Feeding biomechanics of juvenile red snapper (Lutjanus campechanus) from the northwestern Gulf of Mexico

Ecomorphological studies are predicated on identifying patterns among morphology, behavioral performance and ecology(Wainwright, 1994), and have been utilized to test a variety of hypotheses concerning the relationships between feeding performance and foraging ecology among teleosts (e.g. Clifton and Motta, 1988; Wainwright,1996). Ecomorphological studies are also useful in examining the functional consequences of ontogenetic changes on morphology and diet shifts in teleosts (e.g. Osenberg et al.,1988; Hyndes et al.,1997; Hunt von Herbing,2001; Graeb et al.,2005; Monteiro et al.,2005). Ontogenetic shifts reduce competition through intraspecific(Hernandez and Motta, 1997; Hyndes et al., 1997; Soto et al., 1998) or interspecific (Mittelbach et al.,1992; Huskey and Turingan,2001) resource partitioning. Furthermore, such shifts can reduce predation (Werner and Gilliam,1984) and maximize growth rates(Olson, 1996; Post, 2003) of teleosts during early life development. Fast growth is most advantageous during larval and juvenile stages when individuals are most vulnerable to predation(Werner and Gilliam, 1984; Post, 2003). Therefore,improving a juvenile's ability to take advantage of the most abundant or high energy food source over ontogeny may increase individual fitness and enhance recruitment potential (Olson,1996; Persson and Brönmark, 2002; Post,2003).

Early life history studies that focus on the interaction between skull development, feeding mechanics, and their ecological consequences, are an important way to address critical questions in the ecomorphology of fishes. This study used juvenile red snapper (Lutjanus campechanus Poey 1860), to investigate changes in skull development and feeding biomechanics to provide an ecomorphological explanation of divergent early life history patterns. Larval red snapper settle out of the water column at approximately 16 mm (Rooker et al., 2004)and are attracted to complex habitats, which serve as essential nursery grounds for juveniles (Szedlmayer and Howe, 1997). The settlement patterns within these habitats remain unclear. Significantly higher recruitment occurs on shell ridges (on-ridge) in the northeastern Gulf of Mexico(Szedlmayer and Conti, 1999)and on adjacent mud habitats (off-ridge) in the northwestern Gulf of Mexico(Rooker et al., 2004). Juvenile growth rates are significantly higher in off-ridge areas around Freeport Rocks (Rooker et al.,2004; Geary et al.,2007), suggesting that available prey resources may differ between off-ridge mud bottoms and on-ridge shell ridges. Therefore we asked, `Do juveniles respond to prey availability by altering feeding morphology or modulating feeding behavior?'

Although red snapper larval development(Collins et al., 1980; Pothoff et al., 1988; Drass et al., 2000) and diet of both adults and juveniles have been examined(Bradley and Bryan, 1976; Moran, 1988; Ouzts and Szedlmayer, 2003; Szedlmayer and Lee, 2004),their feeding mechanics and behavior have not been investigated. Therefore,this research explored the relationships between morphology and feeding kinematics within the context of trophic ecology of juvenile red snapper using biomechanical modeling, kinematic behavioral performance tests and shape analysis. We hypothesized that juvenile red snapper settling onto different habitats would exhibit a divergence in skull morphology and/or feeding biomechanics that may be correlated to reported juvenile diet patterns(Szedlmayer and Lee, 2004). To test these hypotheses, we performed experiments involving juvenile red snapper collected from different habitats and carried out biomechanical modeling of their jaws. This allowed us to determine how interactions between feeding ecology and functional morphology may influence growth and settlement patterns by testing for significant differences in jaw morphology, lever mechanics,kinematics and phenotypic plasticity among juvenile red snapper across three size classes (≤3.9, 4.0–5.9, ≥6.0 cm SL) and between two nursery habitats (on-ridge and off-ridge).

