The mammalian masticatory rhythm is produced by a brainstem timing network. The rhythm is relatively fixed within individual animals but scales allometrically with body mass (Mb) across species. It has been hypothesized that sensory feedback and feed-forward adjust the rhythm to match the jaw's natural resonance frequency, with allometric scaling being an observable consequence. However, studies performed with adult animals show that the rhythm is not affected by jaw mass manipulations, indicating that either developmental or evolutionary mechanisms are required for allometry to become manifest. The present study was performed to tease out the relative effects of development versus natural selection on chewing rate allometry. Thirty-one dog breeds and 31 mass-matched non-domestic mammalian species with a range in Mb from ∼2 kg to 50 kg were studied. Results demonstrated that the chewing rhythm did not scale with Mb among dog breeds (R=0.299, P>0.10) or with jaw length (Lj) (R=0.328, P>0.05). However, there was a significant relationship between the chewing rhythm and Mb among the non-domestic mammals (R=0.634, P<0.001). These results indicate that scaling is not necessary in the adult animal. We conclude that the central timing network and related sensorimotor systems may be necessary for rhythm generation but they do not explain the 1/3rd to 1/4th allometric scaling observed among adult mammals. The rhythm of the timing network is either adjusted to the physical parameters of the jaw system during early development only, is genetically determined independently of the jaw system or is uniquely hard-wired among dogs and laboratory rodents.

Mastication, like locomotion and respiration, is a rhythmic behavior found among many if not most mammalian species. According to neurophysiological investigators, the rhythm is produced by a population of brainstem cells known as a central timing network (CTN) or central rhythm generator (CRG); these cells form a subcomponent of a larger masticatory central pattern generator (CPG) network (Lund, 1991; Nakamura and Katakura, 1995). The traditional model of the masticatory CPG circuitry placed CTN cells within a fairly localized brainstem region (Nakamura and Katakura, 1995) but further investigations suggest that the rhythmicity is more diffusely represented (Enomoto et al., 2006; Lund et al., 1998; Nakamura et al., 1999; Tanaka et al., 1999; Tsuboi et al., 2003; Wu et al., 2001). Also, although the masticatory CPG is believed to be highly conserved (Lund et al., 1998; Wainwright, 2002), it is likely that many underappreciated taxonomic specificities in CTN design and location exist (Alfaro and Herrel, 2001).

The masticatory rhythm produced by the CTN is surprisingly fixed within individual animals (Carvalho and Gerstner, 2004) and within species (e.g. Dellow and Lund, 1971; Horio and Kawamura, 1989; Morimoto et al., 1985; Wainwright, 2002). However, the masticatory rhythm scales allometrically with body mass (Mb) across mammalian species groups (Druzinsky, 1993; Gerstner and Gerstein, 2008). It is unclear how and why the scaling relationship exists but its existence suggests that the relatively fixed rhythmic output of the CTN is somehow linked broadly across the mammalian class to morphological size or mass parameters.

Bonner and Horn review several possible explanations for the presence of allometric scaling in general (Bonner and Horn, 2000). On the one hand, biologists such as D'Arcy Thompson (Thompson, 1942) have argued that ‘mathematical–physical explanations [are] sufficient to enforce optima automatically with size changes’ (Bonner and Horn, 2000). For the school of structuralism, natural selection and genetics are often irrelevant in scaling (Bonner and Horn, 2000). With respect to the observed relationship between mastication and Mb, the mathematical–physical explanation would probably involve an underlying geometric or physical principle that requires further explication as applied to rhythmic jaw movement production.

On the other hand, Bonner and Horn prefer the argument that natural selection is responsible for manifestations of allometric scaling (Bonner and Horn, 2000). In their argument, allometric scaling emerges because fitness is greater for individuals that manifest the appropriate or optimal scaling than for animals manifesting variation that does not scale.

One way that allometric scaling of the masticatory rhythm could occur involves sensory systems. Sensory feedback, feed-forward or both could exploit information about mass-relevant parameters to adapt the masticatory rhythm accordingly. Such sensory information could be used in relatively fast time scales, i.e. chew-by-chew or bite-by-bite, or in developmental time (Pearson, 2000) to affect the observed scaling. In either case, allometry would result from physiological adaptations to sensory activity levels in physiological time scales. The implication is that an archetypal mammalian CPG network, appropriately coupled to sensory systems, would be necessary and sufficient to result in the observed allometric scaling of chewing rate across species.

