How to create a numeric variable with NA or a number depending on another variable in an R dataframe

I have a dataframe (df) with three numeric score variables (New, Middle, Old). I need to calculate difference scores between the New and Middle (Diff1), and between the New and Old (Diff2). If the Middle score is NA, both difference scores need to be NA. The difference scores will be put in a new variable

Difference 1 is easy. DIFF1 <- New – Middle Always gives the correct answer.

Difference 2 works correctly when (1) the middle score is present, or the middle score is NA, and newest or oldest score is also NA.

When the middle score is NA and the oldest and newest scores are present, difference 2 is always numeric. I cannot get difference 2 to be NA.

I have tried all the permutations of naming the variables (e.g., df$New) and of bracketing that I can think of, with no success. I have also tried using == and = instead of <- . I also verified (using Rcmdr) that the variables I think are NA are in fact NA.

IN DATA

Old     Middle      New     
2483        NA      585     

CODE

Ifelse (is.na(df[,"Middle"]),   
       (df[,"Diff2"] <- NA ),  
       (df[,"Diff2"] <- with(df,New - Old)))  

ACTUAL OUT DATA

Diff1(New – Middle) Diff2(New – Old)  
NA                  -1898  

DESIRED OUTPUT DATA

Diff1(New – Middle) Diff2(New – Old)
NA                  NA