Background

Sound producing spiny lobsters (Stridentes: George and Main 1967; Parker 1883) have a large sound producing apparatus located on the antennular plate, which averages 28% carapace width and 19% carapace length and, in some species of Panulirus, reaches 38% carapace width and 27% carapace length (Figs. 1, 2). In contrast, non-sound-producing genera (Silentes: George and Main 1967; Parker 1883) lack a sound producing apparatus, but still have an antennular plate, which averages only 7% carapace width and 4% carapace length (Figs. 1, 2). Stridentes make a “rasp” sound by rubbing an extension off the base of each antenna (the plectrum) over the files located on the antennular plate below the eyes (Fig. 1) (Moulton 1957; Meyer-Rochow and Penrose 1976; Mulligan and Fischer 1977; Patek 2001a, 2002). The plectrum consists of a series of soft tissue ridges that rub over the macroscopically smooth surface of the file; “stick and slip” friction between these two surfaces results in a series of broad frequency bandwidth sound pulses called the rasp (Patek 2001b). In contrast, the Silentes have, in place of the plectrum, a joint articulation that restricts the joint to rotations with one degree of freedom (Patek 2002).

The Stridentes produce sound during interactions with predators (Lindberg 1955; Smale 1974; Meyer-Rochow and Penrose 1976; Mulligan and Fischer 1977) which most likely increases their chances of escape by causing the predator to pause momentarily (Alexander 1958; Masters 1979, 1980; Lewis and Cane 1990; Sargent 1990). Defensive signals capitalize on unexpectedness or novelty to deter predators (Edmunds 1974; Jetz et al. 2001). Predators more effectively learn to avoid dangerous prey when coupled with a conspicuous signal (Gittleman and Harvey 1980; Gamberale and Tullberg 1996; Speed 2000), and conspicuous signals can act as reliable indicators of defended prey (Sherratt 2002).

Antennae are spiny lobsters' primary weapons against predators (Lindberg 1955; Kanciruk 1980; Cobb 1981; Spanier and Zimmer-Faust 1988; Kelly et al. 1999). Spiny lobsters forcefully grasp and rake predators with the basal antennal segments; the antennae are sharp and covered with spines. Both Stridentes and Silentes have been observed in pod formations and spread their antennae as a “pincushion” of spines for protection against predators (Kanciruk 1980; Kelly et al. 1999). When in their burrows, spiny lobsters extend their antennae out of the burrows and, when walking in the open, they hold the antennae over the carapace and abdomen (Cobb 1981; Spanier and Zimmer-Faust 1988; Kelly et al. 1999).

The size of the antennae and sound producing apparatus are especially relevant in their respective roles in predator deterrence. An increase in size of the sound producing apparatus should permit production of louder, higher pulse number, or higher pulse rate signals that are more likely to startle a predator, much in the same way that bright colors, distinct color patterns, and noxious chemicals provide jarring and memorable signals to potential predators (Guilford 1990; Jetz et al. 2001; Sherratt 2002). Larger antennae are stronger and have the capacity to produce more force without structural failure (Wainwright et al. 1976; Alexander 1983). Furthermore, larger appendages allow space for greater muscle cross-sectional area which contributes to a greater capacity for generating force (although see Schenk and Wainwright 2001; Taylor 2001 for discussion of these size-muscle relationships). Hence, in both systems, increases in size are likely to increase predator deterrent performance.

