Energetic costs of feeding in 12 species of small-bodied primates

Abstract

There are no comparative, empirical studies of the energetic costs of feeding in mammals. As a result, we lack physiological data to better understand the selection pressures on the mammalian feeding apparatus and the influence of variables such as food geometric and material properties. This study investigates interspecific scaling of the net energetic costs of feeding in relation to body size, jaw-adductor muscle mass and food properties in a sample of 12 non-human primate species ranging in size from 0.08 to 4.2 kg. Net energetic costs during feeding were measured by indirect calorimetry for a variety of pre-cut and whole raw foods varying in geometric and material properties. Net feeding costs were determined in two ways: by subtracting either the initial metabolic rate prior to feeding or subtracting the postprandial metabolic rate. Interspecific scaling relationships were evaluated using pGLS and OLS regression. Net feeding costs scale negatively relative to both body mass and jaw-adductor mass. Large animals incur relatively lower feeding costs indicating that small and large animals experience and solve mechanical challenges in relation to energetics in different ways.

This article is part of the theme issue ‘Food processing and nutritional assimilation in animals’.

1. Introduction

Understanding the role of energetic costs in the feeding system is important for relating selective pressures to morphology and behaviour and facilitates broader comparisons between biomechanical systems. While feeding energetics have been the subject of numerous studies [1–14], there are no comparative, empirical studies of the energetic costs of feeding in mammals. Here, we test how the energetic costs of feeding vary in relation to body mass and jaw-adductor muscle mass and with food properties in a comparative sample of 12 non-human primate species ranging in size from 0.08 to 4.2 kg.

Mammals have a specialized feeding behaviour, mastication, and a wide range of food procurement and processing behaviours. They possess a suite of specialized structures of the oropharynx that function during ingestion, mastication, food transport and swallowing [15]. The mammalian feeding system reflects adaptation for a power stroke or slow closing phase of the gape cycle that commonly involves lateral jaw movements to precisely occlude the postcanine teeth and is characterized by asymmetric activation of the jaw elevator muscles ([16–22]; but see [23]).

As in other animals, mammalian feeding and digestion are overlapping processes. Moreover, digestive costs anticipate food ingestion [24,25]. These overlaps make it difficult to study mastication in isolation using a non-invasive, naturalistic experimental set-up in non-human mammals. In a recent study on humans, the energetic costs of mastication were isolated from other feeding behaviours, anticipatory digestive costs, and resting metabolic rate (RMR) using respirometry and electromyography with an odourless, flavourless gum as the chewing substrate [26]. This study found that stiff gum incurred higher masticatory costs than elastic gum (15% increase versus 10% increase compared to initial RMR), though the potential contribution of digestive costs could not be evaluated because postprandial metabolic rates were not reported. Similarly, Laird et al. [27] found that tough/stiff foods were associated with higher mastication work costs in humans measured using jaw kinematics and electromyography.

We model the net energetic cost of feeding as a function of time, the tools used to accomplish that behaviour (e.g. activation of striated muscle, jaw-muscle leverage, tooth morphology) and substrate (i.e. the food). Using this framework, net feeding costs per unit time should be strongly influenced by activation of the jaw-adductor muscles because they generate the force for food breakdown. As indicated by studies on human mastication and others on specialized feeding behaviours [2,4,9,14,26,27], the net feeding costs should also correlate with the properties of the food item reflecting the feeding substrate.

In this paper, we determine the scaling relationships of net feeding costs to body mass, jaw-adductor mass and food properties. Non-human primates are a particularly appropriate mammalian clade to contextualize the energetic costs of mammalian feeding as their feeding morphology, biomechanics and diet are well studied. The jaw musculature and other aspects of oropharyngeal anatomy of primates are often described as generalized and lacking the specializations of many other mammalian clades, such as Carnivora and Rodentia [22,28]. Moreover, primates are a model group for in vivo and theoretical studies of the biomechanics of mammalian feeding, and there is a wealth of data on muscle activation, joint kinematics, bone strain and isometric bite force (e.g. [29–37]). Similarly, the foods primates eat in the wild and in the laboratory have been well studied (e.g. [38–41]), and can be quantified in terms of their mechanical and geometric properties.

