identified in most cases, there is still some possibility that intact seeds might behave slightly differently compared with the aged seeds used in this study.

Conducting the experiments with the three-point bending tool, the very hard seeds were divided into two half spheres, preserving the fruit as a whole (Fig. S4). With all other tools the very hard seeds and their fruits were crushed or cracked, respectively.

Our experimental results fail to support the hypothesis that variations in the width and morphology of the crushing surfaces of the mandibles in P. ostrinus will have an influence on seed-cracking efficiency. The crushing surface morphology of the mega morphs is no more efficient at cracking the hardest seeds than the crushing surface morphology of the small morphs. By focusing on the crushing surfaces, we isolate morphology that is closely associated with seed reduction. However, the morphological variation of the crushing surfaces we observed appears to have little influence on this presumed function. Given that the polymorphism is apparently tied to feeding and resource acquisition (Smith, 1987, 1997; Navalón et al., 2019), we propose that other morphological, biomechanical or genetic mechanisms may be responsible for triggering and maintaining the polymorphism expressed on the food-processing surface of the mandible.

For the three-point bending tool, the work-to-fracture values were lower when cracking the hard₂ and very hard seeds, in the same range for the hard₁ and hard seeds, and higher for the soft seeds, compared with the values obtained with the other tool types (Table 1). The three-point bending tool was built to create the three-point bending effect on the very hard seeds, which fit on top of the triangular ridges of the tool. The soft seeds, however, fit between those ridges, creating a lateral restriction, which probably explains the higher work values, because the two tools may have come into contact during the cracking. For the very hard and hard₂ seeds, crushing surfaces similar to those of the three-point bending tool seem to be advantageous. It is possible that the space between the upper and lower beak (when beaks are closed) resembles a three-point bending or pestle-and-mortar structure, at least for the mega morph and potentially for the large morph (Lucas, 2004). Also, the ridges on the mandible may be advantageous for initiating a crack in the seeds.

Although small beak tools can crack large seeds in our experimental setting, this does not mean that the smaller morphs possess the muscle mass required to generate such force. The relatively wide gape required for cracking the very hard seeds would be a challenge for certain birds, especially for the small and large morphs, but apparently possible for the mega morph; the size of large food particles eaten is correlated with gape width in some birds (Wheelwright, 1985). From our observations, it is questionable whether the small morph would be able to produce enough force to crack the very hard seeds with such a wide gape, as the adductor muscles might be too stretched to be contracted to effectively close the beak, and the angle between the force vectors might be too wide to take effect in cracking the seed. Furthermore, the properties of the keratinous beak are certainly also important, but the keratin sheath was not considered in our experiments. In vivo observations of the three morphs feeding on the three seed types would be helpful to support our findings.

Foraging efficiency is dependent on beak morphology; for example, beak depth influences bite force and deep beaks are crucial for husking hard seeds (Benkman, 1987). The beak morphology and size correspond to optima for times of limited food supply. The beak size and associated musculature also influence seed-husking ability and represent the upper limit for the largest and hardest seeds (Benkman, 1987, 1993). The optimal husking groove width of the palatine grooves of the keratinous beak are essential for securing seeds in the beak while seed husking (Benkman, 1987, 1993).

The individual husking technique of the finches should also be considered. Increased seed size makes husking and cracking more difficult; thus, the number of positioning movements before cracking increases with the size of the seed, and, therefore, seed husking time increases with seed size and hardness (Benkman, 1987; van der Meij et al., 2004). Large seeds are more difficult to position than small seeds, which affects the number of cracking attempts. There are also more cracking attempts when the size of the seed does not fit the husking grooves well, which may also result in a loss of the seed (van der Meij et al., 2004). The small morph, for example, would probably drop a seed that is too large or hard to crack rather than risk beak fracture in the cracking attempt. It should be noted that pre-cracked seeds might get picked up by another bird, and would then likely require fewer cracking attempts, because multiple cracking attempts lowers the strength of the seed husk (van der Meij et al., 2004). In this case, the force to eventually crack the seed, particularly the very hard seeds, might be slightly lower than gathered from our experiments (van der Meij et al., 2004).

A comparative study on finches and sparrows found that finches are capable of feeding on comparatively large/hard seeds in comparison to sparrows (Benkman and Pulliam, 1988). This means that sparrows have to intake more (small) seeds and finches are more vagile, covering large areas in search of seeds. Because small and large finches eat very different seed sizes, resource partitioning is more pronounced in finches than in sparrows, and the diets of small and large sparrows overlap (Benkman and Pulliam, 1988). This supports the observations in P. ostrinus.

Differences in size between morphs may also be correlated with the increased muscle mass in the skull necessary to crack harder seeds. In finches, beak dimensions and head width are correlated with bite force (van der Meij and Bout, 2004, 2006; Herrel et al., 2005a,b), which is dependent on the cross-sectional area of the jaw adductor muscles (Bowman, 1961; van der Meij and Bout, 2004, 2008). Similar relationships between beak size and bite force can be observed in P. ostrinus, but body size variations are independent of beak size variations (Grant and Grant, 2002). Finite element studies of mechanical stress in the beaks of Darwin's ground finches (Soons et al., 2010, 2012, 2015) suggest that beak morphologies evolve due to selection for avoidance of fractures and are optimized to resist natural loading. Thus, different beak shapes and crushing surface morphologies are ideal for mitigating risk of fracture with the species' preferred diet (Soons et al., 2015). Beak size in the large and mega morphs, which is independent of overall body size (Clabaut et al., 2009), may simply be to reinforce the beaks when cracking the much harder seeds.

It is likely that these three factors – beak size, muscle size and bite force – correlate with each other in the African seedcrackers, promoting the different cracking forces/abilities. Size-independent factors such as the depression angle of the bill (van der Meij and Bout, 2008) can influence the bite force as well. A thick rhamphotheca, the keratinous sheath that surrounds the jaw bones and forms the actual beak, a strong downward inclination of the upper beak and a strongly developed jugal have been shown to be important features in increasing bite force (Bowman, 1961; van der Meij and Bout, 2008; Soons et al., 2012). It is possible that the differences in beak size between the Pyrenestes morphs are caused by differences in jaw musculature, which may reflect and influence differences in feeding performance between these morphs. Hence, soft tissues features, rather than the rhamphothecal morphology, may be driving selection and the maintenance of the polymorphism in African seedcracker finches.

Our results show that the relationship between the beaks' crushing surface width and curvature and feeding ecology is more complex than anticipated, which was also demonstrated in a study on modern birds (Navalón et al., 2019). However, Pigot et al. (2020) showed that the avian trophic diversity could be described by a traitspace using only four dimensions. Beak shape in seed eaters may be subject to selection to resist feeding forces and reduce fracturing (Soons et al., 2015); thus, understanding how beak polymorphisms evolve will help to explain the patterns of selection on beak size and shape in natural selection.

We thank M. Dury and A. Povey from the School of Earth Sciences workshop (University of Bristol) for constructing beak tools and S. Powell (University of Bristol) for photography. We thank A. Genbrugge, who provided the CT scans; T. B. Smith, who collected and sent the skulls of the birds and the seeds; and L. Sineux for the photograph of the finch. T. B. Smith and A. Herrel kindly provided helpful comments and discussion. We also thank D. Sustaita and one anonymous reviewer for their constructive comments on the manuscript.