Peripersonal space: why so last-second?

Abstract

A vast range of neurophysiological, neuropsychological and behavioural results in monkeys and humans have shown that the immediate surroundings of the body, also known as peripersonal space (PPS), are processed in a unique way. Three roles have been ascribed to PPS mechanisms: to react to threats, to avoid obstacles and to act on objects. However, in many circumstances, one does not wait for objects or agents to enter PPS to plan these behaviours. Typically, one has more chances to survive if one starts running away from the lion when one sees it in the distance than if it is a few steps away. PPS makes sense in shortsighted creatures but we are not such creatures. The crucial question is thus twofold: (i) why are these adaptive processes triggered only at the last second or even milliseconds? And (ii) what is their exact contribution, especially for defensive and navigational behaviours? Here, we propose that PPS mechanisms correspond to a plan B, useful in unpredictable situations or when other anticipatory mechanisms have failed. Furthermore, we argue that there are energetic, cognitive and behavioural costs to PPS mechanisms, which explain why this plan B is triggered only at the last second.

This article is part of the theme issue ‘Minds in movement: embodied cognition in the age of artificial intelligence’.

1. Introduction

The theoretical frame of embodied cognition has proposed that the body shapes our cognitive abilities. Over the years, it has been applied—with more or less success—to various cognitive domains, including high-level capacities such as social cognition, language and mathematics. However, if there is one area in which there is clear evidence that the body plays an essential role, it is in the perceptual domain. More precisely, a vast range of neurophysiological, neuropsychological and behavioural results in monkeys and humans have shown that the relatively close surroundings of the body, also known as peripersonal space (PPS), are processed in a unique way and that the extent of PPS depends on one’s bodily abilities [1]. These findings provide conclusive evidence that the way we hear and see the outside world can depend on our bodily location and capacities. What remains unclear, however, is why the world is processed in a special way only at such a relatively short distance from the body.

Three distinct roles have been ascribed to PPS mechanisms ¹ :

• —a defensive role, to react to threats;
• —a navigational role, to avoid obstacles; and
• —an appetitive role, to act on objects.

However, in many circumstances, one does not wait for objects or agents to enter PPS to plan these behaviours. In particular, one has more chances to survive if one starts running away from the lion when one sees it in the distance than if one can almost feel its breath on one’s face. Even if the lion is still 2 m away, it is most probably too late. Or consider navigation in a packed environment, such as a classroom full of tables and chairs. Graziano and Cooke [4] claim that this requires the PPS neural network to monitor the proximity of objects but we are not like old-fashioned robots with only short-distance receptors that wait until they are next to the table to get around it. We plan our path ahead computing the easiest and possibly most straightforward trajectory when entering the room because we have visual access to all the potential obstacles long before they are next to us. PPS mechanisms would make most sense in shortsighted creatures but human beings are not such creatures. Although PPS mechanisms may not be exclusively about what is immediately next to our limbs, they are still primarily concerned by the region of space that is relatively close to us and when it comes to action, space is time, the time to prepare for us to act and the time for us to perform the movement. The crucial question is thus twofold: (i) why are these adaptive processes triggered mostly at the last second? And (ii) what is their exact contribution, especially for defensive and navigational behaviours? In order to answer those questions, we shall consider not only the benefits of PPS processes, but also their potential energetic, cognitive and behavioural costs. It is only if we consider the trade-off of PPS mechanisms that we can explain when they are worth it, and in turn, why their emergence has been selected through evolution. In this article, we shall first briefly review the distinctive signatures associated with PPS. We shall then argue that many other mechanisms appear more advantageous for self-protection and propose that PPS mechanisms only correspond to a plan B, especially useful in unpredictable environments. We shall conclude by explaining why this plan B is triggered so late.

2. The signatures of peripersonal space

Since its original discovery by Rizzolatti et al. [5,6], there has been a booming of research on PPS, especially in the last 10 years. Our objective here is not to provide an exhaustive review of this ever-expanding literature, this has been performed elsewhere (e.g. [7]), but rather to give the gist of the most significant trademarks of PPS processing across multisensory, unisensory, sensorimotor and attentional dimensions. Beforehand, it should be noted that the experimental paradigms can vary to such an extent that one may wonder whether they all investigate the same notion of PPS [8]. PPS is classically described in spatial terms, typically as approximately less than 50 cm from the hand and 1 m from the trunk, but this metric definition has been questioned [9,10]. The precise distance from the body can vary depending on the body part, the sensory modality, the context and the species. Notwithstanding these differences, one can still assume that PPS primarily emphasizes objects and events in the relative proximity of the body.

