Our brain constantly integrates the afferent information our senses provide, such as vision, touch, or audition. Numerous studies show that participants generate a more precise perception of an object when they perceive it with several modalities (Gepshtein and Banks, 2003; Alais and Burr, 2004; Bresciani et al., 2006; Nardini et al., 2008; Helbig et al., 2012). The maximum likelihood estimation (MLE) model theorizes how the brain can use noisy afferent information and maintain a stable and coherent perception through multisensory integration (Ernst and Bülthoff, 2004). This model postulates that sensory afferents are weighted according to their reliability: a noisy afferent signal is variable and thus considered less reliable and used less than a reliable signal (Ernst and Banks, 2002; Ernst and Bülthoff, 2004). Ernst and Banks (2002) conducted one of the first studies showing that healthy volunteers integrate sensory information according to the MLE model. In their study, participants had to estimate the height of an object while either seeing it, touching it, or seeing it while touching it (multisensory condition). Participants’ estimations were more precise when visual and tactile information were available than when only visual or tactile information was present, as predicted by the model. This result confirms the benefit of multisensory integration. Moreover, the authors calculated the weight of each sensory information in the visuotactile percept based on the precision of the information (see Method for more details). They showed that, as predicted by the MLE, the more the visual information was degraded (i.e., increasing its variability), the less it was used (i.e., smaller weight) compared to tactile information (Ernst and Banks, 2002). Thereafter, similar results have been reported in several tasks with various modalities (Gepshtein and Banks, 2003; Bresciani et al., 2006; Nardini et al., 2008; Ronsse et al., 2009; Fetsch et al., 2010; Helbig et al., 2012; Chancel et al., 2016; Sexton et al., 2019). This way of integrating information seems essential to generate a coherent body representation, as shown by studies in which the degradation of somatosensory information of a given body part is accompanied by perturbations in the representation of this body part (Lackner, 1988; de Vignemont et al., 2005; Chancel and Ehrsson, 2023).

One way to study sensory information integration is with temporal-order judgment (TOJ) tasks (Morein-Zamir et al., 2003; Ley et al., 2009; Maiworm and Röder, 2011). In these tasks, participants must judge the order of two successive stimuli. The delay between the stimuli varies, and psychometric functions can be generated to describe the participants’ answers according to the delay. Two outcome variables are then extrapolated from the function: the variability (SD, standard deviation) and the accuracy (M, mean) of the percept. Stimuli can be unisensory (e.g., two sequential visual or tactile stimuli) or multisensory (e.g., two sequential visuotactile stimuli). The precision and the accuracy of the multisensory percept are compared to the precision and the accuracy of the unisensory percepts, and the weight of each sensory modality in the multisensory percept can be calculated. In a study using a TOJ task with pairs of auditory, tactile, or audiotactile stimuli, tactile stimuli were degraded to test the effect of different noise levels on audiotactile integration (Ley et al., 2009). Using psychometric functions, the authors found that the empirical data followed the MLE model, with a more precise percept in the multisensory condition compared to the unisensory percepts and with an effect of the increased tactile noise on the weight attributed to tactile information (Ley et al., 2009). This suggests an optimal (i.e., according to the MLE model) integration of artificially noisy sensory information in healthy participants. But what if sensory information is noisy for a long rather than a short period? Does multisensory integration remain optimal in such conditions? Studying multisensory integration in clinical populations that show unisensory integration alterations can help answer these questions.

