The dual function of prokinesis in the feeding and locomotor systems of parrots

Vertebrate skulls range in intrinsic mobility from essentially akinetic (e.g. some mammals, turtles and crocodilians; Ferreira et al., 2020; Preuschoft and Witzel, 2002) to modestly kinetic (e.g. some lizards and early tetrapodomorphs; Arnold, 1998; Herrel et al., 1999; Lemberg et al., 2021) to highly kinetic (e.g. some snakes, birds and fishes; Bout and Zweers, 2001; Lee et al., 1999; Westneat, 2005). Although such kinesis can occur at various locations throughout the cranium, many avian species exhibit prokinesis, in which significant motion occurs between the neurocranium and upper beak (Fig. 1). Prokinetic excursions of the beak are powered by protractor muscles, which pull the quadrate in the anteroposterior plane (Bout and Zweers, 2001), rotating the otic (quadratosquamosal) joint. This movement then shifts palatal elements anteriorly (Wilken et al., 2020), producing rotation at the craniofacial hinge (Bock, 1964; Navalón, 2019; Navalón et al., 2019; Olsen and Westneat, 2016). Anatomically, prokinesis is associated with a complex suite of features (Holliday and Witmer, 2008) that varies between avian lineages and can include a synovial craniofacial hinge or bending zone at the nasal-frontal hinge, and robust protractor musculature (Bock, 1964; Navalón, 2019; Navalón et al., 2019).

Prokinesis has been termed a key innovation in avian evolution (Gussekloo and Bout, 2005; Gussekloo et al., 2001), but its adaptive significance remains largely unclear. Most existing functional studies of prokinesis in birds have focused on its role in feeding and drinking (Dawson et al., 2011; Gussekloo and Bout, 2005). However, several elements of the prokinetic feeding apparatus are also used in additional behaviors, to the point that the beak is sometimes thought of as an avian ‘surrogate hand’ (Bhullar et al., 2016). The multifunctionality of this region of the Bauplan therefore raises the question of how prokinesis additionally contributes towards non-feeding behaviors. As a result, elucidating whether prokinesis has broader utility than previously appreciated is essential to developing a more comprehensive understanding of the evolutionary pressures and drivers shaping the origin and maintenance of cranial kinesis.

To evaluate the potential role of prokinesis outside the feeding system, we experimentally investigated the craniofacial hinge of the rosy-faced lovebird (Agapornis roseicollis). These parrots employ a distinctive tripedal gait when climbing vertical structures (Dilger, 1960; Granatosky et al., 2022; Gupta et al., 2020), using the beak not only to stabilize themselves (Dilger, 1960; Reader et al., 2021; Young et al., 2023), but also to generate comparable propulsive forces to the hindlimbs (Young et al., 2022). Therefore, although prokinetic motion of the craniofacial hinge has been inferred to accommodate the mechanically tough diet of parrots (Munshi-South and Wilkinson, 2006; Renton, 2001; Fig. 1), it is possible that this mode of cranial kinesis may also play an essential role in climbing. Here, we conducted a comparative analysis of the movements and force-generating potential of the parrot mandible and maxilla (via flexion of the craniofacial hinge) during both feeding (forceful jaw adduction) and tripedal locomotion. Ultimately, we discuss the implications of our results for understanding the broader evolution and neuromuscular control of cranial kinesis.

To assess patterns of force and excursion of the craniofacial hinge during locomotion and feeding, we conducted two separate experiments. Data were collected from between four (simulated feeding trials) to six (locomotion trials) young adult rosy-faced lovebirds [Agapornis roseicollis (Vieillot 1818) body mass range: 44.9–53.0 g; mean±s.d.: 48.8±2.8 g], all of which were free from any observable anatomical or behavioral pathologies. Animal protocols were approved by the New York Institute of Technology Institutional Animal Care and Use Committee (Protocol number: 2021-MG-03).

