| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
MOLECULAR, CELLULAR, AND DEVELOPMENTAL BIOLOGY |
* Agricultural Engineering, Technology and Food, Institute for Agricultural and Fisheries Research, 9820 Merelbeke, Belgium; Animal Husbandry and Welfare, Animal Sciences, Institute for Agricultural and Fisheries Research, 9090 Melle, Belgium; Group of Evolutionary Biology, Department of Biology, University of Antwerp, 2020 Antwerp, Belgium; and Terrestrial Ecology Unit, Department of Biology, University of Ghent, 9000 Ghent, Belgium
1 Corresponding author: annelies.vannuffel{at}ilvo.vlaanderen.be
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Key Words: developmental instability • directional asymmetry • measurement error • statistical power • trait selection
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Because FA is affected by environmental, genetic, or both, factors (such as inbreeding or directional selection), it could indicate environmental pollution or genetic deterioration (Eeva et al., 2000; Gomendio et al., 2000; Lens et al., 2002; Hoffmann and Woods, 2003). Fluctuating asymmetry has been associated with reduced performance, fecundity, and competitive ability and with increased susceptibility to parasitism and predation (Polak, 2003). Studies of FA are applicable to a broad array of disciplines and may even aid in optimizing the performance of livestock (Forkman and Corr, 1996; Møller and Manning, 2003; Garland and Freeman, 2005; Tuyttens et al., 2005).
In addition, FA may be useful in determining the suitability of rearing conditions as experienced by animals during their development (Møller et al., 1995; Tuyttens, 2003; Tuyttens et al., 2005; Knierim et al., 2007). The link between FA and animal welfare has been substantiated by empirical studies on birds. For example, FA has been reported to be positively related to leg or gait problems (Møller et al., 1999; Sanotra et al., 2001), corticosterone response to restraint (Satterlee et al., 2000), duration of tonic immobility (Møller et al., 1995; Campo et al., 2000), cannibalistic behavior (Yngvesson and Keeling, 2001; Cloutier and Newberry, 2002), heat stress (Yalçin et al., 2001; Yalçin and Siegel, 2003), inappropriate light conditions (Møller et al., 1999), reduced fecundity (Forkman and Corr, 1996; Møller and Swaddle, 1997; Nestor et al., 2000), high stocking density (Møller et al., 1995), and parasitic load (Brown and Brown, 2002; Bize et al., 2004).
Not all studies, however, indicate a consistent linkage between FA and welfare (Tuyttens, 2003; Knierim et al., 2007). Such studies may suffer from procedural deficiencies, in particular the lack of a proper protocol for measuring FA and inappropriate statistical routines (Lens et al., 2002; Klingenberg, 2003a; Knierim et al., 2007). Hence, the strength of evidence linking FA with welfare is difficult to estimate.
In published studies of poultry, FA was commonly measured by calipers in living or dead birds, whereas the use of photographs or X-ray was rare. Caliper measurements of soft tissue, however, may have low repeatability due to possible patterns of short-term changes in asymmetry (e.g., by hormonal influences; Manning et al., 2002). On the other hand, such short-term changes could contain important information about changes in the environment of an individual. However, this application has not been studied in poultry yet. Moreover, because FA is a relatively weak signal and often measures less than 1% of the trait size (Palmer, 1994), both measurements and statistical analysis require great accuracy.
First, true FA has to be distinguished from measuring error (ME). Otherwise, the level of FA will be overestimated [ME accounts for 25 up to 76% of the apparent variation between trait sides (Palmer, 1994; Swaddle, 2003)]. To estimate ME, repeated measurements are required (Møller and Swaddle, 1997) and should be conducted without memory of earlier measurements or identity of the individuals.
