The concept of bisimulation originates from Johan van Benthem's early work on correspondence theory, where it was introduced under name of p-relation, and later as a zigzag relation. Within that context, bisimulations have allowed for a now famous characterization of conventional modal logic as a fragment of first-order logic. Independently, and around the same time, bisimulations had also been discovered in computer science, where they have been used, notably by David Park, Matthew Hennessy, and Robin Milner, to deal with processes that are behaviourally equivalent, meaning that their behaviours cannot be distinguished by some external observer. Since then, the concept has been adapted many times, both in logic and in computer science, to suit various modal languages or process calculi. Most, if not all, of those languages and calculi can be captured in institution theory using the so-called stratified institutions introduced by Răzvan Diaconescu, Petros Stefaneas, and Marc Aiguier in order to deal with models with states. In this talk, we will revisit the classical notions of bisimulation and bisimilarity. Then, taking inspiration from coalgebraic studies of the subject, we will rephrase them in the abstract context of stratified institutions. We will re-examine the conventional relationship between bisimilarity and elementary equivalence, discuss some institution-theoretic abstractions of other key related concepts, and show how they work together to bring forth several Hennessy-Milner results for stratified institutions.