Theoretical foundations: improvisation, the BwO, and embodiment in VR

Improvisation is often described as the ability to spontaneously adjust to emergent situations in embodied interaction (Schön Reference Schön1983). Cecil Taylor, the pioneering free-jazz pianist, captures this when he writes that improvisation is “the space where patterns and possibilities converge” (Taylor 1987). In dance, Rothfield (Reference Rothfield, Guillaume and Hughes2011, 213) put it another way: “Improvisation is an affirmative relationship to the concrete moments of chance to be found in the passing moment [...] It is the moment-by-moment activity whereby one body becomes another.” Finally, in linguistic anthropology, Duranti and McCoy (Reference Duranti, McCoy and Stanlaw2020:797) consider “a combination of conformity and creativity, unconscious habits, and conscious monitoring” to lie at the heart of improvisation. Within VR gaming, improvisation takes on an added ludic dimension, as players must adjust their bodily actions and gestures to the affordances of a sometimes unpredictable, often confounding, and occasionally hostile virtual environment. These modifications are not exclusively attributable to personal skill or practice, since they are influenced by the technological interface and depend on the actions of other player-participants, which may either restrict or facilitate a player’s ability to improvise. VR thus presents a site of embodied improvisation where agency is contingent on both the players’ actions and the system’s affordances.

In this account, my use of improvisation draws selectively from each of the formulations above, since I treat every improvisatory move as one that carries traces of prior practice while simultaneously opening the way for new interpretation, which means that it operates with both anaphoric recall and cataphoric potential. I therefore understand improvisation as the ongoing reorganization of embodied capacities in situations where participants respond to shifting constraints, variable affordances, and emergent alignments, with each adjustment shaped by the history of previous actions and by the possibilities it creates for what follows.

Deleuze and Guattari’s Body without Organs, although developed within a metaphysical account of how organization takes form, provides a vocabulary that allows us to describe moments when capacities are redistributed across physical, perceptual, and virtual planes.Footnote ⁵ The BWO is, thankfully, not an actual body with its organs removed. In fact, it does not have to be a human body (it can be a social body, a machinic body, a linguistic body (system), or a digital body). We can think of the BWO as a conceptual plane that allows us to consider a body as a dynamic assemblage of intensities and potentialities. The BwO transcends the usual boundaries of the body, allowing for fluid reconfigurations of bodily capacities in response to immersive stimuli and situational demands.

The BwO is developed by Deleuze and Guattari within a philosophical project that describes the conditions under which organization takes shape, which means that the term operates at the level of potentiality rather than at the level of observable behavior. I use it differently, since my concern involves the specific moments when players reorganize their capacities across physical, perceptual, and virtual planes in response to technological or interactional disruptions. My goal is to show how the BwO, treated as a figure for shifting distributions of capacity, can illuminate the concrete reassemblies through which players regain coherence after perceptual breakdowns in virtual environments. In their work, Deleuze and Guattari (henceforth D&G) conceptualize the organism as the body in its traditional form, structured by notions such as identity, meaning, and purpose. The organism is the body under constraint, in which each organ plays a fixed, predetermined role. For D&G, any organism can enter into a relation with the BwO when its habitual organization loosens and its capacities are redistributed across a plane of immanence.

The BwO then refers to an abstract field where affective intensities and connections form outside the conventional structures that organize bodies, identities, or functions. The BwO represents the affective spark,Footnote ⁶ the unrealized potential of the body—what the body could be if it were freed from constraints imposed by both nature and society. D&G encourage us to question the organs as to their function and purpose. What would an organ be like if it had a different function?Footnote ⁷ My interpretive answers below also serve to introduce the concepts of deterritorialization and reterritorialization, which are further defined in the next section.

Consider the following cases:

1. 1. While we typically use our mouths for eating and speaking, the mouth’s primordial function is to seek food and explore the environment. As babies, we instinctively approach objects to mouth them, testing for edibility long before we have learned to speak and before we are systematically prevented from putting everything in our mouths. At that point, we are not aware of objects qua objects, but we engage with them by feeling their surfaces, interpreting their shapes, and assessing their potential through direct interaction. This is a sort of deterritorialization of the mouth’s typically assumed function.