Second, bilateral differences should reflect true FA rather than directional asymmetry (DA) or antisymmetry (AS). Directional asymmetry occurs when there is a consistent bias toward greater development of 1 side (e.g., human lungs, avian heart). Antisymmetry occurs when the natural tendency for 1 side to develop more is variable and unpredictable [e.g., oversized signaling claw in male fiddler crab, Uca musica (Palmer and Strobeck, 1986)]. Because DA and AS do not have an a priori definable state of symmetry, questions arise about the separation of asymmetry caused by developmental instability from that which has a genetic basis (Leamy, 1984; Palmer and Strobeck, 1986; Klingenberg, 2003b). Graham et al. (1993), on the other hand, have suggested that DA, AS, or both, could reflect the effects of stress as well. Although a few empirical studies have confirmed this (Lens and Van Dongen, 1999), the use of traits showing DA or AS is often discouraged. We refer to Palmer and Strobeck (2003), Klingenberg (2003b), and Van Dongen (2006) for recent in-depth discussions and warnings. The use of mixed regression analysis (Van Dongen et al., 1999a) or a 2-way mixed effects ANOVA (Palmer and Strobeck, 1986) allows one to statistically correct for DA if present, whereas no comparable routines are currently available for correction of AS. Both models are based on the same principles and result in the same estimate of FA. Nonetheless, it is preferable not to use traits that show DA, because this correction makes 2 essential assumptions that cannot be tested explicitly: (i) the average state of asymmetry used as a correction factor reflects the optimal state for each individual (Van Dongen, 2006), and (ii) the left and right sides are equally susceptible to developmental noise, that is, that the left and right sides have the same developmental instability (Klingenberg, 2003b). The 3 different types of asymmetry (FA, DA, AS) may co-occur within a single population, which can be investigated statistically using mixture analyses (Van Dongen et al., 1999a). On the other hand, negative kurtosis may indicate AS (Palmer and Strobeck, 1986; Knierim et al., 2007). Hence, if other traits are available that do not exhibit DA or AS, these will be preferred, because they raise fewer concerns about inferences based on patterns of bilateral variation among samples (Palmer, 1994).
Third, organism-wide developmental instability is estimated more reliably when analyses integrate measurements on multiple traits. If several traits yield a concordant pattern of variation, even if very subtle, one has greater confidence that the pattern is true rather than an artifact of something peculiar to a particular trait (Zakharov et al., 1991; Palmer and Strobeck, 1992). So, it is recommended to measure multiple traits on each individual and to combine them into a suitable composite FA index (Leung et al., 2000) in which only those showing no significant correlation in signed FA are preferred. Otherwise, estimates of developmental instability are not independent, because correlated traits might be affected by similar perturbations during development (Klingenberg and Zaklan, 2000; Klingenberg, 2004). In addition, scaling may be necessary if traits are size-dependent (Leung, 1998; Swaddle, 2003).
The objective of this paper is to develop a decision protocol for the study of FA in broiler chickens. Several methods for measuring bilateral asymmetry are compared, and a decision protocol for the selection of suitable traits for estimating organism-wide FA is described. Elevated levels of ME can have profound power consequences (Van Dongen, 1999). We therefore performed a simulation study exploring power consequences of the inherent differences in degree of ME between different measuring methods and the direct effects on required sample sizes, number of repeated measurements, or both.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Measuring Methods
Bilateral asymmetry in each bird was measured using 7 different methods (roughly ordered by increasing complexity, time, and money resources necessary to prepare the traits and to conduct the measurements): (i) living broilers, (ii) intact carcasses, (iii) X-rays of wings and legs, (iv) bone weight, (v) bone dimensions, (vi) digital bone photographs (automatic), and (vii) digital bone photographs (manual). The initial choice of the measurements on traits that were presumed to be bilateral was based on the literature, experience, and the presence of obvious measuring points. Depending on the method, from 10 to 29 different types of measurements were conducted (Table 1
). Due to fragility problems of the defleshed fibula, measurements on the fibula were restricted to widths. Photographic illustrations and specific details of these measurements can be found in the appendix or on http://www.ilvo.vlaanderen.be/documents/fluctasbroilers.pdf.