However, evidence indicates that mass-related adaptations in chewing rate do not occur in such fast time scales. In acute experiments on guinea pigs, experimentally evoked rhythmic jaw movements maintained a constant rhythm when up to 15 g of weight were added to the mandible, in which the manipulation increased the jaw mass (Mj) by nearly an order of magnitude (Chandler et al., 1985). These results indicate that sensory feedback is used to modulate muscle recruitment in response to varying loads on the jaw. The result is that chewing rate remains relatively constant within individual animals despite variation in load or Mj. A similar finding was made for studies of primates, wherein it was discovered that rate modulation rather than time modulation occurred probably to render chewing rate relatively invariant (Ross et al., 2007).

More recently, we demonstrated that nearly doubling Mj in adult rats did not lead to load-dependent changes in licking rate over 12-week time periods (Carvalho and Gerstner, 2004). We also showed that individual animals could be reliably identified by mean licking rates over this time period as well. This suggests that sensory information does not provide feedback capable of modifying oral motor rhythm rates in adult animals over relatively long (multi-month) time scales.

However, the above experimental studies of rats and guinea pigs evaluated adult animals. Whether the chewing rate–mass relationship emerges in evolutionary time scales or is a result of ontogenetic processes operating on proximate mechanisms involving CPG coupling with sensory information cannot be well demonstrated in these studies.

In the present study, we attempted to tease apart this issue by evaluating chewing rate scaling among dog breeds versus among matched, non-domestic, mammalian species. Dog breeds are derived from a common wild ancestor, and they show considerable variation in Mb for a single species (Parker et al., 2004; Vila et al., 1997). In our study, dog breeders can be thought of as having participated in an experiment to vary the independent variables of Mb and jaw length (Lj) while remaining blind to, i.e. relatively disinterested in artificially selecting for the dependent variable of interest, chewing rate. Consequently, dog breeds of varying sizes provide an important resource to examine whether the mammalian CTN and sensory pathways provide both necessary and sufficient explanation for allometric scaling between masticatory rhythm and Mb that may emerge in developmental time scales.

Importantly, any natural selection pressure operating specifically on the relationship between Mb and chewing rate, which would be so important to Bonner's and Horn's argument (Bonner and Horn, 2000), has been relaxed among dog breeds but not among the non-domestic mammalian species we have included in the study. This allows for a test of the contributory role of natural selection in allometric scaling. If canine chewing rate is capable of responding and adapting to mass-related parameters in developmental time scales as a result of properties inherent in an ‘archetypal’ mammalian masticatory CTN, or as a result of mathematical–physical imperatives (Thompson, 1942), then chewing rate and Mb or Mj should scale among dog breeds as it does among other mammals. By contrast, if chewing rate does not scale among dog breeds, this would argue that the mammalian masticatory neuromotor system is not sufficient to explain the allometric scaling observed among other mammalian species.

Also, by matching the Mb of dog breeds with the Mb of non-domestic mammalian species, we attempt to control for statistical and biological noise sources as well. In other words, by using the non-domestic mammalian species as a control group, we can exclude the contributions from noise sources to the negative results observed in the dog data.

Animals

Healthy pure-bred adult dogs from American Kennel Club (AKC) registered breeders, rescue organizations or dog shows held in Michigan were chosen as subjects. Between four and seven individuals were studied per breed, and a total of 31 breeds were used in the study (Table 1). All individuals were pure-bred according to AKC standards, and ranged in age from 4 to 168 months. Exact ages were not recorded for the dogs at one AKC dog show; however, the handlers all confirmed that their dogs were adult dogs over 12 months in age.

This investigation reports negative results for some analyses. Therefore, it was important to demonstrate that these negative results were not due to statistical noise, i.e. variability in the scatter plot, overwhelming the relatively narrow range of Mb represented by the dog breeds, e.g. cf. p. vii and p. 38 in Calder (Calder, 1996). To this end, each dog breed was matched to a non-domestic mammalian species with a similar Mb. This was done by first calculating the mean Mb for each dog breed in this study. Next, a non-domestic mammalian species was selected from previously tabulated sources (Gerstner and Gerstein, 2008), such that each dog breed was closely matched to a given mammalian species according to Mb. This was done while remaining blind to the chewing rates of the dog breeds and non-domestic mammalian species. Once the matching was done, the previously reported chewing rates for the matched, non-domestic, mammalian species (Gerstner and Gerstein, 2008) were used in the study.

Chewing task

Individual dogs were fed by their owners or the investigator while the investigator videotaped the chewing sequence. Dogs were fed treats that owners typically fed their dogs, which included dry chicken jerky, small pieces of crushed rawhide or dry dog food treats. The standardization of bite size was considered; however, in cats, larger bite sizes are associated with longer cycle durations (Thexton et al., 1980). If a similar relationship were to exist in the case of dogs, then standardized bite sizes would result in relatively fast chewing rates in large breeds and relatively slow chewing rates in small breeds, a relationship opposite to that observed to exist between mammalian Mb and chewing rate (Druzinsky, 1993; Gerstner and Gerstein, 2008). Thus, it was decided to allow dogs to receive bite sizes to which they were acclimated and which were commercially designed for breed sizes. This meant that larger dog breeds typically received larger bite sizes than smaller dog breeds.