People with chronic pain often show alterations of the unisensory integration of somatosensory stimuli, such as a hypersensitivity (i.e., lower detection threshold; Westermann et al., 2011; Achenbach et al., 2020) or hyposensitivity (i.e., higher detection threshold; Rommel et al., 2001; Geber et al., 2008; Maier et al., 2010; Hochman et al., 2013; Hilgenberg-Sydney et al., 2016; Evdokimov et al., 2019; Sobeeh et al., 2023) to tactile stimuli, and a worse tactile acuity (Pleger et al., 2006; Harvie et al., 2017; Martínez et al., 2019; Kim et al., 2020; López-de-Uralde-Villanueva et al., 2020). These alterations are sometimes accompanied by distortions of the body representation (Moseley, 2008). In a study in which tactile perception was measured with the two-point discrimination (TPD) test, individuals with low back pain had a worse perception (i.e., higher threshold) than pain-free controls at their lower back (Moseley, 2008). Moreover, when they were asked to draw their lower back as they felt it (as opposed to how they saw it), the drawings were distorted compared to the control group (Moseley, 2008). Alterations of the body representation in chronic pain populations can manifest themselves as a perception of changes in the size (Lewis et al., 2007; Lotze and Moseley, 2007; Peltz et al., 2011; Valenzuela-Moguillansky, 2013; Dieguez and Lopez, 2017; Brun et al., 2019; Menten et al., 2022; Scandola et al., 2022), the shape (Moseley, 2008; Valenzuela-Moguillansky, 2013; Menten et al., 2022), the weight (Valenzuela-Moguillansky, 2013; Brun et al., 2019), the existence (Moseley, 2008; Valenzuela-Moguillansky, 2013; Dieguez and Lopez, 2017; Menten et al., 2022) of the body part or in the feeling of ownership over the body part (Lewis et al., 2007; Valenzuela-Moguillansky, 2013; Scandola et al., 2022). These body representation distortions could stem from alterations in the multisensory integration of somatosensory information in individuals with chronic pain. Indeed, according to the MLE model, somatosensory information should be considered unreliable in these populations and, therefore, used less (i.e., smaller weight) than other sensory information, such as vision. This weighting strategy could allow for the generation of precise perceptions based on noisy somatosensory signals.

Fibromyalgia (FM) represents an excellent model to test this hypothesis. This syndrome affects about 2–3% of the population (Branco et al., 2010; Wolfe et al., 2013) and is defined by widespread chronic pain accompanied by various symptoms affecting sleep, cognition, and motricity (Häuser et al., 2015; Häuser and Fitzcharles, 2018). People with FM also show signs of alterations of the unisensory integration of somatosensory information, such as higher tactile detection thresholds (Üçeyler et al., 2015; Evdokimov et al., 2019), higher TPD thresholds (i.e., poorer tactile acuity; Martínez et al., 2019), and referred sensations (i.e., sensations arising from outside the stimulated site; Martínez et al., 2019) following tactile stimulation. In two recent studies, participants with FM and pain-free participants received pairs of temporally close tactile stimuli on the hand and were asked to indicate when they perceived them as two separate stimuli (Gunendi et al., 2019; Tertemiz and Tepe, 2022). In both studies, participants with FM needed inter-stimuli delay on average two to four times longer than the controls to perceive the stimuli as separate. Moreover, a correlation was found between the altered temporal discrimination threshold and several FM symptoms, such as the intensity of pain, the severity of various somatic and cognitive symptoms, and the level of impairment (Gunendi et al., 2019). In addition to these alterations, individuals with FM experience body representation distortions (Valenzuela-Moguillansky, 2013; Menten et al., 2022; Scandola et al., 2022), which could stem from modifications of the weighting of somatosensory information. However, no study has directly tested multisensory integration in this population.

This study aimed to assess visuotactile integration in people with FM compared to pain-free individuals (CTRL). Objective 1 was to evaluate the precision and accuracy of the unisensory percepts (visual, tactile) and compare them across groups. We expected that the participants with FM would show less precision and accuracy in the tactile percept than the visual percept compared to the CTRL participants. Objective 2 was to assess the benefit in precision and accuracy of having access to several sources of sensory information and compare this benefit across groups. We hypothesized that CTRL participants would exhibit a stronger benefit in precision, and possibly in accuracy, than FM participants. Objective 3 was to determine the relationship between the observed precision of the multisensory percept and the precision predicted by the MLE model in individuals with FM and CTRL participants. We expected a positive correlation between the observed and the predicted precision in CTRL participants but not in participants with FM, suggesting a sub-optimal visuotactile integration in the latter group. Finally, Objective 4 was to compare the weight given to each source of sensory information in the FM group vs. the CTRL group. We expected that participants with FM would rely less on tactile information, as shown by a smaller weight attributed to this information, compared to the CTRL group.

