[R] anova on data means

[Thomas Miller]

I am transitioning from SAS to R and am struggling with a relatively simple
analysis.  Have tried Venables and Ripley and other guides but can't find a
solution.

I have an experiment with 12 tanks.  Each tank holds 10 fish.  The 12 tanks
have randomly assigned one of 4 food treatments - S(tarve), L(ow), M(edium)
and H(igh).  There are 3 reps of each treatment.  I collect data on size of
each fish at the end of the experiment.  So my data looks like

Tank  Trt   Fish   Size
1      S     1      3.4
1      S     2      3.6
....
1      S    10      3.5
2      L     1      3.4
....
12    M     10      2.1

To do the correct test of hypothesis using anova, I need to calculate the
tank means and use those in the anova.  I have tried using tapply() and by()
functions, but when I do so I "loose" the treatment level because it is
categorical.  I have used 
Meandat<tapply(Size,list(Tank, Trt), mean)

But that doesn't give me a dataframe that I can then use to do the actual
aov analysis.  So what is the most efficient way to accomplish the analysis

Thanks

Tom Miller