(a). Multisensory dimension

Since the earliest studies, the multisensory property of PPS has been emphasized as an almost necessary definitional aspect. While perception is overall a multisensory affair [11,12], PPS involves a peculiar kind of interaction that occurs between external senses (vision and audition) and somatosensory senses (touch and nociception). These multisensory effects are anchored to the location of the various parts of the body, as signalled by proprioception. Pioneering work in macaques first reported multisensory neurons that responded to tactile stimuli presented on the animal’s body, as well as to visual stimuli mostly—though not exclusively—when located in its close vicinity [5,6,13–15]. This inspired research in right-brain-damaged patients suffering from tactile extinction. It was found that their awareness of contralesionally tactile stimuli was lost when tactile stimuli were concurrently applied ipsilesionally, but also when visual or auditory stimuli were concurrently presented ipsilesionally, provided they were within a distance of roughly 30 cm from their body [16–18]. Since then, a variety of multisensory studies have explored PPS extent, boundaries and malleability. While cross-modal congruency paradigms mostly focus on the behavioural costs resulting from visual/auditory distracting stimuli over tactile targets (19; but see [20] and for review [21]), dynamic multisensory stimulation paradigms emphasize the benefits in reporting tactile targets when visual/auditory stimuli are concurrently presented at closer compared with farther distances (e.g. [22,23]). Strikingly, this type of multisensory interaction follows the body and its parts when they move in space. Perceptual effects generated by multisensory processing in PPS are literally anchored to the body, which constitutes the origin of PPS reference frame [24–26].

Evidence thus converges in showing that both the neural activity and the behavioural responses to visual/auditory stimuli are enhanced when occurring near the body [27–30]. Although part of PPS mechanisms may be hard wired [9], it has been suggested that these multisensory effects could also be explained in terms of association learning mechanism. Repeated exposure of visual/auditory stimuli soon to be followed by tactile stimuli may have tuned some neurons to be activated for both. On this view, these contingencies have been probabilistically generalized to an extended space around the body in which they have a higher probability of occurring [31,32].

(b). Unisensory dimension

Whatever the effect on tactile processing (either cost or benefit), proximity may also modify visual processing per se. The so-called near-hand effect, for example, consists in a purely visual detection advantage for near-hand compared with far-hand stimulus locations, in the absence of multisensory stimulation [33,34]. Shape discrimination also benefits from proximity, as shown by Blini et al. [35] (figure 1). Participants had to decide as fast and accurately as possible which shape they saw, when presented close (within PPS) or far (outside PPS). Importantly, retinal size correction was applied because in everyday life far objects appear as smaller. However, despite the fact that equating the retinal size of close and far shapes made the latter appear bigger, participants were faster in discriminating shapes appearing smaller but close to them. This advantage was also observed when the shapes were embedded within a two-dimensional Ponzo illusion of depth, indicating that even illusory proximity could profoundly affect fundamental aspects of visual perception.

Figure 1.

Results from the series of experiments (1–5) assessing unisensory visual shape discrimination near and far from the body (from Blini et al. [35]: box-and-whisker plots depict the mean facilitation in response time as a function of distance. Note the significant advantage for close over far shape presentation is present in a three-dimensional virtual environment (experiment 1), as well as in a two-dimensional version (experiment 2) when only monocular depths indices are available. The close advantage in shape discrimination is also present despite retinal correction making far objects bigger (experiment 3) and is further enhanced when shape size respects the natural scaling with distance (experiment 4). When shapes were presented at six different depths (experiment 5), sigmoidal curves for mean accuracy (left panel) and mean response time (RT) (right panel) predicted the advantage as a function of distance (from D1 for the closest to D6 for the farthest position). Error bars show standard errors of the mean. The y-axes refer to the odds of providing a correct response (accuracy) and the relative RT advantage observed with respect to participant-specific mean performance. From Blini et al. [35].

