Video games typically offer to their players new, unknown places to explore. Some famous online games allow the players to discover and interact in and with whole worlds (e.g. League of Legends) or even (partially) create them (e.g. Minecraft). Players therefore have spatial positions in the virtual environment of the game that they can refer to, depending on the perspective they take: for example from inside the game world or from an observer’s perspective. The term embodiment notably describes the player’s experience of ‘living’ the virtual world and displaying a feeling of presence in it. Gee puts it as follows "players inhabit the goals of a virtual character in a virtual world" (Gee, 2008: 259). Online games offer not only a world where avatars and other characters can move in, but also different spaces and channels of interaction between the players. Our corpus consists of triadic interactions in an online collaborative game which has been designed for the purpose of research. Players are part of a submarine’s crew and have to repair some damages, while preventing the submarine from sinking since water flows into the different rooms at the end of each player’s turn. Before starting the game, we gave the players the opportunity to explore the game interface via a tutorial. During the game, they either see the plan of the submarine from a bird’s eye perspective (when it is not their turn) or they see the room they are currently in, from an immersive point of view (when it is their turn). In the first, the three players are represented by coloured circles allowing to locate them in the submarine and they can see the water level in the different rooms. The players can communicate with each other via their webcam image representation in the ‘talking head’ format (e.g. Morel & Licoppe, 2008). Drawing on a qualitative, interactional analysis of screen recordings of players from several groups, we show five different dimensions of embodiment in the context of this game: (i) verbal display of the players' projection of themselves and each other in the virtual submarine space, for example by referring to their feet being under water, (ii) embodied explanations concerning the two dimensional submarine plan, for example by pointing gestures and iconic spatial representations (iii) embodied display of engagement and activity by variation of the players' distance to the screen, (iv) clear gaze orientation towards the webcam image of one of the two co-players as a display of address (in case of the speaking person) and attention (in case of the listening person), and (v) mimicry (also behavioural alignment) with another player's body movement. We explore how these types of embodiment can contribute to group cohesion and collaboration.