42 TROP. LEPID. RES., 22(1): 42-52, 2012 VAN HOOK ET AL.: Standardized measurement of monarch wings A STANDARDIZED PROTOCOL FOR RULER-BASED MEASUREMENT OF WING LENGTH IN MONARCH BUTTERFLIES, DANAUS PLEXIPPUS L. (NYMPHALIDAE, DANAINAE) Tonya Van Hook 1,5, Ernest H. Williams 2, Lincoln P. Brower 1, Susan Borkin 3, and Julie Hein 4 1Biology Department, Sweet Briar College, Sweet Briar, VA 24595; 2Biology Department, Hamilton College, Clinton, NY 13323; 3Invertebrate Zoology, Milwaukee Public Museum, Milwaukee, WI 53233; 45904 Cedar Creek Place, Sturgeon Bay WI 54235 5Corresponding author, Email: [email protected], Tel: (423) 914-0842 Abstract - Standardized measurements using well-defined landmarks are the most effective means to reduce measurement error. We describe such a protocol for monarch forewings based on single measurements with a ruler to the nearest 1.0 mm. Analysis of this protocol showed that it provides excellent intra-observer repeatability, excellent to substantial inter-observer repeatability, and similar wing length estimates as those of calipers at 0.1 mm, as long as sample sizes are > 30. In addition, our study showed that males and females differ in wing length; different observers differ in their measurements and in their measurement error; and wings shrink slightly when dried. We make these recommendations for study of monarch wing lengths: 1) males and females should be analyzed separately; 2) live butterflies should be measured after cooling and dead butterflies should be measured before they are dried; 3) measurements should be restricted to the right forewing; 4) the standard protocol should be practiced and calibrated until measurements are repeatable within and among measurers; 5) the samples should be mixed among all observers when possible to mitigate relative biases; and 6) names, handedness, measurement error, and archived raw data should be reported. Widespread adoption of this protocol will increase the comparability of wing length data from various investigators. Similarly based standardization of measurement would benefit wing measurement of all Lepidoptera. Key words: standardized measurement protocol, butterfly wings, body size, methodology INTRODUCTION has been the most common measure. We found six distinct Forewing length is the most commonly used measure of base points from which forewing straight-line measurements body size in Lepidoptera (Miller 1977, 1991), and wing length originated, and often no landmarks were recorded (Table 1). is often used in studies of monarch butterflies (Table 1). Wing We also found differences in how this measurement was taken, length is correlated with wing width and area (Altizer & Davis including from left or right forewings (sometimes with the left 2010) and with other body measurements such as antennal length and right forewings averaged), from dorsal or ventral surfaces, and thorax width (Arango 1996). Furthermore, wing length is from intact butterflies, and from wings that had been removed a better indicator of body size than wet or dry mass because it from the body. Measurements have been taken while the does not vary with lipid or water content. Finally, wing length butterflies were hand-held or lying on a surface, from live or is easier, faster, and less expensive to obtain than lean body dead individuals, and from dried specimens. Finally, we found mass, the other commonly used measure of lepidopteran body no records of how damaged or worn wings were measured. size (Miller 1977), and it does not require killing the butterflies To reduce inconsistencies both within and among studies, or prolonged handling or storage times. we describe a specific forewing measurement method based Although protocols for forewing measurement have been on well-defined morphological landmarks that can easily be described three times (Beall & Williams 1945; Donham & learned and used by both scientists and amateurs. Additionally, Taylor 1996; Oberhauser et al. 2009), the authors used different we address five questions regarding the protocol: 1) Should landmarks, provided few specific details of how measurements males and females be analyzed separately?; 2) Is recorded were to be made, and did not provide an alternative measure wing length affected by who makes the measurement?; 3) that could be used when wing tips are missing or frayed. Does wing length decrease due to water loss from drying and, Furthermore, none of the existing protocols have been if so, does this bias measurements taken using the standard consistently adopted, thereby limiting the value of comparative protocol?; 4) Do ruler measurements to the nearest 1 mm wing length data. differ from caliper measurements to the nearest 0.1 mm?; and Our review of the literature shows considerable variation in 5) Can forewing cell length be used to estimate total forewing the device, precision of measurement, landmarks, and methods length when neither forewing can be measured due to wing tip used to measure forewing length (Table 1). Rulers, calipers, fraying or damage? Our standardized protocol describes both an optical device [see Williams 1943], and computer programs the method of measurement and recommendations based on that measure scanned images have been used. The landmarks our answers to these five questions. Widespread adoption of employed include: from wing tip to wing tip (Dively et al. this protocol would greatly increase the comparative value of 2004), hind wing length (Herman 1988; Herman et al. 1989), monarch wing length measurements by increasing repeatability wing area (Altizer & Davis 2010; Davis et al. 2007; Davis of measurements within and among observers (Francis & 2009), and the longest straight-line distance from the forewing Mattlin 1986; Bailey & Byrnes 1990; Gordon & Bradtmiller base to the apical margin (forewing length; Table 1), which 1992; Ulijaszek & Kerr 1999; Harris & Smith 2009). VAN HOOK ET AL.: Standardized measurement of monarch wings TROP. LEPID. RES., 22(1): 42-52, 2012 43 MEASUREMENT PROCEDURE forewing surface so that the numbers face right side up but backwards (see Fig. 2). The right forewing should be measured (Fig. 1) unless it is The total length of the red plus green arrows in Fig. 3 deformed or part of the wing tip is frayed or missing, in which indicates the correct line of measurement. Once the leading case the left forewing should be measured. If neither forewing edge of the ruler is correctly positioned on the wing tip, one can be measured directly, the right forewing cell length should must check to be sure that the 1 cm mark did not shift away be measured (see below). The forewing cell length can be used from the basal landmark while rotating the ruler. Readjust if in the regression equation (provided in the Results) to estimate necessary. Using 1 cm rather than 0 cm as the starting point total wing length of the damaged or deformed forewing. The for measurement increases repeatability of measurements, method used to measure the right forewing cell length is but one must remember to subtract 1 cm before recording the described below. length. Record all measurements to the nearest whole mm. If a Regardless of handedness, one should hold the butterfly in measurement appears to fall exactly between two consecutive the right hand with the