Morphological Evolution of the Scapula in Tree Squirrels, Chipmunks, and Ground Squirrels (Sciuridae): an Analysis Using Thin-Plate Splines

EvoIUlion.47(6), 1993, pp. 1854-1873

MORPHOLOGICAL EVOLUTION OF THE SCAPULA IN TREE SQUIRRELS, CHIPMUNKS, AND GROUND SQUIRRELS (SCIURIDAE): AN ANALYSIS USING THIN-PLATE SPLINES

DONALD L. SWIDERSKI Museum ofPaleontology, University ofMichigan. Ann Arbor, Michigan 48109

Abstract.-The mammalian scapula, like many bones, is a single structural element that serves as an attachment site for several muscles. The goal ofthis study was to determine whether the scapula evolves as an integrated unit, or as a collection of distinct parts. Shape differences among the scapulae of tree squirrels, chipmunks, and ground squirrels were described using thin-plate spline analysis. This technique produces a geometric description of shape differences that can be decom­ posed into a series of components ranging in scale from features that span the entire form to features that are highly localized. Shape differences among tree squirrel scapulae were found only in large­ scale features, indicating spatially integrated shape change. Chipmunks and ground squirrels differ from tree squirrels in several features, but shared differences reflectingdivergence oftheir common ancestor were found only in the small-scale features. Divergence of ground squirrels from the common ancestor involved some large-scale changes but was dominated by small-scale changes. Divergence of chipmunks was dominated by large-scale changes. Thus, the scapula evolved as an integrated unit during some transitions but as a collection of distinct parts during others. These results suggest that evolutionary patterns of the postcranial skeleton may be as complex as the patterns that have been described for skulls and feeding mechanisms.

Key words.-Integration, morphometries, scapula, thin-plate spline.

Received April 6, 1992. Accepted March 16, 1993.

Ideally, dissection ofanatomy into parts would Thorington 1982, 1984; Hafner 1984; Hight et allow morphologists to predict the units ofevo­ al. 1974; Moore 1959). The closest relatives of lutionary change, the functional elements ex­ the terrestrial lineage appear to be the tree squir­ posed to selection. In reality, this analysis is rels ofthe conservative tribe Sciurini Burmeister, confounded by the complex network of inter­ 1854, which includes Sciurus and Tamiasciurus connections among anatomical parts. The mam­ (Hafner 1984). The relationships of most sub­ malian scapula is a typical example: it is a single genera and many species groups of terrestrial bone, but at least a dozen muscles attach to it. sciurids have also been resolved; figure 1 illus­ Thus, the entire scapula could be expected to trates the relationships ofthe species used in this evolve as a single large unit, or the various mus­ study and represents the combined results ofsev­ cle attachment areas could be expected to evolve eral studies [within tree squirrels (Hight et al. as several smaller units. 1974; Moore 1959), within chipmunks (Leven­ Support for both evolutionary models can be son et al. 1985), and within ground squirrels found in previous studies ofscapular shape dif­ (Gerber and Birney 1968; Nadler 1966)]. ferences. Lehman (1963), Stein (1981), and Tay­ All chipmunks and ground squirrels are less lor (1974) report that localized enlargement of arboreal and more fossorial than tree squirrels specific muscle attachment areas distinguish var­ (Broadbanks 1974; Elliot 1978; Jones et al. 1983; ious specialists from related generalists. Oxnard Lechleitner 1969; Woods 1980). However, many (1968) reports that several orders of mammals of the nesting and foraging traits that comprise exhibit parallel patterns ofshape differences that "arboreality" and "fossoriality" appear to have reflect general factors acting on the whole scap­ undergone multiple independentchanges in both ula. The goal ofmy study is to determine which chipmunks and ground squirrels (Swiderski pattern characterizes the evolution of scapular 1991a). For example, in each genus, the species shape in the Sciuridae. with the largest and most complex burrow sys­ Within the Sciuridae, chipmunks (Tamias) and tems are distantly related (Tamias striatus and ground squirrels (Spermophilus) are members of Ta. minimus;Spermophilus columbianus and Sp. a monophyleticgroup derived from tree squirrels tridecemlineatus). Thus, there are several parallel (Bryant 1945; Ellis and Maxson 1980; Emry and functional transitions in this group. Comparison 1854

© 1993 The Society for the Study of Evolution. All rights reserved. EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1855

oil oil ;, ;, u 0 'c III 0 .! oil oil .. ;, "'0 0 ;, oil 0 > oil - .. oil ;, QI ;, • ·u .;;; oil oil e oil Co ! ;, ;, c: 0 .s ;, oil e .2 oil 0 0 Co 0 ;, "'0 oil oil oil QI QI .. ;, ;, E E .~ III oil e ~ :0 0 QI "'0 •Co 0 oil u u 0 .: QI oil 0 E E oil oil QI .. oil 0 ;, .~ 0 0 QI ;, '0 QI ·c Gi .s 0 III 0 • ..: "'0 ~ :i c E -; -; E .~ 0 'Q. Of ·e 0 U 'c 'e 0 .. .. 'e QI .s- .. u oil 0 ·u 0 Co I- VI vi vi vi I- ~ ~ ~ ~ VI vi vi vi vi vi

FIG. I. Phylogenetic relationships ofsciurid species. (See the text for references.) ofthe several transformations ofscapular shape ture the spatial organization Thompson intended should reveal whether there is a general pattern to illustrate. Bookstein (1989, 1991) presents a of integrated change. new technique, the analysis ofthin-plate splines, The currently popular approach to analysis of which produces a rigorous quantitative analysis integration uses multivariate statistics to identify of the spatial organization ofshape change. suites ofcovarying distance measurements (e.g., The analysis of thin-plate splines describes Cheverud 1982; Gould and Garwood 1969; Ris­ shape change by interpolating between the rel­ ka 1986; Zelditch and Carmichael 1989). This ative displacements of discrete points, land­ multivariate statistical approach can be traced marks, presumed to correspond between forms to Olson and Miller (1958), and a bivariate ver­ (Bookstein 1989, 1991). The specific interpola­ sion can be found in Huxley's (1932) analysis of tion function is a mathematical expression for allometry. The conceptual roots lie in D'Arcy the deformation of theoretically idealized thin Thompson's (1917) analysis of shape change, steel plates. This function is not intended to model which sought simple patterns oftransformation biological processes but is chosen because it is affecting entire structures. Thompson's Carte­ smooth and consistent across the entire form. sian grid deformations provided graphic illus­ Consequently, the analysis produces an exact, tration ofglobal components ofshape change but reproducible, and geometric description ofshape were not based on quantitative analyses. How­ differences. In addition, the description can be ever, Bookstein and coworkers (Bookstein et al. decomposed into a series of geometric compo­ 1985; Bookstein 1991; Zelditch et al. 1992) have nents. Each component is a weighted combina­ criticized the traditional statistical analyses of tion oflandmarkdisplacements. The mostglobal distance measurements for their inability to cap- components are combinations that represent 1856 DONALDL. SWIDERSKI

8 patterns of change in the mean shapes of each geometric component. The historical patterns are 7 the basis ofthe subsequent discussion ofthe in­ tegration of scapular shape changes.

