1: \begin{abstract} 

2: Couplings play a central role in the analysis of Markov chain convergence and in the construction of

3: new Markov chain Monte Carlo estimators, diagnostics, and variance reduction techniques. Tight

4: bounds and efficient methods require an appropriate choice of coupling, which can be hard to find

5: when the set of possible couplings is intractable. To address this challenge for the

6: Metropolis--Hastings (MH) family of algorithms, we establish a simple characterization of the set of

7: all MH transition kernel couplings. We then extend this result to describe the set of maximal

8: couplings of the MH kernel, resolving an open question of \citet{OLeary2020}. We conclude with a

9: series of examples to build intuition. Our results represent an advance in understanding the MH

10: kernel and a step forward for coupling this popular class of algorithms.

11: \end{abstract}

12: