ABSTRACT
However, in some surveys there are a very large number of responses to multiplechoice questions. The 1991 Census of Population provides a good example of this phenomenon. It contains several multiple-choice questions, some of which have a large number of possible responses. As an example, the census question on occupation may be categorized into 17 different groups; see, for example, Table 86 in OPCS and GRO(S) (1992). If this were also to be subdivided by housing tenure this would quadruple the number of categories. To investigate each response category in tum would require 68 maps to be drawn. On a normal computer screen, or on a reasonably sized sheet of paper, this would require the maps to be extremely small. The ease with which geographical patterns may be spotted using this approach is also open to question. Although Tufte
(1990) demonstrates that showing 'small multiples' of a map (or other kind of diagram) is often a superior form of presentation to attempting to illustrate several variables on the same map, the notion of 'small' may well not apply here. Rather than viewing these as individual maps. each map may become sufficiently small to be considered as an individual symbol in an overall diagram. Cnfortunately, the similarity in the shapes of such symbols/maps then draws more attention to the global arrangement of items (in this example the arrangement of the 68 demographic maps on a page or a VDU) than to their content. This can perhaps be interpreted in terms of Bertin's ideas of 'associativity' for certain types of map symbols (Bertin, 1983, p. 67).