Locomotor trials were conducted following methods established in Young et al. (2022). Substrate reaction forces were collected from the parrot beak during vertical climbing on an instrumented vertical runway (0.61 cm×0.15 m). The instrumented portion (0.05 m×0.08 m) was connected via a three-dimensional printed platform to a calibrated, small-load force plate (HE6X6; Advanced Mechanical Technology, Inc., Watertown, MA) and flush mounted to the surrounding runway with a small separation. The entire surface of the runway was covered in rubber shelf liner to provide traction and Procapture (Xcitex Inc., Woburn, MA) was used to synchronize force (sampled at 1250 Hz) and video data (125 Hz) from a laterally mounted, high-speed camera (XC-1 mol l⁻¹; Xcitex Inc., Woburn, MA).

Kinematic variables for locomotor trials were collected using DeepLabCut (Mathis et al., 2018), a high-throughput point estimation program that employs machine learning. Approximately 200 frames of parrots climbing vertically were extracted from the video. Markers were placed on the base and tip of the upper and lower beak, as well as on the superior aspect of head, the eye, the front of the wing (Fig. S1). We also placed four constant calibration landmarks. The neural network was trained to label the remaining frames from the sample. A custom script written in MATLAB (MathWorks, Natick, MA) subsequently calculated angular and linear measurements using the digitized points from each video frame. Further measurements unable to be obtained through DeepLabCut, such as the angle between the parrot beak and the substrate, were identified manually within ImageJ (https://imagej.net/ij/); these values were used to calculate the linear distance traversed by the neck and beak within a single stride. The onset of a stride was considered the point of touchdown by the beak (see Movie 1). For each video, the length of stance phase was determined by the number of frames between touchdown and lift-off of the beak, converted into seconds using the video frame rate. Excursions of the maxilla and mandible were calculated by subtracting the minimum angle from the maximum angle measured for the maxilla and mandible, respectively, during stance phase. To determine the relative contribution of the maxilla and mandible toward propulsion during vertical climbing, we further calculated the coefficient of stance phase (CSP) – a metric of any individual joint's contribution towards forward propulsion– for each component following Schmidt and Fischer (2000). We first calculated effective angular displacements (EADs) as the difference in angle for each component between touchdown and lift-off. Subsequently, EAD was divided by total amplitude (i.e. the difference between the maximum and minimum angles exhibited by each joint during a single stride) to determine CSP (Schmidt and Fischer, 2000).

To assess the force-generating potential of the maxilla via flexion of the craniofacial hinge, only trials containing a clear isolated beak force were analyzed for locomotor trials. All substrate reaction forces in the propulsive/braking plane (i.e. vertical movements parallel to the substrate) were analyzed and run through a low-pass Fourier filter at 15 Hz in a custom-written MATLAB script. The area under the force–time curve was then used to calculate propulsive impulses, measured in body weight seconds (%BWS) using each individual's body mass.

While it is impossible to completely isolate the propulsive contributions of the cranio-cervical system (e.g. mandible, maxilla and neck), based on visual observation of parrot climbing, there were trials in which movement of the craniofacial hinge preceded that of the neck. We quantified the amount of time craniofacial hinge movement preceded neck movement using custom-written R script to isolate the propulsive forces attributed to craniofacial movement. Briefly, using the tracked kinematic profiles of the craniofacial hinge and neck, the code first established whether the trial was suitable for analysis by determining if the craniofacial hinge did indeed move prior to the neck. This was done by first isolating the first five points from the craniofacial hinge and neck and assessing whether a significant negative relationship between these joint movements and the proportion of the stride (i.e. suggesting joint flexion). This was repeated iteratively on a sliding five-point window until a significant negative relationship was observed. In trials in which a clear distinction between craniofacial flexion was observed prior to neck flexion, the time lag between movement at the joints was calculated to isolate the corresponding propulsive impulse. It should be noted this method only isolates the propulsive forces of the craniofacial hinge prior to neck flexion. As such, any propulsive impulse calculated from this method should be considered a minimum of the potential locomotor forces attributable to the craniofacial hinge.