MATERIALS AND METHODS Shape Coordinates.- Twelve points on each scapula were digitized from video images (fig.2). Most of the points are intersections of ridges or ends ofridges; others are corners. All points were easily identified in all specimens. These consis­ tently recognizable points may be presumed "landmarks," points that correspond across shape changes between taxa (Bookstein et al. 1985; 11 Bookstein 1991). Landmarks are the points at which the biological parts are sampled (Zelditch and Bookstein ms) and the points to which ex­ planations ofshape change are anchored (Book­ stein 1990). Scapulae are not perfectly flat, thus some in­ formation is lost or distorted in the two-dimen­ sional representation ofthese three dimensional objects. In this study, I oriented each scapula with the plane defined by landmarks I, 8, and 9 4 parallel to the plane of focus. The distortion FIG. 2. Scapular morphology and landmarklocations caused by projecting all landmarks onto this plane for the tree squirrel, Sciurus carolinensis: (I) ventral will have greatest effect on the landmarks ofthe end of spine; (2) dorsal end of acromio-c1avicular ar­ acromion and metacromion, which are separated ticulation; (3) ventralend of acromio-c1avicular artic­ laterally from the blade. However, the construc­ ulation; (4) flexure marking the boundary between acromion and metacromion; (5)ventralend of caudal tion of the scapula is fairly simple (two nearly margin of metacromion; (6)dorsalend ofmetacromi­ flat surfaces separated by a strut), and many po­ on-spineboundary; (7)caudalend of vertebralborder, tential effects of changes in lateral position can on teres fossa; (8) intersection of axillary ridge and be easily predicted. Tilting ofthe spine will pro­ vertebral border,between teresfossa and infraspinous fossa; (9) intersection of spine and vertebral border, duce identical antero-posterior displacements of between infraspinous fossa and supraspinous fossa; (10) all landmarks on the acromion and metacromi­ projection ofsubscapular ridge to vertebralborder;(II) on. Tilting of the acromion and metacromion dorsal end of anterior marginal ridge; (12) transition will appear to produce proportionate narrowing between blade and articular process, on the cranial of these structures. Changing the height of the edge. spine will affect the apparent size oftheacromion and metacromion relative to the blade. These transformations affecting the entire form either "distortions" must be kept in mind when inter­ uniformly or in monotone gradients. The other preting shape changes localized to the acromion components represent a series oftransformations and metacromion. affecting progressively smaller regions, down to For each specimen, digitized landmark coor­ highly localized transformations affecting the im­ dinates were transformed to shape coordinates mediate vicinity of only a few closely spaced (Bookstein et al. 1985; Tabachnick and Book­ points. Thus, the components ofthe spline cap­ stein 1990). Landmarks I and 9 at the ends of ture both large-sale, integrated changes, and the scapular spine were designated end points of small-scale, localized changes (Zelditch et al. a baseline and assigned fixed coordinates: (0, 0) 1992). and (0, I). The result ofthis procedure is a new In this study, I use decompositions of thin­ set of 10 pairs of standardized coordinates. (To plate splines to examine evolution of scapular preserve the dorso-ventral orientation of the shape in chipmunks and ground squirrels. Phy­ scapular spine, landmarks 1 and 9 were assigned logenetic relationships are used to infer historical coordinates on the y-axis). The transformation EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1857 takes all specimens to the same baseline orien­ manner that minimizes the amount oflocalized tation and length; shapes of landmark configu­ information implied by the description (Book­ rations are not altered. The term "shape coor­ stein 1991, pp. 318-319). In the scapula, for ex­ dinates" refers to the standardization of scale ample, narrowing at the tip of the acromion is that results from designating a fixed baseline interpreted as localized to this region only if it length. In this study, shape coordinates were used is not integral with changes throughout a larger in the construction ofmean shapes for each spe­ region. The interpolation ofz-axis displacements cies. Shape coordinates were computed for all between landmarks is translated back to the x.y individuals, and the means of the x and y co­ plane, producing a picture ofa two-dimensional ordinates were computed for the landmarks that deformation. This two-dimensional interpola­ are not assigned to the baseline. tion is one realization of the Cartesian grid de­ Bookstein (1991, App. 2) and Goodall (1991) formations proposed by D'Arcy Thompson have demonstrated that construction of mean (1961). (For computational reasons, the x and y shapes depends negligibly on the choice ofbase­ displacements are analyzed separately and the line. The particular choice of baseline is mainly two splines are combined to produce a single for graphical convenience and plays no role in picture ofshape changes.) any ofthe analyses described below. Depictions The deformation ofthe starting form into the of shape changes as displacements or deforma­ final form can be decomposed into affine and tions may appear to be expressed in an arbitrary nonaffine components, as implied above. The coordinate system, but the computations actu­ nonaffine component can be decomposed fur­ ally are coordinate free. Specifically, they are ther, into a series ofprogressively more localized Procrustes-normalized displacements, not dis­ components. The number of these components placements with respect to a baseline (Bookstein is three fewer than the number of landmarks. 1991, p. 324). The form of each component is determined by Thin-Plate Splines. - Descriptions of differ­ the configuration of landmarks in the starting ences in scapular shape among species were pro­ form, and represents the canonical form, or duced by the method ofthin-plate splines. This "mode," ofrelative landmark displacements for method models shape difference as a deforma­ shape changes at that scale oflocalization. These tion between landmarks; technical details are modal forms are called "principal warps" in ref­ provided in Bookstein (1989, 1991). The ap­ erence to the bent steel plate. The contribution proach used in the analysis by splines can be of change at the scale of a particular principal described by a physical metaphor in which the warp to the realized landmark displacements in landmarks ofone form are located on an ideal­ the x,y plane is expressed as a vector, called a ized thin steel plate. The relative displacements "partial warp." Thus, principal warps are geo­ in the x and y directions, which would produce metric terms in which morphological differences the second form, are visualized as if they were can be described and partial warps are the values vertical displacements, along the z-axis. Some assigned to these terms. sets of z-displacements would elevate or rigidly Figure 3 is graphical representation of the re­ rotate the steel plate; these represent affine trans­ lationship between principal warp and partial formations in which parallel lines on the two­ warp. Figure 3A is a starting arrangement of dimensional starting form remain parallel after landmarks, numbered as in figure 2. A principal deformation (e.g., square to parallelogram). Oth­ warp of this configuration, localized to the ven­ er sets of displacements would require bending tral portion ofthe scapula is shown in figure 3B: of the steel plate, equivalent to bending of par­ the numbers in this panel are the relative z-axis allel lines on the starting form. Most real defor­ displacements that will occur in the steel plate if mations include both an affine transformation there is a deformation at this scale. If (0.026, and bending as complementary components. 0.076) is the partial warp score, figure 3C depicts Steel plates bend in a manner that minimizes, the realized landmark displacements in the x.y over the whole plate, both the magnitude of plane, and figure 3D shows the equivalent Car­ bending and the physical energy required to pro­ tesian grid deformation. duce that bending. (Bending and bending energy In the analyses that follow, the principal warps are both zero for the affine component.) The for­ are numbered in order ofincreasing localization mula that describes the steel plate can also be or decreasing scale. Principal warp 1 (WI) is the used to describe landmark displacements in a nonaffine component oflargest scale. The affine 1858 DONALD L. SWIDERSKI

