Marginal event features do not determine episodic binding

Consider two equally weighted sets of event triples: $\{(A,L,1),(B,R,2)\}$ and $\{(A,R,1),(B,L,2)\}$. Their separate distributions of item, place and time are identical, but $P(L\mid A)$ is respectively $1$ and $0$. Thus identical marginal knowledge does not determine the joint event representation. A behavioral test that can be solved from separate feature lists does not by itself establish episodic binding.