Balisage Paper: Documents Cannot Be Edited

Responses that Deny Documents are Strings

These responses reject the standard definitions of a document as a kind of string or relevantly similar entity, such as graph, relation or OHCO. For this to be a plausible response a suitable alternative definition must be proposed.

The Materialist Strategy: This strategy holds that a document is not a string, but a concrete arrangement of a quantity of matter and energy. On this view modification of a document does literally occur: modification consists in physical changes to the material document, with the document preserving its identity across these changes (otherwise it would be destroyed rather than modified).

Against the Materialist Strategy: The identification of a document with a particular concrete arrangement of matter and energy instead of a string appears inconsistent with many of the things that we say about documents. For instance we speak of the document even when there may be many physical instances, intending not to refer to any one of them, or the set, but rather to something they all represent or instantiate. Whether this apparent reference to an abstract object can be paraphrased away remains promissory. Documents certainly have physical representations, and these representations are often materially involved in scenarios of putative document modification. But whether a document can be defined as one of, or even the class of, these separate individual representations remains to be seen. ^[4]

The Social Object Strategy: This response posits as the modifiable document a social object which is constituted by (but not identified with), one string at one time and another string at another time, the changes being determined by social (including institutional and linguistic) circumstances. This approach abstracts from the physical and is consistent with the common belief that documents can be modified. The document presumably maintains a coherent identity across various textual changes and may be associated as well with any number of different physical representations. A theoretical basis for this strategy can be found John Searle's work [searle95] although Searle's theory appears to be primarily one of natural facts that in certain social circumstances "count as" social facts rather than natural objects that that in certain social circumstances count as social objects. Barry Smith has defended a theory of social objects that allows social objects to exist even without natural objects that "count as" or constitute them [smith03]. Another related perspective is Brian Cantwell Smith's notion of holding objects in the "middle distance" [cantwellsmith96]

Against the social object strategy: The social object strategy preserves the intuition that documents are modifiable, but at a considerable cost. First, it requires an ontologically challenging entity, social objects that cannot be identified with physical objects, abstract objects, or even mental states. There is also a corresponding new distinctive metaphysical relation as well: constitution, the relationship that obtains between the social object and the different strings that constitute that object at different times. That this relation cannot be simple identity is evident from the fact that it is not transitive: while the document qua social object may be constituted by one string at one time and a different string at another time, this does not imply that the two strings are themselves identical. Searle's notion of "counting as", if applied to social objects, may avoid some of the traditional problems with constitution, but it will still be hard to reconcile social objects with a naturalistic view of the world.

Responses that Deny Documents can be Modified

These strategies accept the first assertion of the triad, that a documents is a string, and rejects the third assertion, that documents can be modified. If such an approach is to be plausible it must provide a convincing alternative account of what is happening in the situations which we would ordinarily describe as "modifying a document".

The title of this paper notwithstanding, denying document modification does not necessarily require claiming that sentences like "John edited a document" never express facts about the world. These sentences may be considered idioms. As such they sometimes communicate true assertions, but they are notliterally true. Consider again the sentence "The average plumber has 3.2 children". This sentence might indeed be used to make a true assertion. And if it does make a true assertion then it is certainly reasonable to say that the sentence is a true sentence. However we would not conclude from the truth of the sentence "The average plumber has 3.2 children" that there is therefore an entity in the world which is the average plumber and which actually has some fractional number of children, as a naive Russellian interpretation of the definite description would entail. What is being denied in the rejection of the third assertion of the triad is not the truth of sentences like "John edited a document", but rather that such sentences imply a claim such as the one suggested by the first order formula:

The New Document Strategy: This response maintains that the modification of a document is actually the creation of a new document.

Against the New Document Strategy: The new document strategy has new strings created with each document modification. This seems peculiar, contrary to general notion of strings, and generally difficult to reconcile with a naturalistic view of the world. Consider the string "13571". Was that string created? How did that happen? And when? Can it be destroyed? Can it be re-created? Unless strings are material objects the conceptual apparatus of creation and destruction seems entirely metaphorical.

The Selection Strategy: This approach holds that in a typical scenario of alleged modification a new but already existing string is selected for attention. When we say that a document has been modified we mean that a different string has been selected for the purpose at hand by some particular person or persons. The physical infrastructure of analog or digital document development and publishing, in combination with social conventions and practices, is a system for recording and communicating which string is currently distinguished in this way.

Against the Selection Strategy: The selection strategy identifies documents with already existing strings. But this seems strange. Either these already existing strings came into existence at some point in the past or they have always existed. If the former then this strategy has no advantages over the new document strategy; it must still somehow account for how strings can come into existence, from what materials and in what causal circumstances. But if the latter, if strings have always existed, then assuming only that the further requirements for being a document are non-contingent (such as matching a production) we will have documents, and not just strings, existing eternally -- apparently even before cognitive agents. This seems peculiar.

A Strategy that both Redefines Document and Denies Modification.

Strictly speaking the strategy we are about to take up denies only the third assertion of the triad (documents can be modified). However because it proposes a substantially new definition of document it has much in common with strategies that deny the first assertion and so we are placing it in a separate category.

The String-in-a-role Strategy: This strategy holds that a document is a string in a particular communicative role. The string itself may be an uncreated and pre-existing entity, but the strings which are documents need not always have been documents. Being a document is a property that strings have only in particular contingent social/linguistic situations. So on this account documents have not always existed, even though documents are strings, and strings have always existed. This is because while a string which is a document has always existed, that string has not always been a document -- a string becomes a document only in the appropriate social circumstances.

Although this strategy, like the strategies that reject the first assertion of the triad (documents are strings) involves a new approach to the definition of document, it affirms rather than rejects the first assertion of the triad. After all, if a document is a string in a communicative role, then a document is a string. However, the string-in-a-role definition of document is unlike definitions such as the one in the XML specification in that it places a contingent rather than necessary constraint on strings satisfying the definition. If a document is defined as a string with certain purely formal constraints (as it is in the XML specification, where it must match a particular production in a grammar), then the things which are in fact documents are documents necessarily. This is because not only is it impossible for a string to cease to be a string or have been anything other than a string, it is also impossible for a string that matches a particular production to have ever failed to match that production or to fail to match that production in the future.

According to the string-in-a-role strategy the property of being a document is what Guarino and Welty refer to as a "rigid" property [guarino00]. Guarino and Welty define rigidity using modal logic and model theory, but the basic idea is simple: a property is rigid if and only if nothing that has that property could have failed to have that property, or could come to lose that property, (and still exist). For example, being a person is rigid because the things that are persons could not have been anything but persons and cannot cease to be persons (although they can cease to be). But being a student is not rigid because the things that are students (i.e. persons) might not have been students and can cease to be students (without ceasing to be). According to Guarino and Welty rigid properties indicate types, fundamental kinds of things, while non-rigid properties indicate roles that things of some particular type may enter into. On this view a document is not a type of thing, but a role that things of some type or other have in particular contingent circumstances.^[5]

The redefinition of document as a "string-in-a-role" is not itself a response to the inconsistency of the triad; if a document is a string-in-a-role then it is still a string, and strings cannot change. The string-in-a-role strategy rejects the third assertion, and denies that, strictly speaking, documents change.