6 3 8 9 o· . .7· . 10

11 .20 '65 • •6

.38 .1 .30 12 35 • •5 .-2 .2 ·-34 4 3 8 -25 A

... ..: ....,.....,.....; r :"

.;.... .;

...: : o C FIG. 3. Graphical illustration ofthe relationship between principal warps and partial warps: (A) configuration ofscapular landmarks in a starting form; (B) a principal warp shown as coefficients (x 0.0 I) ofrelative landmark displacements; (C) a partial warp ofthe same principal warp shown as vectors oflandmark displacements; (D) the same partial warp shown as deformation ofa Cartesian grid. component can be considered at infinite scale and Bookstein (1990). (For instructions to obtain and referred to as principal warp O. The analysis the current version, write [email protected].) by thin-plate splines was implemented using F. Comparisons. - Each species named in figure J. Rohlfs program TPSPLINE, availablewith Rohlf I is represented by a mean scapular shape: the EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1859 means ofthe shape coordinates for the 10 land­ 8 ...... 9 . marks that are not on the baseline. Sample sizes for most species are 20 adults; the exceptions are Tamias amoenus (N = 18) and Spermophi/us spi/osoma (N = 19). All samples are drawn from museum collections and include individuals from ...... :: . as many geographic localities as possible, pro­ viding relatively broad representation of intra­ : : : : : : . specific variation. All species means are described as deforma­ tions ofthe same starting form, the mean ofthe .....: : : : : : : : five tree squirrels representing the outgroup (fig. 4). This form is not used to "root" the phylo­ genetic analysis, nor is ittreated as a hypothetical taxon representing a primitive morphology. The outgroup mean is merely a standard to which all ...... : : : . other shapes will be compared, a reference shape :11 that does not include prominent idiosyncratic .... 6 : : : : . features. Because there is only one starting form, there will be only one set ofprincipal warps. All species will be described as partial warps of the same principal warps. Each species can be compared with every other species simply by comparing the partial warps. The distributions of partial ..... 1: .. warps ofthe nonaffine components are presented ,.....,.....:... -: ..... :. e' .': .... :.....,.....:..... as scatter plots oftheir x and y values. The dis­ tribution ofloadings for the affine component is . . .. presented as a scatter plot in polar coordinates. : : . ··:·····:······:····2·····:·····:·····: This component ofdeformation converts a circle to an ellipse. The ratio of the ellipse's major to : 4: 3 its minoraxis is a measure oftheamount ofaffine : : : - : : . change, "anisotropy."The orientation ofthe ma­ jor axis relative to a reference axis ofthe starting FIG.4. Landmark configuration for the starting form, the grand mean ofthe five tree squirrel species. In this form indicates the direction of greatest elonga­ and following figures, lines connecting landmarks il­ tion of the starting form. Anisotropy and ori­ lustrate approximate scapular outlines. entation ofthe affine component are plotted fol­ lowing the advice of Bookstein (1991, p, 213). This paper focuses on features of the mean ofthe tests adapted to determine the significance scapular shapes ofdistinct species. Often statis­ of differences between species (including tical tests are performed to indicate that a dif­ MANOYA, from which Hotelling's P is de­ ference between means is large enough to be worth rived), are intended to test for the effect of a reporting, that is, "significant." However, the tests particular factor (Snedecor and Cochran 1967; that have been used, a t-test for single variables, SYSTAT 1992). Size and fossoriality are appro­ or Hotelling's T2 for multivariate cases, are not priate factors, butspecies is not. Systematists and appropriate for the evaluation ofdifferences be­ morphologists cannot rely on statistical formulae tween taxonomic groups. The t-test and various to determine whetherdifferences between species multiple-group adaptations are designed to de­ means are biologically important or phylogenet­ termine whether different samples represent the ically informative. same population (Snedecor and Cochran 1967). However, samples from different species are REsULTS known to be from different populations; a z-test Principal Warps. - The principal warps ofthe in this context measures only whether the sample starting form are shown in figures 5, 6, and 7. sizes are sufficient to detect the difference. Many These principal warps are geometrically orthog- 1860 DONALD L. SWIDERSKI

-13 e e T •.. -6 - e-8 e •

e25 1 - -

e 24 e e e 21e 5 ­ e-21 T .... e e-40 -31 T i A B c FIG. 5. Largest scale principal warp (WI) of the mean tree squirrel (starting form): (A) as coefficients (xO.OI); (B) as displacements parallel to the y-axis;(C) as displacements parallel to the x-axis. The line segments in this figure reflect only the proportional displacements oflandmarks from their positions in the starting form. onal components of the nonaffine part of any izable geometric features, they are a particularly deformation of this form. The combinations of attractive set for analyzing the evolution ofshape. landmark displacements indicated by these warps The largest scale principal warp (WI) of the do not express covariances; but rather the local­ starting form reflects, as usual (Bookstein 1991), ization of shape change by geometric region. displacement of the central landmarks relative When the affine term is included, the principal to the dorsal and ventral ends (fig. 5A). Oriented warps are a complete set offeatures for describing in the j-direction, the partial warp would de­ deformation from this starting form to any end­ scribe relative lengthening ofone end. Figure 5B ing form. Because the principal warps are local- shows the central landmarks moving dorsally and

-26 6 73 e 36 e -68 e -13 e e e e e e -31 22 -66 e60

e-52 e-24 e-2 6e -r» 6e

e7 e10 e- 4 e-8 e14 e-1 41e -12e 3e e-14 e9 e4 e e-13 e e-1 e ·-3 7 -8 1

A B C FIG. 6. Principal warps of the mean tree squirrel representing transformations of the blade: (A) W2, (B) W5, (C) W8. Coefficients xO.OI. EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1861