thorax sandwiched between the thumb millimeter lines (i.e., exactly at 0.5 mm), then, following our and forefinger and the right wings facing upward, as illustrated standard protocol, it should be rounded to the nearest EVEN in Fig. 2. A firm but gentle pressure on either side of the thorax whole number. This method produces unbiased rounding. provides a stable platform for ruler placement while forcing the When both forewing tips are damaged, the forewing cell wings into the closed position, greatly reducing the likelihood may be measured instead (this should be noted). The forewing of escape when measuring live butterflies. cell is enclosed by a series of wing veins and is forked, as Figs. 3 and 4 show the precise landmarks and proper line of outlined in yellow on Fig. 3. The cell length is measured from forewing measurement. To measure the right forewing, locate the same base landmark used to measure the total forewing the single white spot on the forewing at the forewing-thorax length (from the base of the green arrow in Figs. 3 and 4). The junction. This white spot, magnified and labeled as white spot distal landmark is defined as the intersection of the two wing #1 in Fig. 4, can be easily differentiated from the several white veins that create the tip of the distal fork that sits farthest from spots on the thorax by gently grasping the forewing along the the apical (front) margin of the wing (noted by the left tip of the leading margin and rotating the wing slightly upward toward green arrow in Fig. 3). Black wing scales surrounding the wing the head of the butterfly. The correct spot moves with the veins make this point difficult to locate precisely, thus requiring wing. Be sure NOT to include any part of the thorax in the both good lighting and practice. When possible, we recommend measurement. Place a transparent ruler so that the face of the observing the wing under a large self-standing magnifier for ruler lies against the wing surface (the measurement is read this purpose. The green arrow in Fig. 3 marks the proper line through the backside of the ruler). The 1 cm rule line should be of measurement. Cell measurements should be taken to the carefully aligned over the side of the white spot that is closest nearest 0.5 mm in order to use our regression equation (see to the thorax. This ensures that the entire spot is included in Results) to estimate total wing length to the nearest 1 mm. the measurement (i.e., do NOT measure from the center of the spot). We suggest carefully placing a piece of masking tape at METHODS the 1 cm mark of the ruler to make it easier to align this mark with the base landmark on the wing. Once the basal point is set, Collecting and handling the butterflies gently rotate the ruler’s leading edge across the margin of the A total of 56 wild adult monarchs were netted at Newport State apex (wing tip) until the maximum length is located. Pivoting Park, Door Co, WI, on 16 Jun 2009. They were immediately the ruler on the basal landmark while gently pressing it against placed individually in glassine envelopes and stored with ice the wing surface requires considerable dexterity with the left packs, and the butterflies were killed within three hours by hand. Be sure not to press so hard that the wing surface bows. placing them in a standard cooler containing dry ice. We refer This pivoting technique should be practiced until the observer to ‘fresh’ butterflies as either alive or dead but not yet dried for obtains the same values in repeated measurements. storage or chemical analyses. In this study, all fresh butterflies A clear ruler must be used for two reasons. First, it is were measured dead but un-dried. They were shipped on ice important to stabilize the ruler by laying it across the forewing among the observers and then stored inside a storage bag in a surface. This helps maintain the 1 cm mark at the precise base freezer. For dry measurements, butterflies were dried in a forced position while rotating the leading edge of the ruler along the draft oven at 60˚C for 16 hours, the typical drying regime used wing tip and helps keep live butterflies immobilized. The ruler for lipid analysis (Brower et al. 2006). Dried butterflies were should not be held above the front margin of the wing because shipped and stored in their original glassine envelopes inside a it cannot be braced from this position and butterflies can more large plastic storage bag with desiccant. easily escape. Secondly, when the right forewing is facing Whether measured fresh or dry, butterflies were removed from up, the wing tip from which the measurement is taken faces their respective storage containers in batches of five, measured, to the left (see Fig. 1). Thus, the measurement must be read and returned immediately. Following the measurement protocol from the right to the left. Because rulers read from left to right, described in the section above, all measurements were taken the only way to achieve both proper ruler stabilization and the with wings intact on the body. With the exception of Question 2, ability to take the measurement from right to left, is to read the which addresses differences in measurement among observers, measurement through the backside of a clear ruler. Therefore, a single experienced observer (TVH) took all measurements to the front surface of the ruler should be placed against the right avoid introducing inter-observer differences. All measurements 44 TROP. LEPID. RES., 22(1): 42-52, 2012 VAN HOOK ET AL.: Standardized measurement of monarch wings sex and observer as fixed factors on the average of the three repeated measurements for each butterfly. Averaging the replicate values reduces measurement error for assessing the significance of the main factors (Yezerinac et al. 1992). Technical error of measurement (T.E.M.), an estimate of absolute measurement error expressed in the units of measurement (Mueller & Martorell 1988), was used to quantify and compare intra-observer and inter-observer variation in measurement. Intra-observer T.E.M. was calculated as the square root of the variation of the first two repeated measurements of individual butterflies averaged across all butterflies (Mueller & Martorell 1988 and references within). Only the first two measurements were used so that intra-observer error could be compared directly to inter-observer error. Inter-observer T.E.M. was calculated using the first set of measurements taken by each of our two observers (see WHO Multicentre Growth Reference Study Group 2006). A variance ratio test was used to compare Fig. 1. Monarch in the position from which wing length measurements are taken: wings together, right wing facing up, and butterfly facing to the right. the variances in the measurements of the two observers (Zar 1996). Because the impact of measurement error depends on how much of the overall phenotypic variation in wing length is introduced by the process of measurement, we also report percent error (percentage of total phenotypic variation in wing length due to T.E.M.). We also report the reliability coefficient, R (R= 1 – (variance due to T.E.M.)). It represents the proportion of total variance that is due to true variation in forewing length. 