|
All broilers were handled following the same sequence. First, living broilers were measured for asymmetry using digital calipers (accuracy 0.01 mm, method 1: living broilers). Traits of the head and wings were measured while the bird was standing on a table. The bird was placed on its back on the lap of the measurer to measure traits of the legs and feet. Next, the broilers were slaughtered by cervical dislocation. In the second method (intact carcasses), the same traits as measured on the living broilers and 4 additional traits were measured on the intact carcasses with a digital caliper (accuracy 0.01 mm). The head, wings, and legs were separated from the carcass and frozen in a fixed position to take comparable (between left and right sides) X-rays (46 to 48 kV, 350 mA, 0.01 s, and 1 m height) of these frozen limbs. The length and width of 9 skeletal structures were measured using a ruler (accuracy 0.5 mm, method 3: X-ray). Subsequently, 14 bones of the limbs and the head were macerated with sodium perborate tetra hydrate according to the method described by McDonald and Vaughan (1999) and weighed once on an analytic balance (accuracy 0.1 mg, method 4: bone weight). The length and the width of the bones were measured using a digital caliper (accuracy 0.01 mm, method 5: bone dimensions). Finally, 2 digital photographs were taken from the bleached macerated bones on a black background. Each photograph contained the left and right bone of a trait and a caliper (necessary for the calibration). The lengths and bone areas on these photographs were measured using image-analyzing software Optimas 6.5 [Media Cybernetics, Silver Spring, MD; method 6: digital bone photographs (automatic)]. These measurements were conducted by automatic outlining of the trait based on the color contrast between bones and background. The length measurements of the skeletal structures were also done manually [method 7: digital bone photographs (manual)]. Start and end points of the measurements were manually determined on the digital photograph in the image-analyzing program.
For each method, measurements on an individual always started with the left side of the traits followed by the right side. On each individual, all traits were measured twice to determine the ME (except for method bone weight, in which the accuracy of the analytic balance was used as ME) with a time interval of at least 1 h. On the X-rays, replicate measurements were taken from the same X-ray, whereas on the digital photographs, they were performed on new images to include all sources of ME.
Statistical Analysis
Before the analyses, outliers were detected as observations deviating more than 3 SD from the mean. They were consequently omitted from further analysis, because they were presumed erroneous. Because the number of outliers was very small (45 out of a total of 12,600 measurements), elimination was assumed not to affect the outcome of the analysis. Routinely omitting outliers in studies of FA may be incorrect, because highly asymmetrical traits are expected to occur in a small group of individuals and contain valuable information (Van Dongen et al., 2005). Including outliers inflates estimates of signal-to-noise ratios and repeatabilities, but they should be removed cautiously. It can therefore be considered to be a conservative approach to remove observations with extraordinary asymmetry levels during the trait selection process. In a final analysis, the handling of outliers should be done with great care.
Bilateral Trait Asymmetry Analysis.
For each method, traits suitable for FA analysis were selected according to the following criteria: (i) a significant level of FA, (ii) absence of DA and AS, (iii) no between-trait correlation in signed FA values, and (iv) a high signal (FA)-to-noise (ME) ratio. A high signal-to-noise ratio was not put forward as primary selection criterion, because traits with relative low signal-to-noise ratios could turn out valuable if DA and AS is absent. High power can be achieved by increasing the number of repeated measurements (Van Dongen, 1999).
Fluctuating asymmetry was estimated using a mixed regression analysis with restricted maximum likelihood parameter estimation by the following procedure. First, true FA was separated from ME by estimating variance components of the random side effect (Van Dongen et al., 1999a). The significance of FA was then tested by comparing the likelihood of the models with and without the random side effect (likelihood ratio test). Second, the presence of a DA component was tested by F-statistics after Bonferroni correction for the number of traits (with Satterthwaite’s correction for degrees of freedom). Third, unbiased estimates of individual FA (signed FA) levels were calculated as the random effect slopes of the individual regression lines (Van Dongen et al., 1999a; Van Dongen, 2000). The presence of AS was assumed when a kurtosis < –1 of the distribution of signed FA was detected. Finally, developmental independence of the trait-measurements was evaluated by testing for correlations in signed FA between traits (Van Dongen et al., 1999a; Klingenberg, 2004). Traits that showed a significant correlation in their signed FA value with a higher-ranked (according to their signal-to-noise ratio) trait were eliminated.
Repeatability and Power Analysis.