Previous results have indicated that chewing rate does not vary considerably with changing food hardness (Thexton et al., 1980; Yamada and Yamamura, 1996). Therefore, the hardness of the treat was dependent on the typical treat given to each dog and left up to the discretion of the owner.

Chewing sequences were videotaped (JVC Digital Video Camera, model GRDVL720U, Aurora, IL, USA) with auto-iris ‘on’ and digital shutter speed=60 Hz. Animals were videotaped in the sagittal or frontal view, and the camera zoomed to focus mainly on the head and neck. A priori, it was determined that single chews would be omitted from analysis, and that only bursts of at least three consecutive and rhythmical chews would be used in analyses. Therefore, individual animals were videotaped so that bursts of at least three consecutive chews were obtained. The investigator continued to videotape the dog chewing until ∼30 total analyzable chews, i.e. 30 chews occurring in bursts consisting of >3 chews, were recorded.

Data collection

Canine chewing sequences were recorded to digital videotape (Sony premium Mini DV, Tokyo, Japan, or Panasonic Mini DV, Secaucus, NJ, USA). The videotaped clips were digitized at a rate of 29.97 frames s–1 using Final Cut Pro 6.06 (Apple Inc., Cupertino, CA, USA) on a Mac OS X system (Apple Inc.) and a Sony Digital HD Videocassette Recorder to convert the data from Mini DV to a digital format that could be used in Final Cut Pro. Based on previous work with cats (Thexton et al., 1980; Yamada and Yamamura, 1996) and on preliminary canine data obtained in this study, this capture rate was considered adequate for chewing cycle duration (CD) calculations.

Data analysis

CD were determined by watching the digitized video sequences, frame-by-frame and recording the total number of frames, Nf, and the total number of chewing cycles, Nc, in the sequence. A chewing cycle was defined by the period of time between consecutive maximum jaw openings. The mean CD for a sequence was calculated by dividing the ratio, Nf /Nc, by the frame rate, 29.97 frames s–1. When multiple chewing bursts were analyzed for a given animal, a weighted mean CD was calculated for that animal. Each breed was represented by at least four animals, and the breed mean CD was calculated by using the mean CD for each animal sampled in the given breed.

Many allometric studies use Mb as the independent variable, and in this study, certain analyses used Mb as well. However, many dog-breed heads have unique morphologies that may vary independently of Mb. Furthermore, it has been demonstrated that jaw lever arm lengths correlate with CD among primates (Ross et al., 2009). Therefore, a mean Lj was also obtained for each breed. Lj was measured by the investigator directly on the dogs at the time that chewing data were videotaped. The jaw was measured in mm from the angle of the ramus or gonion to the anterior-most point on the mandible or pogonion (Fig. 1).

Statistical methods

Mean CD, Lj and Mb for each breed were log-transformed. Linear regression analyses were used to study the relationships between log CD, log Mb and log Lj. Because a major focus of this investigation involved intraspecific studies, we assumed a normal bivariate distribution and used ordinary least-squares regression as suggested by previous allometric investigations (Wehner et al., 2007) [see also p. vii in Calder (Calder, 1996)].

Table 1 contains the dog breeds, number of dogs (N) and means for Mb, CD and Lj. One s.d. is shown in parentheses for each of the means. Also shown in Table 1 is a list of the non-domestic mammalian species, arranged according to the dog breeds to which they were matched during the selection phase of the study (see ‘Animals’ in Materials and methods). Mammalian Mb and CD are shown as reported in previous work (Gerstner and Gerstein, 2008). Note in Table 1 that three dog breeds had relatively large s.d. associated with CD, viz. Afghan (284), Keeshond (577) and Siberian Husky (427). In all three instances, this was due to one individual dog, each of which was characterized by a relatively long-duration mean CD. Respectively, the breed means dropped to 390 ms, 366 ms and 603 ms for Afghan, Keeshond and Husky when these three dogs were removed. Certain analyses, below, were performed first with and then without these three individual dogs.

Fig. 2 compares CD with Mb for the 31 dog breeds. Log CD was not significantly related to log Mb (R=0.2992, d.f.=29, P>0.10; scaling exponent=0.0691). However, after removing the one Afghan, one Keeshond and one Siberian Husky dog responsible for inflating the s.d. for these breeds (Table 1), there was a significant correlation for the comparison between canine log CD and log Mb (R=0.440, d.f.=29, P<0.014). Nevertheless, it should be stressed that this significant correlation only accounted for about 19% of the variation between CD and Mb. In this latter case, the equation of the line was y=0.0668x+2.308. The 95% confidence interval (CI) on the slope=0.0149–0.119.