Our sample was chosen to include generalized species that exhibit a substantial range of the variation in dental morphology, jaw leverage and diet seen in extant primates in addition to species with highly specialized diets and/or oropharyngeal morphology. Dietary specialists in this study include Daubentonia madagascariensis—a gnawing specialist [42], Hapalemur griseus—a bamboo specialist [43], Propithecus coquereli—a folivore/granivore [44–47] and Sapajus apella—a hard-object specialist [48]. This sample provides a framework to examine feeding costs in relation to body size and jaw-adductor size and to explore the influence of food geometric and material properties (FGPs and FMPs, respectively, table 1). We test the following hypotheses.

Table 1.

Commonly used abbreviations.

abbreviation	definition
E	elastic modulus, stiffness (the relationship between elastic stress and strain, MPa)
FGP	food geometric property
FMP	food material property
FMR	metabolic rate during feeding (J s–1)
IMR	initial metabolic rate (J s–1)
mandible length	the length of the mandible measured parallel to the long axis of the corpus from infradentale to the posterior margin of the ascending ramus (mm)
minimum diameter	minimum diameter of initial food size (mm)
maximum diameter	maximum diameter of initial food size (mm)
netFMRIMR	FMR-IMR (J s–1)
netFMRPMR	FMR-PMR (J s–1)
PMR	postprandial metabolic rate (J s–1)
R	toughness (work needed to propagate a crack through a food item, J m−2)

2. Hypotheses

Hypothesis 1: Scaling of net feeding costs (J s⁻¹, dependent variable) is proportional to body mass (kg, independent variable). H₀: isometry (feeding costs ∝ body mass); expected slope = 1.0.

The two alternative hypotheses are negative allometry (slope < 1.0) and positive allometry (slope > 1.0). Negative allometry will indicate that feeding costs decrease as body mass increases.

Hypothesis 2: Scaling of net feeding costs (J s⁻¹, dependent variable) is proportional to adductor mass (g, independent variable). H₀: isometry (feeding costs ∝ jaw-adductor mass); expected slope = 1.0. Testing H2 makes the assumption that the jaw-adductor muscles comprise the largest source of feeding costs as measured here and/or that the contribution of activation in the other muscles (e.g. facial and limb muscles) to net feeding costs scales similarly.

There are two alternative hypotheses. If, as adductor mass increases, proportionately less jaw-adductor mass is activated during feeding, then the expected slope for net feeding costs relative to body mass is less than 1.0 (negative allometry).

By contrast, the observation of positive allometry (an observed slope that is greater than 1.0) will indicate that, as jaw-adductor mass increases, proportionately more jaw-adductor mass is activated during feeding.

Hypothesis 3: Scaling of net feeding costs (J s⁻¹, dependent variable) proportional to adductor mass (g, independent variable) varies between FMP categories. H3A₀: no difference in slope between FMP categories. If H3A₀ is not rejected, then H3B₀: no difference in intercept between FMP categories.

Alternatively for H3A₀, slope differences among FMP categories will indicate one of two things: (i) higher slopes for tough and stiff foods will indicate that, as jaw-adductor mass increases, feeding costs increase at a proportionally greater rate for tough and stiff foods; or (ii) higher slopes for soft and elastic foods will indicate that, as jaw-adductor mass increases, feeding costs increase at a proportionally lower rate for tough and stiff foods.

Alternatively for H3B₀, if tough/stiff foods incur higher feeding costs compared to soft/elastic foods regardless of jaw-adductor mass, then we expect that foods with high FMPs will have intercepts that are elevated above those foods with lower FMPs.

Hypothesis 4: Scaling of net feeding costs (J s⁻¹, dependent variable) proportional to adductor mass (g, independent variable) varies between FGP categories. H4A₀: no difference in slope between FGP categories. If H4A₀ is not rejected, then H4B₀: no difference in intercept between FGP categories.

Alternatively for H4A₀ and H4B₀, large diameter foods may incur higher feeding costs compared to smaller foods (due, for example, to the gape/bite force trade-off due to relatively poor muscle performance at larger gapes [8]). For H4A₀, if large foods incur higher feeding costs than small foods and these costs increase as jaw-adductor mass increases, then large diameter foods will have higher slopes than small diameter foods. For H4B₀, if costs increase for large foods regardless of jaw-adductor mass, then large foods will have higher intercepts than small foods.

Tests of H3 and H4 are exploratory and attempt to discern patterns of variation in net feeding costs associated with two FMPs (toughness and stiffness) and two FGPs (maximum and minimum diameter of initial food size). Two caveats are that the test foods were not designed to control for the effects of multiple biomechanical factors on feeding costs and were selected to minimize but not eliminate manual processing and ingestive behaviours.

3. Methods

(a) . Species and metabolic trials

Data were recorded from 33 animals comprising 12 species of strepsirrhine and haplorhine primates (0.08 kg to 4.2 kg; electronic supplementary material, tables S1 and S2) at the Duke Lemur Center and the Division of Laboratory Animal Resources at Duke University in Durham, North Carolina. Each animal was trained to sit and feed inside a custom-built, clear Plexiglas chamber in a climate-controlled research room (20.5–23°C). Chamber dimensions allowed animals to sit upright and move but space was limited to prevent locomotor movements such as leaping. Measurements of an initial metabolic rate (IMR, J s^–1), a metabolic rate during feeding (FMR, J s^–1), and a postprandial metabolic rate (PMR, J s^–1) were collected using open-circuit indirect calorimetry (respirometry) systems. The respirometry data for the Microcebus murinus, Loris tardigradus, Saimiri sciureus and a subset of Lemur catta animals were measured using a custom-designed modular system (Sable Systems, Inc., Las Vegas, NV, USA) [49,50]. Data for all other animals were measured using a Field Metabolic System (Sable Systems, Inc., Las Vegas, NV, USA). All measurements were digitally converted (UI2 software, Sable Systems, Inc., Las Vegas, NV, USA), and the digital data were relayed to a computer using Expedata (Sable Systems, Inc., Las Vegas, NV, USA). Simultaneous video recordings (30 fps) were captured using a Sony Handycam, synchronized with the respirometry data using Expedata, and used to define the durations of the pre-feeding, feeding and postprandial periods. This project was approved by the Duke University IACUC committee (#A222-15-08). All experiments were performed in accordance with relevant guidelines and regulations (electronic supplementary material, table S2). Commonly used abbreviations are provided in table 1.

During the pre-feeding period of each trial, the animal entered the recording box and was given time to reach a stable metabolic plateau (figure 1). IMR was determined during this period. Pre-feeding is followed by a feeding period, from which FMR is determined, and a postprandial period, from which PMR is determined (figure 1). Body mass was measured prior to each trial. Metabolic measurements were analysed for a 1–14 min time span of each behaviour (initial, feeding and postprandial) after O₂ consumption had reached a steady-state plateau. The PMR was measured within 5–15 min of the end of feeding. Respirometry variables were calculated using established formulae [51,52] and converted to total metabolic cost as 20.1 J ml O⁻¹₂, assuming a negligible contribution from anaerobic glycolysis, and then converted to metabolic rate as J s^–1 or J s^–1kg^–1. In order to minimize stress (which can elevate metabolism: [53]), animals were acclimated to the recording set-up and were not restrained. During training sessions, animals were fed and thus learned to associate the recording sessions with feeding. Animals were not fed prior to a trial, but it is possible that animals ate food in their home cages during the night (i.e. food left over from the afternoon feeding that occurred the day before). Propithecus coquereli and Hapalemur griseus are folivores and have a greater than 24 h gut transit time [54,55]. Thus, it is not possible to confirm that all animals were post-absorptive at the start of a trial.

Figure 1.

Sample respirometry trial of Persephone (female, Lemur catta) eating apple. (Online version in colour.)

When not feeding, an animal typically sat on the recording chamber floor with occasional movements and grooming. Movements were evaluated qualitatively from the video recordings in order to choose periods with minimal body movement over which to calculate metabolic costs (electronic supplementary material, table S3). Our qualitative observation is that body movements (e.g. pacing, grooming, re-positioning) were more common during the pre-feeding period than during the feeding and postprandial periods. Three criteria were used to admit a trial for analysis. First, if qualitative assessment of the animal's behaviour during the trial showed that the subject moved frequently inside the recording chamber, the trial was dropped. Second, trials in which the IMR did not reach a steady-state plateau were dropped. Finally, trials in which the IMR or PMR was greater than two standard deviations from the mean IMR or PMR for that animal were dropped.

