The basic idea of a rational law is this. Suppose a belief and a desire cause an action. Then the action should promote the satisfaction of the desire, given the truth of the belief. This is a version of Fodor's point in the initial quotation, which we have made before with respect to our example. Jerry's belief that the cup in front of him contains decaffeinated coffee and his desire to have decaffeinated coffee lead him to will to move in a way that results, if the belief is true, in his desire being satisfied. There is surely something quite appropriate about this. Suppose Jerry's belief and desire led him instead to a movement that results in splashing the coffee into his forehead. That would in some sense be inappropriate. As Fodor points out, it is perfectly conceivable that beliefs and desires, conceived as internal states, should have such inappropriate effects, but such effects of beliefs and desires are not what cognitive psychology is all about--they are not grist for the cognitivist's mill.
It will help discussion to define a relation among four propositions P, R, Q and C, thought of as the contents of a belief, a volition, and a desire, and background conditions: (the belief that) P rationalizes (bringing it about that) R, relative to (the desire that) Q, given (the condition that) C. We'll write this
(P, R
C)
We think of this relation as one of incremental rationality.
If C is given, then if P is true also, bringing it about that R will guarantee (or at least promote) bringing it about that Q.
To see one source of the need for, and role of, the background
condition, consider the fully opaque contents of the tokens in our
basic example. Let P = the proposition that the cup in front of the
owner of b contains decaffeinated coffee, Q = that the owner of
d (a token of ``I drink decaffeinated coffee'') drinks decaffeinated
coffee, and R = that the owner of v picks up the cup in front of
him and brings it to his lips. Then it is not the case that bringing
it about that R is true will bring it about that Q is true, if P
is true--without the further condition that b, d, and v
belong to the same agent. This is our reason for saying that without
some appeal to circumstances, the rationality of laws of cognitive
psychology cannot be understood.
Now consider a law, like our 
, to the effect that a belief
token b of type B and a desire d of type D, in the same agent,
will cause that agent to have a volition v of type V. Such a law
is rational if for every b, d, and v that instantiate it,
(
,
where is the fully opaque content of b,
and where C is the condition that b, d, and v belong to the
same agent (and conditions are normal). When beliefs and desires
cause volitions in accord with such laws, we can say that they
motivate in two senses. The belief and desire cause the volition, and
they rationalize it.
We should say a word about normal conditions. In our conception of cognitive psychology, the specificiation of an environment, which includes specification of a range of normal conditions, will be an important part of the psychological theory for a kind or species of agent. In our conception, agents are attuned to certain environments, and the apparatus of perception and belief is used to pick up and store information about factors that vary within those limits. We see the circumstantial nature of thought as one aspect of this attunement. The way visual information is used to guide our hands, for example, involves attunement to the normal relations between the orientation of limbs and the orientation of eyes. Such attunement only goes so far, however, and that is where the function of systems of storeable, manipulable, relatively context insensitive representations come to the fore--for instance, as they are involved in belief. These points are rather tangential to our main aim in this essay, however.
The conception of a rational law explained a few paragraphs earlier makes
it clear that fully opaque explanations, involving rational laws,
would be fine. By subsuming behavior under causal laws, they would
also subsume it under rational laws. But as we mentioned, we don't
ordinarily use attitude reports to attribute fully opaque
belief--what we say when we say ``Jerry believes that...'' is not
understood on the model ``Jerry 
that...'' We want
then to consider how explanations using ordinary opaque and
transparent attitude attributions could be adequate, and how they can
go wrong.
The basic idea is as follows. When we explain an action in terms of a
belief and a desire, we are basically explaining the occurrence of a
volition of a sort for which the action is a more or less basic mode
of execution. The explanation will be correct only
if it can be unloaded down to an instance of a rational law where the
consequent volition is so executable. To see what this means, let's
work through our example.
Consider this explanation of Jerry's executing movement M, thereby picking up c and bringing it to his lips:
Jerry believes c contains decaffeinated coffee and desires that he drink some decaffeinated coffee.
What does this come to? Here's what:
that c contains decaffeinated coffee in
virtue of 
that the cup in front of him
contains decaffeinated coffee together with the external circumstance
that c is the cup in front of him. He 
this in
virtue of owning a belief b whose fully opaque content is that the
cup in front of the owner of b contains decaffeinated coffee.
that he drink some decaffeinated coffee in
virtue of owning a desire d whose fully opaque content is that the
owner of d drink some decaffeinated coffee. above, the belief and desire cited explain the action.
Now consider the case in which Jerry looks in the mirror. Here he doesn't actually perform any action. But we want to understand why, knowing about his belief and desire, we shouldn't expect an action such as his reaching for the cup by executing movement M.
The type of the volition and the desire are just as before. Further,
Jerry again that c contains decaffeinated coffee in
virtue of that the cup in front of him contains
decaffeinated coffee, in the circumstance in which c is the cup in
front of him. He this, however, in virtue of owning a
token b' with the fully opaque content that the cup in front of the
man whom the owner of b' is watching contains decaffeinated coffee.
The belief b', the desire d, and a volition v of which the
(missing) action would be an execution are not instances of a rational
law. They are not instances of , because the fully opaque
content of the belief is not right. Moroever, if there were a law
linking beliefs of this type, with this fully opaque content, with
desires and volitions of the types of d and v, it would not be
rational. Lucky folks like Jerry would get a drink of coffee, but in
most cases the agent would be knocking over things, pawing empty
space, or irritating irritable gorillas.