Suppose I construct the following 3D array
n = 3;
A = zeros(n,n,n);
A(1:n^3) = 1:n^3;
which gives
>> A
A(:,:,1) =
 1     4     7
 2     5     8
 3     6     9
A(:,:,2) =
10    13    16
11    14    17
12    15    18
A(:,:,3) =
19    22    25
20    23    26
21    24    27
One can see how matlab indexes a 3D array from the above example. Suppose I would like to access (ii = 1, jj = 3, kk = 2) element of this array, which can be done by
>>A(1,3,2)
ans =
16
Alternatively, I can use the following form based on the matlab indexing rule demonstrated above
A(ii + (jj-1)*n + (kk-1)*n^2)
as an example, for ii = 1, jj = 3, kk = 2, I get
>>  A(1 + (3-1)*3 + (2-1)*3^2)
ans =
16
To illustrate the problem, I define the following 3D meshgrid (say for the purpose of index manupulations which is not relevant here):
[j1 j2 j3] = meshgrid(1:n);
If I am not wrong, common sense would expect that
A(j1 + (j2-1)*n +(j3-1)*n^2)
to give me the same matrix based on the above discussions, but I get
>> A(j1 + (j2-1)*3 +(j3-1)*3^2)
ans(:,:,1) =
 1     2     3
 4     5     6
 7     8     9
ans(:,:,2) =
10    11    12
13    14    15
16    17    18
ans(:,:,3) =
19    20    21
22    23    24
25    26    27
From this I see that if you want to get the same 3D array you actually need to use
>> A(j2 + (j1-1)*3 +(j3-1)*3^2)
which is very strange to me. I am posting this issue here to see what other people think about this.