-2 6 e 27 e 3 e e e e -10 o e-14 e-6

e-38 e20 -65e

e 33 e3 8 e45 e30 e26 e-2 e e-2 e e-34 -29 -25

A B

3 4 2 e -1 e -1 e o e e e e e e -2 -3 -2 eO eo

e-l e-l -s- -5e

47 45 -35 e e e e-2 e-39 e63 8e e-54 e-24 e-59 e e-13 e e62 e e34 59 -44 -11 c o E FIG. 7. Principal warps ofthe mean tree squirrel representing transformations centered on the acromion and ventral end ofthe blade: (A) W3, (B) W4, (C) W6, (D) W7, (E) W9. Coefficients xO.O!. the landmarks at both ends moving ventrally. ments oflandmarks at the anterior and posterior Such displacements shorten the dorsal halfofthe extremes ofthe dorsal margin, with opposite dis­ scapula and lengthen the ventral half. Figure 5C placements ofthe landmarks near the middle (fig. shows the same displacements in the x-direction: 6A). This warp also includes displacements of now, the ends move posteriorly and the center the extreme points on the acromion and meta­ moves anteriorly. cromion. W5 is a much more localized warp that Three of the principal warps (yV2, W5, and primarily represents a contrast between displace­ W8) represent transformations of the shape of ments of the landmarks near the middle of the the blade (fig. 6). The principal warp ofsecond­ dorsal margin (fig. 6B). W8, the most localized largest scale, W2, specifies coordinated displace- of these three warps, describes contrasting dis- 1862 DONALD L. SWIDERSKI

8

~" ".... ;. "':' ";' .. tt ...: - ....

... " .. ~ ..:.... .;.

." :

.-:.-' .. 8

... ·0.

FIG. 8. Cartesian grids for selected species as deformations ofthe mean tree squirrel. (A) Sciuruscarolinensis, (B) Tamias striatus, (C) Spermophilus variegatus, (0) Tamiasciurus hudsonicus, (E) Tamias rufus. (F) Sp. tri­ decemlineatus. All figures ofCartesian deformations produced by TPSpline. placements of the two landmarks on the teres whereas on W4, the center includes the posterior fossa, at the posterior end ofthe margin (fig.6C). tip ofthe metacromion. The endsare also slightly Deformations represented by W3 and W4 af­ different; and W4 covers a smaller portion ofthe fect elongate regions centered near the neck and blade. Both principal warps are contrasts be­ ventral end ofthe spine (fig. 7A,B). On W3, the tween displacements ofa center and the ends. center includes the dorsal edge ofthe acromion, The remaining principal warps, W6, W7, and EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1863

W9, complete the series of deformations cov­ 90· ering progressively smaller portions ofthe acro­ mion, metacromion, and ventral blade. On W6, the tips ofthe acromion and metacromion have parallel displacements, whereas landmarks near the acromion-metacromion boundary are dis­ placed in the opposite direction (fig.7C). On both 0 N W7 and W9, the two landmarks at the tip ofthe 00 CZl acromion have opposite displacements (fig. 180· O· 0 0.05 0.1 0.15 7D,E). The two landmarks on the neck and ven­ log onisotropy tral end ofthe spine also have opposite displace­ 0 ments on these principal warps. W7 and W9 dif­ 6. fer in the relationship of displacements at the 6. neck and dorsal edge of the acromion: parallel 6.~ in W7, opposite in W9. W7 also includes dis­ placementofthe landmark representingthe aero­ mion-metacromion boundary. 270· Deformations.-Figure 8 presents Cartesian FIG. 9. Distribution of affine components of defor­ grid deformations (produced by TPSpline) show­ mation from TPSpline, plotted in polar coordinates: ing scapulae of selected species as transforma­ log anisotropy, 28 (following Bookstein 1991). Sym­ tions of the mean tree squirrel (starting form). bols indicate locations of species means: circle, tree Grids for two tree squirrel species (fig. 8A,D) squirrel; triangle, chipmunk; square, ground squirrel. deviate only slightly from straight lines, indicat­ ing small deformations ofthe starting form. Still, direction are two distantly related ground squir­ the deformed grids show that there are different rels, Spermophilus lateralis and Sp. spilosoma. patterns of deformation; the expansions and In contrast, only one chipmunk, Ta. striatus, has compressions ofgrid lines occur between differ­ a small affine transformation that falls within the ent sets of landmarks. The two chipmunks (fig. ground squirrel range. The other chipmunks, 8B,E) may share some features, but the defor­ representing the subgenus Neotamias, have much mation ofTamias striatus lacks the dramatic tilt larger affine deformations, and all of them are ofthe horizontal lines shown in the deformation oriented in the antero-dorsal direction. This ofTa. rufus. The deformations ofthe two ground component ofshape change makes a substantial squirrels appear to be even more similar, differ­ contribution to the distinctness of Neotamias ing primarily in the magnitude of change (fig. scapulae (fig. lOB). 8C,F). These six deformations and those for the The bending component of largest scale, WI, other 10 species were decomposed into their af­ also contributes little to the deformations ofmost fine and partial warp components specifically to tree squirrel and ground squirrel scapulae (fig. identify groups of species that share features of 11, WI). Even in Sciurus carolinensis, the tree mean scapular shape. Scatter plots of the load­ squirrel with the largest loading for this com­ ings for each component are examined for gaps ponent, the shape changes are fairly subtle: the between taxonomicgroups and for historical pat­ central landmarks are displaced dorsally, com­ terns reflecting the transition from arboreal to pressing the dorsal endofthe scapulaand stretch­ fossorial habits. ing the ventral end (fig. l2A). One ground squir­ The affine component makes little contribu­ rel, Sp. spilosoma has the opposite transformation. tion to morphological change except in chip­ The other ground squirrels have deformations in munks (fig.9). Anisotropy values are quite small the same range as tree squirrels. Most chipmunks for tree squirrels, even the largest tree squirrel have relatively large loadings in the -x direc­ anisotropy represents a barely perceptible shape tion. This pattern primarily represents narrowing change (fig. lOA). The anisotropies of most of the supraspinous fossa with extension of the ground squirrels are only slightly larger, and that acromion (fig. l2B). This component separates of one ground squirrel is well within the tree most chipmunks from ground squirrels, but one squirrel range. Directions ofaffine deformation chipmunk, Ta. rufus, is clearly within the ground vary widely in tree squirrels and ground squir­ squirrel range. Also, Ta. striatus is near the cen­ rels, and the only species that share a common ter of the Neotamias distribution. 1864 DONALD L. SWIDERSKI

B

":- ...:.....:.... :...

~ .. ' ." .' .' '.. I: .. :." .. :." ·11 ..":' .. ."~ .' ..: .' . '.:. .' .. ~" .. . .~" ~ .. '

..~' .'

..; .

~ . . .. ' .'

B

FIG. 10. Cartesian grids for affine components. (A) Sciurus carolinensis (log anisotropy = 0.045, 2El = 108°); (B) Tamias rufus (log anisotropy = 0.152, 2El = 246°).