Juvenile red snapper were collected between June and September 2004, and in August 2005, on and off the Freeport rocks shell ridge (Freeport, TX, USA). On-ridge areas were characterized by abundant relic oyster shell; off-ridge sites were characterized by silt and mud. Juvenile red snapper(N=530) were collected using a 6-m otter trawl with 2 cm mesh, 1.25 cm inner mesh, 0.6 cm link tickler chain, and 0.457×0.914 m doors. Trawls were made in 5-min increments at 2.5 knots. Juveniles for kinematics studies (N=17) were sorted by habitat (on-ridge N=8 and off-ridge N=9) and kept in separate `live' wells onboard the research vessel. Additional subjects were anesthetized then frozen and kept for jaw lever analyses (N=230), and shape analyses (N=111). Mass (g)and standard length (SL; cm) were recorded for all juveniles and assigned to the following size classes, small (1.8–3.9 cm SL), medium (4.0–5.9 cm SL) or large (6.0–10.88 cm SL). Collections were made under TAMU IACUC Animal Use Protocol no. 2003-84 and Texas Park and Wildlife Permit no. SPR 0902-243.

Prior to conducting parametric statistical tests, normality of all data was tested using a Kolmogorov–Smirnov test. If normality was not met, data were transformed. Levene's test was used to test the assumption of homogeneity of variances. Bonferroni post-hoc tests were used when the assumption of equal variance was met; Dunnett's t3 post-hoc tests were used in cases where variances were heteroscedastic. All statistical tests were conducted using SPSS 11 (SPSS, Chicago, IL, USA) for a Mac and JMP 6 (SAS,Cary, NC, USA). More specific statistical analyses are listed under each methodological subheading (model of lower jaw lever mechanics, feeding kinematics, and phenotypic variation).

Lever mechanics were used to calculate the trade off between velocity and force (Wainwright and Richard,1995; Westneat,1994; Westneat,2003), and make predictions about the feeding mode of juvenile red snapper. The biomechanics of juvenile red snapper feeding were modeled by investigating the anatomical arrangement of the lower jaw as a third order lever using the program MandibLever 3.2(Westneat, 2003). This model incorporates the influence of closing muscles on lever ratio calculations and creates a set of dynamic output variables over the entire jaw closing. The use of a dynamic model is advantageous since static measurements usually overestimate mechanical advantage because the influence of changing muscle insertion angles is not accounted for(Westneat, 2003). The model,therefore, calculates an effective mechanical advantage (EMA), which is a more accurate measurement of force transmission from muscle to the lower jaw. The model also calculates a variety of other dynamic variables, such as bite force, angular velocity and percent muscle contraction, and these parameters can be used to make predictions regarding fish feeding kinematics.

Morphometric measurements of the lower jaw and associated jaw closing muscles (the A2 and A3 subdivisions of the adductor mandibulae; Fig. 1, Table 1) were taken to the nearest 0.01 cm using either a calibrated eye reticule on a Nikon SMZ1500 stereoscope, or with digital vernier calipers. The following 12 measurements were collected: (1) in-lever A2, from quadrate-articular joint to A2 insertion point on ascending process of articular; (2) in-lever A3, from quadrate-articular joint to A3 insertion point on medial face of lower jaw;(3) in-lever Open, from quadrate-articular joint to insertion of interoperculomandibular ligament on posteroventral margin of articular; (4)out-lever, from quadrate-articular joint to anterior most tip of dentary; (5)A2 length, from origin on ventral margin on preopercle to insertion on ascending process of articular; (6) A3 total length, from origin on preopercle and hyomandibula to insertion on medial face of lower jaw; (7) A3 tendon length, from origin on tapering end of A3 muscle to insertion on the medial face of the lower jaw; (8) A2–joint distance, distance from A2 origin to quadrate-articular joint; (9) A3–joint distance, distance from A3 origin to quadrate-articular joint; (10) A2–A3 ins, distance from A2 insertion to A3 insertion; (11) LJtop length, from the tip of the coronoid process to the anterior jaw tip; and (12) LJBot length, from the posteroventral margin of the articular to the anterior jaw tip. Mass of the A2 and A3 muscles were recorded to the nearest 0.01 g. Assumptions regarding jaw muscle contractile physiology were made following Westneat(Westneat, 2003): maximum shortening velocity, or V_max (10 L s^–1),maximum isometric stress of muscle contraction, or P_max(200 kPa), dynamic contraction velocity of muscle (0.05–0.8 of V_max), isometric force per unit area of muscle(0.05–0.79 of P_max); and a peak jaw opening rotation value based on juvenile red snapper kinematic data (57°). Jaw muscle contraction percentage was calculated as the percent change in length from the open (stretched) position. Morphometric measurements and muscular assumptions were used as inputs in the biomechanical lever model, available free on the web from the second author. A total of 230 simulations of lower jaw closing were run to predict feeding behavior of juvenile red snapper from three size classes and two habitats.