Animals ate raw foods ad libitum from a portion similar in weight to their morning feeding allotment. A feeding bout was defined as the time from when the first food item touched the mouth to when chewing of the last food item ended. Only feeding bouts during which an animal ate from the food items continuously were included in the analysis. The food sample consisted of 20 foods many of which were pre-cut (electronic supplementary material, table S5). All species ate food based on their feeding preferences in the wild and greater than one food type, except L. tardigradus (an exclusive insectivore) (electronic supplementary material, table S1). It was not possible for all species to consume the same foods because of dietary restrictions and preferences of the individual animals. For example, P. coquereli was not allowed fruit. To test H2, bilateral jaw-adductor mass was taken on formalin-fixed, wet specimens from published data and from the collection of C. E. Wall ([56–58]; electronic supplementary material, table S4).

Toughness and stiffness values for each food were collected from the literature when possible, and the remaining foods were tested using a Lucas FLS-II tester (Lucas Scientific) fitted with 10 N and 100 N load cells (electronic supplementary material, table S5). Toughness (R) is the work needed to propagate a crack through a food item, and elastic modulus or stiffness (E) is the relationship between elastic stress and strain [59–61]. Toughness was measured using scissor tests for kale or wedge tests for all other foods. Scissor and wedge tests can produce different results for the same food [62], but it was not possible to use one type of test on all foods. Stiffness values were collected using a membrane test for kale and blunt indent tests for the other foods. We report instant (E_i) rather than the infinite (E_∞) loading regime because of its relevance to foods during chewing [63,64]. Average toughness and stiffness values were grouped into two categories based on their R and E values: high and low ([39,61,65–72]; electronic supplementary material, table S5).

Maximum and minimum food sizes were measured with digital calipers prior to each experiment and standardized by each animal's mandible length. Radiographs of each animal were taken at the Duke Lemur Center and the Duke DLAR. Mandible length was measured parallel to the long axis of the corpus from infradentale to the posterior margin of the ascending ramus [73]. Standardized maximum and minimum food sizes were categorized as high or low (electronic supplementary material, table S5). Categories were determined by dividing the range in half as values were evenly distributed.

(b) . Variables and statistical analyses

The tests of H1–H4 evaluate the net (incremental) cost of feeding (J s^–1 or J s^–1 kg^–1), but we also present data for the gross cost of feeding (FMR, J s^–1) regressed on body mass and adductor mass in electronic supplementary material, table S6. Following Passmore and Durnin [74], gross cost of feeding (FMR) is the cost of movements associated with feeding (food manipulations and food intake) in addition to all other metabolic costs (e.g. digestion, maintenance costs) that occur during the feeding period. We calculate the net cost of feeding (netFMR) in two ways. The variable netFMR_IMR (J s^–1) = FMR (J s^–1) – IMR (J s^–1). The variable netFMR_PMR (J s^–1) = FMR (J s^–1) – PMR (J s^–1). The calculation of netFMR assumes that (i) anticipatory digestive costs are similar during IMR and FMR; (ii) costs associated with maintenance of body posture are similar during IMR, FMR and PMR; (iii) movement inside the chamber is negligible and similar during IMR, FMR and PMR; and (iv) costs associated with stress are similar during IMR, FMR and PMR. Thus, netFMR is a measure of the incremental added cost of food manipulation and food intake (see electronic supplementary material, table S3). The sources of energetic costs associated with the IMR, FMR and PMR periods and sources of error are discussed in electronic supplementary material, table S3.

All analyses and figures were generated in R [75]. Raw values were converted to natural logs for all analyses. Analyses for H1 and H2 were conducted using mean cost values for each species (electronic supplementary material, tables S7 and S8), and the regressions of costs relative to body mass (H1) and adductor mass (H2) were calculated using phylogenetic generalized least squares (pGLS) (R packages ‘caper' and ‘extRemes' [76,77]) and ordinary least-squares (OLS) regressions (R base). Means and standard deviations for the raw data are provided in electronic supplementary material, table S8. For H3 and H4, mean cost values for each species were calculated for each FMP and FGP category and evaluated against adductor mass using OLS regression (electronic supplementary material, table S9). H3 and H4 regressions were done only when seven or more species were represented across categories. Slope and ANCOVA intercept tests employed experimentwise error rates and were conducted in the R packages ‘lsmeans’ and ‘emmeans’ [78,79]. For the pGLS regressions, a consensus phylogeny was built from 10ktrees [80], and pGLS parameters were estimated using lambda (λ) fixed at 0, 1-Brownian motion, and maximum likelihood (ML). Fit of the pGLS models was assessed using Akaike information criterion (AIC), and λ=ML provided the best or equivalent fit for all body mass and adductor mass models. Significance was set at p < 0.05.

