Printer Friendly

Morphometric tools for sexing loggerhead shrikes in California.

Loggerhead shrikes (Lanius ludovicianus) are sexually monochromatic, and therefore are difficult to sex in the field, despite the sexual dimorphism in plumage or mensural characteristics that has been reported for some populations (e.g., Miller, 1931; Slack, 1994; Collister and Wicklum, 1996; Santolo, 2013). Quantitative morphometric features can overlap considerably across populations (Haas, 1987; Collister and Wicklum, 1996; Santolo, 2013), and therefore limit their broader applicability. Qualitative plumage characteristics can also be very useful (Slack, 1994), but often rely upon more or less subjective assessments of character states. In the absence of clear diagnostic characters, quantitative multivariate techniques have shown considerable promise for discriminating between sexes in both loggerhead (Collister and Wicklum, 1996) and congeneric northern shrikes (Brady et al., 2009).

As part of a larger ecomorphological study of shrikes (Sustaita, 2013), we captured, banded, and measured 59 loggerhead shrikes throughout Southern and Central California, USA. We determined the sex of these shrikes from samples of their contour body feathers using well-established and widely implemented deoxyribonucleic acid (DNA)-based methods (Fridolfsson and Ellegren, 1999; Jensen et al., 2003; Horavath et al., 2005; Brady et al., 2009). Our primary motivation was to construct a predictive model on the basis of the morphometric measurements taken in the field to assign sex to five individuals for which we failed to obtain feather samples. Consequently, we expanded our immediate application to develop a more robust, geographically explicit model that could be deployed and modified as necessary by others who work with shrikes in western North America, and perhaps elsewhere.

Brady et al. (2009) found that tail length and the extent of black on the outermost rectrices were the best predictors of sex in northern shrikes (Lanius excubitor) on the basis of their discriminant function analysis (DFA). Collister and Wicklum (1996) found that the distal extent of white color on the primary flight feathers, tail length, and bill depth contributed the most to discrimination between sexes in their sample of loggerhead shrikes on the Canadian prairies. However, the extent to which these features work for other populations of loggerhead shrikes is unclear. Here we test the efficacy of these as well as other variables for predicting sex of loggerhead shrikes in California. Though they did not explicitly include it in their DFA, Brady et al. (2009) reported that the pattern of the white-to-black color transition along the proximal-mid-sixth primary (P6) feather vane was a good indicator of sex in adult northern shrikes; a distinct, horizontal transition was typical of males, whereas a more angled, v-shaped transition was more typical of females. Therefore, we incorporated P6 color pattern directly into our analyses in the form of both qualitative and quantitative variables.

In addition to DFA, we also used a vastly underused technique--binary multiple logistic regression (MLR)--for developing predictive models for sex identification on the basis of standard morphological measurements. Furthermore, we took a novel approach to reducing subjectivity and possible biases in the assessment of feather character states presumed to be useful for distinguishing between sexes in shrikes, by incorporating input from multiple observers.

MATERIALS AND METHODS--Fieldwork took place during two separate bouts between 7 December 2010 and 1 November 2011, in six general regions in California, USA: near Willow Springs/ Rosamond (approximately 34[degrees]52.163'N, 118[degrees]30.346'W) and Maricopa (35[degrees]7.9112'N, 119[degrees]25.436'W) in Kern County; near Barstow/Hinkley (34[degrees]59.360'N, 117[degrees]17.787'W) in San Bernardino County; near Glamis/Niland (33[degrees]10.765'N, 115023.119'W) in Imperial County; near Rio Vista (38[degrees]11.832'N, 121[degrees]46.589'W) in Solano County; and near Moreno (33[degrees]52.291'N, 117[degrees]7.029'W) in Riverside County. We opportunistically captured shrikes that we saw perched along country roads, using a walk-in trap described by Craig (1997) baited with a live mouse. Upon capture, we aged (adult = AHY [after hatch year]/ASY [after second year] and juvenile = HY [hatch year]/SY [second year]; after Slack [1994] and Pyle [1997]), photographed, fitted a permanent U.S. Geological Survey numbered leg band to each shrike, and gently plucked two to three contour breast feathers. These feathers were stored and sealed individually in small pharmacy-brand plastic medicine bags, and transported to a freezer within 24-72 h, where they were maintained frozen at approximately -23[degrees]C to -18[degrees]C for 5-16 months until they were processed for DNA extraction and amplification.

We promptly released birds where they were captured upon obtaining the following morphometric measurements: (1) nalospi (length of bill from the rostral end of the naris to the rostral upper bill tip), (2) bill width (perpendicular to upper bill, at nares), (3) bill depth (along from dorsal culmen to ventral gonys, at nares), (4) claw (chord) length (for digits I-IV, from distal ungual to distal tip of nail), (5) hook length (along upper bill, from tomial notch to hook tip), (6) tomial notch aperture (parallel to length of culmen, from ventral vertex of tomial tooth to hook tip), (7) tarsus length (left leg, from hypotarsus to base of foot pad between digits I and II), (8) tail length (from between central rectrices, to distal tip of longest rectrix), (9) wing chord (right wing, unflattened, from carpometacarpal joint to distal tip of longest primary), and (10) the distal extent of white coloration on the primaries (from the carpometacarpal joint to the distal tip of the longest primary; after Collister and Wicklum [1996]). We used digital calipers (to [+ or -]0.01 mm; Absolute Digimatic, Mitutoyo, Aurora, Illinois) for measurements 1-7, and a 10-cm metric ruler (to [+ or -]1 mm) for 8-10. Finally, we weighed birds (to [+ or -]0.5 g) using a 100g-capacity spring scale (Pesola[R], Switzerland). These variables were log10-transformed before statistical analysis to improve normality and ensure homoscedasticity. All procedures were conducted in accordance with state (California Department of Fish and Wildlife SCP, SC-11001), federal (U.S. Geological Survey Bird Banding Laboratory Banding Subpermit, 22664-N; U.S. Fish and Wildlife Service Scientific Collecting Permit, MB232078-1), and University of Connecticut (Institute of Animal Care and Use Committee #A09-040) regulations.

We extracted and amplified DNA from contour feather calami according to the techniques of Jensen et al. (2003), Donohue and Dufty (2006), and Brady et al. (2009), although with some modification based on our optimization of the polymerase chain reaction (PCR). To enhance extraction of DNA from the feathers, we harvested the proximal ~2-3 mm of the calamus, and then sliced it longitudinally with a sterile razor. We used the Nucleospin[R] Tissue kit (Macherey-Nagel, Germany) according to the manufacturer's instructions (June 2012/Rev. 012), with the exception that we used QIAGEN Buffer AE (Valencia, California) to perform the elutions. We incubated the samples at 56[degrees]C overnight before performing PCR. Before processing the field samples, we tested the DNA extraction and amplification procedures on four frozen shrike carcasses (two loggerhead and two northern shrikes) from our research collections, for which we could verify sex anatomically. We then processed the field samples in two separate batches of seven and 47 samples. We used primers 2550F (5'-GTTACTGATTCGTCTACGAGA-3') and 2718R (5 '-ATTGAAATGATCCAGTGCTTG3') after Fridolfsson and Ellegren (1999) and Brady et al. (2009), which amplify approximately 620 base-pair and 470 base-pair regions of a chromohelicase DNA binding gene intron on the Z and W chromosomes, respectively. We made a master mixture consisting of: 14.37-14.87 [micro]L of molecular-grade H2O, 2.5 [micro]L of Ex Taq[TM] 10x buffer, 2.5 [micro]L Ex Taq deoxynucleotide triphosphate mix (or Promega, depending on the batch), 1.25 [micro]L of 2550F (10 pM), 1.25 [micro]L of 2781R (10 pM), 0.18 [micro]L of Ex Taq polymerase, and 2.5-3.0 [micro]L of genomic DNA extraction. We ran the PCR on an Eppendorf Mastercycler[R] pro (Eppendorf, Hauppauge, New York) with a "touchdown" thermal profile consisting of: 94[degrees]C for 5 min; 94[degrees]C for 1 min, 55[degrees]C for 1 min, and 72[degrees]C for 2 min followed by 10 cycles, decreasing the annealing temperature 1[degrees]C/cycle; 94[degrees]C for 1 min; 45[degrees]C for 1 min and 72[degrees]C for 2 min for 30 cycles; 72[degrees]C for 7 min; and 10[degrees]C hold. Each PCR reaction included a positive (a shrike for which sex was independently verified) and a negative (mixture excluding DNA extraction) control. We visualized PCR products on a 1% agarose gel stained with 2.5 [micro]L of Sybr[R] Safe gel stain (Life Technologies, Grand Island, New York). Each agarose gel included a 1-kb ladder to ensure the correct size of the amplicons as compared with those recovered by Jensen et al. (2003) and Brady et al. (2009).