A Priori Trade-Offs Hypotheses

With the extreme size range of antennular plate, up to 38% carapace width in Stridentes compared to 4% carapace width in Silentes (Fig. 1, 2), we developed the a priori hypothesis that the adjacent antennae would decrease in size to accommodate a larger antennular plate, rather than the alternative hypothesis that the overall body size would increase to accommodate a larger communication structure. We conducted three major phylogenetic comparative tests. First, we examined whether the size of the signal producing apparatus affected signal features, a necessary prerequisite for demonstrating the acoustic function of evolutionary size variation. We tested, both within and across species, whether a longer file resulted in longer signal duration, higher pulse number, or higher pulse rate. We found that file length was associated with pulse number and pulse rate, but not with rasp duration. Second, we tested the size trade-off hypothesis that the antennae would decrease in size with increasing antennular plate size, given the assumption that body size imposed an upper limit to evolutionary size variation (Table 1). The results did not suggest a trade-off, and instead showed the opposite—the sizes of both structures were positively correlated within the Stridentes and the assumption that body space imposed an upper limit to size increases was not supported. Third, we examined whether a gradeshift in antennal size occurred between Stridentes and Silentes due to the presence of the large sound producing apparatus. This gradeshift hypothesis was generally supported: when the effects of body size were removed, Silentes had larger antennae for a given antennular plate size than Stridentes.

Palinurid Phylogeny

To conduct a phylogenetic comparative study, it was first necessary to generate phylogenetic hypotheses for the relationships of palinurid taxa. Palinurid taxonomy is well described, with some contention as to the species level designations (e.g., Sarver et al. 1998). However, the genus and species-level evolutionary relationships remain mostly unresolved. Noncladistic methodology previously was used in dividing extant and fossil palinurid genera into the Silentes, Jasus and Projasus, and the Stridentes, Linuparus, Puerulus, Palinustus, Justitia, Palinurus, and Panulirus (Parker 1883; George and Main 1967). Baisre (1994) used larval and adult characters from representatives of each palinurid genus and constructed a phenogram using cluster analyses; he generally found support for relationships suggested in George and Main (1967). Monophyletic genera are consistently supported (George and Main 1967; Holthuis 1991) with the one exception of whether Jasus verreauxi belongs in the genus Jasus (Brasher et al. 1992; Ovenden et al. 1997). McWilliam (1995) examined relationships within the most speciose palinurid genus, Panulirus, and added larval characters to the adult characters used in George and Main (1967). Ptacek et al. (2001) used 16S rDNA and cytochrome oxidase subunit I (COI) to examine groupings of Panulirus species. These studies generally support the scheme suggested by George and Main (1967), although lacking strong resolution for the suggested subgroupings within Panulirus. A new palinurid genus and species, Palibythus magnificus, recently was discovered which exhibits a fully developed sound producing apparatus closely resembling those of the Stridentes (Davie 1990).

Trees Used in Phylogenetic Comparative Analyses

Phylogenetic comparative analyses were based on trees obtained from 18S, 16S, and morphological data sets (28S results provided minimal resolution; see Results). Because phylogenetic resolution and available taxa varied among data partitions, we performed comparative analyses on phylogenies from different datasets in order to find correlational patterns that were common to all analyses. First we used the most parsimonious tree based on morphological characters. Second, for the 18S and 16S molecular datasets to be used in phylogenetic comparative analyses, we used the posterior probability distribution of trees estimated by Bayesian methods. A Bayesian approach more effectively represents the range of likely tree topologies and branch lengths than the single most parsimonious or maximum-likelihood tree calculations (Huelsenbeck et al. 2000a,b; Lewis 2001). Using MrBayes (Huelsenbeck and Hall 2000; Huelsenbeck and Ronquist 2001), we ran 10,000 replicates with four chains, sampled every 10 replicates, and printed every 100 replicates; the last fifty trees from each analysis were used for calculations of independent contrasts (see below).

Sound Acquisition and Analysis

Rasp sounds were recorded from eight species representing five palinurid genera. Lobsters were handheld in most experiments since rasp sounds are naturally elicited during interactions with potential predators. Palinurus elephas specimens were collected by Cleggan Lobster Fisheries, Ltd. (Galway, Ireland). Palinustus waguensis and Panulirus japonicus specimens were collected by the Seto Marine Biological Laboratories staff and local fishermen in Shirahama, Japan. Justitia japonica, Linuparus trigonus, Panulirus homarus, and Panulirus longipes were recorded at the Okinawa Expo Aquarium, Okinawa, Japan. Panulirus argus specimens were collected using FL DEP permit 99S–428 at Long Key, Florida. Specimens from Ireland and Japan were purchased from fishermen or shared from other researchers and did not require the use of a permit.