The loadings for tree squirrels on W2 are more II, W3). The diversity of chipmunks is even diverse than the loadings on WI; only one species lower, but more importantly, their shared di­ has a shape close to the starting form (fig. II). vergence from tree squirrels is readily apparent. Changes represented by W2 also differentiate All chipmunks, including Ta. striatus, are char­ chipmunks from ground squirrels. Although nei­ acterized by dorso-ventral compression of the ther group has diverged far from tree squirrels, acromion and ventral metacromion (fig. 14A). they have diverged in different directions. Chip­ In contrast, these structures tend to be expanded munks have +y loadings reflecting dorso-ventral along the antero-posterior axis in ground squir­ compression ofthe supraspinous fossa with tilt­ rels (fig. 14B), although some ground squirrels ing ofthe acromion and lower metacromion (fig. retain shapes that are very similar to tree squir­ 13A). Ground squirrels have the opposite trans­ rels. An unusual deformation that combines rel­ formation, and a particularly large - y loading atively large expansions in both antero-posterior distinguishes Sp. tridecemlineatus from the oth­ and dorso-ventral directions characterizes Sp. ers. Chipmunks and ground squirrels both have spilosoma and Sp. tridecemlineatus, the two broad ranges of +x loadings that reflect antero­ ground squirrels representing the subgenus /ctid­ posterior compression of the dorsal end of the omys. blade (fig. l3B). The combination oflarge x and The distribution of loadings on W4 suggests small y loading on W2 distinguishes Ta. striatus that evolution of this feature in chipmunks and from Neotamias. Thus, changes represented by ground squirrels extends a linear trend found in this warp contribute to diversity oftree squirrels, tree squirrels (fig. II, W4). However, the defor­ to divergence ofchipmunks and ground squirrels mations ofthe tree squirrels at the two extremes from tree squirrels, and to divergence of chip­ are trivial; these two species have virtually the munk subgenera. same shape for this feature (fig. ISA,B). Ground The diversity ofloadings for tree squirrels and squirrels have somewhat larger loadings on W4, ground squirrels on W3 is relatively low, com­ but the loadings of most ground squirrels still parable to the diversity oftheir WI loadings (fig. represent rather small deformations similar to EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1865

0.06 6. 6. 0.04 6. 6. 0.02 0.04 0 0 6. 0.04 0 0.00 0 0.02 0+ 0.02 0 0 6. 0 0 0 O -0.02 0 0 0.00 CD+ 0 0.00 0 6. 0 0 6. 00 -0.04 ·0.02 0 '0.02 CD 0 0 0 6. 0 o -0.06 ·0.04 '0.04 6. 6. 6.6. 6. 0 -0.08 0 6. ·0.04 '0.02 0.00 0.00 0.02 0.04 0.00 0.02 0.04 WI W2 W3

6.~ 0.08 0 0.06 0 0.00 6. ~ 0.04 LtJ ~6. o 0.04 0 -0.02 0 C§ IIIJ 6. 0.02 0.02 -0.04 00 0 ~o co 0 0.00 6. 0 0.00 ()Io 6. 0 -0.06 0 6. 6.6. 0 B 0 -0.08 0.00 0.02 0.04 0.00 0.02 0.04 -0.02 0.00 W4 WS W6

0'04~00 o.oo~ ~ 0.02t No. 0.02 ·0.02 ~ O.OO~ 0.00 0al ·0.04 <0 0.00 0.02 '0.02 0.00 0.00 0.02 0.04 W7 W8 W9

FIG. II. Distribution ofpartial warp loadings. Symbols indicate locations ofspecies means: circle, tree squirrel; triangle, chipmunk; square, ground squirrel; +, starting form at the origin. that of Sp. tridecemlineatus (fig. I5C). The ex­ (fig. l6A). Transformations this small probably ception is Sp. spilosoma, which has a much larger do not meaningfully affect scapular shape or transformation along the same trend (fig. l5D). function, they are likely obscured by the contri­ Among chipmunks, Ta. striatus and Ta. ruficau­ butions ofother warps. Most chipmunk loadings dus have loadings similar to Sp. tridecemlinea­ also represent negligible change. The exception tus; the other three species are more similar to is a noticeable narrowing at the dorsal end ofthe Sp. spilosoma. Thus, changes in the feature con­ supraspinous fossa in Ta. rufus (fig. l6B). Only tribute primarily to diversity within chipmunks. two ground squirrels have small loadings indi­ Changes in W5 contribute primarily to diver­ cating virtually no divergence from tree squirrels sity within ground squirrels. Even the largest tree in this feature; the other four species have more squirrel loading represents a minute shape change substantial transformations. One species, Sp. 1866 DONALD L. SWIDERSKI

8 ...... a .. '.' r , "{. ..:.....:... ",""'to'" " ~'. , \ \: \ \ \ : \: ~1tt ~ .... : / I / , " : t: J : , ,. I A 0 • :~.·L······· "4."" . ..;... .:':.. : , '{ '. .. ..: ....; .....; ...... 3 0 '. ,1L·... '; ...... : ......

FIG. 12. Cartesian grids for partial warp 1. (A) Sciurus carolinensis, (B) Tamias striatus. columbianus. is characterized by a unique broad­ in tree squirrels (fig. 11, W6). Loadings for all ening ofthe dorsal end ofthe supraspinous fossa chipmunks indicate a wide metacromion with a (fig. 16C).The other large transformation, found short and thin acromion, relative to the starting in Sp. lateralis, Sp. spilosoma, and Sp. tridecem­ form (fig. 17A). The small variation in theirload­ lineatus, increases the dorso-ventral extent ofthe ings should not be interpreted to mean these supraspinous fossa (fig. 16D). structures are exactly the same shape in all chip­ Low loadings on W6 for tree squirrels indicate munks. Rather, these loadings indicate similar that there is no appreciable variation at this scale deviations from the starting form only after ac-

B B

tt

FIG. 13. Cartesian grids for partial warp 2. (A) Tamias rufus. (B) Ta. striatus. EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1867