We used qualitative and quantitative metrics of the color pattern on P6 to evaluate its efficacy for assigning sex. To ensure greater objectivity in the assessment of the color shape, we asked a group of six ornithologists (none of whom had previously worked with shrikes, and who were completely unaware of our motivation) to rate photos of P6 of each shrike captured in the field. Subjects were simply asked to score feathers as "v-shaped" (= 0) or "horizontal" (= 1), and were shown the photo of P6 in fig. 3 of Brady et al. (2009) as an independent reference (similar to our Fig. 1). We used two different compilations of the subjects' responses to assign each feather to one category or the other, the modal assignment among the six observers (Fig. 2A) and the mean score for each feather across observers (in which values closer to 0 indicated more v-shaped transitions, whereas those closer to 1 indicated more horizontal ones; Fig. 2B). With regard to the former, two cases that were tied among observers (one unidentified and one genetically sexed bird) were excluded from the analysis. We asked an additional (seventh) independent (and also blind to the premise of the study) observer to digitize the P6 black/white color transition on digital images of the right wing of each bird, within which P6 was marked by an arrow (Fig. 1). Images were assigned random numbers, and were digitized in random order, using ImageJ 1.44p (U.S. National Institutes of Health, Bethesda, Maryland). The observer used the angle tool to demarcate the boundaries of the black coloration in the middle of the feather vane, on either side of the rachis to compute the angle (to [+ or -]0.5[degrees]). The observer was asked to situate the vertex along the rachis, and draw each arm of the v as closely as possible to the edge of the black coloration. As with the other variables (above), we used the [log.sub.10] transformation of feather angle for subsequent analyses.

We performed DFA and MLR analyses on the three standard measurements recommended by Collister and Wicklum (1996; tail length, bill depth, and extent of white on the primaries), in addition to either the quantitative or a qualitative metric of P6 color pattern. For the DFA, we performed both standard and stepwise (variable entry based on minimizing Wilks' [LAMBDA]; IBM, 2013) procedures to derive the optimal set for predicting sex (with prior probabilities set to 0.5). We also performed the stepwise procedure on all external linear bill and body metrics (described in main text) to explore additional, potentially useful combinations. We interpreted the jackknifed cross-validated percentages of correctly classified cases (in which each case is withheld from the derivation of model coefficients, one at a time, but then is classified by the predictive equation; IBM, 2013) as a more robust metric of predictive performance. In principle, MLR is similar to DFA, in that the dependent variable is categorical, except that the independent variable(s) might be either quantitative or categorical, or a combination of the two (Tabachnick and Fidell, 2001). We used MLR to compute and compare models on the basis of the quantitative and qualitative P6 variables because its maximum-likelihood framework facilitates the direct comparison among candidate models (Quinn and Keough, 2002). We tested model fit by log-likelihood ratio tests as differences in the -2 x log-likelihood between full and reduced models ([G.sup.2]) or between competing reduced models (Quinn and Keogh, 2002). Collister and Wicklum (1996) suggested that intersexual differences can confound the detection of differences among subspecies, and the converse is likely true. Thus, our initial analyses also incorporated the geographic coordinates of each capture location as additional covariates.

Pyle (1997) described the verticity of P6 as an indicator of age in northern shrikes such that greater verticity of the black-to-white transition was more typical of HY-SY birds. Consequently, we originally used this feature in conjunction with others to age the shrikes in our sample, as recommended by Pyle (1997) for loggerhead, and Brady et al. (2009) for northern, shrikes. In adults (AHY/ASY), these features included, but were not limited to, the presence of uniformly glossy-black wing coverts and flight feathers, relatively darker tertials (but without distinct contrasts between adjacent feathers), and relatively fresh, broad, and truncate outer primaries with white tips. In juveniles (HY/SY), these features consisted of: the presence of molt limits and faded/worn greater coverts (compared with fresher, black inner coverts), freshly replaced black tertials that distinctly contrasted with older, brown secondaries, tapered and abraded outer primary coverts, presence of "juvenal" brown rectrices, and some consideration was given to the presence of vermiculations on the throat and breast (although in northern shrikes this varies by sex [Brady et al., 2009]). All of these features were considered collectively when assigning age, but no bird possessed the full suite of characteristics of one age group or the other. Of the 54 shrikes in our sample sexed by DNA, 14 were considered HY-SY according to these criteria. Although we did not include age in the final predictive models, we tested for the effects of age and possible interactions with sex on both quantitative and categorical P6 color pattern metrics using a two-way analysis of variance and a binary logistic regression, respectively. Finally, we performed separate t-tests along each variable to draw univariate comparisons with studies of other shrikes. We performed all analyses in IBM SPSS Statistics 22.0.0 (IBM Corporation, Armonk, New York).

