The heatmap function in R is supposed to help a human being interpret the relative values of the elements of a matrix. However, it seems not to color cells consistently within a given plot, which is a severe obstacle to interpreting the relative values correctly.
For example, let’s generate some data by concatenating columns of normal random variates:
foo <- cbind(replicate(10,rnorm(10)))
Now if we correlate the columns of foo, we can verify that we get 1’s in the diagonal entries since the correlation of any column with itself is 1:
cor.matrix <- cor(foo)
But when we plot:
heatmap(cor.matrix,Rowv=NA,Colv=NA)
(we’re suppressing the dendrogram reording here, although that doesn’t seem to matter)
the diagonal cells are not colored uniformly, as you can see: 
Can anyone explain what’s happening here?
By default heatmap scales by “row”.