Forces during forceful jaw adduction were measured using a custom-designed, three-dimensional printed mobile force transducer (see Fig. 2). The two bite plates (spaced 10 mm apart) were attached to a dual hinge system that allowed for separate, independent movement of the plates. Tension springs (McMaster-Carr, Elmhert IL) of increasing spring constants (spring 1=183.1 N m⁻¹; spring 2=357.5 N m⁻¹; spring 3=2004.7 N m⁻¹; spring 4=6121.7 N m⁻¹) provided resistance as birds contacted the transducer. Moment arms between the bite point (i.e. in-lever), center of rotation and tension spring (i.e. out-lever) were equal in length to aid in force calculations. Two separate springs of the same tension were attached to the dual hinge system to allow for the isolation of excursion and force production from the maxilla (top plate) and mandible (bottom plate) and the springs were interchanged until trials were collected on each spring condition. Parrots readily bit the two plates resulting in at least 60 trials per spring condition. All trials were filmed at 125 Hz from a lateral view using Procapture (Xcitex Inc., Woburn, MA).

All statistical analyses were conducted using R using the ‘lme4’ package (https://CRAN.R-project.org/package=lme4). Shapiro–Wilk tests determined normality of the data sets and all data were rank transformed prior to any analyses. A series of linear mixed effect models were constructed, holding individual as a random effect to account for individual idiosyncrasies (arXiv.1308.5499). For the locomotor trials, three models were created to compare angular excursion, CSP and EAD between the craniofacial hinge and the mandible. An additional model was created to compare linear distance travelled between the beak and neck. For jaw adduction trials, two models were created to compare the force and excursion between the craniofacial hinge and mandible, controlling for spring tension as a fixed effect. We also created two additional models wherein we drop maxilla versus mandible in the model to assess the importance of this term in predicting force generation and excursion. A final model was created to compare the excursion between the jaw adduction and locomotor trials, holding maxilla versus mandible as a fixed effect. Lastly, excursions of the craniofacial hinge and mandible, and linear distances travelled attributable to the beak and neck, were regressed against stance phase, stride length and velocity to evaluate the influence of these gait characteristics upon gait kinematics.

A total of 79 trials were suitable for analysis, all of which featured visible flexion at the craniofacial hinge. As expected, parrots universally employed a tripedal gait when scaling the vertical surface. Upon touchdown of the beak (onset of stance phase), the craniofacial hinge angle is extended and steadily flexes throughout stance phase as the bird pulls itself upward (Fig. 3A). Meanwhile, the mandibular angle remains relatively stable throughout the stance phase but does increase slightly as the stance phase progresses. The angular excursion of the craniofacial hinge (22.38±5.70 deg) exceeds that of the mandible (9.03±2.14 deg) in a given stance (Fig. 3D; Table 1; P<0.001). As shown via CSP, movements at the craniofacial hinge are geared toward propulsion (CSP=0.71±0.18; Fig. 3E), whereas the mandible contributes a negligible amount toward propulsion (CSP=0.06±0.02), with highly significant differences between the two structures (Table 1, P<0.001). Excursion of the craniofacial hinge was more strongly correlated with velocity (r²=0.39) and stride length (r²=0.47) than that of the mandible (r²=0.23 and 0.40, respectively; Fig. 4). During stance phase, the neck (0.02±0.01 m) traverses twice as far as the beak (0.01±0.00 m), a difference that was highly significant (Fig. 3C, Table 1; P<0.001). The distance travelled by the neck was more strongly correlated with both stance phase stride length (r²=0.53 versus 0.30) and velocity (r²=0.51 versus 0.12) compared with values in the beak (Fig. 4).