A hydrophone (20 Hz–15,000 Hz; HTI-94-SSQ, High Tech, Inc., Gulfport, MS) was placed 0.3 m deep and at least 0.15 m from the lobster when possible (during international travel, some tanks were too small to place the hydrophone at this depth). Rasps were recorded on audio tape (Sony TCD5M audio cassette recorder). We first filtered the signal with a Krohn Hite (Brocton, MA) model 3500 band pass filter (20 Hz and 15000 Hz), then with a low pass 15,000 Hz filter (Stanford Research Systems, Sunnyvale, CA, Model S640). We digitally sampled the filtered signals at 37383.2 points/sec and analyzed them using Signal 3.0 software (Beeman 1992).

We measured the number of pulses, duration, and the pulse rate (number of pulses per second) for each rasp (Fig. 3). The rasps were visually represented using a Kay Elemetric Digital Sona Graph, 20–15,000 Hz frequency range (band pass filtered using Krohn Hite model 3550 filter), and 300 Hz frequency resolution. We analyzed at least ten rasps per individual.

Phylogenetic Comparative Tests of Trade-Offs Hypothesis

The trade-offs hypothesis was evaluated by examining correlated evolutionary change between antennal morphology, signal producing morphology, and signal features (Table 1). First, we tested whether mean and maximum values of rasp duration, pulse number, and pulse rate (as measured above) were positively correlated with file length. Second, we tested whether file length was positively correlated with antennular plate size (the structure on which the file is located; Fig. 1) so that, third, we could test whether there was a negative correlation between the size of the sound producing apparatus (as represented by the antennular plate size) and size of the surrounding antennae. Finally, we tested whether there was a gradeshift of antennal bending strength for a given antennular plate size, between Stridentes and Silentes.

In 41 species, male and female when possible, we measured the dimensions of the file, antennular plate and antennae (Fig. 1; Appendix 1). Male and female values were not significantly different (Patek 2001a), thus species values were averaged across males and females. For each individual, we measured body size as the carapace length (base of rostrum to posterior margin of cephalothorax). Morphometric data were log transformed prior to analysis to remove the dependence of variance on the mean values (Sokal and Rohlf 1981). Antennular plate length was measured from posterior to anterior along the midline. The width of the antennular plate was measured at the proximal most edge of the antennular plate. Antennal width was measured from the lateral joint articulation to the medial junction of the antenna where it meets the cephalothorax (in Silentes, the medial joint articulation). Antennular plate length and width were significantly positively correlated (see Results), so antennular plate length was used as a measure of antennular plate size in subsequent analyses because it had a higher level of measurement accuracy than antennular plate width.

The capacity for antennae to generate force hinges upon their ability to resist failure, specifically failure while bending (buckling). Strength in bending thus provides a measure of the capacity to generate force across species. Strength in bending in a thin walled cylinder is proportional to: I/R = (π/4)*(R⁴ − r⁴)/R, where I is the second moment of area, R is the outer radius of the cylinder, and r is the inner radius of the cylinder (Currey 1970; Wainwright et al. 1976; Alexander 1983) (Fig. 3). We measured R as the radius of the first segment of the antennae. We calculated r as R minus the cuticle thickness. Because most of the 41 species were museum specimens, it was not possible to cut the antennae to measure cuticle thickness. Using a personal collection, we were able to measure cuticle thickness of six species from both Silentes and Stridentes genera. We measured cuticle thickness in the dorsal and ventral surfaces of the antennae and used an average of these values for calculation of r across all species. Silentes cuticle thickness averaged 1.5% of carapace length whereas Stridentes averaged 1% carapace length. If systematic biases resulted from this estimation, strength in bending would be slightly underestimated in Silentes. Thus, our comparative tests would err in the direction against our predicted trade-offs that Silentes have stronger antennae. This calculation of strength in bending assumes that the materials are isotropic and the structure is perfectly cylindrical. Because we used this calculation for comparison across species, and actual values are not being reported, the violations of these assumptions should not systematically alter the observed patterns. We conducted additional comparative analyses using antennal diameter in place of I/R to verify that statistical significance was not unduly influenced by errors in cuticle thickness estimates.

