Chatting with a few friends about their PhD, I realised that I wasn’t the only one to grow a quasi-obsession with a question in their literature. Going into my PhD and true to my sceptical habits, I was wondering whether body representation “actually existed”. The question changed and, I think, matured – however, in a way, it still haunts me, even after moving towards multisensory research. As the body representation research community gathers in Utrecht, I couldn’t waste the opportunity to ask around.
I’m not asking whether body representation exists anymore – at least not when using the broad definition Frédérique de Vignemont proposes – “an internal structure that has the function to track the state of the body and encode it”. Now, I’m wondering what aspects of this definition we, researchers, should be focusing on. Should we see body representation as a structure and the brain as the body’s cartographer, managing cognitive maps of the body? Or should we look at body representation as a function primarily, and seek to understand the way in which the brain adjusts inputs and outputs in the way a piano tuner would adjust the pitch of each note?
Taxonomic approach: The brain as a cartographer, body representation as a map
We have a single physical body, but our experience of it is not homogeneous in different contexts and from one individual to another. From choosing the right shirt size to drawing one’s own silhouette, from expressing emotions about one’s body to going through a narrow opening, a great diversity of tasks and contexts require some form of “tracking the state of the body”. Many authors have attempted to tidy things up with body representation taxonomies, characterised by differences in temporality, conscious experience, locality, cerebral substrate, processing level, or function.
Early on in the literature, these models have aimed at distinguishing relevant contexts in which one body representation would be used over another, so that body representation retains a good explanatory and predictive power despite this heterogeneity. I call this the taxonomic approach: a research strategy whose main goal is to identify the number and properties of distinct body representations. This approach looks at the brain as a cartographer, creating, maintaining, updating and using cognitive maps (“internal structures” as de Vignemont would say). Tasks requiring some form of “tracking the state of the body” would tap on one of the many dynamic maps the cartographer brain holds, each of which unique by its properties and the information it carries.
The cartographer analogy is useful to understand that, just like maps, all body representations are not useful in the same contexts. A map of a city showing the best restaurants, cultural locations, main roads and transport hubs would be very useful for a tourist, but not for a census agent – and reversely, this wealth of information would be counterproductive for the census agent, who would rather need a detailed map of all addresses up to the smallest dead-end (this is what Pitron et al. (2018) discussed as a trade-off between richness and precision, one of the main reasons behind the need for multiple body representations).
The taxonomic approach is particularly fruitful when it’s used to make predictions about the type of deficits an individual will encounter when faced an alteration of a given body representation. For example, the popular body image/body schema dichotomy follows the functional dissociation between perception and action. In this view, a specific body schema disorder could be seen in daily spatial navigation between obstacles, reaching movements, or door passability tasks, but not in perceptual judgments of body size, nor in body satisfaction questionnaires – and vice versa in a specific body image disorder. The main advantage and drawback to the taxonomic field is that lots of models have been proposed to make sense of neuropsychological, clinical, and experimental data. They include Head & Holmes' postural vs superficial schema, Thompson’s 4 dimensions of body image, Longo’s focus on processing level (somatosensation, somatoperception, somatorepresentation), Schwoebel & Coslett’s triadic taxonomy (visuospatial body map, body structural description, body semantics). Subcategories have also been used, e.g., the perceptual vs. attitudinal components of body image; some papers use the same words for apparently dissimilar concepts of body representations, while others have questioned the very idea of a body schema and drawn comparisons with peripersonal space. Yes, it’s a bit of a mess – and even this statement is getting repetitive, with many calls to clarify the taxonomic situation in the past 40 years.
Functional approach: The brain as a tuner, body representation as a function
Adopting a taxonomic approach means confronting the thorny problem of the number and properties of body representations. Another research strategy sidesteps this problem by asking different questions about body representation. De Vignemont's (2020) definition emphasises the importance of the function performed by body representation (i.e., tracking the state of the body). Rather than focusing on cognitive maps, their properties and their relationships, it is possible to study the processes that track the state of the body in a variety of contexts. The functional approach is a research strategy that consists of identifying the cognitive processes involved in a task that requires some form of body representation. When analysing the task, researchers then focus on what information is required and how it is processed to perform it. The functional approach is more “specific” than the taxonomic approach, in the sense that it aims at describing a particular behaviour precisely, without necessarily (at least, in the beginning) seeking to generalise this result to behaviours that are thought to be similar.
The metaphor underlying the functional approach is that of the brain as a tuner, adjusting the length of the strings of an instrument so that it “sounds right” in a given situation. Rather than constructing, maintaining and using maps of the body, bodily cognition is thought of as trying to balance the importance given to sources of information according to the situation and the task in hand. Body representation is conceived as a cognitive function or a set of cognitive functions: information processing itself, rather than as a mental structure resulting from this processing. From this point of view, measures of body representation are not best conceived as probes providing access to the content of a given body representation, but as the results of a combination of information processing linked to the context. Research therefore focuses on correctly identifying the quantitative links between sensory inputs and a specific behavioural output: what processing did this behaviour require? What (probabilistic?) laws can be used to describe the effect of a sensory input in a given task? Answering these questions enables us to better characterise the function we’re studying and its parameters.
A cartographer, a tuner, both, or neither?
Obviously, each approach has its advantages. Regarding the inputs, the taxonomic approach is fairly straightforward: The information involved in a task can be deduced from the type of representation that’s involved. For example, a perceptual judgement probing body image should depend on richer, more diverse information than the sensorimotor body schema. The functional approach makes fewer assumptions about the organisation of body representation, but has the disadvantage of making this kind of predictions difficult. In this respect, the functional approach is also more data-driven than the taxonomic approach: whereas a series of properties allow us to know a priori which representation is being used in a task, and to make predictions accordingly, we need quantitative data on very similar tasks to hope to make any prediction in a functional approach, because it does not specify a priori on what criteria two tasks can be considered similar: we don’t know whether two tasks will mobilise the same functions, based on the same input, parameterised in the same way.
So, what do you think of the cartographer/tuner question? Should we aim at defining and identifying clearly the number and properties of body representations, so that they can our theoretical and methodological choices and have a positive clinical impact? Or should we first gather more data on a variety of contexts and try to see what set of parameters explain an individual’s behaviour, so that we can make precise predictions in specific contexts before trying to generalise our models to other contexts? Maybe this is a false dilemma? I hope you’ll reach out to discuss this or poke me at the BRNet 2025 conference to give me your own “hot take”.