To compare the accuracy of different measuring methods, repeatabilities were estimated as intraclass correlations from 2 repeated measurements of the radius and tarsometatarsus that were measured by all methods and from the 5 traits with the highest signal-to-noise ratio after selection. To examine the effect of reduced measurement accuracy on statistical power, we estimated power for a range of degrees of ME and numbers of repeated measurements. Each power estimate was based on 1,000 simulated samples as outlined in Van Dongen (1999).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
). The procedure for trait selection is illustrated for the intact carcasses method in Table 2. Large differences in signal-to-noise ratio were found both between traits measured by the same measuring method as well as between the same traits measured by different measuring methods. Combined for all measuring methods, a total of 13 traits (10%) were excluded, because variation between sides (true FA) was not significantly larger than variation between repeated measurements within sides (ME; Table 3). The proportion of traits excluded for not meeting this criterion differed between measuring methods (
2 = 99.2, df = 6, P < 0.0001) and was highest for measuring methods with digital calipers. The living broilers measuring method was by far the least accurate, with 40% of the measurements not being significant for FA. In total, 19 traits(15.1%) were excluded because of a significant presence of DA. The proportion of traits with DA differed between methods (2 = 56.4, df = 6, P < 0.0001) and was highest for the digital bone photographs (manual) method followed by methods using calipers (bone dimensions, living broilers, and intact carcasses). Directional asymmetry rarely seemed trait-specific, but whether DA was found in a particular trait depended on the measuring method (Table 1). Because all kurtosis values remained above the threshold level of –1, we presumed that all traits met the assumption of absence of AS. Finally, 77 significant correlations in signed FA were found between traits measured by the same method. The proportion of significant correlations over the total number of correlations, however, was small (<0.11%, Table 3) and did not differ between measuring methods (2 = 9.6, df = 6, P = 0.14, Table 3).
|
|
). The highest repeatabilities were obtained for bones measured by balance (bone weight) and digital caliper (bone dimensions), and the lowest values were obtained for measurements on living broilers. Although FA estimates of tarsometatarsus measured with the intact carcasses method were highly repeatable, those of radius measurements showed much lower values (50% on average; Table 4). However, the reverse pattern arose when these traits were measured by the digital bone photographs (manual) method.
|
; see Table 1 for selected traits in each method). These comparisons showed that repeatabilities were low for the living broilers method and to a lesser extent for the intact carcasses method and that all other measuring methods had high repeatabilities (Figure 1).
|
and http://www.ilvo.vlaanderen.be/documents/fluctasbroilers.pdf illustrate the effect of repeatability (or ME) on required sample sizes and number of repeated measurements. Even when ME was low (e.g., repeatability of 83%, Figure 2) and 4 repeated measurements per trait were obtained on both sides, the difference in FA between 2 populations needed to be doubled to achieve a power of 80% with a sample size of 80 individuals from both populations. When 20 individuals per population were sampled, the difference in FA needed to be 4-fold to achieve the same power. When ME was larger (e.g., a repeatability of 33%, http://www.ilvo.vlaanderen.be/documents/fluctasbroilers.pdf), a sample of over 60 individuals per population was required for 2 repeated measurements, or 40 individuals for 4 repeated measurements.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Trait Selection Criteria
Approximately 15% of trait measurements were excluded because of DA. Two types of DA should be differentiated: DA with a genetic basis vs. DA due to ME. If DA in a trait has a genetic basis, we would expect that DA would be found consistently irrespective of measuring method. For example, face length showed DA irrespective of whether it was measured on living birds or on intact carcasses. However, consistent DA for a trait was rarely found in the present study. Brown and Brown (2002) reported that DA likely reflects a handedness bias particularly for caliper measurements. This is supported by the present study, because the 3 measuring methods with calipers had a relatively high proportion of traits with DA. Surprisingly perhaps, we found an even higher proportion of directional asymmetrical traits for the manual measurements on digital photographs for which we have no explanation. Measurements on soft tissue on living or dead animals are more prone to handling bias, because the pressure applied during handling or measuring may alter the exact position of the measuring points and consequently induce DA (e.g., Helm and Albrecht, 2000). Although we find evidence for the importance of low accuracy causing DA, irrespective of the underlying mechanism (genetic or not), traits showing DA should preferably be avoided. Calipers were positioned so as to avoid injury of live birds, and the orientation of dead or living birds differed depending on the side being measured, which affected precision. All traits exhibiting DA in our bone dimensions method were measurements of width (Table 1). This might be explained by the difficulty in determining unambiguous measuring points in width measurements as opposed to length measurements.