RESULTS--We successfully identified (genetically) the sex of all 54 individuals (33 males, 21 females). Of the 54 cases of known sexual identity, 50 (32 males, 18 females) had data for all included variables, and thus 4 were excluded from analysis. The univariate analyses indicated significant differences along the extent of white on the primaries, claw chord length of the third toe, bill depth, and P6 color vertex angle, and mean P6 feather score; body mass (P = 0.052) and wing chord (P = 0.06) were not significant (Table 1). Two-way analysis of variance revealed no significant effect of age on vertex angle ([F.sub.1,48] = 2.20, P = 0.145), but angles were significantly greater (i.e., more horizontal) in males ([F.sub.1,48] = 29.1, P < 0.0001) after removing the nonsignificant ([P = 0.120] age x sex) interaction term. When the effect of age was excluded altogether, the effect of sex was still significant ([F.sub.1,49] = 40.9, P < 0.0001), as it was when juveniles were excluded from the analysis ([F.sub.1,36] = 25.7, P < 0.0001). Similarly, when the modal binary P6 color pattern score was used, there was no effect of age (logistic regression; model [G.sup.2] = 35.0, B = 1.91 [+ or -] 1.07, df = 1, P = 0.074, n = 54), but males were significantly more likely to be assigned a value of 1 (i.e., horizontal pattern; B = -4.13 [+ or -] 1.15, df = 1, P = 0.0003) after removing the nonsignificant age x sex term. When the effect of age was excluded, the effect of sex was still significant (model [G.sup.2] = 38.3, B = -4.32 [+ or -] 1.12, df= 1, P = 0.0001, n = 54), as it was when juveniles were excluded from the analysis (model [G.sup.2] = 29.2, B = -3.95 [+ or -] 1.19, df = 1, P = 0.001, n = 40).

In the standard DFA including bill depth, tail length, P6 color vertex angle, and extent of white on the primaries, a single discriminant function (Wilks' [LAMBDA] = 0.431, df = 4, P < 0.0001) discriminated between sexes with an 82% success rate (three females were misclassified as males; six males were misclassified as females). This function was most highly laden by P6 color vertex angle (0.785), and secondarily by the extent of white on the primaries (0.318). The stepwise procedure retained only the latter two variables. When a stepwise analysis was performed including all of the variables, only P6 vertex angle and extent of white on primaries were retained (as above), except that the classification rate dropped slightly to 80.4%. Therefore, we used the discriminant equation from the first analysis (including all four predictors) to assign the five individuals of unknown sex:

DF score = [X.sub.i][[beta].sub.i] + [X.sub.n][[beta].sub.n] + [[beta].sub.0]

where [X.sub.i] - [X.sub.n] are the [log.sub.10]-transformed values of each of the four independent variables for each individual for which sex was predicted, [[beta].sub.i]-[[beta].sub.n] are the function coefficients pertaining to each variable (Table 2), and [[beta].sub.0] is a constant. Separate DF scores were calculated using coefficients for males and females, and the set of coefficients that yielded the greater DF score represented the predicted sex. This model resulted in the assignment of three males and two females, resulting in a total sex ratio of 36 males:23 females for the study population. The correct-classification rate improved by 2% when longitude and latitude were included as covariates, but did not alter the assignments of these cases.

All MLR models performed well, with overall (non-validated) percent correct classification ranging from 88% to 94%. The general equation took the form of:

P = [e.sup.z]/1 + [e.sup.z];

where z = [X.sub.i][B.sub.i] + [X.sub.n][B.sub.n] + [B.sub.0] such that [X.sub.i]-[X.sub.n] are the ([log.sub.10]-transformed) values of each of the independent variables (except for binary categorical variables, which are assigned a value of 1 or 0) for each individual for which sex was predicted, [B.sub.i]-[B.sub.n] are the logistic regression coefficients pertaining to each variable (Table 3), and [B.sub.0] is a constant. Individuals with a P > 0.5 were predicted to be male, and those with P < 0.5 were predicted to be female. Although the significance of each predictor varied among models, P6 color pattern (in one form or another; Figs. 1 and 2A and 2B) was significant in every model (Table 3). The modal P6 score resulted in a slightly higher nonvalidated rate of correctly classified cases (91.8%) than the quantitative P6 color angle measurement (88%), and a slightly (but not significantly) better-fitting model ([G.sup.2] = 21.6 vs. 22.5, respectively; [DELTA][G.sup.2] = 0.971, df = 1, P = 0.324). However, the two analyses resulted in slightly different assignment of the five cases of unidentified sexual identity. Whereas the former assigned four to male and one to female, the latter corroborated the original DFA above in the assignment of three males and two females. The simple mean score (Fig. 2B) resulted in the highest percentage of correctly classified cases (94%) and the best-fitting model ([G.sup.2] = 15.5). Including age as an additional (categorical; adult vs. juvenile) independent variable slightly improved model fit ([G.sup.2] = 13.3), but slightly decreased the correct-classification rate (92%), and the effect of age itself was not significant in the model (B = 2.84 [+ or -] 2.30, df = 1, P = 0.218). When juveniles were removed from the sample (resulting in n = 40), model performance decreased to a correct classification of 89.5%. This decrement in performance was not due to the reduction in sample size, however, since 100 random subsets of 40 cases (from the complete data set) resulted in an average correct classification of 94.4% (93.6-95.2%, 95% confidence interval).