Of the 44 participants (20 participants with FM and 24 pain-free controls) recruited, we excluded 11 participants either because (1) their performance in at least one of the unisensory conditions was not in the predetermined range of [60; 90]% (n = 6); (2) the SD was abnormally high in a modality even though the performance score was within the predetermined range (n = 2); (3) the nonlinear curve did not fit the data (n = 1); or (4) the participant discontinued their participation (n = 1). Of the participants whose performance was outside the range, seven scores were below 60% (6 CTRL participants: 4 in the Tactile condition and 2 in the Visual condition; and 1 participant with FM in both the Tactile and the Visual condition) and 1 was above 90% (1 participant with FM, in the Visual condition). In total, the data of 33 participants (15 participants with FM and 18 CTRL) was used for further analysis.

To our knowledge, this is the first study to assess multisensory integration in individuals with FM and confront it with the MLE model. Results show no intergroup differences in the tactile and visual percepts’ precision and accuracy. Contrary to our hypothesis, participants with FM were as precise and accurate as controls in perceiving the unisensory tactile and visual stimuli. No group differences were observed in the BP and the BA in the multisensory condition. The CTRL group benefitted from access to tactile and visual information, as expressed by increased precision. In contrast, in the FM group, the BP was not different from zero. We observed no benefit in accuracy for either group. According to the MLE model, multisensory integration was optimal for both groups because the observed precision of the multisensory percept was correlated to the precision predicted by the MLE model. Finally, weights attributed to each sensory information were not different between the groups. However, a bimodal distribution of the weights highlighted two clusters of participants based on the preferred sensory information (tactile dominant or visual dominant), regardless of the group. We will first discuss the results of the unisensory conditions. Then, the findings of the multisensory condition will be interpreted in three parts: 1-the multisensory benefits in precision and accuracy; 2-the optimality of the integration according to the MLE; and 3-the weighting of each sensory information in the multisensory percept. Lastly, we will outline some limitations.

In the unisensory conditions, participants with FM were as precise and accurate as CTRL participants in perceiving tactile and visual stimuli despite the tactile alterations reported in this syndrome (Hilgenberg-Sydney et al., 2016; Evdokimov et al., 2019; Martínez et al., 2019). This observation contradicts recent studies showing that participants with FM need longer time intervals between pairs of tactile stimuli to perceive them as separate, compared to pain-free participants (Gunendi et al., 2019; Tertemiz and Tepe, 2022). However, in these studies, the participants with FM reported higher pain intensity than in our sample (mean scores of 7.7/10 and 8.6/10 on two different scales in the study by Tertemiz and Tepe, and of 7.3/10 in the study by Gunendi et al., compared to the BPI pain severity score of 5.4/10 reported in the present study). In the study by Tertemiz and Tepe (2022), tactile detection thresholds were also higher for participants with FM than controls, which was not the case in our study. These differences suggest that our sample of participants with FM was not as severely affected, which could explain the divergent findings. When comparing to other measures of tactile perception in individuals with FM, a lack of intergroup difference in tactile perception is however not uncommon (Lim et al., 2016; Palmer et al., 2019; Menten et al., 2022). These studies included participants with FM with similar intensity of pain as our sample (Menten et al.: 5.45/10; Lim et al.: 57.2/100; Palmer et al.: 5.68/10) and as studies with large samples of participants with FM (Staud et al., 2006: 277 participants with FM with a mean pain of 4.6/10; Steiner et al., 2017: 216 FM with a mean pain intensity of 5.95/10; Serpas et al., 2022: 147 FM, mean pain intensity of 5.96/10). This result is consistent with the influence of pain intensity on unisensory tactile perception.

For both groups, the tactile percept was as precise as the visual percept, confirming that the reliability level was comparable between the two modalities. This is particularly important for multisensory integration studies since an apparent asymmetry of the reliability between the modalities could lead to the exclusive use of the most reliable modality and prevent multisensory integration [a phenomenon called sensory capture (Rohde et al., 2016)].

Results show that the accuracy of the tactile percept was better than the accuracy of the visual percept for both groups. A hypothesis for this result could be the transmission speed difference between tactile and visual information. Tactile information conduction is considered faster than visual information, meaning that a visual stimulation must be triggered slightly before a tactile stimulation for participants to perceive them as simultaneous (Spence et al., 2003; Poliakoff et al., 2006; Keetels and Vroomen, 2011; Godlove et al., 2014). Therefore, tactile information may be more accurate than visual information in a temporal perception task such as the TOJ (Crevecoeur et al., 2016).

