5.1.1　Limitations of existing research on visual-verbal relations

The existing models on visual-verbal relations have been reviewed earlier in Section 2.2.3.Since they are all conducted with a logico-semantic perspective，they can be regarded as contributing to the issue of local coherence in multimodal discourses.These existing models，as the review suggests，provide very useful perspectives in examining visual-verbal relations in multimodal documents，but we must admit that the research on how verbal texts and image combine in a coherent whole is still in a stage of fast development and unsettled yet.The existing literature on text-image relations has several major weaknesses：

（1）“The relationships between text and image drawn in the Barthian tradition are not on their own sufficient for incisive analyses of how multimodal documents are functioning or，possibly more useful，not functioning.”（Bateman，2008：157）

（2）A common problem in all the proposals on visual-verbal relations reviewed in Section 2.2.3 is the uncertainty and subjectivity in attributing a specific case to a certain category.The root of this problem is that the terms such as anchorage，complementarity，concurrence，reinforcement and instantiation are not raised on the basis of rigid semantic classification，but are actually discrete semantic labels which can never be exhausted and systematically classified.This situation of labeling is becoming more and more serious since advertisements are becoming increasingly innovative with the development of the advertising industry as a highly innovative and creative professional practice nowadays.In face of the indefinite possibilities of semantic combination of semiotic resources in advertising，the pinning down of the relationship between language and image in a specific case might be a highly challenging and subjective job.There will be some cases that cannot be properly attributed to any semantic labels in the frameworks reviewed above;and it's predicted that the quantity won't be small given the degree of unconventionality in advertising in the information age.For example，in the advertisements in my study the visual-verbal relations tend to be loose and lack cross-reference，and thus are sometimes difficult to be described by a certain semantic label.Thus a more effective way of conceptualizing visual-verbal relations is needed.

（3）Moreover，most of the models introduced in Section 2.2.3 draw upon the concepts of SFL paradigm，especially the relations in clause complex developed by Halliday（1994）for the analysis of the verbal text.While concepts such as elaboration，extension，enhancement and projection work well with relations within a verbal text，it remains doubtful whether they could be also effective in labeling the relationship between a verbal text and a picture whose meaning is always open to different interpretations.The appropriateness of other models must also be investigated with caution.For example，it is uncertain whether RST is appropriate for dealing with multimodality at all since there are substantial differences between how language is processed by speakers and hearers and how visual images are processed by their“readers”（Bateman，2008：156）.Although the production and consumption of any pictorial signs are governed by rules which are socially formed and generally known to individuals who are from the same cultural community，they can always evoke diverse potential interpretations in specific contexts.

（4）In terms of methodology，most of the above-mentioned models are conceived on the basis of selected samples and theoretical reasoning.The limitation is（i）the lack of a representative collection of data，and（ii）the lack of more operational criteria of classifying and measuring these verbal-pictorial relations which make it difficult to conduct empirical triangulation to verify the validity，especially in the respect of reader's mental reaction.

To summarize，in the case of the English learning advertisements which are complex，creatively designed，and highly individualistic discourses，neither the models developed from linguistic-internal relations nor those based on a limited number of rigid semantic categories can fully accommodate the complexity and flexibility of semiotic strategies in modern multimodal advertisements.Since multimodality is an emergent theory and many of the theoretical concepts and tools are still unsettled（Jewitt，2006：3），to explore the inter-semiotic relations in multimodal discourses is the most important，yet the most difficult issue in multimodal discourse analysis（Zhang，2009）.To uncover how different modes interaction to form a coherent multimodal discourse more research efforts need to be made in both theoretical and methodological aspects.