To test the hypothesis that there was no significant difference in the morphology of the feeding apparatus of juvenile red snapper across ontogeny,output parameters were analyzed using multivariate analyses of variance(MANOVA) with size class as a fixed factor, and model output parameters as dependent variables. Significant differences among size classes were determined by post-hoc tests. To test for significant differences between habitats, output parameters were analyzed using multivariate analyses of covariance (MANCOVA) with habitat as a fixed factor, model output parameters as dependent variables and standard length as a covariate.

Feeding kinematic trials were used to validate the predictive biomechanical model output, and compare the feeding biomechanics of juvenile red snapper across size classes and between habitats. Juvenile red snapper were transported to the laboratory and housed in habitat-specific 38–189 l saltwater tanks and maintained at 26°C, 32 p.p.t. salinity, and pH 8.2. Fish were allowed to acclimatize, and then trained to feed from a stationary tube under 500 W of light. During the first collection season, a mass mortality event occurred because of an Amyloodinium ocellateumoutbreak during a hurricane evacuation. Not enough individuals were available to investigate ontogenetic changes; therefore, only a habitat treatment was included in the kinematic analysis. Juveniles used in feeding kinematic trials(on-ridge N=8, off-ridge N=9) all fell within the medium size class (4.0–5.9 cm SL).

Juveniles were positioned laterally in front of the camera using a piece of Plexiglas with a 1 cm² grid as a reference and fed pieces of squid,sized to 50% of the individual's oral diameter, until satiated. Feeding events were recorded using a Redlake PCI Motion Scope high-speed camera at 250 frames s^–1. Three representative feeding events for each juvenile were selected for analysis. Juveniles were then sacrificed with an overdose of methane tricaine sulphonate (MS-222). Feeding events were digitized frame by frame, starting with the onset of strike until mouthparts returned to their starting position, using Motus 8.2 (Vicon, Denver, CO, USA). Digitized points(Fig. 2A) included: (A) the anterior tip of the premaxilla, (B) the anterior tip of the dentary, (C) the dorsal most visible point of the maxilla, (D) the maxilla–premaxilla articulation, (E) the mandible–quadrate articulation, (F) the ventral floor of mouth, (G) the posterior-most point of the orbit of the eye, (H) the first dorsal spine origin, (I) the anterodorsal tip of the opercle at the junction with the preopercle and the hyomandibula, (J) the posterodorsal tip of the opercle, (K) the origin of the first pectoral fin ray. These 11 anatomical landmarks were used to calculate the following 14 kinematic variables: (1) maximum gape (cm), (2) time to maximum gape (ms), (3) maximum gape angle (degrees), (4) time to maximum gape angle (ms), (5) maximum lower jaw rotation (degrees), (6) time to maximum lower jaw rotation (ms), (7)maximum upper jaw protrusion (cm), (8) time to maximum upper jaw protrusion(ms), (9) maximum cranial rotation (degrees), (10) time to maximum cranial rotation (ms), (11) maximum depression of the hyoid (cm), (12) time to maximum hyoid depression (ms), (13) maximum maxillary rotation (degrees), and (14)time to maximum maxillary rotation (ms). Angular velocities and phase timings were also calculated.

Scatter plots of gape distance and gape angle versus closing duration were used to determine the predictive ability of the lever model by comparing data from the lever model and live kinematics. To statistically test the model as an accurate predictor of feeding behavior, we log-transformed the gape, gape angle, and time axes to linearize the curvilinear relationships between time and kinematics, and performed analysis of covariance (ANCOVA) to test whether slopes and/or y-intercepts of the model and video data were significantly different.