3. Does wing length decrease due to water loss from drying? To better detect changes in wing length due to drying, we altered the standard measurement protocol; ruler measurements were taken to the nearest 0.5 mm instead of the nearest 1.0 mm. The right forewing of 31 females and 24 males were measured Fig. 2. Proper hand and ruler positioning for taking standardized right forewing length measurements. The thorax is sandwiched between the right three times fresh and three times after drying the butterflies, for forefingers and the ruler, which is braced by the right thumb. The left hand a total of 330 measurements. We analyzed the averages of the is used to support the wing tip while rotating the ruler to determine the replicate measures using a model 1 ANOVA with water status maximum length. The face of the ruler is placed against the wing surface (fresh vs. dry) and sex as fixed factors. In addition, to assess so that the measurement must be taken through the backside of the ruler (numbers are facing backwards). All standardized wing length measurements whether there was an influence of butterfly size on the amount are taken from this position. The wing length is measured as the straight-line of shrinkage, we regressed the difference between fresh and dry distance between the two red arrows. In this example, the total forewing wing length against fresh length. length is 53 mm (63 mm minus 1 cm because the measurement is always taken from 1 cm ruler mark rather than 0). Proper hand and ruler positioning and measurement protocol are described in the text. 4. Do ruler measurements to the nearest 1 mm differ from caliper measurements to the nearest 0.1 mm? Because we found no effect of sex on measurement, we measured only females for this analysis. Two repeated measures were taken independently, i.e., without knowledge of previous of the right forewing length were made from 31 dried females measurements. Data were analyzed using PASW Statistics 18 using a ruler to the nearest 1.0 mm and electronic digital (PASW 2010). calipers (Mitutoyo Digimatic 150 mm/6 in) to the nearest 0.1 mm, for a total of 124 measurements. We compared the effect 1. Should males and females be analyzed separately?, and of instrument on wing length measurement using a paired t-test 2. Is wing length affected by who makes the measurements? on the averages of the two wing measurements. T.E.M. was Questions one and two were answered using the same calculated from the two repeated measures to compare intra- data set. For these two questions only, the left forewing was observer variability in re-measures when measurements were measured rather than the right forewing as specified in the taken with a ruler compared to calipers. standard protocol. Two observers measured 49 fresh butterflies, 23 males and 26 females, at 1 mm precision three times each, 5. Can forewing cell length be used to estimate total for a total of 294 measurements. To test for the effect of sex forewing length when neither forewing can be measured and observer on wing length, we ran a model I ANOVA with due to wing tip fraying or damage? VAN HOOK ET AL.: Standardized measurement of monarch wings TROP. LEPID. RES., 22(1): 42-52, 2012 45 Ruler measurements were taken to the nearest 0.5 mm instead of to the nearest 1.0 mm for this question. Two repeated measures were taken from both the right forewing and forewing cell of 31 dried females, for a total of 124 measurements. Using the mean of the two measurements from each butterfly, total right forewing length and right forewing cell length were analyzed by linear regression. RESULTS 1. Should males and females be analyzed separately? Male forewing length was on average significantly greater than female wing length (mean = 52.22 mm, s.d. = 1.40 mm for males; mean = 50.95 mm, s.d. = 1.82 mm for females; ANOVA, F = 15.090, p < 0.001) (Table 2). Therefore, wing length 1,94 should be analyzed separately for males and females. 2. Is wing length affected by who makes the Fig. 3. Correct lines of wing measurements: Total forewing length is measured as the longest straight-line distance from the wing base to the measurements? wing tip (green + red arrows). The proper line of measurement depends on Based on the averages of the three repeated measurements, the wing tip shape but approximately bisects the forewing cell (outlined in the two observers measured differently (F = 4.878, p = yellow). Forewing cell length (green arrow only) is measured from the wing 1,94 0.030; see Table 2), while there was no significant interaction base to the tip of the lower prong of the forewing cell, delineated by the wing veins (noted by yellow outline). Notice that both measurements start at the far between observer and the sex of the butterfly (F < 0.001, p = 1,94 right (as pictured) edge of the white spot at the base of the forewing. See Fig. 0.992). The two observers also differed in the variance of their 4 for an enlarged view of the base landmarks used for both measurements. repeated measurements (F = 1.823, p < 0.05), a difference 48,48 reflected in the calculation of their T.E.M. values (observer 1 = 0.23 mm; observer 2 = 0.40 mm; see Table 2). Intra-observer measurement error ranged between 2 and 6% of total population variance for our two observers. Thus, between 94 and 98% of the total variance is true variation in wing length (coefficient of reliability, R = 0.94-0.98). Inter-observer T.E.M. was 0.61 mm, so variance due to differences between our two observers was seven times larger than intra-observer 1 variance and two times larger than intra-observer 2 variance. In summary, observers differed in the variability of their repeated measurements and recorded significantly different wing lengths for the same butterflies. Butterfly sex did not affect observer differences in measurement. Finally, measurements taken by two observers varied much more than those taken by a single observer. 3. Does wing length decrease due to water loss from drying? The mean wing length decreased from 51.78 in the fresh state to 51.42 mm when dried (F = 11.249, p = 0.001), representing 1,106 an average decrease of 0.36 + 0.05 mm due to shrinkage upon drying. This shrinkage is very small, representing 0.7% + 0.1% (mean and 95% C.I.) of the fresh wing length. The amount Fig. 4. Magnification of Fig. 3. Both monarch forewing length and forewing of shrinkage did not differ between the sexes (F = 0.031, cell length are measured from the same base landmark (white spot #1). It p = 0.860), nor did shrinkage differ with wing le1,n10g6th (fresh- is important to first differentiate the forewing white spot from the nearby dry difference regressed on wing length, F = 0.001, n.s.). white spot on the thorax (spot # 2) and the 3 white spots on base of the hind 1,53 wing (spot #s 3, 4, and 5). To locate the correct spot, rotate the forewing Although the amount of shrinkage due to drying is less than slightly toward the head while holding the butterfly at the base of the hind the precision level (0.5 mm) used to measure the butterflies, wing. Only the spot on the base of the forewing will move. It is important to combining fresh and dry wing lengths would increase variance include the entire width of the landmark spot in the measurement as indicated. and decrease the sensitivity of the analysis. The right end of the green line of measurement (noted