The measurement of overall body size in birds
The measurement of overall body size in birds
复制标题
DOI：
10.1093/auk/106.4.666
复制
发表时间：
1989-10
期刊：
影响因子：
--
通讯作者：
J. Rising;K. Somers
中科院分区：
文献类型：
--
作者：
J. Rising;K. Somers
--我们比较了鸟类学研究中常用的许多单变量和多变量身体尺寸测量值，包括八种多变量测量值（来自主成分分析），加上头骨长度、尺骨长度、胫跗骨长度、翅膀长度和重量。分析基于对来自整个物种分布范围 53 个不同地理地点的三只随机选择的雄性和三只随机选择的雌性萨凡纳麻雀 (Passerculus sandensis) 的 26 次测量。八个主成分分析中的六个提供了基本相同的关于身体尺寸的信息。基于原始数据或对数转换数据的方差-协方差矩阵的分析提供了与其他多元大小估计值变化最大的第一个轴。在单变量测量中，尺骨长度、翼长和体重提供的信息与总体尺寸的多变量测量不同。体重比翅膀长度更能代表一般尺寸（即 PC I），但由于繁殖条件的差异，体重对于雄性来说比雌性更好。翅膀长度不能代表身体尺寸。由于每个主成分分析都提供了有关 PC I 上身体尺寸的信息，因此我们鼓励研究人员根据分析目标而不是方法简单性或通用性在各种方法中进行选择。 1988 年 11 月 29 日接收，1989 年 5 月 18 日接受。鸟类学家经常面临测量鸟类体型的挑战。需要测量总体规模来检验预测地理变化模式的假设（例如 Bergmann 或 Allen 规则；James 1970、Johnston 和 Selander 1971、Niles 1973、Fleischer 和 Johnston 1982、Handford 1983、Murphy 1985）。还需要对体型进行估计，以检验有关体型性别二态性进化的假设（例如 Hamilton 和 Johnston 1978、Johnston 和 Fleischer 1981、Fleischer 和 Johnston 1984、McGillivray 和 Johnston 1987、Rising 1987b）。此外，必须按体型对物种进行排序，以测试预测生态群落中共存物种之间的体型比例的模型（例如 Ricklefs 和 Cox 1977、Ricklefs 和 Travis 1980、Haefner 1981、Sabo 和 Holmes 1983、Miles 和 Ricklefs 1984、PullJam 1985、Brown 和 Maurer 1986、Miles et等，1987）。在生理学中，代谢活动的标准测量通常表示为身体尺寸的函数，并且通常有助于检查结构或器官相对于整体尺寸的关系（例如 Fisher 1947、Kendeigh 1976、Blem 1984、Calder 1984、Paladino 1985、Rising 1987a、Packard 和 Boardman 1988）。然而，身体尺寸很难测量。也许衡量整体尺寸的最佳方法是总质量，但有关质量的可靠信息通常很难获得。尽管最近的数据汇编对我们对鸟类质量的了解做出了很大贡献（Clench and Leberman 1978，Dunning 1984），但由于季节和饮食相关的变化，现有的质量数据常常不能令人满意（例如 Niles 1973）。因此，鸟类学家通常使用翅膀长度的测量来估计相对身体大小（例如 James 1970、Lack 1971、Snyder 和 Wiley 1976、Payne 1984、Jehl 和 Murray 1986、Zink 和 Remsen 1986）。可以在博物馆研究的皮肤和活鸟上轻松测量翅膀长度；然而，至少在某些情况下，与其他更精确的测量相比，它对身体尺寸的估计很差（Rising 1988）。即使扣除测量误差，许多难以量化的因素也会影响鸟类的翅膀长度。首先，机翼羽毛容易磨损。因此，随着羽毛逐渐磨损，翅膀长度测量的可靠性会降低。这在性别二态性的研究中可能尤其重要，因为在许多物种中，两性之间的行为差异导致羽毛磨损率的性别差异。其次，个体的翅膀长度每年都在变化——尽管鸟的骨骼已经完全僵化（因此，从这个意义上说，鸟已经完全长大了）。例如，Rising（unpubL 数据）捕获并测量（精确到毫米）野生 Savan666 The Auk 106：666-674。 1989 年 10 月 1989 年 10 月] 多年来测量 667 nah 麻雀 (Passerculus andwichensis) 的体型。同一年内捕获两次或两次以上的雄性翅膀长度测量值的平均值差异为-0.13毫米（范围-4至+2毫米，n = 32，SE = 0.24），但一年以上捕获并测量的雄性翅膀长度测量值的平均差异为+1.17（范围-2至+6毫米，n = 23，SE = 0.43）。符号测试表明，在一个季节内，测量到的个体翅膀长度的积极变化和消极变化一样多。多年来，机翼长度似乎每年增加的情况多于减少的情况（P = 0.05）。因此，翅膀长度在某种程度上随着年龄的增长而增加。由于从翼长数据获得有关身体尺寸的准确信息很困难，人们经常使用个体骨骼的测量（例如 Johnston 和 Selander 1971）或器官重量（例如 Power 1970）来估计身体尺寸。或者，鸟类学家计算了性状组合（例如，许多度量的总和；McGillivray 和 Johnston 1987，Rising 1987b），或多变量度量（例如主成分或判别函数分数；Johnston 和 Selander 1971，Niles 1973，Zink 1986，Rising 1988），这些组合解释了性状之间的协变并提取“大小”轴。关于哪种主成分分析模型能够最好地提取“尺寸成分”，已有大量讨论（Jolicoeur 1963、Mosimann 1970、Mosimann 和 James 1979、Bookstein 等人 1985、Somers 1986、Rohlf 和 Bookstein 1987）。在这里，我们根据经验比较八个不同主成分模型的尺寸轴和五个单变量身体尺寸测量，包括质量（重量）和机翼长度，以确定这些整体身体尺寸估计的相对相似性。
--We compared a number of univariate and multivariate measures of body size used commonly in ornithological research, including eight multivariate measures (from principal components analyses), plus skull length, ulna length, tibiotarsus length, wing length, and weight. Analyses are based on 26 measurements on three randomly selected male and three randomly selected female Savannah Sparrows (Passerculus sandwichensis) from each of 53 different geographic localities throughout the species' range. Six of the eight principal components analyses provided essentially the same information about body size. Analyses based on the variance-covariance matrix of raw or log-transformed data provided first axes that varied most from the other multivariate estimates of size. Among the univariate measures, ulna length, wing length, and body weight contributed information that diverged from the multivariate measures of overall size. Weight better represents general size (i.e. PC I) than wing length, but because of variation in reproductive condition, weight is a far better measure in males than in females. Wing length is not a representative measure of body size. Inasmuch as each principal components analysis provides information about body size on PC I, we encourage researchers to choose among the various approaches according to analytical objectives rather than methodological simplicity or general utility. Received 29 November 1988, accepted 18 May 1989. ORNITHOLOGISTS are frequently faced with the challenge of measuring body size in birds. A measure of overall size is required to test hypotheses predicting patterns of geographic variation (e.g. Bergmann's or Allen's rules; James 1970, Johnston and Selander 