4. Results

(a) . Hypotheses 1 and 2, body mass and adductor mass

All models were significant with pGLS R² values ranging between 0.77 (mass-specific netFMR_PMR) and 0.94 (netFMR_IMR) (electronic supplementary material, table S7). OLS models showed similar results (electronic supplementary material, table S7). The net cost of feeding is strongly negatively allometric with respect to body mass whether the net cost is obtained via subtracting the IMR (pGLS slope = 0.68 ± 0.10) or the PMR (pGLS slope = 0.59 ± 0.14) (figure 2; electronic supplementary material, table S7).

Figure 2.

Relationships between body mass and netFMR_PMR, netFMR_IMR, mass-specific netFMR_PMR and mass-specific netFMR_IMR. All relationships are negatively allometric. (Online version in colour.)

Prior to testing H2, jaw-adductor mass was determined to scale isometrically relative to body mass (pGLS slope = 0.96 ± 0.22; figure 3; electronic supplementary material, table S7). For H2, both the netFMR_IMR and netFMR_PMR models were significant with pGLS R² values ranging from 0.72 (netFMR_PMR) to 0.81 (netFMR_IMR; figure 3; electronic supplementary material, table S7). OLS models showed similar results (electronic supplementary material, table S7). The net cost of feeding is strongly negatively allometric relative to jaw-adductor mass whether it is obtained by subtracting the IMR (pGLS slope = 0.63 ± 0.20) or the PMR (pGLS slope = 0.51 ± 0.20) (electronic supplementary material, table S7).

Figure 3.

Relationships between adductor mass and netFMR_PMR and netFMR_IMR. Both relationships are negatively allometric. (Online version in colour.)

(b) . Hypotheses 3 and 4, food material and food geometric properties

The OLS regressions were significant with R² values ranging between 0.66 (maximum diameter - small) and 0.92 (stiffness - high) (electronic supplementary material, table S10). There were no significant differences between the OLS slopes for toughness (low versus high), stiffness (low versus high), maximum diameter (small versus large) or minimum diameter (small versus large) (electronic supplementary material, table S11). The analysis of covariance (ANCOVA) models to test for differences in intercept showed an interaction for three FMP contrasts (netFMR_IMR–toughness, netFMR_IMR–stiffness and netFMR_PMR–stiffness) and one FGP contrast (netFMR_PMR–maximum diameter) (electronic supplementary material, table S12 and figures S1 and S2), therefore intercept tests were not done. The intercept tests for netFMR_PMR–toughness, netFMR_IMR–maximum diameter, netFMR_IMR–minimum diameter, and netFMR_PMR–minimum diameter were not significant.

5. Discussion

Our results demonstrate for the first time that net feeding costs increase with negative allometry relative to both body mass and jaw-adductor mass. Negative scaling of feeding costs with body mass is similar to that seen for basal metabolic rate (J s^–1) and RMR (J s^–1) [81,82] and for horizontal locomotion costs calculated as Cost of Transport (CoT, J kg^–1 m^–1 where m is distance travelled) [83,84]. These scaling results provide a baseline for future work to examine the relationship between components of energy budgets and the biomechanics and morphology of the feeding system.

For example, energetic costs in the primate feeding system are thought to be a smaller proportion of the energy budget than locomotor costs [85–89], and it has been proposed that the feeding system reflects selection for the precise application of bite forces rather than energy minimization [85,86,89]. We suggest that precise force application is an energy saving mechanism because it concentrates force (which carries a cost to produce) between the food and the teeth but the relationship between precise force application and energetic costs remains to be tested. The precise occlusion that characterizes mammals (via both movement caused by muscle and tooth morphology) can be viewed as a mechanism for both precise force application and energetic efficiency [20,90]. However, mastication is also an opportunity for processing food in order to increase digestibility and digestive efficiency so energy minimization may not always be at a premium [91–96]. Further study of feeding energetics in relation to feeding time, food intake rate, digestive costs, energy inputs and mechanical performance are required to understand feeding costs in the context of energy budgets.

The data presented here suggest body mass scaling of feeding costs is similar to body mass scaling of locomotor costs despite differences in the function of these two biomechanical systems. For horizontal locomotion, the observed slopes for CoT (J kg^–1 m^–1) relative to body mass range between −0.35 and −0.47 for running and walking, respectively ([84]; see also [50,97–99]). Our observed slopes are −0.41 (mass-specific netFMR_PMR, J s^–1 kg^–1) and −0.33 (mass-specific netFMR_IMR, J s^–1 kg^–1).