FIG. 14. Cartesian grids for partial warp 3. (A) Tamias rufus. (B) Spermophilus columbianus. counting for the different deformations of the tribution of loadings on W6. Diversity within larger scale components (affine, W2, and W4). each ofthe three main groups is small, but gaps Consequently, this feature is clearly evident only between groups are substantial and Ictidomys is in the total deformation of Ta. striatus (fig. 8). distinct from other ground squirrels. In tree Ground squirrels have larger W6 transforma­ squirrels, the transformations ofmost species are tions (fig. 17B)and relatively less diversity in the trivial compared with the contributions ofother larger scale features. Consequently, the defor­ warps. The exception is the transformation in mations of W6 contribute substantially to total Tamiasciurus hudsonicus, which is similar to that deformations ofthe ground squirrels (fig.8). The found in chipmunks. This transformation in­ two species with particularly large deformations volves tapering and ventral displacement ofthe of W6 are the two members of Ictidomys: Sp. acromion with narrowing ofthe neck (fig. 19A). spilosoma and Sp. tridecemlineatus. Ground squirrel loadings on W9 indicate a much All species have small loadings on W7 (fig. II, larger dorso-ventral component, especially in W7). Even the largest loadings represent trivial Ictidomys. The large transformations in this sub­ morphological changes. Thus, this warp does not genus also include noticeable tilting ofthe acro­ represent an important component of shape mion tip (fig. 19B), which rotates the acromio­ change in any taxon. All transformations of the clavicular articulation counter to the rotation of ventral end of the scapula are at other scales. the whole acromion represented by W6. Thus, Loadings on W8 also span a small range (fig. the acromio-clavicular articulation surface is II, W8), but there is a tendency for loadings on maintained in a relatively stable orientation this warp to differentiate between tree squirrels through the dramatic changes ofthe acromion's and terrestrial sciurids. Ground squirrels and position and orientation. chipmunks all share reduction ofthe teres fossa (fig. 18). However, the teres fossa also is reduced DISCUSSION in one ofthe tree squirrels, so the two groups are All 16 species were compared with one starting not completely separate. form, so that all 16 could be described using a The distribution of loadings on the most lo­ single set ofterms, a single set ofprincipal warps. calized warp (fig. II, W9) is similar to the dis- For an evolutionary study, the most convenient 1868 DONALD L. SWIDERSKI

8 8 9 .... 9 ...... II,- '.' .. <, .•.•.:.. :•.•. J.•..•.••: "," "': "".:O .. . 10 ...... :... ,....: ..... : " ...K:...... C ...... -\ \! 11 11 ,~ ,...... r J I I I , ,I / I -;...... / :/ I 1 / J 1 /...... :. ft2 :. I J /~~...... r 4." .... "\ ,? • 3 A. '\ " J ~ '" -" 1

FIG. 15. Cartesian grids for partial warp 4. (A) Sciurus carolinensis, (B) Tamiasciurus hudsonicus, (C) Sper­ mophilus tridecemlineatus, (D) Sp. spilosoma. starting form would be the ancestral morphol­ rived features that can be recognized only after ogy. The partial warps scores would represent a phylogenetic analysis (Maddison et al. 1984). net divergence from the ancestor (net because However, the outgroups in this study, the tree the evolutionary path may not have been linear). squirrels, are members ofa lineage that has been Outgroups can be instrumental for inferring an­ described as .morphologically uniform (Hersh­ cestral states, but outgroups may also have de- kovitz 1969; Moore 1959; Musser 1968) and ex- EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1869

8 ...... 9 T ..... ····,.i··i~ .: i..·.: "; .... : ...". '-'.~':': mr.... : :...... :K'':' 11 \: ~11 r ;: / J V : / : .... 1 / ..1- • f. :.... : .. ~. .:..", ...."l 1 \

.. 8 , 9' '; :'··8 ..... :- ':-e... ". '. '. <: .. 0 10 " " :".0:." \ ; . ., ...

A ; \ .' \: 1 .: .1 r l I / / / / 1 /., ,.1: \. -, T. c : ~~-

FIG. 16. Cartesian grids for partial warp 5. (A) Sciurus carolinensis, (B) Tamias rufus. (C) Spermophilus columbianus. (D) Sp. tridecemlineatus. tremely primitive (Emry and Thorington 1984). of landmarks relative to the starting form. The Therefore, I used the mean scapular shape ofthe separation of groups evident in these loadings tree squirrels as a reference shape that might ap­ would be detected equally well by any othercom­ proximate the ancestral shape. plete set of morphological descriptors. The ad­ For each species, the entire set of 10 pairs of vantage ofthis particular set ofdescriptors is that loadings (for the affine component and the nine they are readily interpreted as transformations warps) completely describes the reconfiguration ofgeometric regions ofanatomical structures. As 1870 DONALD L. SWIDERSKI

..'. ...,:.. .~.....~ ..~ ..: ····~··;·T·'··:·· i·····:··,··;·· ...:- ....: ...... ; ....:.....:..•..,.....,.... ; 11 .,.: ....!i.).. . -:....

. ; .....

. .;' .

.... :... "

Flo. 17. Cartesian grids for partial warp 6. (A) Tamias rufus and (B) Spermophilus variegatus. such, these features are particularly amenable to parison ofpartial warp loadings revealed a much biological interpretation, including phylogenetic more complex picture ofscapular evolution than analysis. was suggested by previous morphometric stud­ Analysis of sciurid scapular shapes by com- ies. Leamy and Atchley (1984) and Oxnard (1968) both suggested that the scapula will evolve as a single integrated unit, along a single trajectory. In thin-plate spline analyses, spatially integrated change is represented by components ofthe larg­ er scales, the affine and the first few nonaffine (Zelditch et al. 1992). These components rep­ resent shape features that span all or most ofthe scapula. In tree squirrels, to the extent that scap­ ular shape changes at all, the changes are reported as large scale features (WI, W2, and W3). In chipmunks, there is divergence from tree squir­ rels reported on the three largest scale partial warps, but the shape changes are not limited to large scale features. Diversification within chip­ munks is found at small scales, as well. Scapular shape changes in ground squirrels also occur at both large and small scales. In fact, chipmunks and ground squirrels share several small scale divergences from tree squirrels. Clearly, scapular shape evolution in sciurids is not limited to spa­ tially integrated transformations. Shape changes that occur at different scales may still be considered integrated, ifthere is some FlO. 18. Cartesian grids for partial warp 8, Sper­ evidence of their coordination. Currently, there mophilus variegatus. is considerable debate over the appropriate EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1871

; ", : ; ~ ~ ..." : .... .

.....: ;. j : : : : :' :.: :: ...... ': : :: .

...... ; : : : :: :: . : . :: .

~ ~ . ; ... :...... :..: :.'11 >.>.: .'.!.....: : ..+: :11 :······:·····:····\····f····:·····\·····~····:···· ..... ~.....: :.- ..~ : j.....~ :.

...... ;. ..