In allometric studies, it is recognized that relatively narrow ranges of Mb can have profound impacts on correlations between Mb and a dependent variable of interest [see Introduction to this paper and p. vii and 38 in Calder (Calder, 1996)]. Given that the dog-breeds' range of Mb (Table 1) was relatively narrow compared with that found in many allometric studies, it was important to demonstrate whether this narrow range played a role in the observed lack of significance between Mb and CD among dogs. This was a major reason for analyzing data from the non-domestic mammals representing a similar mass range as the dogs (Table 1). Fig. 3 provides a visualization of the matching achieved between canine and mammalian Mb. Plotted are the paired log Mb for the dogs (x-axis) and mammals (y-axis), along with a least-squares regression line to demonstrate the relative closeness of the matching. In the plot, y=0.9892x–0.0454, R2=0.98. Note that the slope is near unity, the intercept is near the origin and that R2 is high, all of which provide an indication of how closely matched the two groups were in terms of Mb. The means (s.d.) Mb in kg for dogs=22.4 (13.2) and for mammals=22.5 (13.2) were not significantly different (paired t=0.72, P>0.05).

An analysis similar to that depicted in Fig. 2 was performed on the non-domestic mammals shown in Table 1. For these mammals, a highly significant relationship between log Mb and log CD was found (Fig. 4, R=0.634, d.f.=29, P<0.0001). Here, the slope or scaling exponent=0.220 and the y-intercept=1.721. The 95% CI on the slope=0.118–0.321, slightly overlapping the 95% CI for the slope determined for the canine data with the outliers removed; however, the slope for the dogs (0.0668) was over 3 times less than the 0.220 slope for the mammals.

Recent studies have begun modeling relationships between CD and jaw lever arm lengths among mammals (Ross et al., 2009). Fig. 5plots log Lj against log CD for the 31 dog breeds, along with a least-squares regression line. The relationship was not significant (R=0.3283, d.f.=29, P=0.0756; scaling exponent=0.1997, intercept=2.2289). However, when the three outlier dogs, viz. one Afghan, one Keeshond and one Siberian Husky, were removed from the analysis, the relationship was significant (R=0.4411, d.f.=29, P=0.013). In this latter case, the analysis yielded the following allometric equation, y=0.1628x+2.2311. The 95% CI for the scaling exponent=0.0370–0.289. However, only 19% of the variation in CD was explained by variation in Lj.

The 1/4th (scaling exponent ∼0.25) to 1/3rd (scaling exponent ∼0.33 or ∼0.167) allometric relationship between Mb and CD found in most mammals (Druzinsky, 1993; Gerstner and Gerstein, 2008; Ross et al., 2009) was not found among adult dog breeds in the present study (Fig. 2). Without the included mammalian data in the study (Table 1 and Fig. 4), it may have been possible to conclude that the lack of scaling was the result of the increased role of error in a data set where the range of Mb was relatively small. Previous allometric studies of mammalian chewing have involved Mb data representing about five orders of magnitude (Druzinsky, 1993; Gerstner and Gerstein, 2008) whereas the present study involved less than two orders of magnitude in Mb. Although this reduction in the range of Mb was a contributing factor to the reduced correlation observed in mammals [compare Fig. 4 results with larger data sets in Druzinsky, and Gerstner and Gerstein (Druzinsky, 1993; Gerstner and Gerstein, 2008)], it was clearly not sufficient to render the correlation insignificant among mammals. In fact, the correlation between Mb and CD among the mammals size-matched with the dog breeds was highly significant. In this sense, the mammalian results in the present study render the interpretation of the canine negative results more powerful.

It is also important to emphasize that the role of error was probably greater in the mammalian data (Fig. 4) than in the dog breed data (Fig. 2) for several reasons. First, CD and Mb were often not sampled from the same individuals in many of the mammals (Gerstner and Gerstein, 2008) whereas both variables were sampled from the same individual dogs. Second, the foodstuffs eaten were more variable among the mammals than among the dog breeds; hence, the extent to which food parameters play a role in modifying chewing rate (Thexton et al., 1980) would have been greater among the mammals than among the dogs. Finally, each canine datum represented the mean of at least four animals whereas most of the mammalian data were represented by fewer than four individual animals and in many cases by one animal only. For these reasons, the scatter in the mammalian data in Fig. 4 is likely to be inflated relative to the scatter in the dog-breed data in Fig. 2. The fact that a highly significant correlation still remained for the mammalian data underscores the robustness of allometric scaling even under suboptimal sampling conditions. Furthermore, the fact that the scaling was absent from the dog data, except when the ‘outlier’ data were removed is all the more demonstrative that CD–Mb scaling in adult mammals may not be guaranteed under certain artificial selection conditions.