The lower CoT at large sizes is influenced by a range of different mechanisms, including size trends in mechanical advantage, elastic mechanisms, and the details of individual-muscle activation and fibre contractile dynamics (i.e. work, force output) (e.g. [49,50,97–106]). These efficiency mechanisms permit animals to vary their locomotor regimes in terms of speed and gait in response to varying substrates across a large range of body sizes [107]. While the units are distinct (work/mass/distance versus power/mass), there may be some broad commonalities to the mechanical parameters that underlie the feeding cost scaling patterns observed here.

Mastication favours precise force application rather than excursion or velocity, and energy saving mechanisms like elastic energy storage are proposed to have little influence during routine chewing [86,108]. Energy saving mechanisms of the feeding system potentially include such structural features as occlusal morphology (e.g. shearing crest development) and jaw-adductor mechanical advantage over a chewing cycle [20,109–112] that may reduce the force and work requirements for larger species. Larger species may also require fewer muscle activation–relaxation cycles per unit time due to lower chewing rates [113,114], while muscle fibre dynamics likely vary with size. The extent to which these, and perhaps other mechanisms, interact and differentially contribute to the cost of feeding over a range of body sizes requires detailed interspecific study.

It is worth noting that intraspecific scaling of feeding costs (calculated as J kg^–1 per chewing cycle, or ‘cost per chew') relative to body mass in Sapajus apella (tufted capuchins) feeding on a size and weight standardized food indicates slight positive allometry [115]. Future comparisons of these two feeding datasets will determine whether the scaling difference is due to distinct interspecific versus intraspecific scaling patterns or the choice of units for the energetic variable (J s^–1 kg^–1 versus J kg^–1 per chewing cycle).

Jaw-adductor mass scaling of feeding costs indicates that, as jaw-adductor mass increases, proportionately less jaw-adductor muscle mass is recruited. In this sample, body mass and jaw-adductor mass are correlated and scale isometrically (pGLS slope = 0.96). We explored whether high jaw-adductor mass residuals are also associated with relatively low feeding costs in a post hoc set of correlation analyses and found no significant differences in these nine contrasts: all foods, high and low stiffness foods, high and low toughness foods, large and small minimum diameter, and large and small maximum diameter (electronic supplementary material, table S13). Thus, for feeding costs measured in J s^–1, the absolute size of the muscle is the strongest signal for how much volume is recruited. In addition to the energy saving mechanisms discussed above, high feeding costs in relation to jaw-adductor mass might occur because FMPs are independent from jaw-adductor mass. At small jaw-adductor mass, a mealworm has the same FMPs whether it is eaten by a mouse lemur (with 1 g of jaw-adductor mass) or a tufted capuchin (with 124 g of jaw-adductor mass). The expectation is that a mouse lemur will recruit relatively more muscle volume compared to a tufted capuchin resulting in relatively higher feeding costs. Thus, despite the jaw-adductor muscles operating for fundamentally different purposes than locomotor muscles, and the former being optimized for precise force application [86], the feeding system appears to use energy-saving mechanisms as body and adductor mass increase.

These findings also highlight the benefits of large jaw-adductor mass. These benefits include more muscle volume held in reserve to prevent muscle fatigue and bite force reduction [116–119] and more muscle available for behaviours that require high forces and therefore high volume recruitment (e.g. anterior biting due to poor bite force leverage). van Casteren et al. [26] suggest jaw-adductor force production drives net masticatory costs as stiff gum has higher chewing frequency and greater peak electromyography values compared to soft gum. However, they also found that the size of the masseter muscle in humans was not correlated with net masticatory costs. Our finding of negative interspecific scaling of net feeding costs with jaw-adductor mass and body mass is consistent with this discrepancy because it indicates that, as jaw-adductor mass increases, proportionately less muscle mass is recruited, increasing the reserve of muscle mass. In support of this, when human data [26] are included with our sample in an OLS regression of netFMR_IMR and body mass, the slope is lower (0.65; electronic supplementary material, figure S3) than the non-human primate slope (0.70; electronic supplementary material, table S7). Since the gum used in the 2022 study had a much lower stiffness value (0.21 MPa) than our test foods (range = 0.1–19.4 MPa; electronic supplementary material, table S5), this suggests humans in this sample were using a low-volume recruitment of the masseter muscle. We would expect the net feeding costs for humans to be higher than the net masticatory costs recorded during gum chewing, resulting in a higher regression slope, because of higher FMPs and more body movements during feeding. It is noteworthy that the negative allometry in this regression suggests that interspecific negative allometry will be upheld when larger size ranges are included.