Recent work extended these findings to social perception ([2]; see for review [3]). Participants had to discriminate male from female faces depicting happiness, anger or neutral emotions presented close or far (50 versus 300 cm). They showed faster response times for close faces, with concurrent modulation of their physiological indexes (e.g. pupil dilation and heart rate variability). Overall, these findings clearly indicate that, notwithstanding their impact on somatosensory modalities, the processing of visual stimuli in the close vicinity to the observer’s body is different at the unisensory level. In brief, close is better [36]. However, one may wonder about the advantage for survival of such a perceptual boosting in one’s bodily proximity.

(c). Sensorimotor dimension

PPS processes take place at the crossroads between perceptual (visual, auditory, vestibular and somatosensory cortices) and motor-dedicated regions, both in human and non-human primates. Neurons capable of PPS multisensory integration in monkeys have been identified in the ventral intraparietal region (VIP), parietal cortex 7b, as well as in the premotor cortex (PM) and the putamen. In these (and human-analogue) sensorimotor regions, electrophysiological studies in non-human primates [5,6,14,15,37,38] and neuroimaging work in healthy humans [27,39,40] have identified populations of neurons exhibiting multisensory processing. Remarkably, the dual- and tri-modal neurons in these regions show tactile receptive fields on specific parts of the body and visual and/or auditory receptive fields spatially overlapping with and anchored to them. While their extension in depth can sometimes reach a few metres, they typically protrude only a few centimetres in the space around the body. They can thus provide detailed joint maps of the limbs and the space around them to plan effective actions within a body-centred frame: the proximity of the hand to a visual object evokes responses from this PPS network, whereas moving the hand farther away reduces such an activity [27].

Thus, both anatomically and functionally, these multisensory systems seem to be the ideal candidate to link perception and action [41]. When acting, one faces the challenge of transforming object-, body- and environment-related signals into a common reference frame, while factoring in the high number of different starting positions of the hand, and its constant motion, possibly unseen. The PPS machinery, with its hand-centred coding of the surrounding space, offers the advantage of achieving this goal in a quite straightforward fashion, as shown by Brozzoli et al. [42]. Participants had to respond to touches delivered to their hand while it was immobile, or about to start a reach-to-grasp movement towards a 47 cm distant object. Visual distractors displayed on the object were found to evoke stronger multisensory effects when the hand barely started moving, as compared with when movement was not yet initiated. Thus, despite being at the same distance from the hand, visual stimuli become more effective to alter tactile processing in agentive situations (see also [43,44]). More recent work along these lines revealed that actual movements were not even required for multisensory enhancement: the mere planning of an upcoming hand action was sufficient to tune the sensitivity of tactile perception to visual interference, further supporting the notion that PPS regions may serve as sensorimotor interfaces to control acquisitive actions [30].

Direct transformations of spatial coordinates may be useful for the motor control of voluntary acquisitive actions, but they might even be mandatory for fast defensive actions. As we will further develop, for PPS mechanisms to be phylogenetically advantageous in protecting the body from threats and damages, they need to process upcoming stimuli quickly. Evidence supporting this quick processing of nearby events comes from a study using single-pulse transcranial magnetic stimulation (spTMS) over the motor cortex, whereby participants performed a simple button-press motor response while seeing a ball that suddenly fell just above their responding hand, or far from it (figure 2). Applying TMS over the contralateral primary motor cortex, we found that this potentially threatening visual stimulus suppressed corticospinal excitability as early as 70 ms following its appearance, which we interpreted as the sign of proactive suppression of an automatic avoidance-related response. This inhibition reversed to facilitation when the two motor behaviours (the avoidance and the task-related responses) were uncoupled (see also [46]). Importantly, in both cases, the modulation of the motor excitability varied with the distance of the ball from the hand, being maximal at the shortest one, and it was independent from the retinal position of the visual stimulus, thus showing the PPS typical hand-centred pattern. Such a fast modulation of M1 may be granted by a short and rapid pathway from the retina to the motor cortex [47], which could update hand-relevant visual information for the dynamic control of hand–object interactions. As suggested elsewhere [48], such a mechanism might be useful to either catch or avoid objects in a suddenly changing environment, when alternative pre-planned options would fail.