2. 2. We typically use our eyes for seeing, but humans also possess underdeveloped echolocation abilities. People with impaired vision often learn to rely on echolocation to navigate their surroundings. My father, who became blind as an adult, would use a cane to tap the sidewalk and listen to the echoes bouncing off nearby buildings, helping him navigate the crowded streets of Ciudad de México and New York City. His enhanced ability to echolocate didn’t come from formal instruction: it was a reorganization of his senses in response to a changing environment and the changing affordances of his own body. His adaptation represents a form of deterritorialization of the eye’s function, one that shifts the functional “seeing” role to audition and to neural sensory-motor cortical coupling (Flanagin et al. Reference Flanagin, Schörnich, Schranner, Hummel, Wallmeier, Wahlberg, Stephan and Wiegrebe2017).

Deterritorialization and reterritorialization in VR

To understand the dynamic transformation of embodiment in VR, I want to draw upon D&G’s ideas of deterritorialization and reterritorialization. Deterritorialization involves the breaking down of fixed boundaries, systems, or structures. In VR, this process can play out through the way one’s avatar-body breaks from one’s normative expectations—not only for what one’s self looks like (avatars can be customized), but also in terms of what one’s body does.

Reterritorialization, conversely, refers to the process through which new systems, identities, or structures are formed in response to the fractures of deterritorialization. One’s fleshly body adapts to the digital environment and reorients to take on a new sense of stability and definition. For instance, in Population: One, one’s avatar possesses the ability to fly. At first, it is very disorienting. Players who are unfamiliar with this ability may try to flap their arms like birds before realizing that they should glide; before realizing that frantically raising and lowering one’s arms will dangerously reduce altitude and making one an easier target. Eventually, a newbie will reterritorialize their perception, realign their motions to achieve the new results, and learn to fly.

Together, deterritorialization and reterritorialization offer a framework for understanding how VR allows for the continuous transformation of both the avatar and the player’s embodied identity. These processes create a dynamic, evolving experience of subjectivity and intersubjectivity, where the self is constantly remade and redefined through interaction with both the virtual and social worlds, and where a player may make assumptions about another player’s de/reterritorialization based on their level of experience and possibly also based on IRL social factors.Footnote ⁸

The cycle of improvisation starts with a territory

Improvisation must begin somewhere; in the background, there may be a grammatical or spatial structure, a rhythm, or a shared understanding, even if partial or implicit. This grounding context, in Deleuzoguattarian parlance, is a territory—a set of constraints that makes improvisation possible in the first place (Deleuze and Guattari Reference Deleuze and Guattari1987, 311). It might be a musical genre like jazz (Duranti and Burrell Reference Duranti, Burrell, Mantovaglio and Zucchermaglio2005; Romero Tenorio et al. Reference Romero Tenorio, Riccardi and Echeverry2020), the infinitely recursive capacity of Chomskyan syntax (Duranti and McCoy Reference Duranti, McCoy and Stanlaw2020), a shared linguistic, cultural, or musical ideological formation among performers (Hugh Sam Reference Hugh Sam2024), or the literal topography of a climbing wall (Hillawaert, this issue). Territory is never neutral because it has been shaped through repetition, through habits of attention and embodied histories.

Deterritorialization, reterritorialization, and watching out for clams

Against the background of the territory, improvisation begins. It deterritorializes by pulling recognizable elements, such as notes, gestures, phrases, and ideologies out of their usual contexts and setting them on divergent paths. Think of jazz improvisation, and consider this conversation between the legendary jazz performers Kenny Burrell (guitar), Sherman Ferguson (drums), and Jeff Clayton (saxophone) while playing together in Alessandro Duranti’s Structure of Jazz Aesthetics class in 2002. They are discussing whether one of them has produced a sour note:

As we can see from the conversation above, improvisation both leaves something behind and makes something anew, thus providing the engine for reterritorialization. The improviser assembles new alignments and momentary coherences, taking the form of grooves, emergent linguistic variation, or affective logics that hold long enough for something else to happen. As Stover (Reference Stover2017, 6) shows in his analysis of improvisational Cuban Rumba, micro-rhythmic deviations would be misclassified as simple embellishments. They are responses that build on each other and unfold over time, shaped by the melody’s aesthetic and emotional pull.