Kinematic variables were also used to characterize and quantify the feeding behavior of juvenile red snapper between habitats. To test the hypothesis that there was no significant difference in feeding behavior of juvenile red snapper between habitats (P≤0.05), kinematic variables were analyzed using multivariate analysis of variance (MANOVA) with habitat as a fixed factor and kinematic variables as dependent variables. Kinematic profiles were generated for each variable to examine their relationship to one another and identify different phases over a complete feeding event.

Shape variables were collected to investigate differences in body shape of juvenile red snapper from the on-ridge and off-ridge habitats. Lateral images of juvenile red snapper (on-ridge N=56, off-ridge N=55;small N=50, medium N=33, large N=28) were captured using a digital camera. Two-dimensional coordinates were recorded from the following 19 landmarks (Fig. 2B) digitized around the juvenile body perimeter using the program tpsDig (v. 2) (Rohlf, 2005a):(1) anterior tip of the dentary, (2) anterior tip of the premaxilla, (3)anterior-most point of the eye orbit, (4) center of the eye, (5)posterior-most point of the eye orbit, (6) anterior-most point of the frontal bone, (7, 8) anterior and posterior insertions of the dorsal fin,respectively, (9) dorsal origin of the caudal fin, (10) middle of caudal fin insertion where the lateral line terminates, (11) ventral origin of the caudal fin, (12, 13) posterior and anterior insertions of the anal fin, respectively,(14) anterior-most insertion of the pelvic fin, (15) first branchiostegal ray at the body outline, (16) quadrate-articular joint, (17) origin of the first pectoral fin ray, (18) posterodorsal tip of the opercle, (19) anterior-most point of the lateral line. TpsRelw software (v. 1.42)(Rohlf, 2005b) was used to align shape data by rotating, translating and scaling the landmark coordinates, using least squares superimposition. Aligned data were used to calculate shape variables. Significant variations in shape were tested using MANOVA with shape variables as dependent variables, and habitat and size as fixed variables. An eigendecomposition of the effect sum of squares and cross-products (SSCP) matrix was performed and used to calculate the shape variance explained by habitat and allometry. In addition, associated eigenvectors were multiplied by shape variables to yield linear axis scores. TpsRegr software (v.1.31) (Rohlf,2003) produced thin-plate spline transformation grids(Fig. 6), which provided a visualization of shape variation.

Model simulations demonstrated that for the A3 muscle effective mechanical advantage increased 3% (A2 P=0.28, A3 P=0.02, MANOVA) and velocity ratio decreased 6.7% (A2 P=0.22, A3 P<0.001,MANOVA) as body size increased (Fig. 3A,B, Table 2). Muscle force contribution (A2 and A3 P<0.001, MANOVA) and total bite force (P<0.001, MANOVA) significantly increased four to eight times with body size (Fig. 3C,D, Table 2), as expected with an associated increase in muscle cross-sectional area. The potential functional roles of the A2 and A3 muscles changed over ontogeny. A3 muscle force contribution to total bite force was two times greater than the A2 muscle over ontogeny, with a substantial increase over the A2 muscle in large juveniles, suggesting a possible mechanism for shifts in juvenile feeding mode (Fig. 3C,D). As gape increased with size, A3 muscle jaw closing duration significantly decreased 8.4% (A2 P=0.56, A3 P<0.001, MANOVA) resulting in an expected significant 16.2% increase in angular velocity of the A3 muscle(A2 P=0.34, A3 P=0.002, MANOVA; Fig. 3E,F, Table 2). The A2 muscle angular velocity was faster than the A3 muscle in small juveniles but slower than the A3 muscle in medium and large juveniles(Fig. 3F). Percent muscle contraction required to close the lower jaw generally decreased 2% over ontogeny for the A2 muscle and significantly by 8.5% for the A3 muscle (A2 P=0.56, A3 P<0.001, MANOVA; Table 2). The A3 muscle increased in size at a faster rate than the A2 muscle, causing the A3 muscle morphology and function to change to a greater extent over ontogeny than the A2 muscle. As the A3 in-lever became longer, A3 force contribution increased as a result of an increase in mechanical advantage. As the A3 muscle became longer the angular velocity increased, resulting in shorter closing durations and less percent muscle contribution to close the lower jaw, resulting in the A3 muscle assuming the dominant role in lower jaw closing.

