I compare three linguistic models based on different conceptions of nativism to determine whether they can provide revealing accounts of properties of natural languages, and whether the traits that these linguistic models require are evolutionarily plausible.

The model based on Language-specific nativism in a broad sense contains rules, conditions, features and categories—Universal Grammar—to capture the empirical phenomena. Broad UG contains many domain-specific devices that are more descriptive than explanatory, and are unlikely to have evolved as irreducible properties of the brain.

Language-specific nativism in a narrow sense tries to improve evolvability by restricting its core component to the operation set-Merge and a universal lexicon of innate concepts. But the transfer operations required to link the language-invariant expressions with actual expressions of natural languages turn out to be inapplicable. Evolvability is also problematic: the model ends up appealing to mystery on three key issues.

Exapted Language Nativism posits that the human language capacity is not due to a brain development specifically devoted to language but is a side effect of a uniquely human capacity of detachment that enables an array of human-specific cognitive traits. Key among these traits is the capacity to form linguistic signs. Syntactic combinations emerge directly from the capacity to form combinatorial signs, a small natural subset of linguistic signs. The perceptual and conceptual elements of signs, and the extralinguistic cognition of speakers, have grounded prior properties that provide principled explanations of the data. The model has high evolvability since its core capacity of detachment is independently related to other phenotypic effects.

我比较了基于不同自然主义概念的三种语言模型，以确定它们是否可以提供有关自然语言属性的揭示性说明，以及这些语言模型要求的特征在进化上是否合理。

从广义上讲，基于特定语言的本土主义的模型包含规则，条件，特征和类别（通用语法），以捕获经验现象。广泛的UG包含许多特定于域的设备，这些设备的描述性比解释性强，并且不太可能演变为大脑的不可归约特性。

从狭义上讲，特定于语言的本土主义试图通过将其核心部分限制在操作集合并和先天概念的通用词典中来改善可进化性。但是事实证明，将语言不变的表达式与自然语言的实际表达式链接所需的传输操作是不适用的。可进化性也是有问题的：该模型最终在三个关键问题上吸引了人们的神秘感。

语言超自然主义认为人类的语言能力不是由于专门致力于语言的大脑发育而引起的，而是人类超然的分离能力的副作用，这种超能力导致了一系列特定于人类的认知特征。这些特征中的关键是形成语言符号的能力。句法组合直接从形成组合符号的能力中出现，组合符号是语言符号的一个很小的自然子集。符号的感知和概念元素以及说话者的语言外认知，已经奠定了先验属性，可以对数据进行有原则的解释。该模型具有很高的可进化性，因为其分离的核心能力独立于其他表型效应。