Figure 2.

Experimental setup (a) and time course (b) of the spTMS study. Participants held their right hand either to the left (as shown in a) or to the right of a central LED which served as the fixation point. They had to respond to a transient (130 ms) offset of the fixation LED by abducting their right index finger to push a button (black square) while ignoring the distractor ball (red sphere). The distractor ball appeared simultaneously with the go signal, approaching either the left or right (shown in a) side of fixation. Between 40 and 120 ms after the appearance of the distractor ball, a single pulse of TMS was delivered to the hand area of the left hemisphere primary motor cortex, eliciting an MEP in the right first dorsal interosseous muscle. From Makin et al. [45].

(d). Attentional dimension

At this point, one may wonder whether one can provide a unified explanation for the different effects found in PPS besides their mere spatial overlap in relative proximity. For instance, it has been initially suggested that sensory enhancement in PPS possibly results from attention-dependent facilitation [49]. However, growing evidence shows that it reflects a distinct process, though attention and PPS processes are likely to co-occur and thus prove difficult to take apart. In the falling-ball TMS study by Makin et al. [50], to control for overt attention, we asked participants to look at LED-cued positions congruent or incongruent with the hand position, so that the proximity effect of the ball could be defined in reference either to the hand (near the hand, fixation to the opposite side), or to the eyes (near ocular fixation, hand on the opposite side). Similarly, to control for covert attention, we asked participants to fixate centrally and to orient attention at different LED-cued positions. In both control experiments, the rapid inhibition of corticospinal excitability was predominantly hand-centred, depending upon the distance of the ball from the hand, regardless of the locations of both visual fixation and covert spatial attention. These findings call for a genuine PPS-related effect independent from the spatial attentional effect.

Zanini et al. [26] recently brought additional evidence for the separation between PPS- and attention-dependent effects on behavioural performance. In a series of experiments designed to distinguish PPS from the (often conflated) arm reaching space (ARS), they found that PPS was smaller than ARS. They further found that multisensory facilitation closely followed hand location for PPS, but not for ARS. Most importantly, when participants’ attention was diverted elsewhere (they had to perform a visual gap detection task at a location equidistant from the locations where near and far stimuli were respectively projected), participants still showed better multisensory performance for near stimuli. This finding corroborates the distinction between PPS- and attention-dependent modulation of perception.

This dissociation rules out attention as the unifying driving mechanism for increased multisensory facilitation and for improved visual processing in PPS. Instead, we explain the various effects found in PPS by the primary function of PPS mechanisms, namely, their motor function. As described by [51], PPS is the space ‘within which it [the body] can act’. All bodily movements necessarily unfold in the space that immediately surrounds one’s limbs. One may then propose that both unimodal and multimodal effects have been selected because of their respective importance for action. On one hand, mapping external objects in body part coordinates allows for direct sensorimotor transformations. On the other hand, faster and more accurate shape recognition facilitates action preparation. Hence, the two types of sensory signatures do not merely coincide in the same region of space; they share the same functional role.

3. A defensive conception of peripersonal space

The original reports on PPS-related regions (e.g. VIP) described that multisensory neurons were especially sensitive for stimulus displacements. This has been taken as evidence that their function is more reactive (responding to changes in the environment) than active (initiating changes in the environment). More specifically, it has been argued that one of the major functions of PPS mechanisms is ‘to maintain a margin of safety around the body and to coordinate actions that defend the body surface’ [4]. The protective role of PPS is said to generalize to anything to which is ascribed a negative value, from predators to mere obstacles on one’s path. One may also argue that it extends to threats surrounding other individuals in light of the activity of PPS mechanisms when objects approach other bodies [52,53], or vice-versa [29]. Although most evidence comes from the stimulation of visuo-tactile neurons, it is in the auditory modality that the defensive conception of PPS appears as the most plausible. For sighted individuals, auditory information provides limited information for instrumental actions such as reaching or grasping. By contrast, it is an efficient alarm system that gives reliable cues about the presence and the proximity of threats (as long as they are noisy), even in the dark and in the rear space that the animal cannot see.