During stance phase, propulsive forces increased steadily and peaked just prior to midstance, after which forces declined as the bird pulled itself up via shortening of the neck (Fig. 3B). From 13 trials in which clear differentiation between the onset of craniofacial hinge flexion and the onset of neck flexion could be determined (16.5% of total trials), net propulsive impulse of the maxilla indicated a propulsive contribution by the maxilla (18.30±14.48%BWS; Fig. 3F).

Across all spring conditions, the mandible participated in a greater frequency of forceful jaw adduction events than the maxilla, with maxillary contributions declining significantly on spring 4 compared with the other three conditions (spring 1: mandible=188 events, maxilla=165 events; spring 2: mandible=263 events, maxilla=245 events; spring 3: mandible=125 events, maxilla=110 events; spring 4: mandible=134 events, maxilla=12 events). During the lowest spring resistance, both the mandible and maxilla demonstrated high magnitudes of angular excursion. As spring resistance increased, angular excursion of both the mandible and the maxilla decreased significantly (mandibular: spring 1=15.23±7.60 deg, spring 2=10.07±2.67 deg, spring 3=7.52±3.27 deg, spring 4=2.49±1.42 deg; maxillary: spring 1=9.63±4.16 deg, spring 2=5.83±3.25 deg, spring 3=3.61±2.20 deg, spring 4=0.7±0.54 deg; all P<0.001), whereas force production increased (mandibular: spring 1=0.36±0.7 N, spring 2=0.82±0.38 N, spring 3=2.16±0.94 N, spring 4=2.22±0.77 N; maxillary: spring 1=0.25±0.06 N, spring 2=0.74±0.39 N, spring 3=1.27±0.81 N, spring 4=0.99±0.26 N; Table 2, Fig. 5; all P<0.001).

As angular excursions during forceful jaw adduction were significantly influenced by spring resistance, we limited statistical comparison between systems (i.e. locomotion versus jaw adduction) to only the lowest resistance spring. Such a comparison revealed that movements in both the maxilla and mandible were significantly greater during locomotion (Table 1,P<0.001). However, while maxillary movements were significantly greater during locomotion, mandibular movements were greater during forceful jaw adduction (Fig. 6, Tables 1–3, all P<0.001). Force-generating capabilities between the two systems could not be compared meaningfully as we were only able to isolate the force-generating potential of the maxilla up to the beginning of neck movement. Since in most trials, neck and mandibular movement happened in concurrence, and sometimes even preceded maxillary movement, the contributions of the neck or mandible could not be isolated from those of the maxilla. For these methodological limitations, we could not directly compare the force-generating potential between tripedal climbing and forceful jaw adduction.

Our findings reveal that maxillary flexion at the craniofacial hinge contributes to propulsion during vertical climbing in parrots. Notably, the CSP of the maxilla while climbing is comparable to those of limb joints in locomotor systems, including the wrist joint of a two-toed sloth (Choloepus didactylus, CSP=0.67; Nyakatura and Fischer, 2011) during horizontal movement along a ‘treadpole’ (arboreal analogue of a treadmill) and the shoulder of the common treeshrew (Tupaia glis, CSP=0.70; Fischer et al., 2002; Schilling and Fischer, 1999) during horizontal movement along a treadmill, in which the majority of joint excursions contribute to forward progress. By contrast, movements of the mandible do not appear to be propulsive (CSP=0.06±0.02).

While the neck musculature is likely to be responsible for most of full body propulsion during vertical climbing, we further demonstrate that in trials for which neck flexion and maxillary flexion could be separated (13 trials), the maxilla is solely responsible for the initial 18.30±14.48%BWS of propulsion. This propulsive impulse represents a minimum value (as maxillary flexion likely continues in parallel with neck flexion later in the stride) and further studies are necessary to elucidate the exact contributions of these systems during vertical climbing.