Deterritorialization and reterritorialization thus operate in tandem. To leave one state is to begin constituting another, even if it does not yet have a name, and even if, in retrospect, it is evaluated as part of a signature aesthetic. Simultaneously anaphoric and cataphoric, this duality invites a reevaluation of the role of history, embodied experience, and artifacts in identity, emphasizing fluidity, multiplicity, and assemblages over fixed corporeal boundaries, in line with digital and posthumanist paradigms (Braidotti Reference Braidotti2013; Milesi Reference Milesi, Herbrechter, Rossini, Grech, de Bruin-Mole and Muller2022; Herbrechter Reference Herbrechter2021).

A concrete example: in the popular VR game Gorilla Tag (Another Axiom), players’ avatars are legless gorillas (See Figure 1 below) who can jump, climb, avoid, and tag other gorillas using only their arms and hands.Footnote ⁹ The limited movement capabilities of the avatars force players to improvise within the constraints of the game’s physicality, offering a clear instance of the BwO as it becomes perceptible through practice. In this sense, the BwO is not an entity at work but a plane of immanent potential through which new configurations of movement and perception emerge whenever the body’s habitual organization is unsettled.

Figure 1.

A legless gorilla in Gorilla Tag.

I once witnessed a group of kids playing “IRL Gorilla Tag,” without any gaming devices by pretzel-crossing their legs, locking them into position, and chasing each other around the room surprisingly quickly in a similar fashion to how the gorillas move in the VR game. Thus, a real-life improvisation mimicked the actions of a VR game, showing that in the absence of the digital environment, the biophysical body can improvise and perform actions that transcend its normal boundaries. This type of adaptation—what D&G would call a line of flight—is not exclusive to VR. Any adult who has witnessed children improvising games and porting them between the digital realm and the IRL realm (playing Angry Birds with their own bodies as projectiles while knocking down living room cushions) may agree with the author that children at play uniquely exhibit the shifting distributions of capacities, the flow and reorganization of possibilities implied by the BwO.

Case study A: Population: One and the failure of re-territorialization

In the early 2020s, one of the leading VR games was Population: One, a first-person shooter (FPS) with battle royale mechanics. Players are flight-dropped into a virtual environment where they must fight until only one competitor or team remains. The game is typically played in squads, and each squad may be composed of strangers from various parts of the world, adding an unpredictable linguistic and social element to the experience.

When we recorded the data in 2021, we found 29 character skins, featuring both male and female representations, along with one scarecrow. Hattie (female) and PJ (male) are the names of the foundational avatar skins that players start with before others can be unlocked through in-game purchases or level progression. While the diversity of skins aims for gender and racial balance, players often defaulted to assuming male underlying identities, with players addressing each other using male terms, such as men, bros, or dudes despite female skins, all while making sexist comments about female avatars, reflecting broader social and gendered dynamics that exist within the metaverse (see Frenkel and Browning Reference Frenkel and Browning2021).

Despite evident game design intentions to be gender-neutral, there are documented differences in how male and female players experience VR. Research has shown that male participants tend to report a higher sense of presence in virtual environments, feeling more immersed and physically present in the virtual world compared with female participants (Felnhofer et al. Reference Felnhofer, Kothgassner, Beutl, Hlavacs and Kryspin-Exner2012). This difference highlights the potential for gendered experiences of embodiment within VR, a phenomenon that can influence how players engage with their avatars and the virtual environment. In Population: One, female players often adjust their communication and gameplay to either accommodate or push back against sexist norms. These resulting improvisations modify posture, voice, and strategy, revealing how gender is actively negotiated in virtual space.

In the context of VR, embodied parallelism also plays a crucial role in how players experience the virtual world. Bollmer and Suddarth (Reference Bollmer and Suddarth2022) argue that VR requires a parallelism between the real-world gestures and movements of the player and the virtual avatar to produce a coherent sense of embodiment. Otherwise, players may experience actual physical discomfort, such as nausea and headaches, or fail to connect with the avatar in a meaningful way, as is often the case for newbie players unfamiliar with the controls and mechanics of VR. A failure in embodied parallelism can also lead to social crosstalk, as we will see below, because new users are often sidelined until they become more proficient at controlling their avatars.