Two main series of results support the defensive conception. First, at the motor level it has been found that the activation of PPS mechanisms triggers defensive responses that are body-part specific [45,54–59]. Second, at the sensory level, the negative value of stimuli has been shown to significantly modify the extent of PPS, especially if participants display anxiety [58,60–65]. These findings suggest that PPS mechanisms are sensitive to whatever has been appraised as potential threats by the amygdala [66] to monitor them and to react to them, even though we shall note some limitations to this conception.

The most convincing evidence comes from Michael Graziano’s research in monkeys. With his team, he found that electrical stimulation of the regions embedding multisensory neurons in the precentral gyrus and in the VIP elicited defensive movements spatially tuned to body parts similar to those induced by a puff of air directed at the animals [55–57]. For instance, monkeys rapidly lifted their hand into the space near the side of their head as if to block an impending impact when the stimulation was applied on the region that contained face-centred tactile neurons or they brought their hand behind their back when it was applied on the region that contained hand-centred neurons. In humans, it has also been found that the perception of threats in PPS can trigger defensive responses. The eye-blink reflex can be elicited by a strong stimulation of the median nerve at the wrist (the hand-blink reflex, HBR) and this reflex is substantially enhanced with the proximity of the hand to the eyes [58]. It has also been found that a 300 ms burst of white noise near the hand reduces corticospinal excitability, which resembles that found during the presentation of noxious stimuli, unexpected events and potential threats [45,54]. In addition, the amplitude of motor-evoked potentials evoked by TMS was enhanced when a sound was heard within a distance of 60 cm from the hand with respect to sounds occurring further away [59].

Unfortunately, these are the only studies that investigate the defensive function of PPS at the motor level. Although PPS has been described as a ‘motor space’ from the beginning [67], little has been done to directly investigate its purely motor effects. ² Instead, most studies focus on its multisensory dimension. The question then is no longer whether PPS mechanisms contribute to defensive movements, but whether they contribute to threat monitoring. Interestingly, the presentation of visual stimuli in PPS can interfere not only with tactile processing, but also with nociceptive processing [68,69]. Even for tactile processing, it has been shown that it is differentially affected depending on the affective value of the surrounding stimuli. Auditory-tactile facilitation effects are found for stimuli at larger distances when participants hear, looming towards them, the sound of screaming individuals or fearful animals, or abstractly unpleasant sounds [60–62]. These effects are enhanced if participants feel anxious or display anxiety traits [58,61,63–65]. For instance, seeing approaching spiders induces visuo-tactile facilitation effects earlier in participants with high levels of arachnophobic fear [64]. Although brain regions can modulate indirectly these responses, these findings are consistent with the hypothesis that PPS mechanisms are used to monitor what has been appraised as immediate threats by the amygdala. On this view, if the extent of PPS ‘stretches’ farther away in threatening contexts, it is to give more time for the animal to react. However, one may ask, is it not already too late to hear or see the predator coming if it is already close to us?

4. Plan B hypothesis

The defensive conception of PPS has been inspired by Hediger’s [70] work on the direct link between spatial processing and defensive animal behaviours (e.g. [4,71]). However, what Hediger and later research have shown is that there is a spatio-temporal gradient of defensive responses, some of them starting very early on, long before the threat enters PPS and long before there is any emergency. According to the threat imminence continuum hypothesis, there are at least three phases, each associated with distinct affective, motor and neural signatures [72,73]. In pre-encounter threat stage, predators may be around but they are not yet detected. The animal is vigilant and acts with caution. It displays ‘anticipatory anxiety’. In post-encounter threat stage, predators have been detected but they are not attacking. The animal now displays ‘encounter anxiety’ and initiates stereotypical behaviours like freeze or flight. When the threat becomes closer in time and space and the attack is a few seconds away, the animal shows a fear response. It typically engages into flight, freeze or fight depending on the context. Finally, in the circa-strike threat phase, when the predator attacks, the animal typically displays panic responses and tries to defend itself [73]. Likewise, humans react differently depending on the imminence of danger. In one study, participants were in a two-dimensional maze in which they had to avoid a ‘virtual predator’ that had the capacity to chase, capture and cause pain of high or low electrical shock. As the virtual predator grew closer, brain activity shifted from the ventromedial prefrontal cortex, associated with decision making, to midbrain structures involved in defensive responses and pain anticipation, especially for high-pain predators [74].

