on its intelligibility who have the burden of proof. (One presents ademandfor a
definition when one doubts that the target in question is well understood.)
Ideological and methodological primitiveness are not equivalent. For one thing,
whether something is methodologically primitive might be something that is relative
to a context; in one kind of metaphysical inquiry, it might be OK to appeal to
grounding without defining it, but in a different kind of context, perhaps not.
(Perhaps in the metaphysics of ordinary objects or social kinds, it doesn’t really
matter whether grounding is in any way primitive, but it does in“fundamental”
metaphysics.) But even setting aside this difference, there are others. Ground might
be methodologically primitive but not ideologically primitive. But if this is so, then
one has the dialectical right to theorize in terms of grounding in advance of any
attempt to define“ground,”even though there is a definition to be had. Ground
might be ideologically primitive but not methodologically primitive. But it would be
unfortunate if this were so: if it is not methodologically primitive, then one must
attempt to define or analyze the notion before employing it, but if it is ideologically
primitive, no such attempt will succeed. Perhaps we should refrain from using
expressions that suffer from this combination.
Finally, what is it to say that grounding is metaphysically primitive? For me, to say
that grounding is metaphysically primitive is just to say that the grounding relation is
a perfectly natural relation, i.e., is among those properties and relations that are most
real.^4 Perhaps this is to impose an external framework that grounding theorists might
be hostile to. I want to avoid begging questions here. For now, it would be good to have
alternative characterizations of metaphysical primitiveness that do not have this
consequence, even if at the end of the day this characterization is the best one to adopt.
Three alternatives suggest themselves. First, if we think, along with Schaffer, that
all entities enter into grounding relations, then we could say that grounding is
metaphysically primitive 1 if and only if nothing grounds the grounding relation.^5
The other two alternatives assume that grounding is a relation between facts or
propositions. Second, we might say that a relation is metaphysically primitive 2 just in
case there are ungrounded facts that have that relation as a constituent.^6 Third, we
(^4) Joshua Spencer has suggested to me that if grounding is a perfectly natural relation, there cannot be
duplicates whose parts stand in opposing grounding relations. (For example, it is not possible that I am
grounded in my parts while my duplicate grounds his parts.) I do not regard this as a cost but rather a
benefit of the proposal. This relies on Lewis’s (1986: 59–63) idea that duplication should be understood in
terms of 1–1 correspondences between parts that preserve perfectly natural properties and relations, which
I accept. 5
In section 8.2, there will be a brief discussion of whether Schaffer now intends to take grounding to be
maximally general in this way. 6
Bennett (2011a) and deRosset (2013) deny that grounding is metaphysically primitive 2. Alex Skiles has
suggested to me that we should make further distinctions here, corresponding to the different ways in
which a property or relation can appear as a constituent in a fact. It might appear as apredicative
constituentand it might appear as asubject constituent. (Note the difference between the fact thata
groundsband the fact thatgroundingis a relation.) Corresponding to these two ways of appearing in a fact
are twofiner-grained ways of understanding metaphysical primitiveness. This is an interesting suggestion,