3 Results

The CQP queries worked best for all subtypes of nominal ellipsis in both languages. There was a high recall for nominal ellipsis, but unfortunately precision was still relatively low; more sophisticated queries have to be designed in the future. Some subtypes of verbal and clausal ellipsis were easy to query with CQP (eg sluicing) others were more difficult to spot. Nominal ellipsis occurs mainly in certain text types (eg texts with many adjectives and nominal style or limited space for printing) but also in the context of certain topics (involving numerals, comparisons, contrasts…). Written discourse in general offers more possibilities for nominal ellipsis due to structural complexity, lexical density, nominalization and longer noun groups. Nominal ellipsis is assumed to be more frequent in German because English can use substitution with ‘one’instead and avoids nominal ellipsis when it could lead to ambiguity due to the morphological characteristics of the language. Surprisingly in our corpus, nominal ellipsis is more frequent in both spoken and written English than in German. Verbal ellipsis often co-occurs with proper names or personal pronouns and therefore also depends on the text topic and the level of interaction in discourse. Verbal ellipsis typically occurs in text types with otherwise rich verb phrase structures to avoid verbal repetition; some subtypes require hypotactic or parallel structures, often involving contrasts between two or more members of a semantic category (eg The parents ate cake, and the children [] cookies).

Due to the lack of exact correspondence between the English and German verbal system, there are more differences between English and German verbal ellipsis than with regard to nominal ellipsis. Filler words, redundancies, anacolutha and less clear sentence boundaries in spoken language make queries for spoken registers more difficult than for written registers. Often larger parts of texts have to be taken into account to determine whether a certain structure is an ellipsis, an anacoluthon or a sentence break, regional variation or simply an error where people forgot to complete a sentence. The special syntax of spoken language and differences between English and German morphology (high frequency of zero derivation/word-class ambiguities in English, declension of adjectives/pronouns as ellipsis remnants in German) have to be considered in queries as well as the particularities of ellipses as syntactically incomplete or–without an appropriate context–even deficient structures. Tagging errors resulting from these untypical syntactic patterns are another aspect that has to be taken into account when formulating CQP corpus queries. Table 2 shows the normalized frequencies of ellipsis subtypes per 100,000 words found in 4 German and English registers of GECCo.