Linear embeddings of contractible and collapsible complexes

(1) We show that if a presentation of the trivial group is "hard to trivialize", in the sense that lots of Tietze moves are necessary to transform it into the trivial presentation, then the associated presentation complex (which is a contractible 2-dimensional cell complex) is "hard to embed in $\mathbb{R}^3$", in the sense that lots of linear subdivisions are necessary. (2) For any d, we show that all collapsible d-complexes with n facets linearly embed in $\mathbb{R}^{2d}$ after less than n barycentric subdivisions. This is best possible, as cones over non-planar graphs do not topologically embed in $\mathbb{R}^{3}$.