Trends in changes in muscle function were observed over ontogeny. To examine if these trends indeed led to an adult feeding mode, adult red snapper(N=3) were also modeled. These results from adults were only used to make general qualitative comparisons of the potential dynamic actions of the A2 and A3 muscles during lower jaw closing. They were not used for any statistical comparisons. Overall, the A3 muscle was larger than the A2 in adults, both in length and cross-sectional area(Table 3) according to the model. The A3 muscle contributed more force to overall bite force, had higher effective mechanical advantage, and thus a lower velocity ratio than the A2 muscle (Table 3). Total duration of the A3 muscle in lower jaw closing was shorter than the A2 muscle(Table 3). Since both muscles rotated through the same gape, the angular velocity of the A3 muscle was higher and the percent muscle contraction required to close the lower jaw was smaller than the A2 muscle (Table 3). The trends observed in large juvenile A2 and A3 muscle function are consistent with data from adult model simulations, suggesting that when juveniles reach approximately 6 cm in length they switch to their adult feeding mechanism.

Model simulations were carried out in which a habitat effect was tested and size classes were pooled to determine if A2 and A3 muscle function differed between habitats. The cross-sectional areas of the A2 and A3 muscles in juveniles did not differ significantly between habitats (A2 P=0.32,A3 P=0.36, MANCOVA; Table 2). Muscle length differed between habitats: the A2 muscle was 31.1% longer in off-ridge juveniles (1.03±0.06 cm) and the A3 muscle was 10.7% longer in on-ridge juveniles (1.21±0.04 cm; Table 2). Effective mechanical advantage (EMA) was significantly less (6.3–20.8%; A2 P<0.001, A3 P=0.002, MANCOVA), and the velocity ratio was significantly greater (3.3–20%; A2 P<0.001, A3 P=0.02, MANCOVA) in on-ridge juveniles compared with off-ridge juveniles. The A2 muscle exhibited lower EMA and greater velocity than the A3 muscle for both habitats (Fig. 3G,H). The A3 muscle force contribution to bite force(P=0.86, MANCOVA) and the total bite force (P=0.59, MANCOVA)was not significant between habitats. The A2 muscle force contribution to bite force was significantly greater (1.1×) on-ridge (P=0.01,MANCOVA). Off-ridge juveniles exhibited 12.2% smaller gapes (A2 P=0.17, A3 P=0.09, MANCOVA), closing durations of 5.8–20.1%, which were significantly longer (A2 P<0.001, A3 P=0.04, MANCOVA), 15–30.4% slower angular velocities (A2 P<0.001, A3 P=0.03, MANCOVA), and 5.7–20% greater percent muscle contractions (A2 P<0.001, A3 P=0.04,MANCOVA) required to close the lower jaw compared to on-ridge juveniles for both muscles (Fig. 3I–L). The closing duration and percent muscle contraction of the A2 muscle was shorter in on-ridge juveniles (79.49±1.67 ms and 18.82±0.39 ms,respectively) and larger in off-ridge juveniles (99.4±1.57 ms and 23.5±0.37 ms, respectively) compared with the A3 muscle(Fig. 3J,L). The angular velocity of the A2 muscle was greater than the A3 muscle in on-ridge juveniles(4.25±0.01 deg. ms^–1 and 4.07±0.01 deg. ms^–1, respectively) and smaller than the A3 muscle in off-ridge juveniles (2.96±0.01 deg. ms^–1 and 3.46±0.01 deg. ms^–1, respectively; Fig. 3K). From these results,the A2 muscle appears to be the dominant muscle in on-ridge juveniles for fast closing, whereas the A3 muscle appears to be the dominant muscle for fast closing in off-ridge juveniles.

