Language learning, communication and the emergence of phonotactic constraints

所有的口语都是通过组合语音来造词的。例如，英语单词“Tap”、“Pat”、“apt”、“at”和“Tat”都以不同的组合使用相同的3个发音。这种组合性有助于人类语言非凡的表达能力，使我们能够从几十个语音中创造出数千个不同的单词。但并不是语言中的所有声音组合都是平等的。例如，英语中的单词都不能以序列“tp”开头或结尾。这种对可允许的声音序列的限制称为音韵。虽然音位制约可能是每种语言特有的，但有些限制会发生在多种语言之间。一个被认为是普遍存在的限制被称为相似位置回避：由元音分隔的辅音不太可能有相同或相似的发音位置(即，产生辅音的口腔部分)。例如，在英语中，像“bop”和“mob”这样包含嘴唇产生的两个辅音的单词比预期的要少。为什么语言有音素，而不是只允许最大限度的自由重组？为什么某些限制条件会反复出现，比如避免相似的地点？文献中一个突出的主张是，解释在于学习偏见，即我们比其他人更容易学习某些组合规则的倾向。这种偏见被认为是每个学习者基本音系学知识的一部分，迫使他们偏爱更简单或更容易产生/感知的模式。这一解释已经在许多研究中得到了验证，这些研究考察了新的音向限制学习起来有多容易。然而，结果并不总是与预测一致，即跨语言常见的模式应该比不常见的模式更容易学习。在这个项目中，我们探索了一种新的方法来解释一些音位制约的普遍存在。我们的方法与以前的说法在两个重要方面不同。首先，我们认为沟通的影响是一个关键因素。除了学习，语言也被用来交流，在交流中工作得很好的声音序列应该会激增。成功的交流要求我们通过最大化词语之间的对比来最大限度地减少混淆。有一些证据表明，单词中相似辅音的重复会导致单词的错误识别，这表明这种影响可能是类似位置回避等限制因素背后的影响。第二，我们认为音位限制是对语言中单词形状的概括，随着单词被学习并被用来在说话者之间进行交流，其组成可以朝着特定的方向变化。我们预测这种变化的方向反映了学习(即哪些类型的单词更容易学习)和交流(即哪些类型的单词被更准确地识别)之间的相互作用。我们将通过3组实验来检验这一理论，测试违反相似地点回避的单词是否更好(或更差)地被成人和婴儿学习，他们是否更容易被错误识别，当他们在不同的人群中学习和使用时，他们在词汇中的数量是否会减少。我们的项目将为语言学习和交流如何解释总体上的音向性限制，特别是类似的地方回避提供重要的见解。该项目的发现还将有助于我们理解语言和认知科学的一个更大的核心问题：人类语言的基本特征来自哪里？

All spoken languages make words by combining speech sounds. For example, the English words "tap", "pat", "apt", "at", and "tat" all use the same 3 sounds in different combinations. This combinatoriality contributes to the extraordinary expressive power of human language, allowing us to create thousands of distinct words out of a few tens of speech sounds. But not all combinations of sounds in the language are equally permitted. For instance, no words in English can begin or end with the sequence "tp". Such constraints on permissible sound sequences are known as phonotactics. While phonotactic constraints can be idiosyncratic to each language, some occur across multiple languages. One constraint that is thought to be universal is called Similar Place Avoidance: Consonants separated by vowels are unlikely to have the same or similar place of articulation (i.e., part of the mouth where the consonant is produced). For example, there are fewer than expected words in English like "bop" and "mob" that contain two consonants produced by the lips. Why do languages have phonotactics at all, rather than just allowing maximally free recombination? And why do certain constraints such as Similar Place Avoidance crop up repeatedly? One prominent claim in the literature is that the explanation lies in learning bias, or the tendency for us to learn certain rules for sound combinations more readily than others. Such biases are thought to be part of every learner's underlying knowledge of phonology, enforcing a preference for patterns that are simpler or easier to produce/perceive. This explanation has been tested in many studies that examined how easily new phonotactic constraints can be learned. However, the results have not always been consistent with the prediction that patterns that are commonly found across languages should be easier to learn than those that are uncommon. It appears, then, that learning bias is not the only reason why some phonotactics recur in the world's languages.In this project we explore a new way to explain the ubiquity of some phonotactic constraints. Our approach is different from previous accounts in two important aspects. Firstly, we consider the impact of communication as a critical factor. As well as being learned, languages are also used to communicate, and sound sequences that work well in communication should proliferate. Successful communication requires us to minimise confusability by maximising contrasts between words. There is some evidence that the repetition of similar consonants within a word can lead to misidentification of the word, suggesting that this effect may be behind constraints like Similar Place Avoidance.Secondly, we consider phonotactic constraints to be generalisations about the shapes of words in a language, the composition of which can change in a particular direction as words are learned and used to communicate between speakers. We predict that the direction of such change reflects the interaction between learning (i.e., which types of words are easier to learn) and communication (i.e., which types of words are more accurately recognised).We will examine this theory through 3 sets of experiments, testing whether words violating Similar Place Avoidance are better (or worse) learned by adults and infants, whether they are more (or less) likely to be misidentified, and whether they decrease in number in a vocabulary when they are learned and used in communication across different groups of people.Our project will provide important insights into how language learning and communication account for the presence of phonotactic constraints in general, and Similar Place Avoidance in particular. Findings from the project will also contribute to our understanding of a larger question central to language and cognitive sciences: Where do fundamental characteristics of human language come from?