Abstract: Word frequency distribution is important for cross-situational word learning. Previous studies mainly investigated cross-situational word learning based on the uniform distribution. However, word frequency distribution usually follows the Zipfian distribution, a type of skewed distribution, in reality. Previous studies have relatively ignored cross-situational word learning based on skewed distributions, such as the Zipfian distribution. Few studies that focus on the skewed distribution have fierce controversies about whether the Zipfian distribution improves or inhibits cross-situational word learning.
Therefore, the present review first proposes a macro perspective of cross-situational word learning, which holds that the learning effect of each vocabulary is dependent on the learning effects of other vocabularies in the same learning environment. From the macro perspective, the present review first proposes the skewed learning advantage effect that is based on the mutual exclusivity strategy. Mutual exclusivity strategy refers to the following word learning methods. Specifically, learners assume that no two words have the same meaning. Then, learners eliminate the meaning competition of new words from the matching hypotheses that are established according to the co-occurrence law through the anchor point of the acquired words, thereby reducing the meaning ambiguity of the new words.
The skewed learning advantage effect believes that the effective use of mutual exclusivity strategy is the key to generate the Zipfian frequency promotion effect. Specifically, in the Zipfian distribution, learners can quickly learn high-frequency words. Then, learners make the learned high-frequency words as anchor points and use mutual exclusivity strategy to reduce the meaning ambiguity of new words. Thus, learners reduce the number of matching hypotheses in the memory system, thereby reducing the memory burden and improving the efficiency of cross-situational word learning. Moreover, if learners can repeatedly use the mutual exclusivity strategy, they may quickly learn multiple words, which is named word spurt. More importantly, the Zipfian distribution provides more opportunities for learners to use the mutual exclusivity strategy.
To further explain the mechanism of mutual exclusivity strategy and develop the skewed learning advantage effect. The present review examines the influence of language learning history, participant age, and situational ambiguity on the use of the mutual exclusivity strategy, as well as the word learning effect. The findings of some important studies suggest that these factors can influence the tendency of learners to establish matching hypotheses. For example, monolinguals are more likely to establish a one-to-one word-meaning matching hypothesis than bilinguals, while bilinguals are better at establishing and maintaining multiple matching hypotheses. Therefore, learners of different language learning histories have different tendencies of establishing matching hypotheses, resulting in different mechanisms of using mutual exclusivity strategy. In addition, different language learning histories contribute to the skewed learning advantage through different mechanisms.
To sum up, the present review firstly systematically reviews the studies of cross-situational word learning in the Zipfian distribution and explores the mechanism of word frequency distribution. Secondly, based on the mutual exclusivity strategy, the present review proposes the skewed learning advantage effect to respond to the controversy of previous studies. Finally, the present review develops the skewed learning advantage effect by deconstructing the mechanism of the mutual exclusivity strategy. In future research, how related factors, such as syntax, Chinese corpus, and social cues, influence skewed learning is another important question that needs to be investigated.

Key words: cross-situational word learning, language learning, macro-learning perspective, skewed learning, mutual exclusivity strategy, word spurt

摘要： 词频分布对跨情境词汇学习至关重要。然而,以往研究主要考察了基于均匀分布的跨情境词汇学习,相对忽视了真实情境中基于Zipfian分布等偏态分布的跨情境词汇学习,而仅有的研究对Zipfian分布究竟促进还是抑制跨情境词汇学习争论激烈。为此,这项研究率先建立了学习机制的宏观视角,提出单个词汇的学习效果取决于学习环境中其他词汇的整体学习效果,据此首次提出基于互斥性策略的偏态学习优势效应,认为互斥性策略的有效使用是产生Zipfian频率促进效应的关键。

关键词: 跨情境词汇学习, 语言学习, 宏观学习视角, 偏态学习, 互斥性策略, 词汇爆炸