Upon cross-validation of the mean P6 score model (using the jackknife procedure described for DFA, above), the percentage of correct classification was 88%, and the inclusion of geographic coordinates did not affect performance. This model assigned four of the unsexed cases to male and one to female, and thus disagreed with the DFA regarding the sex of one individual of unidentified sex (designated male by MLR and female by DFA). Forward and backward stepwise procedures resulted in different combinations of predictors, none of which achieved greater practical performance (i.e., correct-classification rates) than the base models.

DISCUSSION--We tested the utility of both quantitative and qualitative assessments of the angle of the black-to-white color transition along the proximal mid-vane of P6. We found that both methods substantially improved predictive models for identifying sex in our California sample of loggerhead shrikes, for which sex was determined by DNA. our best models correctly predicted sex 86-88% of the time, upon cross-validation. Furthermore, we attempted to make the categorical assessment of color pattern less subjective by incorporating the independent assessments of seven observers blind to our goals. Nevertheless, our models did not perform quite as well as those of Brady et al. (2009) for adult northern shrikes, or other recent applications for other relatively monomorphic birds of prey (e.g., Donohue and Dufty, 2006), which achieved ~97% correct classification. This could be due to the male-biased sex ratio of our sample, or simply that the morphometric variables used are more intersexually discriminative in these other species. Although the age variation inherent in our sample might present a confounding factor, retaining putative juveniles in our sample without explicitly accounting for the effect of "age" seems to have improved model predictive performance.

Studies of other species of shrikes have reported conflicting patterns of sexual dimorphism. Cramp and Perrins (1993; summarized by Gutierrez-Corchero et al., 2007b) found no significant differences between sexes among standard morphometrics (e.g., lengths of the wing, bill, tail, tarsus, and extent of white on the primaries) in great grey shrikes (L. excubitor) in the Netherlands (Gutierrez-Corchero et al., 2007b), whereas Brady et al. (2009) found significant differences in two of these (wing [chord] and tail lengths) for northern shrikes (L. excubitor borealis) in Wisconsin, USA. Infante and Peris (2004) found no overall differences along these metrics in their Iberian Peninsula sample of the southern grey shrike (Lanius meridionalis), although they did report longer primaries in males than in females. Conversely, Gutierrez-Corchero et al. (2007a, b) found significant differences between sexes for their substantially larger number of samples from the Canary islands (2007a) and northern Spain (2007b). Their multivariate and univariate analyses indicated that males were significantly larger than females along several variables (e.g., wing length, third primary length, extent of white on the primaries, extent of white on the rectrices, and tail length). Also, they emphasized the importance of accounting for age differences in detecting sexual differences (Gutierrez-Corchero et al., 2007b). Taken together, these studies underscore the importance of establishing different criteria for different populations, as well as accounting for possible age differences. Although we detected no significant effect of age or small-scale geographic variation, ultimately a larger sample is required to fully disentangle the potential effects of age, sex, and population differentiation on this character in the context of the classification analyses.

Our results corroborated the observations of Brady et al. (2009) regarding the utility of the angle of the black-to-white color transition along the proximal middle of P6, as this was the most important predictor of sex among the four variables (Fig. 1). In fact, we found that even qualitative assessments of the vertex of the color transition (i.e., v-shaped vs. horizontal) considerably improved model performance. When the P6 variable was excluded from the analyses (in any form), the percentage of correctly classified cases dropped to 69.2% for the DFA and 73.1% for the logistic regression. The comparative logistic regressions indicated that the qualitative (group) assessment of the P6 color pattern (Fig. 2) performed similarly, if not better, than the quantitative digital image measurements for discriminating between the sexes (Fig. 1). It is important to note that there was disagreement among the various forms of the P6 metric (vertex angle, modal score, mean score), as well as the different forms of analysis (DFA and MLR), resulting in conflicting assignments for one case of unidentified sex. However, in seven of nine disagreements between DFA and MLR predictions regarding cases of genetically known sexual identity, the MLR was correct. For this reason, in addition to its superior cross-validation results, we used the MLR's assignment for the case in conflict. Ultimately the merits of one form of classification analysis or the other should be vetted with a larger sample size. Nevertheless, we recommend the inclusion of a qualitative or quantitative metric of P6 feather color pattern in classification analyses on the basis of field morphometrics (or digital photos of the flight feathers). Furthermore, Slack (1994) suggested that the color of the rachis and barbs (typically brown in females) could also be useful for distinguishing between sexes; these could easily be incorporated as additional predictors in an MLR, if and when rachis and barb colors are scored.