by green arrow) marks the exact location from which both the forewing length and forewing cell lengths originate. (The yellow lines delineate the forewing cell as shown in 4. Do ruler measurements to the nearest 1 mm differ Fig. 3 and are not used as a landmark; see text for further explanation). from caliper measurements to the nearest 0.1 mm? 46 TROP. LEPID. RES., 22(1): 42-52, 2012 VAN HOOK ET AL.: Standardized measurement of monarch wings Table 1. A non-exhaustive literature survey of measurements of monarch wing size. A dash (-) means that the information was not reported. Device: R = ruler, D = digital, C = calipers, O = optical device. Side: R = right, L = left, AV = average of left and right. Precision: Precision level of measurement (mm). Landmarks: a = dorsal surface from the proximal costal forewing corner to the most distant point in the wing apex, a* = some measurements were taken using a special optical device without disturbing museum specimens, b = ventral forewing length measured from the wing attachment to the apex of wing, c = white spot on forewing base to the apex of wing, d = white spot on thorax to the apex of wing, e = thorax to longest extension, f = distal tip of left forewing to distal tip of right forewing, HW = hind wing measured, Area = area measured. Comparison: 1 = sexes, 2 = correlated physical traits, 3 = correlated behaviors, 4 = larval rearing conditions, 5 = differences across overwintering season, 6 = population comparisons, 7 = generation comparisons, 8 = differences across years at single location, 9 = differences across fall migration phase at single location, 10 = other comparison. We assume a single observer made the measurements when a paper had a single author. The table shows extreme heterogeneity in measurement protocols published between 1945 and 2010. Land- Multiple Source Device Side Precision marks Observers Comparison Alonso et al. 1997 R R 0.5 b No 2,3,5 Altizer 2001 C - - b No 10 Alitzer & Oberhauser 1999 C - - b - 2 Altizer et al. 1999 C - - b - 10 Alonso et al. 1997 R R 0.5 b No 2,3,5 Altizer & Davis 2010 D R - b - 1,2,6,5,10 Altizer et al. 2004 - - - b - 10 Arango 1996 C 0.05 c No/Yes 2,4,6,7,8 Beall 1946 - R 1 a No 1,2,6,7,8,9,5,10 Beall & Williams 1945 O + R R 1 a* Yes 6,9 Borland et al. 2004 - - - b Yes 1,2,8,9,10 Bradley & Altizer 2005 D - - Area - 10 Brindza et al. 2008 C L 0.1 b No 1,6,8,9 Brower et al. 2006 R - 0.5 - - 8 Brown & Chippendale 1974 - - - - - 2 Calvert & Lawton 1993 - - - - - 1,2,5,9 Davis 2009 D - - Area No 2 Davis et al. 2007 D AV . Area - 2,3 Dively et al. 2004 - . - f - 4 Dockx 2002, 2007 - R - c No 1,4,6 Eanes 1978 - - - - No 2 Frey et al. 1998 - AV 1 b - 1,2,3,10 Gibo & McCurdy 1993 - R 0.5 - - 9 Herman 1988 R - 0.5 HW No 1,7 Herman et al. 1989 R - 0.5 HW - 1,5,6,7 James 1984 - R - - No 1,7,8 Jesse & Obrycki 2000, 2004 - R - d - 4 Knight 1998 R R 0.5 c No 1,6 Lavoie & Oberhasuer 2004 C - 0.1 b - 4 Leong et al. 1993 R R 1 - - 1,2,3 Leong et al. 1995 R - 1 - - 3,5 Levine et al. 2003 - - 1 e - 5 Lindey & Altizer 2009 D - . Area - 4 Malcolm et al. 1989 - R - - - 1,4 Oberhauser 2004 - - 0.1 - No 2 Oberhauser & Frey 1999 - - - - - 3 Solensky & Oberhauser 2004 - - - - - 3 Solensky & Oberhauser 2009 - - - Area - 2,3 Tuskes & Brower 1978 - R - - - 1,5 Van Hook 1993, 1996 - - 0.5 b No 1,2,3,4,5 York & Oberhauser 2002 - - - b - 4 VAN HOOK ET AL.: Standardized measurement of monarch wings TROP. LEPID. RES., 22(1): 42-52, 2012 47 Measurements taken with a ruler to the nearest 1 mm and Table 2. Summary statistics for forewing length measurements of our Door with calipers to the nearest 0.1 mm gave similar descriptive County, WI sample. Sex and observer effects are shown. All measurements results; the overall mean right forewing lengths, calculated from were taken with a ruler to the nearest 1 mm. Mean and SD are based on 2 repeated measurements for each butterfly. Technical error of measurement the means of two repeated measures for all 31 butterflies, were (TEM) and percent measurement error (%ME) are based on the variance of 2 50.98 mm at the 0.1 mm level of precision and 51.06 mm at the repeated measurements for each observer (intra-observer) and the variance of 1.0 mm level (Table 3). Based on a paired comparison of ruler the 1st measurement for 2 observers (inter-observer). _____________________________________________________________________________ and caliper averages for each butterfly, there was no significant Intra-Observer Inter-Observer effect of measurement device on forewing length measurement Observer 1 Observer 2 Observer 1 vs 2 (t = 1.669, p = 0.106). The T.E.M. for ruler measurements was ________________________ _______________________ ____________ 31 Mean SD TEM % ME Mean SD TEM % ME TEM % ME 0.25 mm compared to 0.17 mm caliper measurements (see Table _____________________________________________________________________________ Males 52.58 1.36 0.21 2.3 51.86 1.38 0.39 7.7 0.66 21.0 3). This difference in variability of repeated measurements Females 51.31 1.80 0.24 1.8 50.59 1.80 0.42 5.6 0.57 10.4 was not statistically significant (variance ratio test: F = Both 51.91 1.71 0.23 1.7 51.18 1.75 0.40 5.5 0.61 12.8 30,30 _____________________________________________________________________________ 1.0220, p > 0.05; Table 3), and measurement error using either Mean comparison of the sexes, p < 0.001 device represented less than 2% of the total variance. Thus, we Mean comparison of the observers, p = 0.030 Variance differs between observers, p < 0.05 conclude that measurement of wing length by ruler with 1.0 mm precision is sufficient for most studies (but see discussion). Table 3. The effect of measurement device on the forewing descriptive statistics, Post-hoc statistical power analysis (Soper 2011) of our technical error of measurement (TEM), and percent of total variance due to repeated forewing length data measured with a ruler to the nearest 1 mm measures (% ME) for 31 dried females measured by a single observer with calipers at 0.1 showed that sample sizes of 30 individuals (15 in each group Tmambl aen 3d. aT rhuele er fafte 1c.t0 o mf mm.e aMseuarne manedn St Dde avriec bea osend t ohne tfhoer eawveirnagge ds eosfc trhiep ttiwvoe rsetpaetiastteidc s, for 2 factor analyses) provide sufficient power to detect large technical error of measurement (TEM), and percent of total variance due to measurements for each butterfly, while TEM and % ME are based on the variance of the and medium effects (power = 0.86 to 0.92 and 0.48 to 0.61 for repeated measures (% ME) for 31 dried females measured by a single observer two repeated measurements. with calipers at 0.1 mm and a ruler at 1.0 mm. Mean and SD are based on the one-tail and two-tail hypotheses for large and medium effects, a_v_e_r_a_g_e_s_ _o_f _t_h_e_ t_w__o_ r_e_p_e_a_t_e_d_ m__e_a_s_u_r_e_m__e_n_ts_ _fo__r _e_a_c_h_ _b_u_tt_e_r_fl_y_,_ w__h_il_e_ _T_E_M__ _a_n_d_ __ respectively, at alpha = 0.05). Much larger sample sizes would % ME are based on the variance of the two repeated measurements. be necessary to detect very small forewing differences among Measurement Mean a SD TEM % ME b populations. Procedure ________________________________________________________________________ 5. Can forewing cell length be used to estimate total Calipers at 0.1 mm 50.98 1.91 0.17 0.8 forewing length when neither forewing can be measured due to wing tip fraying or damage? Ruler at 1.0 mm 51.07 1.89 0.25 1.8 Using the mean of the two repeated measures for the right ________________________________________________________________________ forewing length and right forewing cell length, we found these a Comparison of means, p = 0.106, n.s. measurements to be correlated significantly (R2 = 0.75, p < b Variance ratio test of variance between measurement devices: p > 0.50, n.s. 0.01, n = 62; Fig. 5). We conclude that the forewing cell length, based on the mean of two repeated measures to the nearest 0.5 