A comparison of model output to kinematic results tested the predictive accuracy of the jaw lever model. The overall relationship between jaw closing duration and gape displacement (Fig. 4A), as well as a plot of the average model with several representative individuals (Fig. 4B), show a close agreement between model and live kinematic data. Model and video data were not significantly different for slope(P=0.21, ANCOVA) or intercept (P=0.08, ANCOVA). Similar plots of jaw closing duration against gape angle(Fig. 4C,D) show that the rate of angle change was greater in the model simulations than in living kinematics, although, similar jaw closing curve shape and gape angle values were observed, particularly when several representative kinematic plots were compared to the model (Fig. 4D). For gape angle, model and video data were not significantly different for slope (P=0.81, ANCOVA) but had significantly different intercepts (P=0.02, ANCOVA). Overall, the lever model was an accurate predictor of gape displacement and gape angle in juvenile red snapper, with expected higher variability in the living fish kinematics.

On-ridge juveniles (92%) would approach and engulf prey items from a distance using a single explosive jaw movement. Jaw protrusion and hyoid depression began after mouth opening and reached their maxima after maximum gape was achieved (Fig. 5A). Maxillary rotation and cranial rotation began to increase at the beginning of the feeding event and reached their maxima after maximum gape was achieved(Table 4). The hyoid and jaw tips returned to their original positions after the mouth had closed(Fig. 5A).

Off-ridge juveniles (37%) generally captured the prey item and momentarily held it between the jaws, resulting in a prey transport cycle to move the prey to the pharyngeal jaws. The hyoid depression began increasing at approximately the same time as the initial gape displacement and increased to its maximum,which occurred after the prey transport gape maximum(Fig. 5B). After the initial gape, jaw protrusion decreased and then increased again during the prey transport gape, reaching a second maximum along with maximum hyoid depression(Fig. 5B). After the prey transport was achieved, the mouth closed and the hyoid, jaw tips, cranium and maxillary returned to their starting positions simultaneously(Fig. 5B).

On-ridge juvenile red snapper expressed larger and faster jaw movement compared with off-ridge juveniles. Maximum displacement variables were significantly greater in on-ridge juveniles (all P<0.001, MANOVA). Maximum angular variables were greater in on-ridge juveniles, significantly for maximum cranial rotation (P=0.02, MANOVA) and maximum maxillary rotation (P=0.02, MANOVA). Time to maximum displacement and angular variables did not differ significantly between habitats (all P>0.05, MANOVA). Maximum angular velocities were faster in on-ridge juveniles for all angles, significantly for maximum lower jaw rotation velocity (P=0.03, MANOVA). Prey capture time was shorter in on-ridge juveniles (P=0.86, MANOVA).

Shape analysis further supported morphological and behavioral differences in juvenile red snapper throughout ontogeny and between habitats. Lateral body morphology of juvenile red snapper significantly differed across size(P<0.001, MANOVA) and also differed between the two habitats(P=0.01, MANOVA). Habitat effect accounted for 1.6% of the total morphological variation and size effect accounted for 9.1%. The effect of habitat was small in magnitude, but high in significance. Thin plate spline transformation grids illustrate changes along the shape axis(Fig. 6). The habitat effect axis indicated that off-ridge juveniles had a deeper head and body than on-ridge juveniles (Fig. 6). The size effect axis indicated that as juvenile body size increased the head and body became deeper.

