1.4.1　The reason for using the English learning advertisements as data

The choice of the English learning advertisements as the sample for studying discourse coherence is based on considerations in the following aspects：

Firstly，in the respect of research approach，large and principled collection of natural texts is highly beneficial because it enables discourse analysts to respond to how language，or expressive resource in multimodal sense，is actually used，rather than relying on intuitions or artificial examples，and have access to a richer basis for constructing theoretical models（Bateman，2009：59）.Therefore，the English learning advertisements as natural authentic data can provide a sound basis for the theorization and analysis in this book.

Secondly，in terms of methodology，because discourse features vary greatly from genre to genre，a representative sample of discourses is needed to ensure that the theorization is based on detailed，sound empirical analysis.As Bateman（2009）proposed，the main task for the immediate future for multimodality was to undertake detailed empirical analyses with respect to more representative samples of multimodal documents.These documents collected need to be explicitly situated with respect to some selected set of dimensions drawn from multimodal genre.Therefore，altogether 97 English learning advertisements were collected to form a medium-size corpus of such representative multimodal discourses.

Thirdly，besides the social meaning mentioned at the beginning of the book，the English learning advertisements are also complex in content and semiotic configurations，that is，they can accommodate the multiple layers of coherence resources and various semiotic strategies.Therefore，they serve as good materials for a systematic analysis and empirical research of coherence construction in multimodal discourses.