consider the way conceptual networks relate to mathematical formalisms. My position on material inference is derivative from Wilfrid Sellars.29 However, rather than summarizing the Byzantine structure of his language games, rules, roles, language entrance and departure transitions, and quotes , I will begin with an example from a very different field, the ongoing discussions concerning rights: of women, minorities, gays and lesbians, transgendered persons, fetuses, animals, and future generations. Conceptual connections hinge on word--word relations. Thus, 'right' has complex relations to other concepts. A right may be exercised, enjoyed, given, claimed, demanded, asserted, insisted upon, secured, waived, or surrendered. A right may be contrasted with: a duty, an obligation, a privilege, a power, or a liability. Such connections are determined by analyzing usage.
Ontological inferences involve word--world connections. For a starter consider the assumption that the only real possessors of rights are those beings who are capable of exercising or surrendering a right, and who also recognize related duties, claims, and obligations. Such word--world assumptions are tested by their consequences. This assumption would deny rights to animals, fetuses, and probably future generations. Those who find such consequences unacceptable propose other ontological bases, or word--world connections, as bases for rights: rationality, sensitivity to pain, or possession of an immortal soul. These have different consequences in according rights to the irreducible comatose, the cognitively impaired, fetuses, and animals. (See White, 1984, chap. 6).
Our concern here is not with the substantive issues, but with patterns of argumentation. At the forefront of development the process of argumentation is complex and multilayered. When once controversial issues become matters of widespread intersubjective agreement, then they are effectively treated as matters of fact. In most of the Western world, for example, there is no longer any need to argue that slavery is a fundamental violation of basic human rights or that possession of civil rights does not depend on belonging to a particular class, caste, sect, or sex. Accepting ontological assumptions as objective facts can reduce complex arguments to bumper-sticker slogans. This is a general feature characterizing developing dialogs. As Gilbert Harman (1973, 1980) has argued, any attempt to regulate belief acceptance by Bayesian norms, or even highly simplified Bayesian-type schemas, would make life impossibly difficult. Each new fact or bit of evidence would require reconditionalizing the relative strengths of a network of interrelated beliefs. This is a practice with little survival value. It quickly overloads our information storage and processing capacities and makes practical decisions almost impossible. A practice with a much higher survival value is an all or nothing acceptance practiced for most basic beliefs. This accords with Davidson's well-known contention that normal discourse presupposes the acceptance as true of a large but amorphous body of claims.
A similar conclusion stems from Kyburg's (1985) account of the confirmation of quantitative laws. The measurements used to test quantitative laws inevitably involve errors and require a theory of error. Even the simplest measurements presuppose a network of assumptions, which are also subject to measurement, testing, and error. To avoid regress into a quagmire of unending testing scientific practice treats established quantitative laws as a priori commitments of a body of knowledge and acceptance of a new quantitative law as a matter of practical certainty relative to that body of knowledge. In a similar spirit Shapere
29 See Sellars (1963), Essays 1, 8 and 11 and (1974), Essays 5, 9, 14, and 17.