This study provides data suggesting a possible mechanism for a transition from a juvenile to an adult feeding mode. Modeling data suggests that the potential function of the A2 and A3 adductor muscles changed ontogenetically. In small juveniles, the A2 muscle dominated lower jaw movement, whereas the A3 dominated in large juveniles. The switch from A2 muscle dominance to A3 muscle dominance at 6 cm SL in juvenile red snapper marks the transition to an adult feeding mode. This ontogenetic change of the A2 and A3 muscle function is supported by the fact that the A3 muscle in adult red snapper is also the dominant lower jaw closing muscle. The A2 and A3 muscle function in large red snapper juveniles was similar to data from a benthic foraging wrasse(Cheilinus trilobatus) (Westneat,2003). The jaw lever model data also suggests that ontogenetic changes in morphology had an impact on feeding kinematics of juvenile red snapper and possibly diet. Lever ratio data demonstrated that, over ontogeny,the lower jaw movement became slower and more forceful as size increased. The lever ratio values of small and large juvenile red snapper are consistent with values for suction feeders and biters, respectively. Measurements of lever ratios have successfully predicted a diet of small, soft prey items and a diet of larger, harder prey items, respectively, in other teleosts(Barel, 1983; Westneat, 1994; Westneat, 2004; Wainwright and Richard, 1995). Therefore, this study strongly suggests that an ontogenetic shift in morphology occurred that enabled large red snapper juveniles to exploit harder prey types. Adult red snapper lever ratios demonstrated that the slow but forceful lower jaw movement is indeed the mature feeding mode, which correlates well with hard prey items of the adult diet (i.e. crabs and mollusks) (Bradley and Bryan,1976; Moran, 1988; Ouzts and Szedlmayer, 2003). Dietary data from studies of juvenile red snapper(Bradley and Bryan, 1976; Szedlmayer and Lee, 2004) and other lutjanids (Rooker, 1995)suggest a shift towards the adult feeding mode that corresponds with the morphological and biomechanical switch observed in this study in juveniles at 6 cm SL; it is known that a change in diet of lutjanids initiates movement to deeper water (Rooker, 1995; Cocheret de la Morinière et al.,2003; Szedlmayer and Lee,2004). Over ontogeny juveniles developed the morphological capability to consume a broader range of prey, which would allow juveniles to begin to occupy a wider ecological niche(King, 1971; Liem, 1980; Luczkovich et al., 1995),become more opportunistic feeders and therefore could effectively move into adult populations in deeper waters.

Juvenile red snapper from different habitats (on-ridge vsoff-ridge) exhibited differences in their feeding mechanics, feeding behavior,as well as their head and body morphology. The lower jaw lever model of juvenile red snapper from the two habitats demonstrated significant differences in their potential feeding capabilities. The mechanical advantage of the lower jaw from on-ridge juveniles was similar to values from suction feeders, whereas mechanical advantage of the lower jaw from off-ridge juveniles was similar to values from biters(Wainwright and Richard, 1995; Westneat, 2004). Therefore,the lower jaw lever model prediction that off-ridge juveniles have an increased capability to crush harder prey types, or bite off pieces of larger prey, whereas on-ridge juveniles have an increased suction capability to capture small, soft prey appears to be supported. In addition, results from feeding kinematic trials of juvenile red snapper also supported predictions of the jaw lever model. Actual feeding events of juvenile red snapper demonstrated that on-ridge juveniles expressed kinematic profiles typical of suction feeders (e.g. Liem,1980; Van Leeuwen and Muller,1984; Svanbäck et al.,2002), whereas off-ridge juveniles exhibited a more manipulative,biting behavior. Furthermore, in captivity, off-ridge juveniles were observed actively biting the prey given to them, as well as each other. Captive off-ridge juveniles would approach other juveniles and bite them to remove large pieces of flesh, or completely bite them in half. Anecdotally, we observed that fish prey identified in off-ridge juvenile stomach contents were large pieces of fishes, not whole fishes. This is consistent with the behavior observed in captivity. Captive on-ridge juveniles were observed using suction to capture prey given, and were rarely seen biting each other.

