This is certainly not a reasearch question, but I got no answers on math.stackexchange so I duplicate the question here.
Siebenmann on page 1 of his manuscript gives an example of an ENR space (euclidean neighborhood retract) whose group of homeomorphisms is not locally contactable. Take the sphere $S^3=R^3\cup\{\infty\}$ and a sequence of wild non-cellular arcs $A_n$ which are copies of an arc in a unit ball in $R^3$ translated by $(4n,0,0)$. Then factorize by sending each arc to a point. The resulting space is said not to be locally contactable since there are homeomorphisms arbitrary close to the identity which permute images of $A_n$ and none of these homeomorphisms is isotopic to the identity. The nonexistence of isotopy is not clear. Could anybody point where (and how) are exactly the wildness and non-cellularity used?