1 INTRODUCTION

Time-to-contact (TTC) perception refers to the ability of an observer to estimate the remaining time before an object reaches a point in the environment. Accurate TTC perception could be critical for catching prey or predatory avoidance for animals (Frost & Sun, 2004), but is also likely to be involved in human daily activities like safely crossing a street or catching a ball (e.g., Baurès et al., 2014). Generally, observers are found to rely principally on the inverse relative expansion rate of the object's image on the retina (“tau”) or its derivative (“tau-margin”) for their estimation (Bootsma & Oudejans, 1993; Lee, 1976), independently of most of the other trajectory parameters (e.g., Yan et al., 2011).

However, it can happen that the object is lost from sight, for example if one car occludes the approach of another one, or for a pedestrian willing to cross a street and who cannot look at cars approaching from both directions at the same time. In laboratory, this particular case of estimating the TTC of an occluded object has been termed a prediction motion (PM) protocol. In this task, having seen the initial part of an object's trajectory prior to occlusion, the participant is required to make a response (e.g., button press) that coincides with arrival time of the now unseen object at a specified location (e.g., Bennett et al., 2010). Two alternative strategies have been proposed to explain the PM performances (DeLucia & Liddell, 1998; Makin, 2018): first, the observer could estimate the TTC during the visible part of the motion, and count down the TTC during the occlusion (clocking strategy). Alternatively, they could run a mental simulation of the object's position (cognitive motion extrapolation), and would update it at a certain speed until the estimated arrival. The task therefore might be temporal (estimating the object's TTC), or spatial (mentally simulating an object at a moving position). For the sake of clarity, we will hereafter refer to a TTC estimation task, even if we fully acknowledge that the observer might use a spatial strategy rather than a temporal one. In the PM protocol, various parameters of the task can vary, such as the occlusion duration (e.g., Baurès & Hecht, 2011), movement direction (in the fronto-parallel plan, e.g., Landwehr et al., 2014, or looming object, e.g., Yan et al., 2011), or observers' characteristics as age (e.g., Benguigui et al., 2008). In terms of performance, participants are generally found to overestimate short occlusion times, and conversely to underestimate long occlusion times, with a linear transition between those points (e.g., Bennett et al., 2010).

The neural substrates of this TTC perception involve activations in area V1 (e.g., Coull et al., 2008) and in the MT/V5+ complex (Bosco et al., 2008; Delle Monache et al., 2017; Dessing et al., 2013; Schenk et al., 2005). Depending on the task demands (e.g., manually intercepting the object) and context (e.g., implied gravity acceleration), activations have been detected in the superior parieto-occipital cortex (Dessing et al., 2013), temporoparietal junction (Delle Monache et al., 2017; Indovina et al., 2005), and insula (Indovina et al., 2005). Interestingly, it has been shown that when the object approaches the observer directly, primary sensorimotor areas, the inferior parietal lobule, the ventral premotor cortex, and the supplementary motor area (SMA) are engaged (Coull et al., 2008; Field & Wann, 2005). Whether this cortical network is left-lateralized (Assmus et al., 2003; Coull et al., 2008; Field & Wann, 2005), right-lateralized (O'Reilly et al., 2008), or balanced between both hemispheres (Billington et al., 2011) remains an open question. Overall, these results are in agreement with previous findings suggesting that attention is directed to the object requiring the most urgent behavioral response (Franconeri & Simons, 2003; Lin et al., 2008).

However, it remains unknown whether such activations reflect the TTC estimation process or rather the preparation/execution of the motor response it triggers. As pointed out by Vaidya et al. (2019), brain imaging methods (e.g., MRI or EEG) measure correlation between brain activity and behavior. As such, they should not be interpreted as a direct and causal demonstration of the brain areas that affect the behavior. Such causal roles require studies that involve lesion or perturbation including brain stimulation or transcranial magnetic stimulation (Vaidya et al., 2019).

Our aim was to use both types of causal approach to determine the brain areas actually engaged in TTC estimation of a looming object. We conducted two studies, one with 40 brain-lesioned patients and one with 15 patients performing language, motor, and TTC tasks during awake surgery and brain mapping through cortical electrostimulation.