Although Pyle (1997) recommended the P6 color transition as a criterion for ageing northern shrikes, Brady et al. (2009) have more recently established a greater role for this feature for determining sex. Unfortunately the utility of this feature for ageing loggerhead shrikes remains equally unclear. The effect of age on quantitative and qualitative vertex angle was not significant in our analyses. However, our sample of juveniles was relatively small. Furthermore, there is no firm, objective criterion for assessing age as there is with sex (i.e., DNA), which could undermine predictive models. Thus, we suggest that further work be directed toward understanding the role of P6 color transitions (as well as those of other feathers) for ageing loggerhead shrikes, or toward establishing separate criteria for adults and juveniles, as recommended by Brady et al. (2009).

One other interesting point that emerged from our analysis is that the combination of predictors that were individually significant (on the basis of t-tests; Table 1) was not the "optimal" combination for predicting sex. For example, there was no significant (univariate) difference in tail length between sexes in our sample, despite Santolo's (2013) finding that males in his California sample of loggerhead shrikes had significantly longer tails than did females. Nevertheless, tail length was an important predictor in our models, which is generally consistent with Santolo's (2013) findings. Another example further illustrates the advantage of applying multivariate techniques to parse complex associations among variables, which can ultimately be more informative: the difference in the color patterning of P6 might not be completely independent of the extent of white on the primaries, which is consistently greater in males across species of shrikes (Collister and Wicklum, 1996; Harris and Franklin, 2000; Gutierrez-Corchero et al., 2007a, b). Although the P6 vertex angle was not directly correlated with the extent of white on the primaries (r = 0.102, P = 0.476, n = 54), there was a significantly greater extent of white on the primaries in those birds that were judged by the majority of observers to have horizontal vs. v-shaped P6 color transitions ([t.sub.48] = 2.296, P = 0.026). Despite the uncertainty regarding the precise role of the white wing patch in shrikes (but see Cade [1962]), these results suggest that the distad recession of the black coloration might be one mechanism by which the observed greater extent of white in males is achieved. We suggest that further study be directed toward understanding how these feather color patterns vary concomitantly with sex and age, and across populations.

Because we sampled from potentially panmictic populations, we cannot be certain of the extent to which these indices work for individual subspecies. We captured shrikes within the resident ranges of three putative subspecies (Lanius ludovicianus gambeli, L. l. sonoriensis, and L. l. nevadensis; Miller, 1931; Yosef, 1996). However, previous work by Chabot (2011) has suggested instead that these and other mainland California and Baja subspecies likely pertain to a single gambeli genetic cluster. In addition, we cannot exclude the possibility that our sample contained wintering birds from yet other subspecies. Regardless of the subspecific representation in our sample, some degree of geographic population differentiation is inherent. Nevertheless, the fact that our models performed reasonably well in light of the potential interpopulation (and ontogenetic) variation in our sample is encouraging; given the extensive zones of intergradation (Miller, 1931), such intermixing in shrike samples is more likely the rule than the exception.

Although molecular techniques are becoming ever more widely accessible, they still require resources and expertise to which many ornithologists and wildlife managers may not have access. Thus, subspecific and population-specific predictive models of sex identification are critical not only for establishing sex of shrikes in-hand, but also for understanding patterns of geographic variation in sexual dimorphism. Loggerhead shrike populations have shown significant declines in California, and are considered a Bird Species of Special Concern (Humple, 2008). Thus, the ability to distinguish sexes, particularly outside the breeding season when breeding condition cannot be assessed, has great potential for understanding population demography.

We are grateful to C. Simon and B. Goffinet for the use of their lab spaces and resources for processing the DNA samples. S. Craig graciously provided training and insights for capturing and handling shrikes. Special thanks to S. Sewell, K. Kietzer, and M. White for access to study sites, and to N. Peterson, T. Scott, and G. Santolo for their consultations. We especially thank the members of the UConn Ornithology Research Group and N. Broccoli for their participation in the P6 feather survey. C. Field and C. Elphick provided valuable comments on an earlier version of the manuscript. Special thanks to R. Brady and D. Collister for their constructive and thoughtful reviews, and to S. Craig, R. Yosef, and G. Santolo for their comments on the revised manuscript.


BRADY, R. S., J. D. PARUK, AND A. J. KERN. 2009. Sexing adult Northern Shrikes using DNA, morphometrics, and plumage. Journal of Field Ornithology 80:198-205.

CADE, T. J. 1962. Wing movements, hunting and displays of the Northern Shrike. Wilson Bulletin 74:386-408.

CHABOT, A. A. 2011. The impact of migration on the evolution and conservation of an endemic North American passerine: Loggerhead Shrike (Lanius ludovicianus). Ph.D. dissertation, Queen's University, Kingston, Ontario, Canada.

