reshape problem: id and variable names not being recognized

Hi everyone. Long time listener, first-time caller here.

I have a data set that's been melted with the excellent reshape package, but
I can't seem to cast it the way I need to.

Here's the melted data's structure:

str(mdat)

'data.frame':   6978 obs. of  4 variables:
$ VehType : Factor w/ 2 levels "Car","Truck": 1 1 2 1 1 2 1 1 1 1 ...
$ Year    : Factor w/ 6 levels "2003","2004",..: 5 1 5 6 6 2 2 3 2 5 ...
$ variable: Factor w/ 1 level "mpg": 1 1 1 1 1 1 1 1 1 1 ...
$ value   : num  22.4 21.5 22.6 22.4 25 ...

For the purpose of testing, I have stripped out all the variables except for
"mpg."
Casting it without specifying any ids or variables works fine:

cast(mdat,,mean)

 VehType Year      mpg
1      Car 2003 22.03623
2      Car 2004 21.94160
3      Car 2005 21.77286
4      Car 2006 21.49105
5      Car 2007 21.38180
6      Car 2008 21.56873
7    Truck 2003 16.91461
8    Truck 2004 16.88771
9    Truck 2005 17.19801
10   Truck 2006 17.48225
11   Truck 2007 17.40694
12   Truck 2008 17.74042

I should then be able to make a crosstab of the means by writing a formula,
right? It fails, though:

cast(mdat, VehType ~ Year | mpg, mean)

Error: Casting formula contains variables not found in molten data: mpg

When I make the same table by using "variable" instead of the name of my
variable, it works:

cast(mdat, VehType ~ Year | variable, mean)

$mpg
 VehType     2003     2004     2005     2006     2007     2008
1     Car 22.03623 21.94160 21.77286 21.49105 21.38180 21.56873
2   Truck 16.91461 16.88771 17.19801 17.48225 17.40694 17.74042

Why can't it find the "mpg" variable when I call it explicitly?

Thanks,
Matt Frost

       [[alternative HTML version deleted]]