We corrected for phylogenetic nonindependence of species data points by calculating independent contrasts (Felsenstein 1985b) using Compare version 4.4 (Martins 2001). Branch lengths only were calculated for the molecular datasets; branch lengths were set equal to one for morphological datasets, and, as recommended by the program documentation, branch lengths at unresolved nodes were set equal to a small value. Compare software does not allow missing data, hence each dataset used a different subset of taxa due to missing data or taxa in some datasets. We accounted for body size by inputting the morphological variables and body size measures into Compare software to calculate the regression line. We then calculated the residuals for each variable in contrasts space and used those residuals as the body size corrected values for further independent contrasts tests (Garland et al. 1992). Rasp signal features were not always significantly correlated with body size (Patek 2001a), thus these values were not corrected for body size. We calculated average Pearson product moment correlation coefficients for contrast values calculated across 50 Bayesian trees in the molecular datasets and the most parsimonious trees in the morphological phylogeny. We used two tailed tests and set the significance level at P < 0.05.

To test whether Silentes and Stridentes have different antennal bending strength for a given antennular plate size, we used a modification of a nonparametric rank test suggested in Garland et al. (1993). We regressed size-corrected contrasts of antennal bending strength on antennular plate size (also size-corrected with independent contrasts) and calculated the residuals from the regression line. For rank tests, the absolute values of the residuals were used. We identified the independent contrast calculated between the Stridentes clade and Silentes clade. The probability of the ranking of this contrast value was calculated as the rank score divided by the number of residuals. The test was conducted on the 12 most parsimonious trees of the morphological dataset and ten trees of the 16S Bayesian dataset (chosen arbitrarily as the last ten trees generated by MrBayes). The 18S dataset was too small to provide statistical power for this test.

Palinurid Phylogeny

A heuristic search of the 79 morphological characters yielded twelve most parsimonious trees (tree length: 224; consistency index: 0.4911; retention index: 0.8014) (Fig. 4A). These and subsequent tree descriptions exclude uninformative characters.

We obtained 21 new complete 18S rDNA sequences, 11 new complete 16S rDNA sequences, and 44 new complete and partial 28S rDNA sequences (GenBank Accession numbers are listed in Appendix 1). Technical difficulties precluded obtaining all gene regions for all species. Also listed are sequences that were used in these analyses but which were obtained from GenBank.

The 18S dataset consisted of an edited, unambiguous alignment of 1852 base pairs and yielded 297 most parsimonious trees (informative characters: 109; tree length: 282; CI: 0.7374; RI: 0.8120) (Fig. 4B). For 16S tree searches, we assumed the best-fit model and parameter values as: Tamura-Nei (1993) model; NST = 6; Gamma shape = 0.9171; proportion of invariant sites = 0.6793.

16S rDNA analyses resulted in 57 most parsimonious trees (126 informative characters: tree length, 554; CI: 0.415; RI: 0.682) based on an edited, unambiguous alignment of 407 base pairs (Fig. 4C). For 16S tree searches, we assumed the best-fit model and parameter values as: HKY model; NST = 2; Gamma shape = 0.5547; proportion of invariant sites = 0.4717.

28S rDNA is a very large region so we sequenced four smaller areas within the 28S gene (ddff, ee, mm, and vx) and concatenated these regions into a single analysis. Three most parsimonious trees were found with an alignment of 2274 characters (139 informative characters; tree length: 725; CI: 6330; RI: 0.6022) (Fig. 4D). For 28S tree searches, we assumed the best-fit model and parameter values as: Tamura-Nei + Invariant sites + Gamma rate heterogeneity; NST = 6; Gamma shape = 0.6441; proportion of invariant sites = 0.4325.