In the present study, AS was not found in any trait. In the literature, the presence of AS is often tested by determining whether the distribution of the signed asymmetry deviated significantly from normality. Such deviations from normality could arise for other reasons next to AS as well, including leptokurtosis, which may arise from real FA (Palmer and Strobeck, 1992). Because we used a kurtosis threshold of –1 to specifically detect the presence of AS, this difference in statistical approach could be an explanation for the difference in the proportion of traits showing AS between our study and other studies (Campo et al., 2000; Yalçin et al., 2001).
Few correlations in signed FA value between traits per method were found. Most developmental correlations occurred between traits of the wing and leg (13%) and between traits of the wing and foot (10%). A considerable percentage of wing (16%) and leg traits (19%) were inter-correlated, and 13% of leg and foot traits were correlated. Van Dongen et al. (1999b) mentioned that interdependent development is more common between traits with a common function (wing-wing, leg-leg, and foot-leg). Nevertheless, 23% of the correlations appeared between traits with no common function (wing-leg or wing-foot).
Measurement Accuracy
In general, more complex measuring methods yielded more accurate measurements of FA. Measurements on living poultry are noninvasive and relatively easy and fast to perform, but approximately 40% of the traits were excluded because FA did not significantly exceed ME. Moreover, when measuring living animals, caution must be taken to avoid injuries. Contrary to living and dead animals, bones or images of bones can be measured repeatedly over time. However, it should be noted that to be able to measure bones, further manipulations (X-raying or defleshing) are required that are time-consuming and that may induce additional ME. Because measurements were done twice on the same set of X-ray photographs, not all possible sources of ME were included in our study. Given that levels of ME for this technique might be largely underestimated, we cannot formulate conclusions, about its usefulness in the study of FA on the basis of our dataset. Contrary to the findings of Yngvesson and Keeling (2001), who also measured living birds, in the present study, the repeatability of radius and tarsometatarsus FA measurements on living birds were low. Although FA estimates of tarsometatarsus length measured on intact carcasses were highly repeatable, those of radius showed much lower values. The reverse pattern arose when these traits were measured manually with image-analyzing software. This indicates that the best measuring method may differ between traits. Besides this trait-method interaction, the results confirm that more repeated measurements are required on living broilers (Palmer, 1994; Møller and Swaddle, 1997; Yngvesson and Keeling, 2001).
Power Analysis
The power analysis clearly shows the importance of repeatable measurements when designing FA experiments. As has been shown in other studies (Van Dongen, 1999; Knierim et al., 2007), the statistical power when comparing levels of FA is low. Applying measuring methods with low repeatabilities (e.g., 33% as compared with 83%) requires a substantial increase in samples size, number of repeated measurements on each individual, or both. As expected, obtaining additional measurements on both sides can contribute relatively largely to power when ME is large, whereas the benefit of larger sample sizes apparently is irrespective of the degree of ME (see also Van Dongen, 1999, for comparable findings). Because measurements on living birds had low repeatability, we recommend that roughly a sample of 60 individuals per treatment should be measured twice. When only 40 individuals are measured per treatment, as suggested by Palmer (1994), 4 repeated measurements on both sides are required. When repeatability is high (as for all our measuring methods on defleshed bones), a sample size of 20 to 40 individuals may be appropriate, especially if each side is measured 4 times. Based on these simulations, we suspect that sample sizes may have been inadequate in some published studies on FA in poultry.
|
APPENDIX LANDMARKS OF MEASUREMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Landmarks of Measurements on Skeletal Structures on the X-Rays of the Limbs of Broilers
The head, wings, and legs were separated from the carcass, and limbs were frozen in a specific position: wings were stretched out, and the legs were put in such a way that the joint between the femur and the tibiotarsus and the joint between the tibiotarsus and the tarsometatarsus formed an angle of 90°. The feet and toes formed 1 line with the tarsometatarsus.