Conversely (and across all tension conditions), movement of the mandible (as opposed to maxillary flexion) was the primary driver of force generation during jaw adduction. Maxillary excursions remained consistently less than those of the mandible, nearly disappearing in the highest resistance condition. In light of these findings, we suggest that during feeding, movement of the mandible is the primary contributor to breaking the food, whereas the maxilla is used both to manipulate the food item in position (mobility) and provide a solid anvil to resist movement of the mandible (stability). Again, future studies exploring oral food manipulation are necessary to explicitly explore this hypothesis. Thus, while maxillary excursions dominate during locomotion, this trend reverses to mandible-dominant excursions during forceful jaw adduction, demonstrating the dual functionality of cranial kinesis in both feeding and locomotor behaviors.

Incorporating the feeding system into locomotor cycling represents a remarkable innovation with regards to neuromuscular control of locomotion, given that movements of the hindlimbs are regulated by spinal central pattern generators whereas muscles powering cranial kinesis are regulated by cranial nerves. Although the three-dimensional, parallel four bar linkage system of the avian skull limits the degree of freedom of the kinetic apparatus (Olsen, 2019; Olsen and Westneat, 2016), coordination of multiple elements in the avian skull is necessary to achieve desired postures and kinematics (Olsen, 2019). Thus, it is necessary for parrots to coordinate several muscles across the maxilla and mandible to maintain function during both locomotion and forceful jaw adduction, a consideration made more complex by the differing levels of excursion required for these two behaviors.

The mobility of the craniofacial hinge of the parrot far exceeds the mobility of the craniofacial hinge of other birds (Fig. 1). Rotation/flexion of the upper beak has been previously quantified in various birds via in silico simulations, cadaveric manipulations and in vivo recordings (e.g. Hesperiphona vespertina, Gallus domesticus, Columba livia, Zonotrichia albicollis, Anas platyrhynchos, Bubo virginianus), exhibiting roughly 5–15 deg of rotational capacity (Bout and Zweers, 2001; Heuvel, 1991; Hoese and Westneat, 1996; Olsen and Westneat, 2016; Van Gennip and Berkhoudt, 1992; Zusi, 1967). By contrast, during the stance phase of climbing, maximum excursions in rosy-faced lovebirds reached almost 40 deg (mean±s.d.=22.38±5.70 deg; Fig. 3D), a degree of cranial kinesis that approaches the extreme mobility observed in suction-feeding teleost fishes with 20 or more independently mobile cranial elements (Westneat, 2005). Although excursions of the craniofacial hinge during feeding are more conservative, with a maximum rotation of almost 20 deg (Figs 5 and 6). this mobility still dwarfs the documented excursions of the craniofacial hinge in other birds. The increased mobility of the craniofacial hinge in parrots relative to other birds can likely be attributed to the morphology and development of the joint (Tokita, 2003; Tokita, 2003; Manafzadeh, 2023). The ‘pseudoprokinetic’ craniofacial hinge of parrots arises from within the nasal capsule rather than between the nasal capsule and the frontal, as seen in prokinetic birds (Tokita, 2003). While most birds exhibit some degree of flexibility at the craniofacial hinge (Cost et al., 2017, 2020; Tokita, 2003; Zusi, 1984), the craniofacial hinge of parrots is a simplified, single synovial joint, as opposed to a complex nasal-frontal suture that allows for bending, as seen in other birds (Bailleul and Horner, 2016; Tokita, 2003).