COLLISTER, D. M., AND D. WICKLUM. 1996. Intraspecific variation in Loggerhead Shrikes: sexual dimorphism and implication for subspecies classification. Auk 113:221-223.

CRAIG, S. H. 1997. What goes around--gets caught! An improved trap for shrikes. North American Bird Bander 22:124-125.

CRAMP, S., AND C. M. PERRINS. 1993. The birds of the Western Palearctic. Oxford University Press, Oxford, England.

DONOHUE, K. C., AND A. M. DUFTY, Jr. 2006. Sex determination of red-tailed hawks (Buteo jamaicensis calurus) using DNA analysis and morphometrics. Journal of Field Ornithology 77:74-79.

FRIDOLFSSON, A.-K., AND H. ELLEGREN. 1999. A simple and universal method for molecular sexing of non-ratite birds. Journal of Avian Biology 30:116-121.

GUTIERREZ-CORCHERO, F., F. CAMPOS, AND M. A. HERNANDEZ. 2007a. Sexual dimorphism in an insular southern grey shrike subspecies Lanius meridionalis koenigi. Ardeola 54:327-330.

GUTIERREZ-CORCHERO, F., F. CAMPOS, M. A. HERNANDEZ, and A. AMEZCUA. 2007b. Biometrics of the Southern Grey Shrike Lanius meridionalis in relation to age and sex. Ringing & Migration 23:141-146.

HAAS, C. 1987. Eastern subspecies of the Loggerhead Shrike: the need for measurements of live birds. North American Bird Bander 12:99-102.

HARRIS, T., AND K. FRANKLIN. 2000. Shrikes and Bush-Shrikes including Wood-Shrikes, Helmet-Shrikes, Flycatcher-Shrikes, Philentomas, Batises and Wattle-Eyes. Helm Editions, London.

HORVATH, M. B., B. MARTINEZ-CRUZ,J. J. NEGRO, L. KALMAR, AND J. A. GODOY. 2005. An overlooked DNA source for non-invasive genetic analysis in birds. Journal of Avian Biology 36:84-88.

HUMPLE, D. 2008. Loggerhead Shrike (Lanius ludovicianus) (mainland populations). Pages 271-277 in California Bird Species of Special Concern: a ranked assessment of species, subspecies, and distinct populations of birds of immediate conservation concern in California (W. D. Shuford and T. Gardali). Western Field Ornithologists and California Department of Fish and Game, Camarillo and Sacramento, California.

IBM. 2013. IBM SPSS advanced statistics 22. IBM Corporation, Armonk, New York.

INFANTE, O., AND S. J. PERIS. 2004. Sexual dimorphism in the Southern Grey Shrike Lanius meridionalis in the central west of the Iberian Peninsula. Ardeola 51:455-460.

JENSEN, T., F. M. PERNASETTI, AND B. DURRANT. 2003. Conditions for rapid sex determination in 47 avian species by PCR of genomic DNA from blood, shell-membrane blood vessels, and feathers. Zoo Biology 22:561-571.

MILLER, A. H. 1931. Systematic revision and natural history of the American Shrikes (Lanius). University of California Publications in Zoology 38:11-242.

PYLE, P. 1997. Identification guide to North American birds. Slate Creek Press, Bolinas, California.

QUINN, G. P., AND M. J. KEOUGH. 2002. Experimental design and data analysis for biologists. Cambridge University Press, Cambridge, England.

SANTOLO, G. 2013. Weights and measurements for American Kestrels, Barn Owls, and Loggerhead Shrikes in California. North American Bird Bander 38:161-162.

SLACK, H. E. III. 1994. Age and sex related characteristics of the Loggerhead Shrike (Lanius l. ludovicianus) in Coastal Mississippi. North American Bird Bander 19:84-89.

SUSTAITA, D. 2013. Biomechanics of feeding in Loggerhead Shrikes. Ph.D. dissertation, University of Connecticut, Storrs.

TABACHNICK, B. G., AND L. S. FIDELL. 2001. Using multivariate statistics. Fourth edition. Allyn and Bacon, Boston.

YOSEF, R. 1996. Loggerhead Shrike (Lanius ludovicianus). Pages 1-27 in The birds of North America. Volume 231 (A. Poole, and F. Gill, editors). The Birds of North America, Inc., Philadelphia.

Submitted 12 November 2013.

Acceptance recommended by Associate Editor, Eddie K. Lyons, 16 June 2014.