Bootstrap values and Bayesian consensus values were consistent across the four analyses in supporting a Jasus + Projasus clade, the monophyly of the Panulirus genus (with the exception of the 18S dataset in which P. homarus, inflatus, and versicolor fell outside the genus) and the Palinurus genus. In the three maximum-likelihood tree topologies (Fig. 4B–D), the Silentes occur within the Stridentes. The morphological phylogeny (Fig. 4A) suggests a monophyletic clade of Stridentes. Bootstrap and Bayesian values do not preferentially support either of these topologies. Support is found for inclusion of Palinurellus in the family Palinuridae and, based on the 16S and 18S datasets, may form a clade with Jasus and Projasus species. Palibythus magnificus falls within the Palinuridae in all datasets.

Phylogenetic Comparative Tests of Correlated Signal and Morphological Change

First, we tested whether file length was correlated with signal features. The file/antennular plate complex noticeably varies across the palinurids, ranging from no file in Silentes to an elongate file in the Panulirus genus (Figs. 1, 2). Rasp duration, number of pulses, and pulse rate (pulses/second) were documented across eight palinurid species (Table 2). Average pulse number was positively correlated with file length across a size range of Panulirus argus individuals (Table 3, Fig. 5). Across palinurids in the morphological and 16S datasets, file length and average rasp duration were negatively correlated; pulse rate was positively correlated with file length; maximum number of pulses and file length were positively correlated (Table 3 and Fig. 5). No significant relationships were found when using the 18S dataset, possibly caused by extremely long branch lengths of Panulirus homarus that forced this species outside of the Panulirus clade for this particularly small dataset (seven taxa) (Fig. 4B).

File length was tightly correlated with antennular plate size, hence antennular plate size was used as a measure of the size of the sound producing mechanism (Table 4; Fig. 6). Antennular plate length and width were positively correlated (Table 4). Finally, we tested whether the size of signal producing morphology (antennular plate length) correlated negatively with bending strength of the surrounding antennae, as predicted by the trade-offs hypothesis. Antennular plate length and bending strength (I/R) were positively correlated in one of three datasets across all palinurids (Table 4; Fig. 6) and positively correlated in all datasets across Stridentes (Table 4; Fig. 6). Antennular plate length and antennal width followed the same pattern as those found for I/R (Table 4).

The comparison of antennular bending strength for a given antennular plate size between Stridentes and Silentes (both variables were corrected for body size using the independent contrasts method as discussed above) (Fig. 7), yielded significant differences in all 16S trees (10 trees; 28 residuals; average P = 0.036) and nonsignificant results in the morphological trees (12 trees; 40 residuals; average P = 0.11). In the 16S trees, the contrast value between Stridentes and Silentes always ranked first. The contrast value ranked third in ten of the morphological trees; the contrast in the remaining two trees ranked fourth and twenty-first.

Signal Feature and File Length Variation

Both between species and within Panulirus argus, the number and rate of pulses increased with file length (Fig. 5; Table 3), and hence increased with the overall size of the antennular plate (Fig. 6; Table 4). Intuitively, as the file length increased, the plectrum could move a greater distance, and hence a higher number of pulses could be produced. In contrast, a somewhat counterintuitive result was the weaker yet statistically significant finding that rasp duration decreased with increasing file length. These results were unexpected given that rasp duration equals pulse number divided by pulse rate. As such, in order for rasp duration to decrease, one would expect that pulse number would decrease and/or pulse rate increase. We offer two possible interpretations of these results.

The first interpretation is that there were behavioral differences in the use of the sound producing apparatus. Specifically, plectra (Fig. 1) could be moved simultaneously to maximize the pulse rate with an overall decrease in rasp duration. Alternatively, plectrums could be moved sequentially to maximize rasp duration with an overall decrease in pulse rate. Because these two strategies were not distinguished in the analyses, our findings that file length was correlated with decrease in average rasp duration and increase in average pulse rate suggest that the strategy of simultaneous plectrum movement was used more often than the sequential activation. In either strategy, the maximum number of pulses should increase with file length, which was supported in our results. Interestingly, these two types of behavioral strategies have been observed in pulsed frog sounds (Cocroft and Ryan 1995).