....~... ~ .... ""',

.: ...-

.. ' .., ....: ...... FIo. 19. Cartesiangrids for partial warp 9. (Left) Tamiasciurus hudsonicus, (right) Spermophilus tridecemli­ neatus. method for incorporating phylogenetic relation­ parallel transformations in chipmunks and ships when determining precise correlations of ground squirrels that separate the highly fossorial character transformations (cf. Felsenstein 1985; species (Tamias striatus and Ta. minimus; Sper­ Maddison 1990; Martins and Garland 1991). mophilus columbianus and Sp. tridecemlineatus) However, congruent phylogenetic distributions from less fossorial species. These results do not ofcharacter state transformations can be used to mean that scapular shape is not influenced by indicate historical coordination (Donoghue 1989; function; but the variety ofpatterns does indicate Bjorklund 1991). Tests for constant proportion­ that no one factor has a strong effect. ality can be considered after historical coordi­ This is not the first study to show that a seem­ nation has been demonstrated. Examination of ingly cohesive skeletal structure has not evolved the scatter plots ofthe partial warps reveals only as an integrated unit. Van der Klaauw (1948­ two features with similardistributions. Both plots 1952) and Dullemeijer (1958) demonstrated that suggest that there may have been a small shape the skull evolves as though it contains several change in the common ancestor of chipmunks discrete units, and even the bones that comprise and ground squirrels, and that later changes in a the skull do not always evolve as single elements. different direction separated ground squirrels Theirobservations and theories provided the ba­ from the common ancestor and Ictidomys from sis for Moss's functional matrix theory of skull the other ground squirrels. This is the only evi­ evolution (Moss 1962; Moss and Saletijn 1969). dence ofhistorical coordination ofscapular shape Dullemeijer also argued that changes in the func­ changes in sciurids. tional relationships ofskull bones (e.g., changes Leamy and Atchley (1984) and Oxnard (1968) in their mechanical connections to other bones also suggested that scapular shape changes will or to associated soft parts) would result in changes reflect functional shifts. However, the preceding in the historical associations of evolutionary discussion of scapular shape changes indicates changes. Indeed, dramatic changes in the evo­ that there are nine historical patterns for 10 shape lutionary patterns of fish feeding mechanisms features. None ofthese patterns is associated with have been attributed to changes in the mechan­ the evolution offossoriality. No feature exhibits icallinkages among the bones (Liem 1973, 1980; 1872 DONALD L. SWIDERSKI

Schaefer and Lauder 1986; Sanford and Lauder Bookstein, F. L. 1989. Principal warps: thin-plate 1989). splines and the decomposition of deformations. IEEE Transactions on Pattem Analysis and Ma­ More unusual are comparable studies of co­ chine Intelligence II :567-585. ordinated evolution in the postcranial skeleton --. 1990. Introduction to methods for landmark (e.g., Emerson 1988; Swiderski 1991 b). The lack data. Pp. 215-227 in F. J. Rohlf and F. L. Book­ of examples prompted discussants at a recent stein, eds. Proceedings of the Michigan Morpho­ Dahlem conference to remark on the conserva­ metries Workshop, 1988. University of Michigan Museum of ZoologySpecial Publication 2, Ann Ar­ tism of the postcranial skeleton relative to the bor. skull and jaws (Lauder et aI. 1989). This remark --. 1991. Morphometric tools for landmark data: is based, in part, on the perceived stability of geometry and biology. Cambridge University Press, bone-bone articulations in vertebrate limbs. New York. Bookstein, F. L., B. Chernoff, R. L. Elder, J. M. Hum­ Linkages between bones are hypothesized to pre­ phries, G. R. Smith, and R. E.Strauss. 1985. Mor­ dict correlated transformations of the linked phometries in evolutionary biology. Special Pub­ bones (Lauder 1981; Emerson 1988). Thus, be­ lication 15. Academy of Natural Sciences cause the femur always articulates with the pel­ Philadelphia, Philadelphia, Pa. vis, transformations of the femur and pelvis Broadbanks, H. E. 1974. Tree nests of chipmunks with comments on associated behavior and ecol­ should always be coordinated. However, ifeach ogy. Journal ofMammalogy 55:630-639. bone is composed of several independently Bryant, M. D. 1945. Phylogeny ofNearctic Sciuridae. evolving regions, their articulation may predict American Midland Naturalist 33:257-390. only which regions of the bones are correlated. Cheverud, J. M. 1982. Phenotypic, genetic, and en­ vironmental integration in the cranium. Evolution Changes in the locations of the articulation, or 36:499-512. in the positions of muscle or ligament attach­ Donoghue, M. J. 1989. Phylogenies and the analysis ments could produce new suites ofcoordinated of evolutionary sequences, with examples from seed shape changes. The complexity of postcranial plants. Evolution 43:1137-1156. skeletal evolution could be much greater than is Dullemeijer, P. 1958. The mutual structural influence ofthe elements in a pattern. Archives Neerlandaises generally appreciated. de Zoologie 13, supplement 1:74-88. The analysis of sciurid scapulae also demon­ Elliot, L. 1978. Social behavior and foraging ecology strates the utility ofthin-plate splines for studies of the Eastern Chipmunk (Tamias striatus) in the of morphological transformations. The region­ Adirondack Mountains. Smithsonian Contribu­ tions to Zoology 265:1-107. alization of the warps-that is, their ability to Ellis, L. S., and L. R. Maxson. 1980. Albumin evo­ localize changes to distinct geographic regions of lution within New World squirrels (Sciuridae). a starting form-is not just aesthetically attrac­ American Midland Naturalist 104:57-62. tive. By recourse to theirgeometry, shape change Emerson, S. B. 1988. Testing historical patterns of can be decomposed into biologically meaningful change: a case study with frog pectoral girdles. Pa­ leobiology 14:174-186. components. In addition, reference to the start­ Emry, R. J., and R. W. Thorington, Jr. 1982. De­ ing form facilitates phylogenetic analysis. scriptive and comparative osteology of the oldest fossil squirrel, Protosciurus (Rodentia: Sciuridae). ACKNOWLEGMENTS Smithsonian Contributions to Paleobiology 47:1­ F. L. Bookstein and M. L. Zelditch provided 35. --. 1984. The tree squirrel Sciurus (Sciuridae, considerable assistance with learning the use and Rodentia) as a living fossil. Pp. 23-31 in N. Eld­ description ofthe warps. Videodigitizing equip­ redge and S. M. Stanley, eds. Living fossils. Spring­ ment was provided by D. Erwin. I would like to er-Verlag, New York. thank the following curators for their time and Felsenstein, J. 1985. Phylogenies and the compara­ tive method. American Naturalist 125:1-15. the loan of their specimens: A. C. Carmichael, Gerber.J. D., and E. C. Birney. 1968. Immunological Michigan State University Museum; P. Myers, comparisons of four subgenera of ground squirrels. University ofMichigan Museum ofZoology and Systematic Zoology 17:413-416. B. Patterson, Field Museum. I would also like to Goodall, C. R. 1991. Procrustes methods in the sta­ tistical analysis of shape. Journal ofthe Royal So­ thank the three anonymous reviewers for their ciety ofStatistics Series B 53:285-339. efforts and constructive comments. Gould, S. J., and R. A. Garwood. 1969. Levels of integration in mammalian dentitions: an analysis of correlations in Nesophantes micrus(Insectivora) LITERATURE CITED and Oryzomyscouesi (Rodentia). Evolution 23:276­ Bjorklund, M. 1991. Evolution, phylogeny, sexual 300. dimorphism, and mating system in the grackles Hafner, D. J. 1984. Evolutionary relationships ofthe (Quiscalus spp.: Icterinae). Evolution 45:608-621. Nearctic Sciuridae. Pp. 3-23 in J. O. Murie and G. EVOLUTION OF SCAPULAR SHAPE IN SQUIRRELS 1873