Department of Ecology & Evolutionary Biology, University of Connecticut, 75 N. Eagleville Road Unit 3043, Storrs, CT 06269-3043 (DS, CLO, JCV, MAR)

Present address of DS: Department of Ecology & Evolutionary Biology, Brown University, Box G-W, Providence, RI 02912

Present address of CLO: Computational Biology Institute, George Washington University, Innovation Hall, Suite 305, 45085 University Drive, Ashburn, VA 20147

Present address of JCV: Royal Botanic Gardens Edinburgh, 20A Inverleith Row, Edinburgh EH3 5LR, Scotland, United Kingdom

* Correspondent:

TABLE 1--Descriptive statistics and results of univariate t-tests (a)
for differences in morphometric measurements (b) between the sexes
of Southern and Central California populations of loggerhead shrikes.

                        Males         Females        Total
                        (n = 33)      (n = 21)

                        Mean   SD     Mean   SD      Mean   SD

Body mass (g)           49.3   3.56   47.3    4.31   48.5   3.96
White on primaries **   57.2   2.54   55.0    2.46   56.3   2.71
Wing chord              99.2   2.93   97.6    3.12   98.5   3.08
Tail length             98.7   3.44   96.9    3.13   98.0   3.41
Tarsus length           28.6   0.82   28.2    0.74   28.5   0.80
Claw chord D1            6.8   0.53    7.0    0.45    6.9   0.50
Claw chord D2            4.9   0.39    4.8    0.47    4.8   0.42
Claw chord D3 *          6.3   0.54    6.6    0.52    6.4   0.55
Claw chord D4            4.5   0.43    4.4    0.41    4.5   0.42
Nalospi                 12.1   0.74   12.2    0.60   12.2   0.69
Bill depth *             8.4   0.19    8.3    0.19    8.4   0.20
Bill width               6.4   0.29    6.4    0.24    6.4   0.27
Hook length              3.0   0.62    3.0    0.39    3.0   0.54
Mean P6 (c) score        0.7   0.29    0.1    0.17    0.5   0.38
  (0-1) ***
P6 vertex angle         91.6   23.6   60.1   10.3    80.4   24.7
  ([degrees]) ***

(a) Asterisks indicate significant differences between the sexes
(* P < 0.05; ** P < 0.01; *** P < 0.001); measurements without
asterisks did not differ significantly between sexes (P > 0.05).

(b) All measurements are in millimeters unless indicated otherwise.

(c) P6 sixth primary flight feather.

TABLE 2--Fisher's linear discriminant function
coefficients (b) for assigning loggerhead shrikes
of uncertain sex to males or females, on the basis
of morphometric and digital image measurements of
genetically sexed individuals (n = 54).

Independent variable X                Male        Female

([log.sub.10]) P6 (a) vertex angle        213.1       189.9
([log.sub.10]) Bill depth               7,698.1     7,632.2
([log.sub.10]) White on primaries       4,189.1     4,111.8
([log.sub.10]) Tail length              6,578.6     6,601.2
Constant                              -14,014.5   -13,820.4

(a) P6, sixth primary flight feather.

TABLE 3--Multiple logistic regression parameters
(B [+ or -] SE) resulting from three alternative
models (sixth primary flight feather [P6] vertex
angle, P6 modal score, and P6 mean score) for
assigning loggerhead shrikes of uncertain sex
to males or females, on the basis of genetically
sexed individuals (n = 54).

Independent variable X       Vertex angle         Modal score

                                B         SE         B         SE

([log.sub.10]) P6              39.46    16.08 *
  vertex angle
P6 pattern (horizontal =                             7.68     2.69 *
  1, v-shaped = 0)
P6 mean score (0-1)
([log.sub.10]) Bill depth      12.22     51.64     185.54    76.48 *
([log.sub.10]) White on        94.38    45.79 *     59.62     40.23
([log.sub.10]) Tail length     46.63     53.38     -83.61     50.67
Constant                     -341.47    173.94    -111.19    106.28

Independent variable X       Mean score

                                B         SE

([log.sub.10]) P6
  vertex angle
P6 pattern (horizontal =
  1, v-shaped = 0)
P6 mean score (0-1)            17.65     7.09 *
([log.sub.10]) Bill depth     245.94   110.52 *
([log.sub.10]) White on        84.47     60.09
([log.sub.10]) Tail length    -98.11     67.86
Constant                     -184.58    159.61

* Parameter was significant at P < 0.05.
COPYRIGHT 2014 Southwestern Association of Naturalists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Sustaita, Diego; Owen, Christopher L.; Villarreal, Juan Carlos; Rubega, Margaret A.
Publication:Southwestern Naturalist
Article Type:Report
Date:Dec 1, 2014
Previous Article:Differential response by bronzed cowbirds to songs of potential hosts in the Lower Rio Grande Valley of Texas.
Next Article:Effects of wildfire on riparian trees in Southeastern Arizona.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters