The Matlab comparison isn't that useful though, because the languages are geared toward different problems. Matlab at its core is linear algebra ("engineering"), while Mathematica is based on symbolic computation ("algebra"). That's a pretty huge difference. They have both branched out I'm sure, but the history remains.
People have a similar confusion over Matlab/Julia vs. R. They're both scientific computing, but R is for statistics. (R has matrices and linear algebra, though they are fairly terrible compared to the data frame operations, and seem second class.)
I would think of each language as having a core data structure:
1) Mathematica/Wolfram - expression/symbol
2) Matlab - matrix/vector
3) R - data frame (columns of heterogeneous type)
Julia is trying to subsume some of R, but it really is almost putting two different (but related) languages under one roof.