Germanic dispersion beyond trees and waves

自从北日耳曼语言和西日耳曼语言的密切关系最初被假设以来，我们看待日耳曼语言内部等级的方式或多或少保持不变。似乎所有相关的语言都被考虑了，它们的相互依存关系和遗传关系被结合了西北日耳曼语和东日耳曼语的标准树模型所描述得最好，前者后来进一步扩展到北日耳曼语和西日耳曼语，而往往忽略了较小的、较不成熟的语言。然而，这种自信总是伴随着对树模型的明显怀疑，因为树模型是描述导致早期日耳曼语言的过程的适当启发式方法。被这样一个模型最好地近似所需的基本语言和社会机制在现实中不太可能发生。树模型通常需要“硬”分裂，即两种语言同时脱离一种共同的原始语言，很少或根本不会受到后来的相互影响。在方言连续体中发现的较慢的溶解过程，其中相关语音社区的异步发散和收敛运动以不同的速度发生，很难用树模型来解释。随后，学者们在整个辩论中提出了几个相反的建议，似乎更好地捕捉到了这些可观察到的原语言消解机制。然而，无论突出程度如何，从波动理论、面变辐射到声门测量学，这些提议都不能完全解释手头的语言数据。为了从计算角度探讨分支学，近年来从生物学等科学中采用了新的方法。这些方法，特别是系统发育学，带来了新的方法和技术，以对照通过这些模型获得的结果来检验先前的假设。就日耳曼语而言，以前的研究要么是在更大的范围内进行的，比如整个印欧语系，要么是较小的案例研究。因此，我们项目的目标将是将已知的计算方法专门应用于日耳曼语，除了经过充分证明的语言之外，还将特别关注广泛的语言。然而，由于即使是这些方法也不能以可能的最佳方式近似数据，因为它们通常还假设树形分解，最终目标将是为这项任务编写一个新的模型，不仅考虑到不同的变化速度，纳入较少得到充分证明的语言，而且还能够模拟方言连续体以及快速迁移。我们打算用于这一模拟任务的技术是基于代理的建模方法，目前几乎完全从政治和社会科学以及经济学的研究中得知。基于主体的模型恰好允许微调所需的专业化，以编程为日耳曼统一的解体提供最好的可能模型。

Since the close relationship of North and West Germanic languages was first postulated, the way we view the internal hierarchy of the Germanic languages has more or less remained the same. It seemed that all relevant languages were considered and their interdependencies and genetic relationships were best described by the standard tree model that incorporates Northwest and East Germanic, the former further branching into North and West Germanic later on, often while disregarding the smaller, less well-attested languages. This confidence, however, was always accompanied by a distinct skepticism of the tree model as an adequate heuristic to describe the processes leading up to the early Germanic languages. The underlying linguistic and social mechanics necessary to be best approximated by such a model are unlikely to occur in reality. Tree models, in general, require 'hard' splits where two languages simultaneously break off from a common proto-language with little to no later mutual influence. Slower dissolution processes found in dialect continua in which asynchronous diverging and converging movements of related speech communities occur at varying speeds can hardly be accounted for by tree models. Subsequently, several counter-proposals have been suggested by scholars throughout the debate that seem to better capture these observable mechanisms of proto-language dissolution. However, regardless of prominence, none of these proposals ranging from wave theories and areal change radiation to glottometrics could yet fully account for the linguistic data at hand.In an attempt to approach cladistics from a computational angle, new methods have been adopted in recent years from sciences such as biology. These approaches, phylogenetics in particular, brought new methods and techniques to test previous assumptions against the results obtained through these models. In the case of Germanic, previous studies were either carried out with a larger scope, such as the entire Indo-European family, or were smaller case studies. The aim of our project will therefore be to apply known computational methods specifically to Germanic with a particular focus on a wide range of languages in addition to the well-attested ones. Yet since even these methods do not approximate the data in the best possible way as they often also assume a tree shape dissolution, the ultimate goal will be to program a new model for this task that not only accounts for different speeds of change and incorporates less well-attested languages but is also capable of simulating dialect continua as well as rapid migration movements. The technique we intend to use for this simulation task is the method of agent-based modeling which is currently almost exclusively known from studies in political and social sciences as well as economics. The agent-based models exactly allow the fine-tuned specialization needed to program the best-possible model for the dissolution of Germanic unity.