In the multisensory condition, the BP was significantly positive for the CTRL group but not for the FM group. Despite this result, we observed no statistically significant difference in the BP between the groups. Together, these data suggest that the benefit of precision in participants with FM may have been masked by intragroup variability.

The analysis of the BA shows that the accuracy was not better in the multisensory condition compared to the unisensory conditions (i.e., no real benefit in accuracy) for both groups. A trend toward a positive BA was nonetheless observed for the FM group; again, intragroup variability might have prevented the BA to reach significance in this group. We expected to detect a benefit in accuracy for the CTRL group to accompany the benefit in precision observed. However, the inspection of the performance in the unisensory conditions shows that the visual and tactile percepts’ accuracy was already excellent for both groups, leaving little room for improvement. It is also important to note that, contrary to the benefit in precision, a benefit in accuracy is not necessary to show multisensory integration. According to Rohde and colleagues, “noise reduction is the most important hallmark of optimal integration” (Rohde et al., 2016). Thus, an improvement in accuracy was considered as a secondary measure, nonessential to determine multisensory integration. A difference of M in the multisensory condition compared to the unisensory conditions can also be used as a complementary measure of sensory weighting: the closer the multisensory M is to one of the unisensory M, the larger the weight of the sensory information is Bultitude and Petrini (2021). In a recent study, participants pointed to visual targets using either visual, somatosensory, or visual and somatosensory information (Bultitude and Petrini, 2021). Healthy participants showed no difference between the M of the multisensory percept compared to the M of the unisensory percepts, which was interpreted as an absence of difference in the weighting of visual and somatosensory information (Bultitude and Petrini, 2021). This also provides evidence for an absence of benefit in accuracy, even though an improvement of precision was observed (Bultitude and Petrini, 2021).

When the precision observed in the multisensory condition was compared to the precision predicted by the MLE model, a positive correlation was found for both groups, which suggests an optimal integration of visual and tactile information. Here, optimality refers to an appropriate integration of the visuotactile stimuli, as defined by the MLE model. Given a certain precision of the unisensory percepts, the model predicts the precision of the multisensory percept. The correlation shows that that observed visuotactile precision follows that predicted by the model.

To summarize, the results in the multisensory condition point to an optimal visuotactile integration for perception in individuals with FM and pain-free controls. This integration was reflected by an improvement in the precision (non-significant for participants with FM) and a positive correlation between the observed visuotactile precision and the precision predicted by the MLE model. This conclusion is supported by studies testing multisensory integration in individuals with chronic pain by generating body illusions. For example, in the rubber hand illusion, participants feel a tactile stroke on their hidden hand while seeing a rubber hand being stroked simultaneously (Botvinick and Cohen, 1998; Tsakiris and Haggard, 2005). The integration of the tactile information felt on the hand with the visual information of the stroked rubber hand leads most pain-free participants to feel that the rubber hand belongs to them (Botvinick and Cohen, 1998; Lloyd, 2007). This illusion type depends on a functioning visuotactile integration (Limanowski, 2021). Several studies found that participants with chronic pain also experience these illusions, suggesting that multisensory integration is not completely disrupted (McCabe et al., 2007; Reinersmann et al., 2013; Don et al., 2016; Brun et al., 2018; Martínez et al., 2018). The intensity of the illusion, however, varied across these studies, leading some authors to suspect a different weighting of the sensory modalities.

Our study showed no such difference since the weighting of visual and tactile information was comparable between the FM and the CTRL groups. Considering the absence of alteration of unisensory integration of tactile information in the FM group, this result is consistent with optimal multisensory integration. Tactile information was not less precise or less accurate in participants with FM and, therefore, was not weighted differently compared to CTRL participants. In fact, for participants with FM, it seemed advantageous to use tactile information more, since a bigger proportion of participants with FM who showed a dominance for tactile information displayed a positive BP. This perceptual benefit of weighting tactile information more was not observed for CTRL participants.

In the group with FM, tactile weight was correlated with general pain intensity, meaning that, in the multisensory condition, participants who experienced more intense pain relied less on tactile information to perceive the visuotactile stimuli. This points to an influence of pain on tactile weight within the FM group. This only partly agrees with our hypothesis since we observed no intergroup difference in sensory weighting. Congruent results were reported in a body-illusion paradigm (Brun et al., 2018). Participants with FM experienced a stronger illusion compared to pain-free controls, and this group difference was explained by the intensity of their pain (Brun et al., 2018). Thus, FM may not be automatically associated with smaller tactile weights, but the presence of moderate to severe pain might be decisive.