In this continuum, it seems that PPS arrives at the very end only. Although it has been suggested that PPS corresponds to the flight zone (distance at which the animal starts fleeing) [4], this zone extends far beyond the majority of receptive fields of PPS-related neurons. Although some of the neurons in PPS regions have receptive fields that extend out a few metres, most have a strong preference for near space, whereas the flight zone can be as large as 130 m for a giraffe or 10 m for a howler monkey [70]. Others have recently proposed that PPS is more closely related to freezing, which consists in reduced mobility, tense body posture and bradycardia [75,76]. However, the main functions of freezing are for the animal to evade detection, scan its environment and prepare its actions. It must thus occur again before the predator enters the PPS. For instance, in rats, freezing is near maximal when the predator is 100–300 cm distant from the animal, and declines abruptly at about 50 cm, when the attack becomes imminent [77]. At this final step, the animal is gradually overtaken by the predator and it is generally left with little choice but attack with emergency characteristics. Hediger [70] then talks of critical distance, which is a space not only for fleeing but also for fighting. This may be possibly the closest equivalent to PPS. Interestingly, if this is the case, then it means that the direction of defensive responses in PPS can be towards the predator, and not systematically away from it.

One may reply that monkeys react with squinting or ducking behaviours after electrical stimulations of PPS neurons, and not with aggressive behaviours [56], but it is hard to predict what they would do in more ecological conditions when facing a predator that cannot be outrun, especially if it is already within PPS. Still, it remains to be tested whether fighting responses do involve PPS mechanisms. Our claim is only that when a predator is close, the best response is not necessarily flying away and thus, contrary to what is sometimes assumed (e.g. [10]), the negative value of the surrounding stimulus should not be equated with avoidance behaviour.

In our contemporary Western society, we rarely encounter predators but there are still many sources of danger. There again, we generally do not wait for the last second to act. Consider the case of looming perception. There are predictive mechanisms that allow computing the time to collision at an early stage, before the looming stimulus enters the immediate surroundings of one’s body [23]. This allows time to plan one’s response (interception or avoidance). Consider also the defensive function of the colliculus, which displays multisensory processing with interesting basic integration principles properties. According to the ‘inverse efficiency’ principle, visual-auditory neurons typically display a stronger (super- or sub-additive) integration of sensory inputs when the latter is weak. For example, seeing a barking dog at a distance (when both visual and auditory signals are weak) makes neurons fire more strongly than when the dog is close [78]. In a nutshell, collicular multisensory integration properties are best suited to detect threat well in advance compared with PPS processing (figure 3).

Figure 3.

Multisensory integration at the level of the single neuron in the colliculus allows for the anticipated detection of threats. The approach of the dog is signalled by audition and vision. 1. When the dog is far and the sensory inputs are weak, multisensory integration is super-additive: the neuronal response is stronger than the sum of each unisensory response. 2. As the dog approaches, unisensory responses become more vigorous, whereas integrated responses get smaller. The computation now becomes additive. 3. When the dog is close, the computation is sub-additive. All enhancements increase the probability of orientation, but the benefits of multisensory integration are proportionately the greatest when cross-modal cues are the weakest, i.e. when the threat is the farthest. From Stein et al. [79].

We thus have a range of mechanisms specifically designed to protect our body. For these mechanisms to be the most efficient, they need to be activated before the last second. Some may be tempted to enlarge the notion of PPS so that it is no longer specific to the immediate surroundings of the body and so that it encompasses these other mechanisms. On this view, PPS is characterized as a graded action field triggered by the ascription of values at various distances from the body [10]. Such an enlarged notion of PPS can secure its major significance for self-preservation, but at the cost of losing the physiological and behavioural specificity of PPS. Furthermore, even within an enlarged conception, one might still fix limits to PPS. Arguably, it is doubtful that seeing a predator 20 m away and initiating a flight response depends on PPS mechanisms. As argued in their original paper on PPS, ‘proximity of the stimulus to the animal was really the condition responsible for the neuron responses’ [5, p. 149]. Rizzolatti et al. meant by proximity 1 m away, but even if one stretches it farther, it still remains that the majority of PPS neurons are sensitive to what is relatively close to the subject.