Defleshing
Fourteen bones of the limbs and the head were macerated with sodium perborate tetra hydrate according to the method described by McDonald and Vaughan (1999). Head, wings, and legs were macerated the first time in 60 g of sodium perborate tetra hydrate in 1 L of hot water (65°C) for a period of approximately 24 h. After removing a large part of the surrounding flesh, the skeletal structures were submerged a second time in 30 g of sodium perborate tetra hydrate in 1 L of hot water (65°C) for approximately 24 h. Skeletal structures were cleaned using a water jet vacuum pump and dissection material. Finally, the skeletal structures were dried for 24 h in a laboratory stove at 50°C.
Weights of Skeletal Structures
Bones were weighed on an analytical balance with a precision of 0.01 mg.
Landmarks of Measurements on Skeletal Structures Using a Digital Caliper
Subsequently, the length and the width (at the narrowest point when the bone was held in a specific position or at the thickest point near the joints) of the bones were measured using a digital caliper to the nearest 0.01 mm. All bones were kept in the same position during measurements.
Automatic and Manual Measurements on Digital Photographs of Skeletal Structures Using Image-Analyzing Software
Digital photographs were taken from the bones that were macerated (left traitside top, right traitside bottom). Using image-analyzing software, the lengths and surface areas of the skeletal structures on these digital photographs were measured automatically. Length measurements of the skeletal structures were repeated manually. Start and end point of the measurements were manually determined on the digital photograph in the image-analyzing program.
|
|
|
|
|
ACKNOWLEDGMENTS |
|---|
Received for publication June 14, 2006. Accepted for publication September 8, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAPPENDIX LANDMARKS OF...REFERENCES |
|---|
Brown, C. R., and M. B. Brown. 2002. Ectoparasites cause increased bilateral asymmetry of naturally selected traits in a colonial bird. J. Evol. Biol. 15:1067–1075.[CrossRef][ISI]
Campo, J. L., M. G. Gil, I. Muñoz, and M. Alonso. 2000. Relationships between bilateral asymmetry and tonic immobility reaction or heterophil to lymphocyte ratio in five breeds of chickens. Poult. Sci. 79:453–459.
Cloutier, S., and R. C. Newberry. 2002. Differences in skeletal and ornamental traits between laying hen cannibals, victims and bystanders. Appl. Anim. Behav. Sci. 77:115–126.[CrossRef][ISI]
Eeva, T., S. Tanhuanpaa, C. Raberg, S. Airaksinen, M. Nikinmaa, and E. Lehikoinen. 2000. Biomarkers and fluctuating asymmetry as indicators of pollution-induced stress in two hole-nesting passerines. Funct. Ecol. 14:235–243.[CrossRef]
Forkman, B., and S. Corr. 1996. Influence of size and asymmetry of sexual characters in the rooster and hen on the number of eggs laid. Appl. Anim. Behav. Sci. 49:285–291.[CrossRef][ISI]
Garland, T., Jr., and P. W. Freeman. 2005. Selective breeding for high endurance running increases hindlimb asymmetry. Evol. Int. J. Org. Evol. 59:1851–1854.
Gomendio, M., J. Cassinello, and E. R. S. Roldan. 2000. A comparative study of ejaculate traits in three endangered ungulates with different levels of inbreeding: Fluctuating asymmetry as an indicator of reproductive and genetic stress. Proc. R. Soc. Lond. B. Biol. Sci. 1446:875–882.
Graham, J. H., D. C. Freeman, and J. M. Emlen. 1993. Antisymmetry, directional asymmetry, and dynamic morphogenesis. Genetica 89:121–137.[CrossRef][ISI]
Helm, B., and H. Albrecht. 2000. Human handedness causes directional asymmetry in avian wing length measurements. Anim. Behav. 60:899–902.[CrossRef][ISI][Medline]
Hoffmann, A., and R. E. Woods. 2003. Associating environmental stress with developmental stability: Problems and patterns. Pages 387–401 in Developmental Instability: Causes and Consequences. M. Polak, ed. Oxford Univ. Press, New York, NY.
