Constructions all the way down? A Problem for Goodmanian Constructivism Jan Westerhoff January 15, 2009 Amongst the different forms of constructivism the Goodmanian variety (also known as irrealism) is one of the most extreme, and one of the most inter- esting. Unlike various localized constructivist theories it does not just claim that scientific theories or social institutions are constructs but that everything is a construct. 1 This universal claim leads to an interesting problem. To see what this problem is, note first that the relation ‘is constructed from’ is an existential dependence relation, like ‘is fathered by’, ‘is caused by’, or ‘is a conglomeration of parts of’. This means that if the antecedent of the relation did not exist, the consequent would not exist either: if there is no father there is no son, if there is no fuel there is no fire, if there are no lego blocks there is no lego structure. An intriguing question concerning dependence-relations is whether they have a foundation, that is whether there is something which is only the antecedent of the relation but does not have an antecedent itself, or at least no antecedent different from it. The discussion about the foundation of the last two examples stretches back to the early days of the philosophy. Both the existence of a first cause and the existence of partless atoms have been debated ever since. For Goodmanian irrealism at least the question of foundation is easily set- tled. Because constructions ‘go all the way down’ there is no unconstructed 1 ‘As nothing is at rest or is in motion apart from a frame of reference, so nothing is primitive or derivationally prior to anything apart from a constructional system.’ (Good- man, 1978, 12), ‘And this, as I have mentioned earlier, goes all the way down. Not all differences between true versions can be thought of as differences in grouping or marking off within something common to all. For there are no absolute elements, no space-time or other stuff common to all, no entity that is under all guises or under none.’ (Goodman, 1983, 107, note 6), ‘ We cannot find any world-feature independent of all versions. [...] No firm line can be drawn between world-features that are discourse-dependent and those that are not.’(Goodman, 1980, 212), ‘The line between convention and content is arbitrary and variable’ (Goodman, 1980, 214). 1