The following excerpt from a Population: One gameplay session captures the above dynamic in action. Transcription conventions are provided in the appendix.

The segment begins with the team materializing outside a tower (as evidenced by Jake’s “hello,” line 2), and members beginning to engage both socially and tactically. Katy reports a technical glitch (“I’m like stuck in the ground,” line 13) and tries to align herself with gameplay (“I think gameplay is starting already,” line 17), yet receives no acknowledgment. Her avatar has literally glitched and its virtual body rendered nonfunctional, but what follows is also a social glitch: she is not taken up as part of the improvisational field, but indirectly chided for her lack of greeting. Katy’s stalled participation reflects a form of deterritorialization, in which her embodied expectations, her avatar’s compromised affordances, and her unratified bids for alignment fail to coalesce into any new configuration, leaving her without the interactional foothold required for reterritorialization.

In contrast, Matt arrives late (line 46), signals his unfamiliarity, and is immediately welcomed. Paul responds with affiliative reassurance (“no worries”), and Jake begins offering technical guidance in response to Matt’s curiosity. Matt’s deterritorialization, marked by his temporal and experiential marginality, is met with smooth alignment and reterritorialization through technical talk and team-building moves, such as Paul’s expert-framed, avuncular adage “shoot – kill – wi:n” (line 47). The group folds him in, even offering tips for next-level hacks. The male players also exhibit pervasive format tying: quoting, recycling, and repeating bits and pieces of each other’s utterances, cooperating in what Charles Goodwin calls “structure-preserving transformations on a public substrate.” (Goodwin Reference Goodwin2013:9)

What this moment makes clear is that improvisation in video gaming does not unfold in a neutral space (see also Gray Reference Gray2014). Processes of de/reterritorialization are not equally available to all players. But can we really say in this case that the marginalization is gendered? Or could it be because Katy is an inexperienced newbie (or both)? In fact, Katy’s exclusion may be substantially a result of a bungled entrance, where she fails to greet the team, and the glitch in her system, which doesn’t allow her to enter the tower where the rest of the team is playing, rendering her clearly audible in her numerous glitch reports but erasing her from visual copresence. She remains stuck in the spot where everyone on her team originally materialized.

I understand Katy’s marginal position as emerging from the convergence of a disrupted entrance, a sustained glitch that prevented her from joining the group’s perceptual frame, and a possible gendered orientation toward her bids for affiliation, with these elements combining to limit her interactional uptake throughout the sequence.

Katy’s view is shown in Figure 2 below:

Figure 2.

Katy’s headset view from her glitched-out, stuck-in-the-ground avatar.Other players on her team have left and are fighting opponents inside the tower in the background.

Figure 2 Long description

The image shows a first-person perspective in a video game. A player's hand is visible in the foreground, holding a weapon. On-screen game statistics are displayed, including scores for two teams, with one team having 17 points and the other 15 points. The total score is 30. The time remaining is shown as 10 minutes and 50 seconds. Health and ammunition stats are also visible, with health at 220. In the background, a game environment is visible, featuring a corridor with doors and another player character standing near the entrance.

Deleuze and Guattari describe deterritorialization as a threshold or edge of transformation (Deleuze and Guattari Reference Deleuze and Guattari1987, 503–506), but crossing that threshold socially requires others to recognize it as meaningful. In this case, Katy’s improvised solutions remain unratified; her glitch is not interpreted as a cue for solidarity or a bid for play. Instead, her recurrent and unanswered glitch reports provide their own parallel narrative in which uncertainty and breakdown are not reabsorbed into the flow of the game but left hanging. She is deterritorialized, but without reterritorialization, her glitch has left her with no new assemblage to inhabit.

Case study B: Richie’s Plank Experience, clams, and response cries in VR

Richie’s Plank Experience (2016) is a VR game designed to simulate the experience of walking along a narrow plank placed high above the ground in a skyscraper setting. You don a VR headset and find yourself on the ground outside a city building. You then ascend in an elevator and step right onto a virtual plank that extends from the edge of the building, seemingly hundreds of feet above the ground. The game is simple in its premise but remarkably effective in its manipulation of embodied perception.