R. Michener, eds. The biology ofground-dwelling ican and Guatemalan grey squirrel Sciurus aureo­ sciurids. University of Nebraska Press, Lincoln. gaster F. Cuvier (Rodentia: Sciuridae). Miscella­ Hershkovitz, P. 1969. Recent mammals ofthe Neo­ neous Publications Museum ofZoology University tropical Region: a zoogeographic and ecological re­ of Michigan 137:1-112. view. Quarterly Review of Biology 44:1-70. Nadler, C. F. 1966. Chromosomes of Spermophilus Hight, M. E., M. Goodman, and W. Prychodko. 1974. franklinii and taxonomy ofthe ground squirrel ge­ Immunological studies ofthe Sciuridae. Systematic nus Spermophilus. Systematic Zoology 15:199-205. Zoology 23:12-25. Olson, E. c., and R. L. Miller. 1958. Morphological Huxley, J. S. 1932. Problems of relative growth, 2d integration. University ofChicago Press, Chicago. ed, Dover, New York. Oxnard, C. E. 1968. The architecture ofthe shoulder Jones, J. K., Jr., D. M. Armstrong, R. S. Hoffmann, in some mammals. Journal of Morphology 126: and C. Jones. 1983. Mammals of the northern 249-290. Great Plains. University of Nebraska Press, Lin­ Riska, B. 1986. Some models for development, coln. growth, and morphometric correlation. Evolution Lauder, G. V. 1981. Form and function: structural 40:1303-1311. analysis in evolutionary morphology. Paleobiology Rohlf, F. J., and F. L. Bookstein, eds. 1990. Pro­ 7:43~42. ceedings of the Michigan Morphometries Work­ Lauder, G. V., A. W. Crompton, C. Gans, J. Hanken, shop, 1988. University of Michigan Museum of K. F. Liem, W. O. Maier, A. Meyer, R. Presley, O. Zoology Special Publication 2, Ann Arbor. C. Rieppel, G. Roth, D. Schluter, and G. A. Zweers. Sanford, C. P., and G. V. Lauder. 1989. Functional 1989. Group report: How are feeding systems in­ morphology of the "tongue bite" in the osteoglos­ tegrated and how have evolutionary innovations somorph fish Notopterus. Journal of Morphology been introduced? Pp. 97-115 in D. B. Wake and 202:379-408. G. Roth, eds. Complex organismal functions: in­ Schaefer, S. A., and G. V. Lauder. 1986. Historical tegration and evolution in vertebrates. Wiley, New transformation of functional design: evolutionary York. morphology of feeding mechanisms in loricarioid Leamy, L., and W. R. Atchley. 1984. Morphometric catfishes. Systematic Zoology 35:489-508. integration in the rat (Rattus sp.) scapula. Journal Snedecor, G. W., and W. G. Cochran. 1967. Statis­ ofZoology 202:43-56. tical methods. Iowa State University Press, Ames. Lechleitner, R. R. 1969. Wild animals ofColorado. Stein, B. R. 1981. Comparative limb myology oftwo Pruett, Boulder, Colo. opossums, Didelphis and Chironectes. Journal of Lehman, W. H. 1963. The forelimb architecture of Morphology 169:113-140. some fossorial rodents. Journal ofMorphology 113: Swiderski, D. L. 1991a. Factors influencing morpho­ 59-76. logical evolution ofthe limbs in fossorial mammals Levenson, H., R. S. Hoffmann, C. F. Nadler, L. Deutsch, (Sciuridae and Soricidae). Ph.D. dissertation. and S. D. Freeman. 1985. Systematics ofthe Hol­ Michigan State University, East Lansing. arctic chipmunks (Tamias). Journal of Mammal­ --. 1991b. Morphology and evolution of the ogy 66:219-242. wrists of burrowing and non-burrowing shrews Liern, K. F. 1973. Evolutionary strategies and mor­ (Soricidae). Journal ofMammalogy 72:118-125. phological innovations: cichlid pharyngeal jaws. SYSTAT. 1992. Statistics, version 5.2 edition. SY­ Systematic Zoology 22:425-441. STAT, Inc., Evanston, Ill. ---. 1980. Adaptive significance ofintra- and in­ Tabachnick, R. E., and F. L. Bookstein. 1990. The terspecific differences in the feeding repertoires of structure of individual variation in Miocene Glo­ cichlid fishes. American Zoologist 20:295-314. borotalia. Evolution 44:416-434. Maddison, W. P. 1990. A method for testing the Taylor, M. E. 1974. The functional anatomy of the correlated evolution oftwo binary characters: Are forelimb of some African Viverridae (Carnivora). gains and losses concentrated on certain branches Journal ofMorphology 143:307-336. ofa phylogenetic tree? Evolution 44:539-557. Thompson, D'Arcy W. 1961. On growth and form. Maddison, W. P., M. J. Donoghue, and D. R. Mad­ Abridged ed. Edited by J. T. Bonner. Cambridge dison. 1984. Outgroup analysis and parsimony. University Press, Cambridge. Systematic Zoology 33:83-103. van der Klaauw, C. J. 1948-1952. Size and position Martins, E. P., and T. J. Garland, Jr. 1991. Phylo­ ofthe functional components ofthe skull. Archives genetic analyses ofthe correlated evolution of con­ Neerlandaises de Zoologic 9:1-361. tinuous characters: a simulation study. Evolution Woods, S. E., Jr. 1980. The squirrels ofCanada. Na­ 45:534-557. tional Museum ofCanada, Ottawa. Moore, J. C. 1959. Relationships among the living Zelditch, M. L., F. L. Bookstein, and B. L. Lundrigan. squirrels ofthe Sciurinae. Bulletin ofthe American 1992. Ontogeny of integrated skull shape in the Museum ofNatural History 118:153-206. cotton rat, Sigmodon. Evolution 46:1164-1180. Moss, M. L. 1962. The functional matrix. Pp. 85-98 Zelditch, M. L., and A. C. Carmichael. 1989. On­ in B. S. Krauss and R. Riedl, eds. Vistas in ortho­ togenetic variation in patterns of developmental dontics. Lea and Febiger, Philadelphia, Pa. and functional integration in skul1s of Sigmodon Moss, M. L., and L. Saletijn. 1969. The primary role fulviventer. Evolution 43:814-824. of functional matrices in facial growth. American Journal ofOrthodontics 55:566-577. Musser, G. G. 1968. A systematic study ofthe Mex- Corresponding Editor: J. Hanken