The challenge is thus to understand how PPS mechanisms may still contribute to bodily protection despite being triggered so late. One may indeed wonder what role is left for them to play given the richness of these other defensive mechanisms and the necessity for them to be activated as early as possible. Here, we propose what we call the Plan B Hypothesis, according to which PPS mechanisms for self-defence are required only when all the other mechanisms have failed, which unfortunately often happens in unpredictable environments. They thus may have originally been selected for two reasons: for bad surprises and as a last resort.

We have seen that we have long-range sensory systems that allow us to detect threats and obstacles at large distances and that we further have a number of anticipatory mechanisms that allow us to plan our actions accordingly. Although some neurons in PPS regions can contribute to longer range processing, eye-centred mechanisms can outweigh PPS hand-centred mechanisms under predictable conditions. Indeed, they are likely to provide more accurate visuospatial information about the relative positions of the object and the hand. However, there are limits to what one can predict, especially under unstable environmental conditions. By definition, last-minute changes cannot be anticipated. They thus require last-minute mechanisms, namely, PPS mechanisms. These mechanisms are specialized in the rapid updating of relevant visual information during response selection and online control of action, which are useful when the position of a target object unexpectedly changes [50]. This is especially true when one is surrounded by other agents. Objects obey basic principles in physics that allow us to predict their trajectory, but agents’ movements are more complex to anticipate. If you need to walk in the subway at rush hour, instead of a room full of furniture, you cannot predict well in advance the other passengers’ paths. Monitoring your immediate environment is then important to react quickly to sudden, unpredictable changes. Predictions can also fail for lack of sensory information. We may not be short-sighted but it does not mean that we have an exhaustive sensory access to all our surroundings. A snake may have been hidden from our sight till our steps get us close to it and it suddenly jumps on us. As recalled above, humans are much quicker at recognizing shapes in close than in far space [35]. This efficiency is directly useful when something suddenly pops out next to one’s body and one needs to identify it to determine whether it is a threat or not. To recap, the unexpected explains why it is beneficial for the organism to have enhanced visual processing in PPS. Finally, even when one is able to accurately predict the threat in advance, one’s responses may be insufficient to fully neutralize it. PPS mechanisms can then be seen as the final stage in a series of attempts to stop the danger, a last resort.

5. A cost–benefit analysis of peripersonal space

At this point, one may wonder why wait so late to trigger the plan B. Would not it be more beneficial if PPS mechanisms operate when the threat is not yet in relative proximity? One might intuitively assume that an animal in which PPS mechanisms are activated when the potential threat is still relatively far away would have more chance to survive but this would be without factoring the cost of these mechanisms. Most studies have focused on their benefits, but rarely on their detrimental aspects. Yet, they do have a price, and it may not be worth it to trigger PPS mechanisms earlier. Here, we can draw a comparison with optimal flight initiation distance (FID), which is computed on the basis of cost–benefit analysis. It has been proposed that optimal FID occurs at the distance when the cost of not fleeing and the cost of fleeing are equal [80]. It depends not only on predation risk, but also on the costs of fleeing, which include the energy it requires, and the loss of opportunities, like eating or mating. The fact is that not all components of fitness can be maximized simultaneously [81]. For instance, short FID is associated with high fecundity, but low survival prospects, whereas long FID is associated with low fecundity but high survival prospects.

We may then ask whether an earlier PPS would be worth the costs. Within the recent trend of predictive coding theories, predictions are said to be pervasive across all the layers of the brain architecture, as if they could be generated for free. However, if they do have a psychological reality, instead of being merely a computational model of what happens in the brain, then they must have a neural and thus energetic cost. Neural computations do not come for free and this applies to PPS computations too. The cost is also cognitive. For instance, when participants are asked to localize a target among distractors, they are slower when the visual display is close than when it is far [82]. In this respect, PPS mechanisms seem to fit well the ‘better safe than sorry’ tenet. Visual search can be less efficient in PPS than in far space because it is more thorough. Relatedly, ongoing work investigated whether perceptual learning could vary according to the proximity of visual stimuli. As in Blini et al. [35], we used the Ponzo illusion to manipulate the perceived distance. Participants performed a visual search task in which they reported whether a specific target object orientation (e.g. triangle pointing downward) was present among distractors. Performance was assessed before and after practicing the visual search task (30 min a day for 5 days) at either close or far distance. Results showed that the performance improved only when trained in far space and not in near space [83]. Arguably, learning is more effective thanks to a greater deployment of attention, which may not be worth its cost in the immediate vicinity of one’s body. Indeed, when an object appears in PPS, investing too much attention is both inefficient and risky. There is not enough time left, from its identification to the appropriate reaction, and there is the risk of increasing the likelihood of failure because of slower, attention-demanding processes.