 mm, can be used as an adequate alternative measure to estimate forewing length when neither right nor left forewing can be measured directly. The regression equation was: forewing Fig. 5. Regression of right forewing length vs. right forewing cell length: length in mm = 1.508 (forewing cell length mm) + 10.744 mm. y = 1.508x + 10.744 (r2 = 0.747). A single observer measured total forewing length and cell length two times each to the nearest 0.5 mm using a ruler, DISCUSSION and all measurements were independent. The averages of two repeated measurements were used for the regression. 56 Sex We found that male wing length was on average 1.27 mm 55 greater than female wing length. Our result was consistent with 54 many (e.g. Beall 1946; Herman 1988; Calvert & Lawton 1993; ) m53 Van Hook 1993, 1996; Knight 1998; Oberhauser & Frey 1999; m Borland et al. 2004; Dockx 2002, 2007; Brindza et al. 2008; (52 h Altizer & Davis 2010) but not all (e.g. Tuskes & Brower 1978; gt51 n James 1984; Frey et al. 1998; Leong et al. 1993; Knight 1998; e Malcolm et al. 1989; Dockx 2002, 2007) findings based on g l50 n various monarch populations. Although our literature review wi49 showed that the sexes are usually considered separately (Table 48 1), this was not always the case (e.g. James 1984 and some 47 comparisons in Brindza et al. 2008). The standard protocol therefore includes the provision that the sex always be recorded 46 and that males and females be considered independently in 24 25 26 27 28 29 30 analyses of wing length. cell length (mm) 48 TROP. LEPID. RES., 22(1): 42-52, 2012 VAN HOOK ET AL.: Standardized measurement of monarch wings Intra- and Inter-observer error using the Standard identify subtle differences and then work together until everyone Measurement Protocol is satisfied that they have minimized those differences. When Measurement error is inversely related to quality of the observers change through time, as is common in monitoring data, and standardization of the measurement procedure is the studies, established observers should help to train the next most effective way to minimize such error (Ulijaszek & Kerr group. 1999). Using the standard protocol, our two observers both Although inter-observer error cannot be factored out during showed ‘excellent’ agreement in their repeated measurements statistical analyses when single measurements are used (Harris (measurement errors less than 10%: see Stokes 1985; Perini et & Smith 2009), we recommend against taking the mean of al. 2005; WHO Multicentre Growth Reference Study Group repeated measurements or dramatically increasing sample 2006). However, they differed in their measurement errors: 2% sizes because they are not necessary. This is because repeated and 6% based on the combined sex sample. Consistent with measurements do not reduce measurement differences among the literature, the observer who was most experienced using observers, and increased sample sizes can magnify them the standard protocol showed a higher level of repeatability (Palmeirim 1998). Instead we recommend putting the time it (Gordon & Bradtmiller 1992; Yezerinac et al. 1992; Tong et al. would take to obtain duplicate measures on all butterflies into 1998; Ulijaszek & Kerr 1999; Kania 2004; references in Perini proper training, practice, and calibration before collecting data. et al. 2005). We conclude that with a single observer, single The effects of any remaining differences in measurement can be measurements with a ruler to the nearest 1 mm are adequate, mitigated when multiple observers are used within a single study but practice using the standard protocol before data collection by dividing each of the butterfly groups of interest among all of begins is necessary to minimize error. the observers. For example, if the question is whether wing The measurements of our observers were consistently length differs between coastal and inland migratory monarchs, different. This difference, combined with differences in the each observer should measure both coastal and inland. size of their measurement errors, increased inter-observer We encourage monitoring groups to teach the standard error compared to intra-observer error. Larger variation protocol and calibration methods. However, calibration among in measurements among observers compared to repeated sites and across time will often be unfeasible. We therefore measurements by a single observer is a common finding even make the following additional recommendations. First, the when the measurement protocol is standardized (Yuan et al. names and handedness of observers should be reported (see 2004; Geeta et al. 2009; Harris & Smith 2009; Muñoz-Muñoz Helm & Albrecht 2000), and raw data should be archived so & Perpiñán 2010; but see Palmeirim 1998; Ulijaszek & Kerr they can be accessed for statistical comparisons. Second, after 1999). training and calibrating, a sub-sample of at least 20 butterflies The effect of any increased variation caused by using should be measured independently twice by all observers. Intra- multiple observers depends on the total variance in wing length observer and inter-observer T.E.M. should be reported, and of the population measured. Our data illustrate this: T.E.M. reassessments should be made at regular intervals (Yezerinac was similar for females and males. However, these errors et al. 1992; see Mueller & Martorell 1988 for methodology to varied over two-fold in total length variation (from 10 to 21% determine T.E.M.). Third, if inter-observer error is greater than for inter-observer error, respectively, Table 2). Even though 10% of total variation in wing length attempts should be made measurement error would be larger for males than females, the to further reduce differences in measurement technique among coefficients of reliability, 0.90 and 0.79 for females and males observers. When inter-observer error is 10% or less or after respectively, still indicate from ‘excellent’ to ‘substantial’ attempts have been made to reduce observer differences, the agreement between our measurers (WHO Multicentre Growth first of the paired measurements used to calculate observer error Reference Study Group 2006). However, because percent can be included in the overall data set to minimize resources measurement error was consistently and substantially smaller and time needed to document reliability. for single observers compared to multiple observers (Table There is no way to assess true bias in measurement, and 2), we recommend that a single measurer be used whenever it would be impractical to compare relative bias across sites possible to minimize relative bias and variation in measurements