2 METHODS

3 RESULTS

3.1 Study 1

Six patients over 40 had impaired, preoperative TTC performances. Figure 1 illustrates the behavioral results from two selected participants, with frontal and parietal brain tumors, respectively. As can be clearly seen, the tumor does not affect the ability to estimate the TTC for the first participant. For each block, as well as the overall average, the CE slowly decreases and switches from positive (overestimation) to negative (underestimation). Moreover, there is no distinguishable difference in CE due to the object's velocity. This is a very different pattern for the participant 2, who was classified as having an impaired ability to estimate TTC. There is no consistency in CE, with erratic differences across the blocks.

The classification of participants with normal (N = 34) or impaired (N = 6) performances showed no marked difference in the mean CE between the two groups (mean CE = −80 ms for the normal group compared to 199 ms for the impaired group). The error distribution is however slightly superior for the impaired group, with nearly 15% of the trials with CE > 1.5 s, compared to only 1.6% for the normal group. Overall, the intertrial variability in the responses was about two times more pronounced in the impaired than in the normal one (SD of 829 ms vs. 424 ms, respectively). Impaired patients therefore appear unable to estimate consistently the TTC of an approaching object.

Table 2 describes the number of patients with normal or impaired performances as a function of the localization of their brain tumor. Most patients with impaired performances had tumors in the right parietal lobe (4/6), although this tumor location represented only six over 40 patients. This marked imbalance was statistically significant (Fisher's exact test, p = 0.01).

TABLE 2. Number of participants with normal or impaired results as a function of brain tumor localization and hemisphere lateralization (left/right)

Brain region	Results
	Normal	Impaired
Frontal	7/6	0/1
Temporal	9/8	1/0
Parietal	2/2	0/4

We examined the MRI scans of the six participants with impaired performances (see Figure 2). In the parietal lobe, the tumors cover regions in the right hemisphere made of the superior and inferior parietal lobules, (including the intraparietal sulcus). The region of higher overlap for these tumors seems to be around the intraparietal sulcus, suggesting that this region is of particular importance in the TTC estimation process. In the right frontal lobe, the tumor covers a region from the posterior part of the superior gyrus to the middle frontal gyrus. Finally, in the left hemisphere, the tumor covers the anterior part of the superior temporal lobe.

3.2 Study 2

In the 15 patients of this experiment, 44 cortical areas in left and right hemisphere were tested with TTC and language tasks. Figure 3 represents examples of the constant error (CE) as a function of occlusion time for some representative patients (1, 2, 8, and 9) during brain mapping (All the graphs obtained during brain mapping are available in the Supporting Information Results S2). For each patient, each line represents the average CE across the three repetitions when a particular brain site is stimulated. The grey squares represent the error range obtained in the presurgery condition for the same patient, the purple squares represent the error range for the age and gender paired control participant, while the blue squares represent the 95% confidence interval of the mean obtained in each corresponding condition during Study 1, for the patients who had no significant deficit in task performances. The ANOVA did not show any difference between the control and patients group, nor as a main effect, F(1, 28) = 0.36, p = 0.552, nor in interaction with any other factors (all p > 0.170). At the individual level, the error range of each patient strongly overlapped with the error range of his/her control participant, indicating that the ability to estimate the TTC was preserved in these patients.

On the basis of these intervals, TTC estimation data during the surgery were classified as:

1. no interference in TTC estimation (Figure 4, white points);
2. interferences in TTC estimation and language tasks (orange points);
3. interferences in TTC estimation and ocular/motor movements detected (green points);
4. interference in TTC estimation and no interference in language or motor/ocular tasks (Specific TTC areas, red points).

Specific TTC areas were all located in the right hemisphere, in the parietal lobe. Within the parietal lobe, six stimulation sites led to an impact on the participant's ability to estimate the TTC. Notably, only the TTC estimation for the shortest occlusion time is impaired while the estimation for the longest occlusion time remains unaffected.

Nonspecific TTC areas (orange or green points in Figure 4) represented a second subgroup. Eighteen stimulation sites led to a strong increase in CE of the TTC estimation. Here the interference pattern differs depending on the hemisphere stimulated. For the participants stimulated within the left hemisphere (11 orange points), and quite distinctly from the others groups, the increase in CE is very general and does not affect only one particular occlusion time. The slope of the CE is also positive, therefore opposite to that observed in normal subjects (e.g., Bennett et al., 2010). Importantly, the stimulation of these areas also induced perturbations in language tasks. As such, for all these sites (orange points, the increases in the TTC interference seem nonspecific to the TTC estimation task (Roux et al., 2015), but more generally a by-product of interfering with language.

