Go through Cummins model;
conclude that all analysis of properties is 
non-mereological

Establish constitution as an Explanatory Relation (ER).
Give examples of constitution explanations
Give examples of ER switching

Then: 
set out examples of explanatory constitutional *hypotheses*:
this is what we sometimes call "rational reconstructions".
For example, in linguistics, with ordered transformations:
people misread these as a causal history of the 
psychological processes the speaker goes through
(for: if they're *not* causal accounts, then what/how
exactly do they explain?)
There is *explanation* going in a theory w/ ordered
transformations; but the ER involved is not causation.
Instead, the ER is constitution.  The transformational
theory provides a (general, unified) theory of why this
or that sentence *counts as* (doesn't *count as*)
grammatical to this speaker/group of speakers.

This is how we explain normativity (w/o explaining,
e.g. the psychological processes involved, which would
appeal to the causation ER): we provide a theory which 
*explains* acceptance/rejection behavior (but doesn't
explain it *causally*).  Examples: Maher, Laudan, Pearl.
What I'm calling "acceptance/rejection behavior" is:
cases of something *counting as* X (or not so counting).
We are faced with specific examples of counting as/
not counting as (grammatical, conversationally cooperative,
ethical, rational), and we want a general, unified theory
which explains this data.  (After all, that's just what
scientific theories do generally: provide a unified, 
general framework in which the various data in question
can be embedded.  Constitutive theories (of the 
reconstructive sort) fit in the mold of scientific theories 
generally.


*****


Hypothesis:
to get ER symmetry, require looping
looping only possible for S-T relations
(hence: no looping for criteria; for functions(?))