because the same butterflies cannot be used. However, relative (Measey et al. 2003; Perini et al. 2005). bias can be reduced through calibration (Kouchi et al. 1999) Our estimates of relative bias and inter-observer variation and is included in inter-observer error. Therefore, when in measurement may have been inflated because our observers inter-observer error remains high or unknown after training, did not calibrate their measurements on the same butterflies calibration, and practice, small significant differences based on prior to the study. The literature suggests that both variation samples measured by different observers should be viewed with in measurement and relative bias could be reduced perhaps caution (Palmeirim 1998). well beyond those found in this study if observers calibrate their measurements after adequate training using the standard Fresh vs. dry butterfly measurements protocol (see Gordon & Bradtmiller 1992; Kouchi et al. 1999; Measured wing length of fresh monarchs was on average WHO Multicentre Growth Reference Study Group 2006). To 0.36 mm greater than measurements after they were dried, with calibrate measurements among observers, all measurers in no significant difference by sex in the amount of shrinkage. a single study or at a single monitoring site and date should Shrinkage was small compared to the 1 mm precision level work as a group to compare their measurement techniques and used in the standard protocol and represented less than 1% of resulting measurements on a single sub-set of butterflies to the total fresh wing length. However, because all sources of VAN HOOK ET AL.: Standardized measurement of monarch wings TROP. LEPID. RES., 22(1): 42-52, 2012 49 variation among measurements are cumulative, we recommend to detect large and medium effects with single observers. We that observers measure wing length before drying when recommend using power analysis to assure adequate sample possible and note this. If fresh wing lengths must be compared sizes, especially when small differences in wing length are to dried wing lengths, we suggest first estimating wet wing important. length from dry wing length by adding 0.7% to all dry wing When different populations are measured by different lengths and interpreting small significant differences in wing observers, the results should be viewed with caution because lengths cautiously. any significant differences may be a product of the observers Water evaporation during long term freezing can presumably rather than the populations (Palmeirim 1998). When fewer than decrease wing length slightly in the same way as drying them 30 butterflies are measured, two replicate ruler measures should for chemical analyses. Since the degree of drying and thereby be averaged to decrease measurement error. Furthermore, we shrinkage will vary, we recommend first drying such specimens recommend that calipers be used in studies whose goal is to for 16 hrs at 60 degree C, taking ‘dry’ wing length measurements, detect very small differences (e.g., when assessing right-left and then adding 0.7% to convert wing length measurements wing asymmetry). Our results indicate that measurements of to fresh equivalents. We know of no empirical evidence or forewing length taken with calipers could be rounded to the theoretical reason that wing length should differ between live nearest 1 mm for comparison to ruler measurements. and just frozen specimens. However, measurements on live Computer-based measurements allow higher levels of specimens may be more variable (see below) because they are measurement precision, may provide better intra-observer more difficult to measure. repeatability (see Muñoz-Muñoz & Perpiñán 2010), and may When museum specimens are measured, not only have the be preferred under some circumstances (e.g. Davis et al. 2007; wing lengths shrunk due to water loss, but the standard protocol Altizer & Davis 2010). However, this method is not suited for must also be altered. Such measurements would not be easy general use because it increases handling time, (thereby adding to standardize across all observers because the dorsal surface stress to live animals), requires special computer programs, and has no white spots to serve as base landmarks; it is harder increases inter-observer biases because of calibration difficulties to differentiate thorax from wing on spread specimens; and (see Muñoz-Muñoz & Perpiñán 2010). Because software specimens are so fragile that they usually cannot be touched. differences make standardization of computer analyses across When wing lengths of museum specimens must be compared observers difficult, we suggest that ruler measurements be taken to those taken using the standard measurement protocol, we in addition to image-based measurements when measurements recommend using the following procedure. First, measure are to be compared to those taken by the standard protocol. a set of fresh monarchs using the standard protocol. Then, after pinning and drying the butterflies at 60 degrees C for Side of body measured and total forewing wing length 16 hours, measure them again to create a regression of fresh estimation vs. pinned (dried and spread) wing lengths. Finally, using the When left and right wing lengths are combined, small errors same measurement protocol and the same observer, measure can result from true asymmetry in right and left wing lengths the museum sample of interest and convert the measurements to (Palmer & Strobeck 1986) or from biases resulting from how fresh equivalents using the established regression. the two sides are measured (Arango 1996; Helm & Albrecht 2000). Therefore, following Beall & Williams (1945) and Measurement device and sample size the prevailing trend in the existing monarch literature (Table We based the standard protocol on ruler measurements 1), the standard protocol restricts measurement to the right because the ruler is the most common measurement device forewing. However, the difference in right-left wing length is reported in the literature, and its widespread use maximizes small compared to the 1 mm precision level used in the standard comparability of wing length measurements across long- protocol (Arango 1996). It is important to measure all butterflies term monitoring programs, citizen scientists, and researchers. in a particular sample, even when the right forewing cannot be Rulers are also easier to use and more affordable than calipers. measured, because wing length may be correlated with other Ruler measurements taken to the nearest 1 mm did not differ factors of interest, such as behavior, age, population source, significantly from those taken with calipers to the nearest etc. (Van Hook 1993; Oberhauser & Frey 1999; Oberhauser et 0.1 mm, nor was measurement error significantly increased. al. 2009). The standard protocol therefore dictates that the left Furthermore, error associated with ruler measurements (less forewing be measured when the right forewing is not intact, and than 2%) was considerably smaller than error introduced by this should be noted. different observers (10-21%). When neither forewing can be measured, the right forewing A researcher’s choice of sample size depends on the question cell length should be measured twice to the nearest 0.5 mm. The at