The game mechanics of RPE are not goal-oriented. The plank simulation has neither a score nor an opponent, nor is the game played in a campaign with others online. All action revolves around your own movements, as you are given the option to walk to the end of the plank, with the in-game avatar tapping directly into your own reactions to height, balance, and physical orientation (See Figure 3). Unlike console video games, where players inhabit characters through external controllers, RPE uses full-body immersion and locomotion to generate its distinctive uncanniness. As in the Population: One game, the player experiences the phenomenon of embodied parallelism (Bollmer and Suddarth Reference Bollmer and Suddarth2022), where the avatar’s locomotion in the virtual space mirrors the player’s real-world movements.

Figure 3.

Looking down at the plank before stepping out.

Figure 3 Long description

The image on the left shows a person indoors wearing virtual reality gear, holding controllers and standing on a carpeted floor. Furniture, including a chair and a table with a lamp, is visible in the background. The image on the right displays a virtual scene of a wooden plank extending over a cityscape, with tall buildings and a cake at the end of the plank.

For this study, a real-life 2 × 4 wooden plank was procured and aligned with the in-game plank (this is such a popular mixed-reality modification that there are in-game instructions for setup). In addition, a small fan on the floor was pointed up at the player and switched on when the virtual elevator doors opened to enhance the realism of the situation. When we pre–tested the scenario, some players fell silent, and, in fear of plummeting to the virtual ground, one of our pre–test subjects crawled on the real floor, while another threw himself at the opposite wall when he saw the spider. These experiences alerted us to the need to provide a padded environment and to establish a buddy system where a friend could catch the player.Footnote ¹⁰ Because we wanted to record player interactions (as well as administer a recall task for a later study), a questionnaire was designed to keep the player conversing fluidly with their friend, and instructions were given to the player to answer in sentences rather than monosyllables. Thus, the friend that the main player was asked to bring along had two jobs: keeping the research participant talking and catching them if they became unsteady on their feet.

Improvisation and agency in Richie’s Plank Experience

Improvisation in RPE manifests physically, linguistically, and in interaction with the technology. Players adjust their walking pace, balance, and posture to account for the heightened sense of danger and the instability of the virtual space. The stakes seem high: organismic integrity is constantly in question, given the looming possibility of simulated dissolution through falling or jumping off the plank. Players must adapt in response to semiotic layering (Mendoza-Denton and Jannedy Reference Mendoza-Denton and Jannedy2011), or what Charles Goodwin (Reference Goodwin2013) refers to as semiotic lamination, where multiple modalities are simultaneously built up across time and action. In this case, semiotic layering includes haptic feedback from the physical wooden plank, the sensation of wind in one’s hair, and the continuing interaction with friends, alongside in-game sensory inputs such as the in-game sounds of the howling wind, creaking wood, spiders trilling, and the helicopter, along with uncanny visuals, such as having a shadow but no feet, and controllers where the hands should be.

This convergence of real-life, in-game, and proprioceptive cues creates a densely layered interactional field, compelling players to improvise moment by moment as they navigate fragmented perceptions and the porous boundary between virtual and physical embodiment. Let’s now look at the transcript of the interaction:

The above interactional sequence from Richie’s Plank Experience reveals that response cries in VR are neither simple affective outbursts nor malformed “clams.” Their interjections are structurally and rhythmically organized components of gaming participation (Aarsand and Aronsson Reference Aarsand and Aronsson2009). These patterned shifts in phonation mark a redistribution of perceptual and vocal capacities under affective pressure, and it is this transient reorganization that I treat as the empirical manifestation of the BwO within the virtual environment.

The f0 contour pictured in Figure 4 is one of three separate and near-identical quasi-lexicalFootnote ¹¹ turns by Milly, each one uttered in an interactional sequence with her friend Fiona, in the sequence ABABAB. Drawing on Goffman’s (Reference Goffman1978) work on response cries and C. Goodwin’s (Reference Goodwin2013, Reference Goodwin2018) theories of cooperation, we can understand these utterances as important to the organization of affect and social action, and to the process of deterritorialization and reterritorialization. Furthermore, M.H. Goodwin, Cekaite, and C.Goodwin’s (Reference Goodwin, Cekaite, Goodwin, Sorjonen and Perakyla2012) analysis of “eeeew” in data from Los Angeles middle school girls shows that this type of utterance functions as a strong negative stance marker of disgust, and is often accompanied by actual physical recoiling. Such an interpretation of a negative, avoidant stance is supported by Milly repeated body torque to turn away from the spider and her use of the controllers to “shield” her face.Footnote ¹²

Figure 4.

