As is well-known, code-mixing is particularly frequent at clause boundaries and with elements expressing pragmatic meaning. However, most of the literature has focussed on switching of simple elements such as conjunctions and discourse markers. This paper, in contrast, analyses clause peripheral switching involving two complex constructions: left dislocations and pseudo-clefts. The data are from English-Spanish bilingual conversations recorded in Gibraltar. A great majority of the bilingual constructions in the corpus belong to a few types occurring with a restricted set of lexical items. A vast amount of such highly recurrent strings in the data confirm the hypothesis that complex multi word strings that are switched together constitute units in code-mixing, i.e. they are processed as single lexical items.

众所周知，代码混合在从句边界和表达语用意义的元素中尤为常见。然而，大多数文献都集中在简单元素的切换上，例如连词和话语标记。相比之下，本文分析了涉及两个复杂结构的子句外围转换：左错位和假裂缝。数据来自在直布罗陀记录的英西双语对话。语料库中的绝大多数双语结构都属于少数类型，这些类型具有有限的词汇项集。数据中大量的这种高度重复的字符串证实了这样一个假设，即复杂的多词串被切换在一起构成了代码混合中的单元，即它们被处理为单个词汇项。