A Stylostatistical Study of Goethe-Schiller-Correspondence and the Authorship Problem of the Lalebuch (1597)

本研究有两个目的：(1)比较一些统计方法(特别是多元分析)在文献学研究中的应用，并估计它们的效率；(2)使用最有效的技术，解决1597年匿名出版的《拉勒布赫》的作者问题。方法论实验的代表性文献学对象是1794-98年歌德和席勒的书信，在这些书信中，德国最伟大的古典作家坦率而友好地交换了关于文学和思想的相同主题和话题的批评意见。相同的体裁，几乎相同的形式和主题，两个完全不同的人格，相同的能力等级等，是文体统计方法进行方法论检验的最好条件。在对句子长度进行考察后，发现句子长度服从对数正态分布，并选取了20个非语境功能词进行多元…更多分析：这些字数是用于G-S对应的判别分析、主成分分析和聚类分析的变量(分为25个歌德样本和26个席勒样本)。主成分分析的成分-分数图显示，一边是G群，另一边是S群：借助判别分析的结果，可以更有效地区分两组。关于拉勒布赫作者问题的结论与霍内格的假设有一些相似之处，即J·费斯查特写了这本讽刺小说《伏尔克斯布赫的和平》(民间故事)，但得出结论的原因截然不同。他认为一年后出现的更多的克隆书《斯柴尔堡》比拉莱布赫更有权威性，并进一步假设存在一个未被发现的原版本。由于缺乏证据和说服力，这些论点不能得到支持。首先，统计分析表明，故事的第一部分和第二部分是不同的，所以我决定将拉勒布赫的文本数据分为两个样本(L1，L2)，并与菲斯查特的作品样本EulenSpiegel Reimenweiβ，Geschichtk-litterung等进行比较分析和统计检验。经过对句子长度分布、从属从句中的二项谓词和词的第一个符号的比较分析和统计检验，主成分分析(25个变量)的成分-分数图最明显地显示，L1被菲斯查特的分组样本包围，而L2与其他样本分开。根据上述数据，费斯查特极有可能写下了有争议的拉莱布赫的第一部分，但第二部分被认为是他留下的未完成的故事，有人在死后对其进行了修订和出版。较少

This research has two purposes : (1) to provide a comparison of some statistical methods (in particular multivariate analysis) applied to philological objects and to estimate their effiency ; (2) using most effective techniques, to solve the authorship question of the Lalebuch, published anonymously in 1597.The representative philological objects for methodological experiments is the collected letters of Goethe and Schiller corresponded in the years 1794-98, in which the greatest German classical Authors exchanged frankly and friendly their critical opinions on the same themes and topics concerning literature and thought. The same genre, almost the same form and theme, two quite different personalities of the same ability-rank etc, are best conditions for methodological examination of styl-statistical approach. After an investigation of sentence-length finding that the sentence-length follows the log-normal distribution, I have selected 20 non-contextual function words for multivariate … More analysis : the count of these words are the variables used for discriminant analysis, principal component analysis and cluster analysis of G-S correspondence (divided into 25 Goethe- and 26 Schiller-samples). PCA with component-score plots shows on the one side G-groups and on the other side S-groups : one can distinguish one group from the other more effectivly with the help of the result from discriminant analysis.The conclusions about the authorship problem of the Lalebuch have rather some similarity to Honneger's hypothesis that J.Fischart wrote this satirical masterpeace of Volksbuch (folktales), but the reasons for the conclusions are quite different. He regards the one year later appeared and more proliferous clone book Schildburger for more authoritative than Lalebuch and assumes further the existence of an unfound original print. For lack of evidence and persuasive power these theses are not to support. The statistical analysis showed above all that the first and the second part of the tale are not homogeneous, so I have decided to divide the text-data of Lalebuch into two samples (L1, L2) and to compare with samples of Fischart's works, namely Eulenspiegel Reimenweiβ, Geschichtk-litterung etc. After the comparative analysis and statistical tests of the sentence-length distribution, of the binomial predicates in subordinate clauses, and of the first symbol of words, the component-score plot of the PCA (25 variables) showed most evidently that L1 is surrounded by Fischart's groupe samples, while L2 stands quit apart from others. On the basis of the mentioned data, Fischart is extremely likely to have written the first part of the disputed Lalebuch, but the second part is supposedly left by him as unfinished story which some one posthumously revised and published. Less