Mean sensory weights should nevertheless be interpreted carefully because the distribution of the weights revealed two clusters of participants, regardless of the group: a tactile-dominant cluster and a visual-dominant cluster. Although the participants integrated visual and tactile information (i.e., no sensory capture), a minority relied on both modalities symmetrically or quasi-symmetrically (i.e., weights in the 0.4 to 0.6 range). This weighting strategy was not explained by demographic characteristics or tactile sensitivity differences (as measured by the tactile detection threshold). A similar weight range was reported in a few other multisensory studies involving (Brun et al., 2018) a diverse clinical population (Brun et al., 2018) or pain-free participants (Maiworm and Röder, 2011). In a visuoauditory task, Maiworm and Röder (2011) reported that about half of the participants integrated information optimally, whereas the other half did not. Thus, they divided participants into an Optimal group, attributing a smaller weight to the auditory signal, and a Suboptimal group, relying more on the auditory signal (Maiworm and Röder, 2011). The authors conclude on an individual’s “hard-coded preference for certain modalities under particular circumstances” (Maiworm and Röder, 2011). This top-down predisposition [sometimes referred to as a prior (Limanowski, 2021)] could originate from an attentional bias towards one modality, which would lead to faster processing times and higher weights (Spence et al., 2001; Werkhoven et al., 2009; Maiworm and Röder, 2011; Limanowski, 2021). Another explanation for the bimodal distribution observed in the FM and CTRL groups could lie in a difference in sensory precision between the clusters. We observed no link between the weighting strategy and tactile sensitivity; thus, variations in visual abilities might have played a role. All participants had normal or corrected-to-normal vision and scored at least 60% in the V condition, but finer inter-subject disparities could have influenced sensory weighting (Ernst and Banks, 2002; Helbig and Ernst, 2007; Fetsch et al., 2012).

A technical limitation of the present study was the inability to customize tactile stimuli intensity between the two sites. The chosen stimulation intensity was readily detectable, though not uncomfortable or painful for all participants, finer sensation differences were still present between Site 1 and Site 2 for some participants. This observation did not lead to differences in the preferred site between the groups (in the T condition, the frequency of answers “Site 1” was 0.52 for both groups), which limits its potential bias on the TOJ results. Another limitation is the variability of the group with FM. Though variability seems intrinsic to FM (Wilson et al., 2009; Rehm et al., 2010; Yim et al., 2017; Bartley et al., 2018) our group’s relatively small sample size may have exacerbated this variability. Our sample size was based on pragmatic considerations, given the prevalence of this population and the associated symptoms (severe fatigue, trouble concentrating over long periods, difficulty moving or walking, etc.). Moreover, a study with a sample size similar as ours (15 participants with FM and 15 pain-free controls) showed statistically significant intergroup differences in the temporal discrimination of somatosensory stimuli (Gunendi et al., 2019). Finally, excluding participants whose unisensory performance was too high or too low was necessary to ensure proper multisensory integration [the elimination rate can reach over half of the participants in some multisensory integration studies (Kostaki and Vatakis, 2018)].

In conclusion, we found no difference in multisensory integration between participants with FM and pain-free participants. Both groups integrated visuotactile stimuli optimally, resulting in a more precise visuotactile percept than the visual or tactile percepts (although intragroup variability might have prevented the improvement in precision to reach significance in the group with FM). The mean weighting of the visual and the tactile information in the visuotactile percept was not different between the groups either, which is consistent with the fact that unisensory perception of the visual and tactile percepts was not different between the groups. In the group with FM, the weight attributed to tactile information was correlated to pain intensity, with a smaller weight given by participants with more intense pain. Overall, these findings point to an absence of alterations in multisensory integration in individuals with FM. However, future research should further investigate the role intense pain may play in weighting tactile information.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by CIUSSS de la Capitale-Nationale, Quebec City, Canada. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

TA: Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft. MS: Funding acquisition, Supervision, Writing – review & editing. CM: Conceptualization, Funding acquisition, Supervision, Writing – review & editing.