hand, population variation, the effect size of interest, mean can then be used to estimate total forewing length based on confidence level needed, and measurement error. However, our regression equation. Estimates from this regression are not measurement error is included in the overall population variance as accurate as direct measurements, of course, so it depends on when single measurements are taken, causing a slight loss in the questions being asked whether regressed estimates should statistical power (Yezerinac et al. 1992). This can be countered be used in a study. (Measurers should note when estimates are by increasing sample size. Based on post-hoc statistical power used on their data sheets, and the percentage of estimated wing analyses, we found that using sample size of at least 30 (15 in lengths should be reported.) Because the forewing cell length is each sample when two population are compared) was sufficient more challenging to measure, excellent lighting conditions are 50 TROP. LEPID. RES., 22(1): 42-52, 2012 VAN HOOK ET AL.: Standardized measurement of monarch wings important, and magnification may be needed. the power of statistical tests. Although total hind wing length may provide a better We emphasize the importance of proper training and regression equation for total wing length compared to forewing sufficient practice before collecting data. A single observer cell length, we recommend against using the hind wing length should be used when possible, but when multiple observers for two reasons. First, when the forewings are frayed or are necessary, inter-observer repeatability may be increased damaged, the hind wings are also often similarly worn (Leong beyond levels found in our study if all measurers work together et al. 1993). Secondly, the hind wing measurement is more to calibrate their measurements. Because observer error difficult to standardize than forewing cell length because the varies and cannot be factored out when single measurements wing margin is feathery and scalloped (see Fig. 3) and requires are used, the names, handedness, and intra-observer and using a different base landmark. inter-observer T.E.M. should be reported, while the raw data should be archived. Observer bias can be mitigated by sub- Measurements taken on live monarchs dividing samples among the observers so that observers and Measurement error may be higher when measurements are factors of interest do not co-vary. When this is not possible, taken on live compared to dead animals because of the difficulty small significant differences should be viewed with caution of holding them (Blackwell et al. 2006; personal experience). due to unavoidable measurement differences among observers Therefore, when monarchs are being preserved, forewing (relative bias). length should be measured after they are killed. Measurements The sex should be recorded along with forewing length and should be taken as soon as possible after collection to avoid males and females analyzed separately. All of the butterflies possible shrinkage in wing length associated with water loss in a sample should be measured; use the left forewing when during long-term freezer storage (see above). However, one of the right forewing apical tip is frayed or missing and use the the merits of using wing length to answer biological questions, means of right forewing cell lengths measured twice to the especially relevant when citizen scientists gather the data, is nearest 0.5 mm and our regression equation to estimate the that it does not require killing the butterflies. When monarchs total wing length when neither forewing can be measured. are measured alive, Donham & Taylor (1996) suggest placing Measurements should be taken before drying dead specimens, them inside clear envelopes to measure them, but we do not and live butterflies should be cooled before measuring. know how this method might influence variability or relative We hope that monarch researchers and monitoring groups bias of measurements. Instead, we recommend cooling the will adopt this standardized measurement protocol. Its butterflies by placing them into glassine collection envelopes widespread use will increase the comparability and usefulness that should be kept inside a sealed plastic bag stored in a cooler of monarch wing length data. The general methods we used or until they are removed for measurement. Cooling keeps the for standardization can be applied to all lepidopteran species to butterflies quiet and prevents them from damaging their wings increase repeatability in wing length measurement. and using energy reserves. If each butterfly must be measured immediately and ACKNOWLEDGEMENTS then released, we suggest practicing the proper technique of holding the ruler over the body for taking measurements while We are grateful to Michelle Hefty for permission to sample firmly holding the thorax between the thumb and forefinger. monarchs in Newport State Park, WI and Ray Sullivan for Together, these techniques stabilize the ruler for more precise field assistance. We also thank Zachariah Dietz, who provided measurements while keeping the butterfly quiescent. Any important statistical guidance, and Gwen Kirschke, who increase in measurement error due to measuring live specimens assisted with the photographs. Barbara Cockrell generously likely increases variation rather than bias in measurement and assisted us by obtaining references. Two anonymous referees therefore should not limit statistical analyses of the data as long provided helpful comments that improved the manuscript. This as sample sizes are > 30 individuals. work was supported by the October Hill Foundation (LPB) and by the Christian A. Johnson Fund of Hamilton College (EHW). Summary of the standard forewing measurement protocol REFERENCES CITED We describe a standard protocol for forewing length measurements based on well-defined landmarks and specific Alonso-Mejía, A., E. Rendon-Salinas, E. Montesinos-Patiño & L. P. Brower methods for handling the butterfly and measurement 1997. Use of lipid reserves by monarch butterflies overwintering in Mexico: Implications for conservation. Ecological Applications device. This protocol provided ‘excellent’ repeatability of 7(3):934-947. measurements when a single observer was used and from Altizer , S. M. ‘excellent’ to ‘substantial’ repeatability with multiple observers 2001. Migratory behaviour and host-parasite co-evolution in natural (see WHO Multicentre Growth Reference Study Group 2006). populations of monarch butterflies infected with a protozoan parasite. Evolutionary Ecology 3(3):611-632. Single ruler measurements to 1 mm should generally provide Altizer, S. M. & A. K. Davis sufficient statistical power as long as sample sizes are robust 2010. Populations of monarch butterflies with different migratory (> 30) and a single observer is used. However, after assessing behaviors show divergence in wing morphology. Evolution T.E.M., power analysis should be used to estimate appropriate 64(4):1018-1028. Altizer, S. M. & K. Oberhauser sample size. When sample size is severely restricted, calipers 1999. Effects of the protozoan parasite Ophryocystis elektroscirrha on or the mean of two repeated measures can be used to increase the fitness of monarch butterflies (Danaus plexippus). Journal of VAN HOOK ET AL.: Standardized measurement of monarch wings TROP. LEPID. RES., 22(1): 42-52, 2012 51 Invertebrate Pathology 74(1):76-88. Cuba. Biological Journal of the Linnean Society 92:605-616. Altizer, S. M., K. Oberhauser & K. A. Geurts Donham, P. & O. R. Taylor 2004. Transmission of the protozoan parasite Ophryocystis elektroscirrha 1996. The Great Dragonfly Challenge: Measuring Monarchs. Dragonfly in monarch butterfly populations: Implications for prevalence and Magazine Sept/Oct (1):21-23. population-level impact. pp. 203-218. In Oberhauser, K. S. & M. J. Eanes, W. F. Solensky (Eds.), The Monarch Butterfly: Biology and Conservation. 