There are many examples in which aspects of the locomotor system have been co-opted into feeding behaviors, ranging from the manual manipulation of tools and food items (Alcock, 1972) to more specialized exaptations (e.g. the use of the pectoral girdle during suction feeding in the largemouth bass, channel catfish and white-spotted bamboo shark; Camp and Brainerd, 2014; Camp et al., 2017; Camp et al., 2020). However, to our knowledge, this is the first documentation of a feeding structure exapted for locomotion. The difficulty of this evolutionary scenario may be attributable to the divergent optimality criteria underlying feeding versus locomotion. While locomotor systems are generally optimized towards generating forces over a wide range of potential joint excursions, the feeding system is tuned towards maximizing forces over a more narrow range displacement range (Anderson and Roberts, 2020; Faltings et al., 2022; Granatosky and Ross, 2020; Granatosky et al., 2019). Integrating musculoskeletal structures originally optimized for force generation during feeding into locomotion therefore presents an obvious challenge relating to their inherent excursion ranges. However, the notably large excursions seen in the craniofacial hinge of climbing parrots clearly overcomes this barrier, while also representing a potential mechanism to simultaneously increase overall stride length while reducing energetic cost of transport (Granatosky and McElroy, 2022; Granatosky et al., 2019; Pontzer, 2007).

This type of study inherently comes with certain limitations that require careful consideration. Firstly, our ability to precisely differentiate between the contributions of the neck and the maxilla in calculating propulsive force is constrained. Although the neck musculature is presumed to play a significant role in overall body propulsion during vertical climbing, our findings indicate that, in instances where neck flexion and maxillary flexion could be isolated, the maxilla alone accounts for the initial propulsion. It is crucial to note that this value represents a conservative estimate, as maxillary flexion likely continues concurrently with neck flexion in later strides. Further research is essential to comprehensively understand the specific contributions of these systems in vertical climbing.

Furthermore, the conventional understanding of biting often assumes an equilibrium of forces between the anvil (maxilla) and the hammer (mandible). Yet, in the design of our apparatus, there is no connection between the two springs or two bite plates, and because the mandible and maxilla can move independently and are not activated by the same muscles or nerves, there is no physical requirement for the two to provide equal and opposite forces. Given these considerations, it is more accurate to refrain from characterizing the observed behavior on our apparatus as ‘biting’ (hence, our use of ‘forceful jaw adduction’ throughout when discussing this movement). Consequently, caution is advised when extrapolating such data to infer bite force potential in parrots. Please see Dickinson and colleagues (2022) for traditional force plate derived measures of bite force.

Related to this aforementioned concern, the design of our apparatus grapples with significant limitations in its capacity to yield precise measures of bite force. The accuracy of our results relies on the assumption that force vectors are perfectly perpendicular to the plate, a condition rendered unattainable because of the curved shape, extreme overbite and the necessitated handling position for the bird. Although, in most instances, parrots positioned the mandible to achieve a force vector orientation more orthogonal than the maxilla, the quantification of the proportions of these placement angles was not undertaken. Consequently, this leads to the production of lower out-forces on the maxilla compared with the mandible, potentially explaining observed differences in output force. While acknowledging this experimental limitation, the substantial disparity in output forces between the mandible and maxilla appears to be a biologically meaningful finding extending beyond the constraints of experimental design. Additionally, the escalation of this discrepancy with increased spring resistance supports our assertions. Considering these considerations, exercising caution is prudent when extrapolating such data to make inferences about bite force potential in parrots.

Seldom in nature do instances exist in which distinctly different cyclic behaviors (i.e. feeding and locomotion) can be directly compared within the same anatomical system. Parrots act as a model system in which the role of craniofacial hinge switches in response to the behavioral context, revealing remarkable and previously unappreciated functional plasticity in prokinesis. Developing a more comprehensive understanding of the drivers and constraints surrounding cranial kinesis and its neuromuscular control is paramount in better illuminating the origins of birds (Benito et al., 2022; Bhullar et al., 2016; Holliday and Witmer, 2008). As a result, we suggest that accurately interpreting the adaptive significance of cranial kinesis both in birds and more broadly in vertebrate evolution will require deliberately investigating the full spectrum of behaviors encompassed by the kinetic apparatus.

We are grateful to Associate Editor Monica Daley and two anonymous reviewers for their valuable contributions to an earlier version of this manuscript. The authors wish to thank the animal husbandry staff at NYITCOM for helping with animal care and use, and the NYITCOM Visualization Center for collecting CT data on Agapornis.