Table 1 shows that the wolf, Canis lupus, had an Mb=32,000 g, which is greater than 24 of the dog breeds. However, the wolf CD=280 ms was of shorter duration than all dog breeds studied. By contrast, a comparison of Figs 2 and 4, the abscissas and ordinates of which have the same ranges, will reveal considerable overlap in the scatter of data points between dogs and mammals. These results demonstrate that the dog breeds' CDs were slower than the wolf CD but that the dog CD range was similar to the mammalian CD range. This suggests that whatever selection pressure may be responsible for the relatively rapid wolf CD is missing, altered or reduced among the domestic dog breeds.

Several possibilities for this finding exist. For one, the wolf data in our study could represent relatively fast-chewing wolves. Perhaps the wolves from which all modern dog breeds are derived were slower chewers, more similar to the mean chewing rate of modern dog breeds.

Alternatively, dog-breed CDs may be ‘drifting’ towards more Mb-typical or mammalian-class-typical CDs. However, it is unclear what artificial selection pressure would have led towards a biased slowing of CD in all dog breeds, both large and small. If random drift in CD has occurred, dog breed CD would be expected to be both greater than and less than the wolf CD. Some other biological variable is likely to be contributing to the across-breed slowing in CD. We hypothesize that it has something to do with development, and we will discuss this hypothesis further, below.

It is important to comment on the statistically significant relationship between CD and Mb and between CD and Lj that occurred when outlier data were removed, and to explore the reason for the relatively low slope of the regression line under these circumstances. When the three outlier dogs were removed from the analysis, the scaling exponent (slope) for the comparison of Mb and CD was 0.0668 (95% CI=0.0149–0.119) for dogs (Fig. 2) whereas the same comparison yielded a slope of 0.220 (95% CI=0.118–0.321) for mammals. These data are summarized in Table 2. Although the 95% CI for dogs versus mammals slightly overlapped, the slope for the dogs was ∼3 times lower than that for the mammals. Importantly, the 95% CI for the dog data did not include the 1/3rd or 1/4th scaling exponent previously reported in mammals (Druzinsky, 1993; Gerstner and Gerstein, 2008); the slope and 95% CI for previously reported mammalian work is presented in Table 2 for comparison, along with results for carnivores specifically (cf. Gerstner and Gerstein, 2008). This suggests that a different scaling rule is involved in the relationship between CD and Mb in the dog data, if one is to accept as valid the removal of the ‘outlier’ individual dogs in order to achieve statistical significance. Moreover, whatever this scaling rule might be, the relationship between Mb and CD for dogs sans outliers accounted for only 19% of the variation compared with 40% of the variation being accounted for in the mammalian relationship (Fig. 4). Clearly, whatever rule governs the Mb–CD relationship, it is much weaker among dogs than among mammals.

By the same token, the scaling exponent for comparison of Lj and CD was 0.163 for dogs sans the three outliers (95% CI=0.0370–0.289), compared with scaling exponents ranging from 0.514 to 0.583 for non-primate mammals and primates, respectively (Ross et al., 2009). The last three rows in Table 2 replicate the Ross et al. data along with the present study's dog data for comparison. Note that the 95% CI for CD and Lj for the dogs versus the Ross et al. mammals does not overlap, and that the slope for the dogs is, again, ∼3 times shallower than for these other mammals (Table 2).

One explanation for these differences may be due to the different developmental trajectories of dogs versus non-domestic mammals. Infant size tends to scale with adult body size in non-domestic mammalian species (cf. Calder, 1996), which means that CD would scale proportionally with both infant and adult Mb in non-domestic animals. By contrast, pre-weaned puppy sizes are relatively similar across dog breeds of vastly different adult sizes (Hawthorne et al., 2004). In fact, using the logistic equation for growth in dog breeds presented by Hawthorne et al., i.e. Mb=a/{1+exp[–(xx0)/b]}, their data for 12 dog breeds [table 1 in Hawthorne et al. (Hawthorne et al., 2004)], and assuming a weaning age of 4 weeks for dog breeds, we calculated a 13.4-fold range of Mb for 4-week old dogs compared with a 30.4-fold range of Mb for adult dogs. In other words, breed size-differences are about 2.3 times greater across adult dog breeds than across puppies of the corresponding breeds. If Mj-related CD adjustments can occur during infant stages only, then one would expect the canine CD–Mb scaling exponent to be 2.3 times less than predicted based on adult dog sizes. In our data, the scaling exponent was 3 times less than expected. This suggests that part of the explanation for the reduced scaling exponent in our canine results compared with other mammalian results (Table 2) may be due to developmental effects.

The brainstem is likely to be more plastic in immature animals when circuits are first being constructed and formed. Previous studies have demonstrated that increasing the mass of the adult jaw does not lead to corresponding changes in CD in adult mammals (Carvalho and Gerstner, 2004; Chandler et al., 1985). This strongly suggests that CD cannot be adjusted to mass-related changes in adult animals. Consequently, if mass-related adjustments in CD can occur in immature animals and brainstems, then the observed CD in adult animals would mark the developmental stage at which mass-associated adjustments in CD are no longer possible. Subsequent jaw growth would then occur in the absence of corresponding CD modifications, after which developmental time point, sensory feedback would result in increased muscle recruitment to adjust to further increases in jaw size. In mammals, CD would appear to be adjusted to mass in the adult animals because of the infant–adult isometric scaling (Calder, 1996). By contrast, jaw growth among dog breeds varies considerably, and this would provide an important explanation for the scaling-exponent findings of the present study. Further developmental studies are in order to test this possibility.

It is also compelling that the relative similarity in puppy sizes across dog breeds apparently represents an adaptation for suckling (Coppinger and Coppinger, 2002). Adapting mammalian CTN output to match infant Lj would make particular sense if, for instance, suckling efficiency is more important than chewing efficiency in terms of fitness or fecundity. Future studies will need to test these hypotheses through comparative developmental investigations to sort out whether there is a developmental time period during which mass-related oral rhythm adjustments occur, whether this time period represents a definitive point in neural ontogenesis across mammals, whether there are significant fitness or fecundity costs associated with mismatches at this developmental stage, etc.

It is also noteworthy that the above discussion allows for the possibility of an archetypal mammalian CTN whose rhythmic output is adjusted to species-typical and individual-specific parameters in developmental time scales. Then, at some point, the rhythmicity becomes fixed, after which point changes in load are associated with adaptations in muscle recruitment (Ross et al., 2007). In this case one important question becomes, at what point in development do load-dependent adjustments in CTN output shift from modifications in rhythmic rates to modifications in muscle recruitment.

It has been argued that a movement's rhythmicity is a product of sensory feedback and feed-forward (Kuo, 2002). The ‘strong’ interpretation of this argument is that a specific rhythm is not actually embodied in the nervous system but is rather an emergent property of feedback and feed-forward between a central oscillatory model and the physical properties of the moving body part. This may be true for human locomotion (Kuo, 2002); however, evidence from adult animal studies strongly suggests that masticatory rhythmicity is produced centrally (Carvalho and Gerstner, 2004; Chandler et al., 1985). Based on the present study's results, we conclude that mammalian oral rhythms probably become centrally embodied during development, in which case there may be an early developmental period during which Kuo's argument is tenable (Kuo, 2002). However, if masticatory rhythmicity proves to be unresponsive to mass-related variation at any point during development, but emerges early and independent of feedback, this would suggest that the CTN is genetically inherited.

The central embodiment and heritability of the masticatory rhythm is supported by much previous work. Food properties do not affect chewing rate in rabbits (Morimoto et al., 1985; Yamada and Yamamura, 1996), humans (Horio and Kawamura, 1989), monkeys or cats (reviewed in Inoue et al., 1989). Licking in rats appears to be controlled by ‘a central timing mechanism that is somewhat impervious to disturbance’ (Travers et al., 1997). Indeed, the oral rhythm generator in rats has been called ‘particularly rigid’ and is little affected by water deprivation, taste, environmental modifications, behavior conditioning or behavior modification experiments (Bures et al., 1988). Oral rhythmicity is not prolonged by increasing Mj in either guinea pigs (Chandler et al., 1985) or rats (Carvalho and Gerstner, 2004). These studies, which have involved numerous mammalian species, provide compelling evidence that oral rhythms in adult animals are centrally generated, nearly impervious to modification and apparently designed to be rigid (e.g. Ross et al., 2007). In light of these studies, the argument has been made by other investigators than ourselves that the rhythmicity is heritable (Kobayashi et al., 2002).

Moreover, given that chewing rate does not necessarily scale to adult Mb, we would argue that either there is an ‘archetypal’ mammalian CTN that adjusts its rhythmicity to the mass of the developing animal, but only up to a specific point in time (as discussed above), or that each mammalian species must possess unique species-specific CTN components that result in CD being scaled to Mb. The unique solutions would be due largely to the unique histories of each phylogenetic lineage or clade. This argument corroborates arguments made elsewhere about vertebrate feeding in general (Alfaro and Herrel, 2001). This has profound implication for health sciences, which rely on a relative few species models to provide insight into the human condition. Future research will be required to address the nature of chewing rate heritability, of homology and analogy in mammalian CTN (Alfaro and Herrel, 2001), and how genetic, epigenetic and environmental parameters during development interact to produce not only species-typical chewing rates but the individual specificity that has also been reported (Carvalho and Gerstner, 2004).

We wish to thank Brian Sackett and Jonathan Gerstein for help with data acquisition and analysis. G.E.G. wishes to dedicate this paper to Duncan, one of the English setters in the study who initially inspired the study and shuffled off this mortal coil during the preparation of the manuscript.

This work was supported in part by NIH grant DE10625 to G.E.G. Deposited in PMC for release after 12 months.

     
  • AKC

    American Kennel Club

  •  
  • CD

    chew duration

  •  
  • CI

    confidence interval

  •  
  • CPG

    central pattern generator

  •  
  • CRG

    central rhythm generator

  •  
  • CTN

    central timing network

  •  
  • Lj

    jaw length

  •  
  • Mb

    body mass

  •  
  • Mj

    jaw mass

  •  
  • Nc

    number of chewing cycles

  •  
  • Nf

    number of frames

  •  
  • s.d.

    standard deviation

Alfaro
M. E.
,
Herrel
A.
(
2001
).
Introduction: major issues of feeding motor control in vertebrates
.
Am. Zool.
41
,
1243
-
1247
.
Bonner
J. T.
,
Horn
H. S.
(
2000
).
Allometry and natural selection
. In
Scaling in Biology
(ed.
Brown
J. H.
,
West
G. B.
), pp.
25
-
35
.
New York
:
Oxford University Press
.
Bures
J.
,
Buresova
O.
,
Krivanaek
J.
(
1988
).
Motor learning
. In
Brain and Behavior. Paradigms for Research in Neural Mechanisms
. pp.
151
-
184
.
New York
:
Wiley and Sons
.
Calder
W. A.
III
(
1996
).
Size, Function, and Life History
.
Mineola
:
Dover Publications
.
Carvalho
T. C.
,
Gerstner
G. E.
(
2004
).
Licking rate adaptations to increased mandibular weight in the adult rat
.
Physiol. Behav.
82
,
331
-
337
.
Chandler
S. H.
,
Goldberg
L. J.
,
Lambert
R. W.
(
1985
).
The effects of orofacial sensory input on spontaneously occurring and apomorphine-induced rhythmical jaw movements in the anesthetized guinea pig
.
Neurosci. Lett.
3
,
45
-
49
.
Coppinger
R.
,
Coppinger
L.
(
2002
).
Part IV. The tail wags the dog
. In
Dogs. A New Understanding of Canine Origin, Behavior, and Evolution
, pp.
271
-
282
.
Chicago
:
University of Chicago Press
.
Dellow
P. G.
,
Lund
J. P.
(
1971
).
Evidence for central timing of rhythmical mastication
.
J. Physiol.
215
,
1
-
13
.
Druzinsky
R. E.
(
1993
).
The time allometry of mammalian chewing movements: chewing frequency scales with body mass in mammals
.
J. Theor. Biol.
160
,
427
-
440
.
Enomoto
A.
,
Han
J. M.
,
Hsiao
C.-F.
,
Wu
N.
,
Chandler
S. H.
(
2006
).
Participation of sodium currents in burst generation and control of membrane excitability in mesencephalic trigeminal neurons
.
J. Neurosci.
26
,
3412
-
3422
.
Gerstner
G. E.
,
Gerstein
J. B.
(
2008
).
Chewing rate allometry among mammals
.
J. Mammal.
89
,
1020
-
1030
.
Hawthorne
A. J.
,
Booles
D.
,
Nugent
P. A.
,
Gettinby
G.
,
Wilkinson
J.
(
2004
).
Body-weight changes during growth in puppies of different breeds
.
J. Nutr.
134
,
2027S
-
2030S
.
Horio
T.
,
Kawamura
Y.
(
1989
).
Effects of texture of food on chewing patterns in the human subject
.
J. Oral Rehabil.
16
,
177
-
183
.
Inoue
T.
,
Kato
T.
,
Masuda
Y.
,
Nakamura
T.
,
Kawamura
Y.
,
Morimoto
T.
(
1989
).
Modifications of masticatory behavior after trigeminal deafferentation in the rabbit
.
Exp. Brain Res.
74
,
579
-
591
.
Kobayashi
M.
,
Masuda
Y.
,
Kishino
M.
,
Ishida
T.
,
Maeda
N.
,
Morimoto
T.
(
2002
).
Characteristics of mastication in the anodontic mouse
.
J. Dent. Res.
81
,
594
-
597
.
Kuo
A. D.
(
2002
).