Waveform, spectrogram, and f0 contour (in blue) for line 33:[i:Ih:y:: ʉ:::::hʌ]/ “eeeeew” duration 2.1 seconds.

Figure 4 Long description

The plot consists of three panels analyzing a voiced segment over 2.1 seconds. The top panel displays the waveform with the x-axis labeled 'Time (seconds)' ranging from 0 to 2.1 and the y-axis labeled 'Amplitude'. The middle panel shows the spectrogram with the x-axis as 'Time (seconds)' and the y-axis as 'Frequency (Hz)', ranging from 0 to 5000. The bottom panel overlays the fundamental frequency contour in blue, indicating pitch variation over time. Notable features include a sustained energy level in the spectrogram and a drop in the fundamental frequency contour towards the end of the segment.

Milly’s vocalizations function as semiotic resources that mediate the tension between her bodily experience of disgust, fear, and excitement and the complexity of the VR environment (See also Mendoza-Denton et al. Reference Mendoza-Denton, Eisenhauer, Wilson and Flores2017). These interjections/response cries are extragrammatical but stand as full utterances (Ameka Reference Ameka1992); some are borderline lexical, exhibiting distinctive pitch contours and communicative respiratory mechanisms. They form a recognizable interactional pattern that recurs throughout the encounter. Their grammatical marginality is mirrored in their analytic marginalization in the literature, though as Ameka (Reference Ameka1992) and more recently Dingemanse (Reference Dingemanse2024) argue, interjections constitute a universal and systematic part of language and play a central role in interaction. Jittery and shimmery vocalizations in the beginning come across as trepidation-laden “[ɯ:::], [ɯ:hhuʔɯ::], [hɯhɯ:::]” (lines 1–4), and establish a rhythmic and acoustic structure that persists like a leitmotif as the game intensifies. Response cries serve as affective discharges and affective scaffolding, showing Fiona where she can provide support and anchoring Milly’s bodily responses across a semiotic partiture (Mendoza-Denton and Jannedy Reference Mendoza-Denton and Jannedy2011) involving rapid-fire visuals, in-game sounds, and real-time peer interaction.

As the VR scenario becomes more intense, particularly with the appearance of spiders and sudden haptic feedback, Milly’s vocal expressions evolve into a patterned series of sequenced cries. These include bivalent lexical-affective hybrids, such as [ʍãʔ]/ “what” and [ha:w]/ “how” (lines 9, 12); and clearly nonpropositional vocalizations, such as [hʉhɘhɘhɘ] (line 11), [h͋i:::] (lines 21, 24, 29), [i::y:: ʉ::::ʌ:] (line 27), [ʉ:::::hʌ] (lines 30, 37). Line 10 notably repeats the phonetic form of the line 9 response cry but sets it in motion in the grammar, where falsetto gives way to modal phonation. This is reminiscent of the jazz conversation we saw above, where a “clam” winds up being repeated and reinterpreted as a purposeful segment. Performing a kind of style shift, here the repetition causes the entire utterance in line 10 to be reinterpreted as anaphoric, bivalent between affective intensity and lexicon. I encourage the reader to listen to these changes through the link provided.That the sounds exhibit format tying and repetition through the reuse of prosodic and phonetic forms across a single speaker’s turns means that they create cohesion without relying on syntactic structure (Goodwin Reference Goodwin, Streeck, Goodwin and LeBaron2011). Across speakers, the repetition of vowel types, intonation contours, and phonation methods establishes cooperative transformation and affective interactional rhythm. Fiona’s responses follow Milly’s phonation and timing. For example, Fiona’s laughter pulses in lines 20 and 28 mirror the vowel height in Milly’s laughter and post-laughter inhalation respectively, echoing pitch, vowel geometry, and number of pulses and rhythm to form a rhythmic, collaborative affective sequence.