1978. Morphological variance and enzyme heterozygosity in the monarch Cornell University Press, Ithaca, New York. butterfly. Nature 276:263-264. Arango, N. Francis, R. I. C.C. & R. H. Mattlin 1996. Stabilizing selection in migratory butterflies: A comparative study 1986. A possible pitfall in the morphometric application of discriminate of queen and monarch butterflies. Unpublished Masters Thesis, analysis: measurement bias. Marine Biology 93:311-313. University of Florida, Gainesville, Florida, i-viii + 1-70. Frey, D., K. L. H. Leong & E. Peffer, R. K. Smidt & K. Oberhauser Bailey, R. C. & J. Byrnes 1998. Mate pairing patterns on monarch butterflies (Danaus plexippus 1990. A new, old method for assessing measurement error in both L.) at a California overwintering site. Journal of the Lepidopterists’ univariate and multivariate morphometric studies. Systematic Society 52(1):84-97. Zoology 39(2):124-130. Geeta, A. H. Jamaiyah, N. M. Safiza, G. L. Khor, C. C. Kee, A. Z. Ahmad, Beall, G. S. Suzana, R. Rahmah & A. Faudzi 1946. Seasonal variation in sex proportion and wing length in the migrant 2009. Reliability, technical error of measurements and validity of butterfly, Danaus plexippus L. (Lep. Danaidae). Transactions of the instruments for nutritional status assessment of adults in Malaysia. Royal Entomological Society of London 97:337-353. Singapore Medical Journal 50(10):1013-1019. Beall, G. & C. B. Williams Gibo, D. L. & D. A. McCurdy 1945. Geographical variation in the wing length of Danaus plexippus 1993. Lipid accumulation by migrating monarch butterflies (Danaus (Lep. Rhopalocera). Proceedings of the Royal Entomological plexippus L.). Canadian Journal of Zoology 71:76-82. Society of London 20:65-76. Gordon, C. C. & B. Bradtmiller Blackwell, G. L., S. M. Bassett & C. R. Dickman 1992. Interobserver error in a large scale anthropometric survey. 2006. Measurement error associated with external measurements American Journal of Human Biology 4:253-263. commonly used in small-mammal studies. Journal of Mammalogy Harris, E. F. & R. N. Smith 87(2):216-223. 2009. Accounting for measurement error: A critical but often overlooked Borland, J., C. Johnson, T. Crumpton, T. Thomas, S. Altizer & K. process. Archives of Oral Biology 54 Suppl. 1:107-117. Oberhauser Helm, B. & H. Albrecht 2004. Characteristics of fall migratory monarch butterflies, Danaus 2000. Human handedness causes directional asymmetry in avian wing plexippus, in Minnesota and Texas. pp. 97-104. In Oberhauser, length measurements. Animal Behavior 60:899-902. K.S. & M.J. Solensky, (Eds.), The Monarch Butterfly: Biology and Herman, W. S. Conservation. Cornell University Press, Ithaca, New York. 1988. Body weight and wing length changes in Minnesota populations of Bradley, C. & S. Altizer the monarch butterfly. Journal of the Lepidopterists’ Society 42:32- 2005. Parasites hinder monarch butterfly flight ability: implications for 36. disease spread in migratory hosts. Ecological Letters 8: 290-300. Herman, W. S., L. P. Brower & W. H. Calvert Brindza, L., L. B. Brower, A. K. Davis & T. Van Hook 1989. Reproductive tract development in monarch butterflies 2008. Comparative success of monarch butterfly migration to overwintering in California and Mexico. Journal of the overwintering sites in Mexico form inland and coastal sites in Lepidopterists’ Society 43(1):50-58. Virginia. Journal of the Lepidopterists’ Society 62(4):189-200. James, D. G. Brower, L. P., L. S. Fink & P. Walford 1984. Phenology of weight, moisture and energy reserves of Australian 2006. Fueling the fall migration of the monarch butterfly. Integrative and monarch butterflies, Danaus plexippus. Ecological Entomology Comparative Biology 46(6):1123-1142. 9:421-428. Brown, J. J. & G. M. Chippendale Jesse, L. C. H. & J. J. Obrycki 1974. Migration of the monarch butterfly, Danaus plexippus: energy 2000. Field deposition of Bt transgenic corn pollen: lethal effects on the sources. Journal of Insect Physiology 20:1117-1130. monarch butterfly. Oecologia 125:241-248. Calvert, W. H. & R. O. Lawton 2004. Survival of experimental cohorts of monarch larvae following 1993. Comparative phenology of variation in size, weight, and water exposure to transgenic corn pollen and anthers. pp. 69-75. In content of eastern North American monarch butterflies at five Oberhauser, K. S. & M. J. Solensky (Eds.), Monarch Butterfly: overwintering sites in Mexico. pp. 299-307. In Malcolm, S. B. & Biology and Conservation. Cornell University Press, Ithaca, New M. P. Zalucki (Eds.), Biology and Conservation of the Monarch York. Butterfly. Publications of the Natural History Museum of Los Kania, W. Angeles County, Los Angeles, California. 2004. Observer error in measurements of nestling wing length in small Davis, A. K. passerines. The Ring 26(2):79-87. 2009. Wing color of monarch butterflies (Danaus plexippus) in Eastern Knight, A. North America across life stages: Migrants are “redder” than 1998. A population study of monarch butterflies in North-Central and breeding and overwintering stages. Psyche:1-5. South Florida. Unpublished Masters Thesis, University of Florida, Davis A. K., N. Cope, A. Smith & M. J. Solensky Gainesville, Florida, viii + 125. 2007. Wing color predicts future mating success in male monarch Kouchi, M., M. Mochimaru, K. Tsuzuki & T. Yokoi butterflies. Annals of the Entomological Society of America 1999. Interobserver errors in anthropometry. Journal of Human Ergology 100(2):339-344. 28:15-24. Dively, G. P., R. Rose, M. K. Sears, R. L. Hellmich, D. E. Stanley-Horn, D. Lavoie, B. & K. S. Oberhauser D. Calvin, J. M. Russo & P. L. Anderson 2004. Compensatory feeding in Danaus plexippus (Lepidoptera: 2004. Effects on monarch butterfly larvae (Lepidoptera: Danaidae) after Nymphalidae) in response to variation in host plant quality. continuous exposure to Cryl-Ab-expressing corn during anthesis. Environmental Entomology 33(4):1062-1069. Environmental Entomology 33(4):1116-1125. Leong, K. L. H., D. Frey, D. Hamaoka & K. Honma Dockx, C. 1993. Wing damage in overwintering populations of monarch butterfly 2002. Migration of the North American Monarchs Danaus plexippus to at two California sites. Annals of the Entomological Society of Cuba. PhD Dissertation, University of Florida, Gainesville, Florida, America 86(6):728-733. x + 119. Leong, K. L. H., E. O’Brien, K. Lowerisen & M. Colleran 2007. Directional and stabilizing selection on wing size and shape in 1995. Mating activity and status of overwintering monarch butterflies migrant and resident monarch butterflies, Danaus plexippus (L.), in (Lepidoptera: Danainae) in Central California. Annals of the