Duffy and Knuth (1990) address a similar point when they suggest that the essentially exploratory type of reading which a net hypertext elicits denies the importance of goals and/or tasks. Reading, like any other information seeking activity, is purposeful: students read texts to pass exams or write essays; technicians read documentation to solve problems, and so on. Few writers aim merely to transfer a body of factual information to the reader, even if the local semantic connections are made explicit. The level of analysis or conceptualisation is simply too low. Thus, Duffy and Knuth suggest in a learning context that "simply learning the links in a database is certainly not going to result in someone thinking like the professor."

An aspect of hypertext systems that is often proposed as an advantage is their potential for supporting very large bodies of information. However, Bereiter and Scardamalia (1987) contrast skilled and unskilled (or novice) writers and differentiate between the associated story re-telling and story transforming abilities. Similarly, teaching and learning in the broadest sense are concerned with the transmission of knowledge not simply the transfer of relatively unstructured information – we may know facts but we understand theories and perspectives. The creation of very large hypertext systems may be of limited value unless they are structured in such a way as to be coherent to the reader.

While a reader might spend considerable amounts of time ‘browsing’ paper documents, this activity is rarely haphazard. Browsing is the process by which scanned material is evaluated for relevance according to a set of needs or interests (which may be capable of only vague definition by the reader), with irrelevant material being quickly rejected. This activity is enabled by the conventions of text layout and typography. A skilled reader need often only swiftly glance at the headings/sub-headings and opening paragraph in order to gain a good impression of the scope of a document.

However, it is not clear that this type of information-seeking can be supported by hypertext systems that entail relatively flat semantic networks or by computer systems which use small sized displays that are undifferentiated typographically. Richardson, Dillon and McKnight (1989) carried out an experimental study in which subjects used either a small or large computer display to search for items of information in a text. While no difference was found in terms of success, a clear preference was shown for the larger display. Of more significance was the finding that, even allowing for the fact that the small display showed less text, the text was manipulated significantly more in the small display condition than in the large display condition. This suggest that the readers’ browsing may have been hampered by the small amount of text displayed at any one time.

The problem of adequately supporting browsing has been recognised in the design of some hypertext systems. For example, KMS has a pronounced hierarchical structure and rapid system response which are designed to support browsing. The consistent structure within and across hypertexts help orient the reader and the overview nature of the frames provides expectancies about what an unread section is likely to contain (in terms of content breadth and depth). Although there has been little experimental investigation of the potential advantages of various hypertext structures, there is evidence to support the adoption of hierarchical structures.

For example, Simpson and McKnight (1990) report an experiment in which subjects were provided with a range of structural cues to support them in accessing a hypertext:

However, if simple hypertext webs are not appropriate vehicles for teaching or learning (transmission of knowledge) they may have an important rôle in the preparation of conventionally structured texts.