For the patients stimulated in the right frontal lobe (7 green points), the interference pattern is similar to the TTC-specific points. However, the stimulation also induced involuntary eye or hand movements, which prevented the participants from performing the task normally (see an example of stimulation in Figure 5 and Supporting Information—brain mapping video).

Finally, stimulation sites that did not lead to an increase in CE on the TTC task are shown as white points in Figure 4. Of those 20 sites, 15 also failed to induce any effect in the other tasks performed during the surgery, while five sites induced errors in the language task (denomination or reading, see Supporting Information Results S1 for the details).

4 DISCUSSION

What brain areas support the TTC estimation of an approaching object that eventually gets occluded? In two studies, we investigated this question by testing participants suffering from tumor in various brain areas, including during an awake surgery. We found converging results in favor of a causal implication of the region around the right intraparietal sulcus in the TTC estimation.

The results of the first study show that two thirds of the patients with right parietal tumors (encompassing the superior and inferior parietal lobules and the intraparietal sulcus) exhibited impaired performances in the TTC estimation task. It is however important to notice that two other patients with a brain tumor in the right parietal area still had correct results. It could be hypothesized that these patients benefitted from brain plasticity that allowed them to achieve correct performance in spite of the tumor. Together, these observations provide an initial indication that the right parietal cortex plays a causal role in TTC estimation process, at least when the object is looming and suggests an impending collision. In contrast, the frontal areas that were activated in previous experiments (Billington et al., 2011; Coull et al., 2008; Field & Wann, 2005) appear not to be involved in the TTC estimation per se, and are more likely to play a role in the motor preparation or execution following the estimation. An alternative interpretation would be that these areas are recruited for both the TTC estimation and the motor preparation, and that the electrical stimulation used in our study does not allow disentangling between these two activities.

To understand what could be the implication of this region in the TTC estimation, it is important to describe the different steps of the task. In the prediction motion paradigm we used, the task requires overt attention to the ball (Baurès et al., 2015), pursuit eye movements (Bennett et al., 2010), a dynamical representation of the ball's position (according to the cognitive motion extrapolation strategy) or TTC (according to the clocking strategy) during the occlusion time (Bosco et al., 2015; Makin, 2018) and a motor response to indicate the estimated arrival time. All these processes are known to involve the parietal area (e.g., Astafiev et al., 2003; Bisley et al., 2011; Harrison et al., 2010). The current results do not allow to identify which of these processes is (or are) impaired by the presence of the tumor, but only that the parietal region is causally engaged in one stage or more of the TTC estimation task. It could be hypothesized that after an early stage in both occipital lobes, downstream TTC estimation processing would relate more specifically to the region of the ventral right intraparietal sulcus involving a few, fine-grained, sub-centimeter cortical territories. The right intraparietal sulcus could also be involved in visuospatial attention and neglect. In functional imaging studies, although results may vary according to the task used, sites are consistently found around the intraparietal sulcus and the inferior parietal lobe, especially with line bisection tasks (Fink et al., 2000). Although this is debated, lesion studies (Committeri et al., 2007) indicate that extrapersonal neglect could depend on a right frontal (ventral premotor cortex and middle frontal gyrus) and posterior sylvian (superior temporal, supramarginal, angular, postcentral gyri) circuit (Mort et al., 2003; Rorden et al., 2006). Electrostimulation (Roux et al., 2011) showed that visuospatial attention is mainly sustained by the region of the right posterior Sylvian fissure or the superior parietal lobule (Vallar et al., 2014). The complete or partial independence of the TTC estimation process from the more global visual attention network cannot be solved by the current study. We assume that the task we used to evaluate the TTC could reveal one aspect of a more global visuospatial network also involved in spatial neglect.

However, as shown by Study 2, electrostimulation of the region around the right intraparietal sulcus led to an increase in CE when the object is occluded for 0.5 s, but not when it is occluded for 1.5 s. Given the object's velocity, a short occlusion time of 0.5 s implies that the object disappears very close in space to the observer's body, which is not the case when the occlusion time is 1.5 s. It therefore seems that the involvement of the parietal area depends on the distance between the approaching object and the observer. It could be that these areas and the underlying visual attention process are only or predominantly engaged if the object is close to the observer (e.g., Serino et al., 2011), and thus within their peripersonal space (PPS). PPS is the space immediately surrounding the body where sensory events typically require fast, appropriate motor responses. It has been shown that participants react faster (Serino et al., 2011) and preferentially allocate their attention (Garrido-Vásquez & Schubö, 2014) to events localized in PPS. Seminal studies in non-human primate have shown that the ventral premotor cortex (vPMc), (e.g., Fogassi et al., 1996), and intraparietal sulcus, specifically in the ventral intraparietal area (VIP; e.g., Avillac et al., 2005), may be responsible in the construction of the multisensory representation of PPS. As such, our results may indicate that the TTC perception depends on a proper representation of the observer's PPS. Incidentally, this might argue in favor of the cognitive motion extrapolation strategy to explain the performance in the PM protocol, according to which an observer simulate mentally the object's motion until its arrival. In this hypothesis, a proper representation of the space in which the (mental) object moves is necessary to simulate a proper motion.