Klingenberg, C. P. 2003a. Fluctuating asymmetry and animal welfare: How far are we? And how far should we go? Vet. J. 166:5–6.[CrossRef][ISI][Medline]
Klingenberg, C. P. 2003b. A developmental perspective on developmental instability: Theory, models and mechanisms. Pages 14–34 in Developmental Instability: Causes and Consequences. M. Polak, ed. Oxford Univ. Press, New York, NY.
Klingenberg, C. P. 2004. Integration, modules and development: Molecules to morphology to evolution. Pages 213–230 in Phenotypic Integration: Studying the Ecology and Evolution of Complex Phenotypes. M. Pigliucci and K. Preston., ed. Oxford Univ. Press, New York, NY.
Klingenberg, C. P., and S. D. Zaklan. 2000. Morphological integration between developmental compartments in the Drosophila wing. Evol. Int. J. Org. Evol. 54:1273–1285.
Knierim, U., S. Van Dongen, B. Forkman, F. A. M. Tuyttens, M. 
pinka, J. L. Campo, and G. E. Weissengruber. 2007. Fluctuating asymmetry as an animal welfare indicator – a review of methodology and validity. Physiol. Behav. doi:10.1016/ j.physbeh.2007.02.014
Leamy, L. 1984. Morphometric studies in inbred and hybrid house mice. V. Directional and fluctuating asymmetry. Am. Nat. 123:579–593.[CrossRef][ISI]
Lens, L., and S. Van Dongen. 1999. Evidence for organism-wide asymmetry in five bird species of a fragmented afrotropical forest. Proc. R. Lond. B. 266:1055–1060.[CrossRef]
Lens, L., S. Van Dongen, S. Kark, and E. Matthysen. 2002. Fluctuating asymmetry as an indicator of fitness: Can we bridge the gap between studies? Biol. Rev. 77:27–38.[Medline]
Leung, B. 1998. Correcting for allometry in studies of fluctuating asymmetry and quality within samples. Proc. R. Lond. B. 265:1623–1629.[CrossRef]
Leung, B., M. R. Forbes, and D. Houle. 2000. Fluctuating asymmetry as a bioindicator of stress: Comparing efficacy of analyses involving multiple traits. Am. Nat. 155:101–115.[CrossRef][ISI][Medline]
Manning, J. T., A. R. Gage, M. J. Diver, D. Scutt, and W. D. Fraser. 2002. Short-term changes in asymmetry and hormones in men. Evol. Hum. Behav. 23:95–102.[CrossRef][ISI]
McDonald, R. A., and N. Vaughan. 1999. An efficient way to prepare mammalian skulls and bones. Mamm. Rev. 29:265–266.[CrossRef]
Møller, A. P., and J. T. Manning. 2003. Growth and developmental instability. Vet. J. 166:19–27.[CrossRef][ISI][Medline]
Møller, A. P., G. S. Sanotra, and K. S. Vestergaard. 1995. Developmental stability in relation to population density and breed of chickens Gallus gallus. Poult. Sci. 74:1761–1771.[ISI][Medline]
Møller, A. P., G. S. Sanotra, and K. S. Vestergaard. 1999. Developmental instability and light regime in chickens (Gallus gallus). Appl. Anim. Behav. Sci. 62:57–71.[CrossRef][ISI]
Møller, A. P., and J. P. Swaddle. 1997. Asymmetry, developmental stability and evolution. Oxford Univ. Press, Oxford, UK.
Nestor, K. E., J. W. Anderson, and R. A. Patterson. 2000. Effects of selection for increased body weight, egg production and shank width on developmental stability in turkeys. Poult. Sci. 79:937–945.
Palmer, A. R. 1994. Fluctuating asymmetry analyses: A primer. Pages 335–364 in Developmental Stability: Its Origins and Evolutionary Implications. T. A. Markow, ed. Kluwer Acad. Publ., Dordrecht, the Netherlands.