The relative roles of feedforward and feedback in the control of rhythmic movements
.
Motor Control
6
,
129
-
145
.
Lund
J. P.
(
1991
).
Mastication and its control by the brain stem
.
Crit. Rev. Oral Biol. Med.
2
,
33
-
64
.
Lund
J. P.
,
Kolta
A.
,
Westberg
K. G.
,
Scott
G.
(
1998
).
Brainstem mechanisms underlying feeding behaviors
.
Curr. Opin. Neurobiol.
8
,
718
-
724
.
Morimoto
T.
,
Inoue
T.
,
Nakamura
T.
,
Kawamura
Y.
(
1985
).
Characteristics of rhythmic jaw movements of the rabbit
.
Arch. Oral Biol.
30
,
673
-
677
.
Nakamura
Y.
,
Katakura
N.
(
1995
).
Generation of masticatory rhythm in the brainstem
.
Neurosci. Res.
23
,
1
-
19
.
Nakamura
Y.
,
Katakura
N.
,
Nakajima
M.
(
1999
).
Generation of rhythmical ingestive activities of the trigeminal, facial, and hypoglossal motoneurons in in vitro CNS preparations isolated from rats and mice
.
J. Med. Dental Sci.
46
,
63
-
73
.
Parker
H. G.
,
Kim
L. V.
,
Sutter
N. B.
,
Carlson
S.
,
Lorentzen
T. D.
,
Malek
T. B.
,
Johnson
G. S.
,
DeFrance
H. B.
,
Ostrander
E. A.
,
Kruglyak
L.
(
2004
).
Genetic structure of the purebred domestic dog
.
Science
304
,
1160
-
1164
.
Pearson
K. G.
(
2000
).
Neural adaptation I the generation of rhythmic behavior
.
Annu. Rev. Physiol.
62
,
723
-
753
.
Ross
C. F.
,
Dharia
R.
,
Herring
S. W.
,
Hylander
W. L.
,
Liu
Z.-J.
,
Rafferty
K. L.
,
Ravosa
M. J.
,
Williams
S. H.
(
2007
).
Modulation of mandibular loading and bite force in mammals during mastication
.
J. Exp. Biol.
210
,
1046
-
1063
.
Ross
C. F.
,
Reed
D. A.
,
Washington
R. L.
,
Eckhardt
A.
,
Anapol
F.
,
Shahnoor
N.
(
2009
).
Scaling of chew cycle duration in primates
.
Am. J. Phys. Anthropol.
138
,
30
-
44
.
Tanaka
S.
,
Kogo
M.
,
Chandler
S. H.
,
Matsuya
T.
(
1999
).
Localization of oral-motor rhythmogenic circuits in the isolated rat brainstem preparation
.
Brain Res.
812
,
190
-
199
.
Thexton
A. J.
,
Hiiemae
K. M.
,
Crompton
A. W.
(
1980
).
Food consistency and bite size as regulators of jaw movement during feeding in the cat
.
J. Neurophysiol.
44
,
456
-
474
.
Thompson
D. A. W.
(
1942
).
On Growth and Form
.
Cambridge
:
Cambridge University Press
.
Travers
J. B.
,
Dinardo
L. A.
,
Karimnamazi
D. H.
(
1997
).
Motor and premotor mechanisms of licking
.
Neurosci. Biobehav. Rev.
21
,
631
-
647
.
Tsuboi
A.
,
Kolta
A.
,
Chen
C. C.
,
Lund
J. P.
(
2003
).
Neurons of the trigeminal main sensory nucleus participate in the generation of rhythmic motor patterns
.
Eur. J. Neurosci.
17
,
229
-
238
.
Vila
C.
,
Savolainen
P.
,
Maldonado
J. E.
,
Amorim
I. R.
,
Rice
J. E.
,
Honeycutt
R. L.
,
Crandall
K. A.
,
Lundegerb
J.
,
Wayne
R. K.
(
1997
).
Multiple and ancient origins of the domestic dog
.
Science
276
,
1687
-
1689
.
Wainwright
P. C.
(
2002
).
The evolution of feeding motor patterns in vertebrates
.
Curr. Opini. Neurobiol.
12
,
691
-
695
.
Wehner
R.
,
Fukushi
T.
,
Isler
K.
(
2007
).
On being small: brain allometry in ants
.
Brain, Behav. Evol.
69
,
220
-
228
.
Wilson
D. E.
,
Reeder
D. M.
(ed.) (
2005
).
Mammal Species of the World: A Taxonomic Reference
, 3rd edn,
2142
pp.
Baltimore, MD
:
Johns Hopkins University Press
.
Wu
N.
,
Hsiao
C.-F.
,
Chandler
S. H.
(
2001
).
Membrane resonance and subthreshold membrane oscillations in mesencephalic V neurons: participants in burst generation
.
J. Neurosci.
21
,
3729
-
3739
.
Yamada
Y.
,
Yamamura
K.
(
1996
).
Possible factors which may affect phase durations in the natural chewing rhythm
.
Brain Res.
706
,
237
-
242
.