Finally, we need to question the actual usefulness of the gains given by PPS processing if it were activated farther away from the body. Consider first its multisensory signature. Objects seen or heard in PPS are encoded in a somatotopic frame of reference relative to the specific body part that they may impact. One way to interpret these results is that the brain generates somatosensory predictions on the basis of what is seen or heard. Though it is perceived as being still outside of the body, one anticipates its contact with the body. Somatosensory encoding of external events allows for more spatially tuned responses. You do not react in the same way if you see a bee next to your head or next to your hand. When the bee is still relatively far away, however, the exact end point of its trajectory is still too uncertain for making it worth computing. Furthermore, your response needs not be as specific. At this stage, the bee is only flying in your direction, and you can simply move away. What matters is only its relation to your whole body, and not to any specific part of it. In the end, activating the PPS machinery at a large distance would thus come at a large cost for little-to-no advantage. Predicting a forthcoming contact makes sense when the object is close by because its probability is at its highest. It suffices for either the object or the subject to move slightly to make it happen. By contrast, when the object is further away, the uncertainty is too high because many things can intervene preventing the contact, or making the contact happen in a previously unforeseen location on your body. We argue here that fine-grained spatial tuning enabled by the somatotopic organization of PPS mechanisms increases the likelihood of survival when the threat is close, but it is not required when it is farther away. Consider now the sensorimotor signature of PPS. In close space, perception can directly be transformed into action: this is efficient insofar as there is no need for intermediary decision-making process. However, if one can afford the time to think things through, because the threat is still remote for instance, it seems that it is better to be smarter, even though slower.

6. Conclusion: picking up berries

PPS mechanisms may be a plan B when it comes to self-defence, but is there no other purpose for which it is the plan A? Is it only a spare wheel, and nothing else? Here, it should be reminded that self-defence in the animal kingdom primarily requires the detection of the predator but its localization can remain only approximate. All that is needed for escape is a rough sense of the direction of danger [84]. It is only when attacking that the animal requires fine-grained spatial information. In many species, however, the same mechanisms are involved in both escape and predation: ‘the neural and muscular machinery that evolved for escape purposes has also been usurped (or perhaps co-evolved) within the same species for the purpose of attack’ [84, p. xv]. This is true of PPS mechanisms too, especially in the visual modality, which is our primary source of information to act on the world. One may tentatively suggest that depending on the external context and one’s inner states, whether there is food around and whether one wants to get it for instance, PPS neurons may switch to a different mode/function, which then becomes more prepotent. Here, we would like to explore the hypothesis that PPS visual mechanisms could be the plan A for fine-tuning appetitive movements. This hypothesis does not entail that the defensive function of PPS mechanisms is only secondary and the appetitive function primary. The contrast that we draw instead is that PPS mechanisms are only required in some circumstances for self-defence, but that they might be more systematically required for appetitive action planning. More precisely, our working hypothesis is that in many circumstances the precise trajectory of the hand and the arm is computed only when the target of the movement is in the immediate surroundings of the effector. Imagine being in front of a bush full of raspberries. You form the intention to pick them up long before you are next to it but it is only when you are very close that you programme the precise hand motion to get to the fruit that seems to be ripe enough. There is no need to plan it before. Further work, however, needs to be done to explore both theoretically and experimentally this plan A hypothesis, which remains highly speculative at this stage.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

This article has no additional data.

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

F.d.V.: conceptualization, writing—original draft, writing—review and editing; A.F.: conceptualization, writing—original draft, writing—review and editing.

Both authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

No funding has been received for this article.