1971, Niles 1973, Fleischer and Johnston 1982, Handford 1983, Murphy 1985). An estimate of body size is also required to test hypotheses about the evolution of sexual dimorphism in body size (e.g. Hamilton and Johnston 1978, Johnston and Fleischer 1981, Fleischer and Johnston 1984, McGillivray and Johnston 1987, Rising 1987b). In addition, species must be ranked by body size to test models that predict size ratios among coexisting species in ecological communities (e.g. Ricklefs and Cox 1977, Ricklefs and Travis 1980, Haefner 1981, Sabo and Holmes 1983, Miles and Ricklefs 1984, PullJam 1985, Brown and Maurer 1986, Miles et al. 1987). In physiology, standard measures of metabolic activity are frequently expressed as a function of body size, and it is often useful to examine the relationship of structures or organs relative to overall body size (e.g. Fisher 1947, Kendeigh 1976, Blem 1984, Calder 1984, Paladino 1985, Rising 1987a, Packard and Boardman 1988). Body size, however, is difficult to measure. Perhaps the best measure of overall body size is total mass, but reliable information on mass is often difficult to obtain. Although recent compilations of data contribute much to our knowledge of the mass of birds (Clench and Leberman 1978, Dunning 1984), the available data on mass are all too often unsatisfactory because of seasonal and diet-related variability (e.g. Niles 1973). Consequently, ornithologists commonly use a measure of wing length as an estimate of relative body size (e.g. James 1970, Lack 1971, Snyder and Wiley 1976, Payne 1984, Jehl and Murray 1986, Zink and Remsen 1986). Wing length is measured easily on museum study skins and living birds; however, at least in some cases, it is a poor estimate of body size when compared with other, more precise measurements (Rising 1988). Even discounting measurement error, many factors that are difficult to quantify affect the wing length of a bird. First, wing feathers are subject to wear. Thus, the reliability of measurements of wing length decreases as the feathers progressively become more worn. This may be especially important in studies of sexual dimorphism, because in many species behavioral differences between the sexes lead to sexual differences in rates of feather wear. Second, the wing length of an individual varies from year to year--even though the bird's skeleton is completely ossified (and thus, in this sense, the bird is completely grown). For example, Rising (unpubL data) captured and measured (to the nearest mm) wild Savan666 The Auk 106: 666-674. October 1989 October 1989] Measurement of Body Size 667 nah Sparrows (Passerculus andwichensis) over many years. The average of the differences of measurements of wing length of males captured twice or more during the same year is -0.13 mm (range -4 to +2 mm, n = 32, SE = 0.24), but the average differences in wing length of males captured and measured during more than one year is +1.17 (range -2 to +6 mm, n = 23, SE = 0.43). A sign test shows that within seasons there are as many positive changes as negative changes in the measured wing lengths of individuals. Among years, wing lengths appear to increase more often from year to year than decrease (P = 0.05). Thus, wing length, to some extent, increases with age. Because of the difficulties of obtaining accurate information about body size from wing-length data, people have often used measures of individual bones (e.g. Johnston and Selander 1971) or of organ weight (e.g. Power 1970) as estimates of body size. Alternatively, ornithologists have computed combinations of characters (e.g. the sum of many measures; McGillivray and Johnston 1987, Rising 1987b), or multivariate measures (such as principal component or discriminant function scores; Johnston and Selander 1971, Niles 1973, Zink 1986, Rising 1988) that account for the covariation among characters and extract a "size" axis. There has been considerable discussion concerning which of the models of principal components analysis best extracts a "size component" (Jolicoeur 1963, Mosimann 1970, Mosimann and James 1979, Bookstein et al. 1985, Somers 1986, Rohlf and Bookstein 1987). Here we empirically compare size axes from eight different principal components models and five univariate measures of body size, including mass (weight) and wing length, to determine the relative similarity of these estimates of overall body size.