A second possibility is that the properties of frictional stick and slip sound production may have varied independently from file length across species due to frictional load, elasticity of component structures, and surface asperities (Patek 2002; Persson 2000; Rabinowicz 1995). These physical parameters determine the critical load and velocity for alternating sliding and sticking friction, hence affecting the pulse rate, pulse duration, and slip duration.

Size Trade-Offs

Counter to our a priori size trade-off hypothesis, both antennae and antennular plate size were positively correlated in Stridentes. The nearly twofold difference in antennular plate size within the Stridentes and no proportional decrease in size of the surrounding antennae suggested that body space did not impose a limit on the combined size of these structures. However, when Stridentes and Silentes were compared, there was general support for a gradeshift in antennal size relative to antennular plate size (Fig. 7). Because the comparison between Stridentes and Silentes involved an even greater, fivefold size differential in antennular plate size, space limits may have been more apparent at this level of analysis.

These kinds of size trade-offs can be interpreted at the level of available body space (Barel et al. 1989; Smits et al. 1996) as well as at the level of the costs and resources required to build the structures during development (Klingenberg and Nijhout 1998; Nijhout and Emlen 1998). The extreme range of antennular plate size in spiny lobsters also suggested a system in which construction costs during development should be considered, particularly the resource intensive costs in constructing the sound producing apparatus, including the file, plectrum, and additional musculature (Patek 2002).

Indeed, the life history of spiny lobsters suggests that constructional costs and resource limitations during development could limit size increases and cause the gradeshift observed between Stridentes and Silentes. For resource competition between body parts to occur, resources must be limited in the time period when the structure develops (Emlen 2001). In spiny lobsters, the apparatus is first visible, but not functional, in the late phyllosoma stage when the larvae are still planktonic (Meyer-Rochow and Penrose 1974). The phyllosoma larvae molt into puerulus larvae and swim toward the continental slope. Notably, resources are limited because the pueruli are not known to feed during this time period and probably rely on stored nutrients (Booth and Phillips 1994; Meyer-Rochow and Penrose 1974). The larvae eventually settle on the substrate and molt into postpuerulus larvae, at which time the file grows in length, file surface features (shingles) increase in number, and the plectrum becomes fully differentiated (Meyer-Rochow and Penrose 1974). Meyer-Rochow and Penrose (1974) recorded sounds produced by the apparatus in postpuerulus larvae (incidental sounds were recorded in the pueruli). During these resource-limited time periods, the larger antennular plate size observed in Stridentes could impose resource costs not present in Silentes, thus causing a size trade-off with the surrounding antennae.

As an alternative explanation to resource-based size trade-offs, the observed gradeshift may be due biomechanical and physiological limits imposed by muscle attachment area and construction of the sound producing antennal joint. The antennal muscles attach over a substantial surface area of the dorsal and ventral cephalothorax (Patek 2002; Paterson 1968). Although greater antennal size provides space for larger muscles, these larger muscles must have sufficient space for attachment in the cephalothorax. Furthermore, as mentioned in the introduction (Fig. 1), Stridentes have lost their medial joint articulation such that the plectrum (a modification of the joint articulation) can translate across the file to make sound (Patek 2002). As a result, the Stridentes have a sliding joint articulation with multiple degrees of freedom, whereas the first antennal joint in Silentes is limited to one degree of freedom of movement. Increases in antennal diameter may be more effective for increasing strength and force in Silentes, but at the cost of manipulability for Stridentes. The differences in joint mechanics could lead to decreased stability of the antennae during force production in Stridentes, thus limiting them to smaller antennae.

