Word-types represent the primary form of data for many models of phonological learning, and they often predict performance in psycholinguistic tasks. Word-types are often tacitly defined as phonologically unique words. Yet, an explicit test of this definition is lacking, and natural language patterning suggests that word meaning could also act as a cue to word-type status. This possibility was tested in a statistical phonotactic learning experiment in which phonological and semantic properties of word-types varied. During familiarization, the learning targets—word-medial consonant sequences—were instantiated either by four related word-types or by just one word-type (the experimental frequency factor). The expectation was that more word-types would lead participants to generalize the target sequences. Regarding semantic cues, related word-types were either associated with different referents or all with a single referent. Regarding phonological cues, related word-types differed from each other by one, two, or more phonemes. At test, participants rated novel wordforms for their similarity to the familiarization words. When participants heard four related word-types, they gave higher ratings to test words with the same consonant sequences, irrespective of the phonological and semantic manipulations. The results support the existing phonological definition of word-types.

中文翻译：

---

从词型中学习的语音和语义线索


词型代表了许多语音学习模型的主要数据形式，并且它们通常预测心理语言任务中的表现。词类型通常被默认定义为语音上独特的词。然而，缺乏对该定义的明确测试，并且自然语言模式表明词义也可以充当词类型状态的提示。这种可能性在统计语音学习实验中得到了测试，其中单词类型的语音和语义特性各不相同。在熟悉过程中，学习目标——单词中间辅音序列——由四种相关的单词类型或仅一种单词类型（实验频率因子）实例化。我们期望更多的单词类型能够引导参与者概括目标序列。关于语义线索，相关的词类型要么与不同的所指对象相关联，要么与单个所指对象相关联。关于语音线索，相关的词类型彼此之间存在一个、两个或多个音素的差异。在测试中，参与者对新词形式与熟悉词的相似度进行评分。当参与者听到四种相关的单词类型时，他们对具有相同辅音序列的测试单词给出更高的评分，而不管语音和语义操作如何。结果支持现有的词类型的语音定义。