A  number of scholars studying language from a functional point of view (cf., Danes, 1964; Firbas, 1964, 1966; Halliday, 1985; and Quirk, Greenbaum, Leech, & Svartvik, 1985) have recognized the role that the linear order of words plays in written texts to compensate for the loss of explicit clues about information focus available through intonation in the spoken language. In spoken English, emphasis is achieved by manipulating intonational contours so that informationally important elements receive nuclear stress in tone units. Since those tone units can be conveyed directly, this creates few problems. In written English, however, those tone units must be inferred from their graphic representations. Consequently, the reader has to rely more heavily on certain conventions of interpretation — especially conventions that relate the structure of the tone unit to the structure and linear ordering of the syntax. While speakers can directly convey their emphases through stress and pitch, the writer must construct a clause, indeed the whole sentence (and text), carefully — with judicious word order, punctuation, and discourse implication (about what the writer considers given or new information for the audience) — to recapture the loss of such explicit markers of emphasis in speech. The loss of prosodic markers of information focus in written English highlights a major problem that arises for all writers who must rely on word order to carry information focus. That problem might be called the poverty of surface syntactic information, i.e., the restricted number of distinctions linear order can make as an information carrying device. Linear order can provide only two possible pieces of information: (1) two constituents can be sequential or not (i.e., linear order can describe adjacency relations) and (2) if the constituents are sequential, their order may be either X - Y or Y - X (i.e., linear order can describe precedence relations).

Through the relationships of adjacency and precedence, linear order – supplemented by morphology and intonation (in speech) – provides information about grammatical relationships of subject, object, etc; about thematic structure of theme/rheme or topic/comment (the psychological subject discussed by Sandmann, 1954); about the participant roles of agent, patient, etc (the logical subject of Sandmann, 1954); and about information structure of given and new (cf., Prince, 1981). As Chafe (1976, p. 27) puts it, "A noun in its sentence plays many roles, or has the potential of doing so."

So part of the difficulty in determining the contribution of linear order to our understanding of language results from the interplay of various language processes. Those linguistic processes conspire to determine the linear order of clausal constituents and the linear order of clauses themselves. Out of context, or in a controlled context, it is possible to isolate the functions of end-focus, thematic prominence, or euphony in determining linear order, but in vivo, as it were, it becomes more difficult to characterize precisely the contribution of individual language processes. Neutralization processes provide an analogous situation to the conspiracy described here. Neutralization rules, at any level of linguistic analysis, eliminate a potential contrast, thereby creating the potential for ambiguity. At some level of analysis, one would want to explain the ambiguity by positing different forms, which are no longer overtly distinguished at the surface level because of a neutralization process. Likewise, different sentence types seem to neutralize some distinctions between grammatical, thematic, psychological, and logical subject, for example, in order to express some distinction that otherwise may be missed. That is, in appropriate contexts, language users may need to be explicitly clear about information structure, or thematic structure, etc, for efficient language processing. For example, when speakers need to be explicit about the packaging of information within a clause, the cleft and pseudo-cleft constructions allow a distinction between given versus known information. And as Pawley and Syder (1983) argue, certain structural strategies (or packages of information) may be better suited in certain communicative contexts because of physical or cognitive constraints, and many of those constraints vary as context varies, arguing against any notion of an ideal relationship between form and function.

To compensate for the poverty of surface syntactic information in written English, writers commonly employ two conventions or strategies of interpretation — end-focus and thematic structure. Two phenonmenon that are difficult to measure in data sets because of their context-sensitive nature, at least while corpus tools tend to be context-free.