There was no statistical support for the hypotheses that scaling of net feeding costs proportional to jaw-adductor mass varied among FMP and FGP categories. We acknowledge the limitations of our data from a mechanical perspective as food size was not controlled in these experiments and foods were not evenly distributed across species. As these data were collected non-invasively without restraint in non-human primates, it was also not possible to isolate masticatory costs from overall feeding costs or to control body movement, digestive costs, and food properties to the extent done for humans by van Casteren et al. [26]. However, trends in our comparisons suggest FMPs and FGPs influence feeding costs. Higher intercepts for tough foods (electronic supplementary material, figure S1A and 1B) suggest that, at a given jaw-adductor mass, feeding costs are higher for tougher foods compared to weaker foods. This is consistent with the human mastication data [26,27]. The higher intercept for the small minimum diameter comparisons (electronic supplementary material, figure S2C and S2D) supports the hypothesis that feeding costs are higher in mechanically challenging foods because the small minimum diameter foods included foods that had high toughness and stiffness (e.g. mealworm). However, the ANCOVA interactions for stiff foods (electronic supplementary material, figure 1C and 1D) suggest relatively lower costs as jaw-adductor mass increases. Our data also suggest feeding costs associated with large foods decrease as jaw-adductor mass increases (electronic supplementary material, figure 2A). In summary, the major pattern is one of interaction among mechanical variables, feeding costs and jaw-adductor size. This suggests that small- and large-muscled animals are solving mechanical problems related to food properties in different ways.

Our measures of FMR and derived variables likely underestimate food preparation costs (i.e. preparing whole foods using various body parts and skeletal muscles) because food preparation by the animal was limited in the experimental design. Primate behavioural studies have not typically quantified how much of feeding time is taken up by food preparation versus food intake, and they may be over-lapping. In the wild, there is the potential for significant added feeding costs during both extra-oral and oral food preparation. Extreme food preparation behaviours, although not measured in this study, occur in many mammals, including several for whom feeding data are presented here (Hapalemur griseus, Daubentonia madagascariensis, Sapajus apella). Diets that include structurally defended foods such as seeds and wood-boring insect larvae, and other difficult to access foods such as bamboo pith and underground storage organs, may require large amounts of muscle force and power output (or non-oral processing) before food intake occurs [40,120] and incur high energetic costs [14,121–123].

6. Conclusion

These results are the first empirical comparisons of the energetic costs of feeding among mammals, and our small-bodied primate sample indicates negative allometry of feeding costs in relation to body mass and jaw-adductor mass. This suggests that, as with locomotion, small and large animals solve mechanical challenges in relation to feeding energetics in different ways. For the feeding system, we suggest that these solutions will be influenced by food properties and will have inter-related morphological, physiological and behavioural components. More precise data are needed in order to determine whether energy minimization or bite force requirements exert stronger selection pressures on the feeding apparatus. Our data were limited to primates in a fairly narrow (80 g to 4 kg) size range, and future studies measuring feeding energetics in a broader taxonomic and size distribution will be able to test whether negative allometry characterizes mammals across their size range. Because of the complex interactions among mechanical and structural variables, future work to determine the effects of FGPs and FMPs will require careful control of these variables.

Ethics

The research using animals complied with animal welfare guidelines at Duke University, USA. The appropriate approval was obtained by the Duke University Institutional Animal Care and Use Committee (DU-IACUC #A222-15-08). Detailed animal welfare information is available in electronic supplementary material, table S2.

Data accessibility

The data supporting this article are provided in the electronic supplementary material [124].

Authors' contributions

C.E.W.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing—original draft, writing—review and editing; J.B.H.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, resources, software, supervision, validation, visualization, writing—original draft, writing—review and editing; M.C.O.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing—original draft, writing—review and editing; M.T.: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, writing—original draft, writing—review and editing; M.F.L.: conceptualization, data curation, formal analysis, investigation, methodology, resources, software, supervision, validation, visualization, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by a National Science Foundation grant to C.E.W., J.B.H. and M.C.O. (grant no. NSF-BCS-1062239).