Visual tracking throughout the visual frontal areas is another component of the TTC estimation process. Within the right frontal lobe, seven stimulation sites, inducing involuntary eye movements (or involuntary hand movement arrest for one), led to the same error pattern within a region between F1/F2. It has been previously shown that pursuing the object with the eye leads to a better TTC estimation (Bennett et al., 2010), and therefore involuntary eye movements through the stimulations are probably the cause of the observed interference in TTC estimation. As Study 1 did not indicate a link between a brain tumor in the frontal lobe (while 14 patients were tested with a frontal brain tumor) and a deficit in the TTC estimation process, we conclude that the frontal lobe is likely not involved causally in the TTC estimation.

Finally, we also found nonspecific TTC interferences when stimulating points in the left cortical hemisphere, also inducing interferences in the language tasks. This questions previous findings (Assmus et al., 2003; Coull et al., 2008; Field & Wann, 2005) of a left lateralization of TTC perception. Electrostimulation of language areas can cause either a global behavior arrest or interfere more specifically with attention and short-term memory while the participants are engaged in automatic tasks (Ojemann et al., 1989). Language mapping with language comprehension tasks (Roux et al., 2015) showed the patients are not conscious of their language interferences induced by electrostimulation in the late stages of language comprehension. Similarly, it could be hypothesized that electrostimulation interfered with an essential component of TTC estimation process, without being in a position to precisely identify this process. It could be hypothesized that estimating the object's TTC share a common attentional component with the language network of left hemisphere and located downstream of the initial TTC processing.

Our work is however not going without limits to acknowledge. For obvious clinical reasons, the stimulation sites for example were all in the same hemisphere than the tumor, and the stimulation sites were not identical for all patients. For these reasons, only three patients were stimulated in the right parietal area, and no statistics at the group level could be performed. However, for those three patients, we have a very consistent pattern for most of the stimulation sites (six sites lead to an increase in the TTC estimation for the shortest occlusion time only, over eight different sites). Our results and conclusions are therefore based on a limited number of participants, and further work should aim at testing more patients who could be stimulated close to the right intraparietal sulcus.

To conclude, our results demonstrate that when an approaching object is on an imminent collision course, within the PPS of the observer, TTC perception cannot be seen as a purely visual task. We here confirm and extend previous MRI studies (Billington et al., 2011; Coull et al., 2008; Field & Wann, 2005), demonstrating that the TTC estimation of an approaching object activates the right parietal lobe, specifically close to the right intraparietal sulcus, if the object has reached the PPS of the observer. Our findings reveal that this area is causally implied in the TTC estimation (Vaidya et al., 2019), and not simply activated by the imminent collision as a mere output following the TTC estimation (e.g., protecting the body from the upcoming collision). Taken together, these results suggest that this parietal network, possibly through its role in the visual attention network and in the representation of the PPS, lead the participant to anticipate the object's arrival. This is congruent with a purpose of this network being to prepare the observer for the incoming collision and to protect themselves or move away from the object's path. It can be wondered if this network is specific to TTC perception, or engaged in more general tasks requiring visuospatial interactions with static or moving objects (Aravind & Lamontagne, 2014). Testing patients with various visuospatial tasks during their awake surgery might answer this question.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

Conceptualization, R.B. and F.E.R.; Methodology, R.B. and F.E.R.; Investigation, R.B., F.E.R., M.F., S.T., C.G., L.P., G.M., and J.P.; Software, M.R.; Formal Analysis, R.B., F.E.R., and J.B.D.; Resources: F.E.R. and M.R.; Writing - Original Draft, R.B.; Writing - Review & Editing, F.E.R. and J.B.D.; Visualization, R.B. and J.B.D.; Supervision, R.B. and F.E.R.; Funding Acquisition, R.B.