Palmer, A. R., and C. Strobeck. 1986. Fluctuating asymmetry: Measurements, analysis, patterns. Ann. Rev. Ecol. Syst. 17:391–421.[CrossRef][ISI]
Palmer, A. R., and C. Strobeck. 1992. Fluctuating asymmetry as a measure of developmental instability: Implications of non-normal distributions and power of statistical tests. Acta Zool. Fenn. 191:57–72.
Palmer, A. R., and C. Strobeck. 2003. Fluctuating asymmetry analyses revisited. Pages 279–319 in Developmental Instability: Causes and Consequences. M. Polak, ed. Oxford Univ. Press, New York, NY.
Polak, M. 2003. Developmental Instability: Causes and Consequences. Oxford Univ. Press, New York, NY.
Sanotra, G. S., J. D. Lund, A. K. Ersboll, J. S. Petersen, and K. S. Vestergaard. 2001. Monitoring leg problems in broilers: A survey of commercial broiler production in Denmark. World’s Poult. Sci. J. 57:55–69.[CrossRef][ISI]
Satterlee, D. G., G. G. Cadd, and R. B. Jones. 2000. Developmental instability in Japanese quail genetically selected for contrasting adrenocortical responsiveness. Poult. Sci. 79:1710–1714.
Swaddle, J. P. 2003. Fluctuating asymmetry, animal behavior, and evolution. Adv. Stud. Behav. 32:169–205.[CrossRef]
Tuyttens, F. A. M. 2003. Measures of developmental instability as an integrated, a posteriori indicator of farm animal welfare: A review. Anim. Welf. 12:535–540.[ISI]
Tuyttens, F. A. M., L. Maertens, E. Van Poucke, A. Van Nuffel, J. Creve, and L. Lens. 2005. Measuring fluctuating asymmetry in fattening rabbits: A valid indicator of performance and housing quality. J. Anim. Sci. 83:2645–2652.
Van Dongen, S. 1999. Accuracy and power in fluctuating asymmetry studies: Effects of sample size and number of within-subject repeats. J. Evol. Biol. 12:547–550.[CrossRef][ISI]
Van Dongen, S. 2000. Unbiased estimation of individual asymmetry. J. Evol. Biol. 13:107–112.[CrossRef][ISI]
Van Dongen, S. 2006. Fluctuating asymmetry and developmental instability in evolutionary biology: Past, present and future. J. Evol. Biol. 19:1727–1743.[CrossRef][ISI][Medline]
Van Dongen, S., G. Molenberghs, and E. Matthysen. 1999a. The statistical analysis of fluctuating asymmetry: REML estimation of a mixed regression model. J. Evol. Biol. 12:94–102.[CrossRef][ISI]
Van Dongen, S., E. Sprengers, and C. Löfstedt. 1999b. Correlated development, organism-wide asymmetry and patterns of asymmetry in two moth species. Genetica 105:81–91.[CrossRef][ISI]
Van Dongen, S., W. Talloen, and L. Lens. 2005. High variation in developmental instability under non-normal developmental error: A Bayesian perspective. J. Theor. Biol. 236:263–275.[CrossRef][ISI][Medline]
Yalçin, S., S. Özkan, L. Türkmut, and P. B. Siegel. 2001. Responses to heat stress in commercial and local broiler stocks. II. Developmental stability of bilateral traits. Br. Poult. Sci. 42:153–160.[CrossRef][ISI][Medline]
Yalçin, S., and P. B. Siegel. 2003. Exposure to cold or heat during incubation on developmental stability of broiler embryos. Poult. Sci. 82:1388–1392.
Yngvesson, J., and L. J. Keeling. 2001. Body size and fluctuating asymmetry in relation to cannibalistic behaviour in laying hens. Anim. Behav. 61:609–615.[CrossRef][ISI]
Zakharov, V. M., E. Pankakoski, B. I. Sheftel, A. Peltonen, and I. Hanski. 1991. Developmental stability and population dynamics in the common shrew, Sorex araneus. Am. Nat. 138:797–810.[CrossRef][ISI]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