In their study of political speech, Mendoza-Denton and Jannedy (Reference Mendoza-Denton and Jannedy2011) analyze how intonation and gesture operate within complementary and supplementary multimodal systems. In complementary systems, modalities such as speech and gesture contribute different components to a single communicative act, as when a speaker says “this city” while pointing to a location on a map. In supplementary systems, multiple modalities reinforce the same communicative content, amplifying its salience through redundancy, as in a raised voice accompanied by a forceful gesture. In Milly’s case, response cries operate within both systems. These cries supplement visible fear in her bodily reactions (shielding her face, shaking off the virtual spiders) while also contributing complementary cues that direct attention to features of the environment, such as spiders or the plank’s edge. These are both emotional expressions and interactional structures that help the player organize perception and action across sensory domains.

The presence of lengthened ingressive phonation at the end of several of Milly’s laugh tokens adds another layer of affective and physiological significance to the response cries. These inhalation events—the high amplitude ingressive segments in lines 21, 24, 29, and 31—were realized with high vowels and tended to emerge at points of high emotional saturation or physical overwhelm. Trouvain et al found in corpus studies that ingressive phases are common at the end of laughter bouts, but hypothesized that “inhalation in laughter is a consequence of unplanned air leakage due to spontaneous vocalization.” (2022, 29). Here, the poetic patterning of the inbreaths signals the dissipation of affective arousal and often accompanies the shift back into verbal activity and modal voice. In Milly’s case, these ingressives do not appear to be unplanned leaks, and are reminiscent instead of the well documented laughter-joining events in the discourse and conversation analytic literature (Glenn Reference Glenn2003), since they participate in the interaction by providing patterns of vowel geometry and timing that are followed by the interlocutor (see also laughter entrainment, as in Ludusan and Wagner Reference Ludusan and Wagner2022). These moments sit at the edge of what Ameka (Reference Ameka1992) and Dingemanse (Reference Dingemanse2024) call the core of language: interjections and vocal expressions that, though often dismissed as peripheral, perform critical interpersonal and affective work. In this context, the ingressive laugh tokens form part of a distributed system of vocal regulation, allowing Milly to process, display, and transition out of overwhelming affective states while maintaining interactional presence within a densely layered VR environment.

While some of Milly’s vocalizations may initially appear as “clams”—that is, malformed or misfired phonatory sequences that deviate from recognizable linguistic or prosodic expectations—they should not be dismissed as performance failures. Rather, these clams mark moments of improvisational opening, where the body exceeds prescripted interactional norms and enters a state of affective experimentation. What might seem like excesses, stumbles, or leakages in the phonatory stream are, in fact, rhythmic and prosodic cues that invite co-participation. Fiona’s echoic responses show that she is attuned not only to Milly’s lexical utterances but also to her nonlexical vocal stylings. These productive disalignments create the space for joint meaning-making, affective resonance, and what Goffman (Reference Goffman1974) would call the transformation of a “serious frame” into a “play frame.” In this sense, clams are better understood as launch points for improvisation, co–constructed in real time through vocal patterning, timing, and embodied responsiveness.

Conclusion: improvisation, presence, and the virtual body

The cases I have traced here suggest that embodiment in VR is itself a form of improvisation under affective constraint. Through the lens of case studies on the games Population: One and Richie’s Plank Experience, we have seen how players inhabit a semiotically layered and precarious terrain where the avatar becomes a BwO, responding creatively to glitches and clams in interaction.

This work argues for a view of interaction that makes room for the incipient, the precarious, and the excessive. In linguistic anthropology, as in linguistics more broadly, we often center our analyses on what we can identify as speech events, and break these down into utterances and tokens. In other words, we center systems of signs that have already occurred, leaving us ill-equipped to account for what is absent, unsaid, or not-yet-realized. This is the realm of negative data. Interrupted gestures, failed alignments, and affective intensities that are never resolved into words (failures of reterritorialization) are difficult to analyze in disciplines that rely on transcription, and in the case of sociolinguistics, on proportional counts of linguistic phenomena. By engaging with de/reterritorialization as an analytic tool, this paper attempts to make legible those zones of affective instability that resist symbolization.