Juvenile red snapper also exhibited phenotypic plasticity in response to differences between habitats. Shape analysis has been used previously to demonstrate the expression of phenotypic plasticity of an organism induced by varying environmental factors (Robinson and Wilson, 1996; Robinson et al., 1996; Svanbäck and Eklöv, 2002; Svanbäck and Eklöv,2003; Doughty and Reznick,2004; Parsons and Robinson,2007). Habitat has been shown to generate resource polymorphism in fish. For example, where fish of the same species occupy different habitats some species developed different body shapes and the ability to consume different prey types (Lavin and McPhail,1986; Ehlinger and Wilson,1988; Malmquist,1992; Robinson et al.,1996; Svanbäck and Eklöv, 2002). In this study, off-ridge juvenile red snapper had deeper heads, whereas on-ridge juveniles had more streamlined heads. Studies of polymorphism suggest that streamlined bodies are associated with midwater feeders and are optimal for high velocity prey capture of elusive prey. By contrast, deeper bodies are associated with low velocity and high maneuverability, and this is optimal for benthic foragers that feed on hard prey (Ehlinger and Wilson,1988; Malmquist,1992; Motta et al.,1995; Robinson and Wilson,1996; Robinson et al.,1996; Walker,1997; Hjelm et al.,2003; Svanbäck and Eklöv, 2003). Controlled experiments in which prey items were manipulated demonstrated a morphological difference in head shape, streamlined versus deep, when fish were fed small, soft prey versusharder prey types, respectively (Meyer,1987; Wimberger,1991; Wimberger,1992; Hegrenes,2001; Parsons and Robinson,2007). Among the selected landmarks from this study, the branchiostegal ray point (15) moved anteriorly, producing a more streamlined head in on-ridge juveniles. Previous studies showed that this point moves in the same direction in red drum (Scianenops ocellatus) fed soft prey items (Ruehl and DeWitt,2007). Therefore, differences in juvenile red snapper head and body shape further supports the hypothesis that on-ridge juveniles consume softer prey items, or smaller items that are ingested whole, most probably using suction. Concomitantly, shape analysis of off-ridge juvenile snapper supports the hypothesis that deeper bodied off-ridge juveniles consume harder prey types, or pieces of larger prey, using biting. These differences in body shape between habitats also suggest that off-ridge juveniles have reached a more developed ontogenetic state faster, since they have taken the shape of larger juveniles. On-ridge juveniles were slower to reach the near-adult ontogenetic stage; similar patterns have been observed in sharpsnout seabream(Diplodus puntazzo) (Kouttouki et al., 2006).

By integrating morphological modeling, kinematic behavioral performance testing and shape analysis, this study has provided new insight into data on the biomechanical development of juvenile red snapper during early life history stages, and its possible trophic consequences. It is probable that developmental modifications in feeding ability of juvenile red snapper,primarily in the ontogenetic changes in A2 and A3 muscle function, resulted in size-related diet shifts, and the capability to consume harder prey types that are more typical of adult diets. The pivotal size for juvenile red snapper development appears to be at 6–7 cm. At this size the hyoid and mandibular arches have fully ossified(Potthoff et al., 1988) and juveniles have attained the morphological capability to fill a wider ecological niche. This allows juveniles to successfully begin competing with larger juveniles and adults, and they can therefore effectively move into the adult population. However, habitat may influence the transition between these ontogenetic stages. Off-ridge juveniles in this study possessed the morphological capability to consume harder prey types, but more importantly fish, earlier than on-ridge juveniles at the same body size. By developing a stronger bite, off-ridge juveniles may compensate for any gape limitations by biting pieces of prey larger than their mouth. A fish diet is high in caloric value, so an earlier switch to piscivory promotes faster growth and survival(Persson and Brönmark,2002; Post, 2003; Graeb et al., 2005), which may explain the higher growth rates of juvenile red snapper reported in off-ridge areas (Rooker et al., 2004; Geary et al., 2007). In general, faster growing fish resulting from an early switch to piscivory represent the population majority within the cohort, and therefore contribute more individuals to the adult population(Olson, 1996; Ludsin and DeVries, 1997; Persson and Brönmark,2002).

The authors gratefully thank Drs Thomas DeWitt and Jay Rooker for their contributions throughout the study. We thank Jay Rooker for assistance in collecting juvenile red snapper. We also thank Joe Mikulas and Ryan Schloesser for their help in the collection and maintenance of juvenile fish. This work was supported by the Texas A&M at Galveston Department of Marine Biology and the Luke and Erma Lee Mooney Student Travel Grant. The development of MandibLever software was supported by NSF grant 0235307.