Ecological factors may also have played a role in patterns of antennal strength across these palinurid taxa; specifically, higher predation pressures in two palinurid genera may provide an alternative explanation to the observed gradeshift. Theories of escalation propose that predator defenses increase in taxa that move to warmer, more productive waters (Vermeij 1987). Response to predation includes a suite of predator defenses, including increased exoskeletal strength, and targeted defenses, such as toxicity. In Figure 7D, the Stridentes and Silentes have distinctly separate distributions of strength in bending relative to antennular plate length. However, Silentes and Panulirus species have overlapping areas of strength in bending (Fig. 7A, D). Interestingly, two separate invasions of relatively predator-rich habitats indicate that predation has been an important selective force in palinurid evolution (Brasher et al. 1992; George 1997; George and Main 1967; McWilliam 1995; Ovenden et al. 1997; Pollock 1990, 1992, 1993; Ptacek et al. 2001). First, the Stridentes genus Panulirus invaded shallow, tropical waters (George 1997; George and Main 1967). The movement of taxa toward warmer waters and complex reef systems was likely accompanied by increased predation pressures and hence increased selection on predator defenses (Vermeij 1987; Pollock 1995). Secondly, major shifts in Jasus (Silentes) distribution may also have led to increased predation pressure and some members of Jasus and Panulirus now inhabit similar ecological habitats (George 1997; Pollock 1995). The similar distribution of strength in bending between Panulirus and Jasus (Fig. 7D) suggests responses to predation by increasing the strength and potential force production of their antennal weapons. With a suitable dataset on the ecology and predation pressures across lobster species, phylogenetic comparative analyses could be used to test whether a statistical association indeed exists between these factors.

Size and predation pressures are also germane to explanations of the observed gradeshift, particularly when the evolutionary origins of communication structures are considered. The evolutionary origins of communication structures often involve “co-option” of existing motor or structural systems (Bailey 1991; Bradbury and Vehrencamp 1998; Lee 1995; Nowicki et al. 1992; Spurway and Haldane 1953); this can result in the communication structure being used in a similar behavioral context as the pre-existing motor system (e.g. spiny lobster antennae are used as defensive weapons and, in a subset of taxa, in acoustic defense). This poses a theoretical twist to the size trade-offs described in previous studies (Emlen 2001; Nijhout and Emlen 1998; Podos 1997, 2001): the communication structure and the structure with which it shares space and function may both increase in performance through increase in size. In other words, size increases of one structure could potentially decrease the size and performance of the other structure, even though they both are used for related behavioral functions. Thus, we add a new dimension to the question of evolutionary trade-offs by examining correlated evolutionary size changes when shared morphological systems are used in related behavioral functions and when both systems increase in performance through increases in size.

Hence, an alternative explanation to size trade-offs for the observed distribution of antennal strength may arise from the shared structure and function of the antennal region. Specifically, the Stridentes and Silentes may respond differently to selection for predator defenses. In particular, the presence of an additional mode of predator defense in Stridentes may result in relatively less selection for antennal strength compared to Silentes. If Jasus and Panulirus clades both encountered increased predation at different time periods in evolutionary history, the selection for predator defenses might have operated differently on taxa with one major defense (antennal weapons) versus two major defenses (antennal weapons plus defensive sound production). Are taxa with antennal weapons plus an aposematic sound producing apparatus better at deterring predators than taxa only having antennal weapons? If yes, then Silentes may be under greater pressure to increase the size and bending strength of their antennae. Thus, one would expect the observed pattern that for a given antennular plate size, Silentes have stronger antennae (Fig. 7), not due to size trade-offs, but instead due to different predator deterrent performance in these two clades.

Conclusions

Morphology plays a key role in the evolutionary origins of communication structures as well as the nutritional and performance limits imposed by the shared structures of other behavioral systems. These interactions are complex and require multiple levels of analysis—within species, within developmental time periods, and across clades. Spiny lobsters offer a particularly interesting system for studies of structural limits on communication evolution because of their relatively simple sound producing mechanism and small number of taxa. The presence of shared structures that are used in different modes of the same behavioral function adds an additional dimension, peculiarly common to communication systems, in which shared structural trade-offs might also be interpreted as shared functional trade-offs. An understanding of correlated evolutionary changes in morphology and behavior is fundamental to explaining morphology's evolutionary role in the remarkable diversity of animal communication systems.