Auditory stimuli extend the temporal window of visual integration by modulating alpha-band oscillations

Reviewer #1 (Public review):

Summary:

Is peristimulus alpha (8-14 Hz) frequency and/or phase involved in shaping the length of visual and audiovisual temporal binding windows, as posited by the discrete sampling hypothesis? If so, to what extent and perceptual scenario are they functionally relevant? The authors addressed such questions by collecting EEG data during the completion of the widely-known 2-flash fusion paradigm, administered both in a standard (i.e., visual only, F2) and audiovisual (i.e., 2 flashes and 1 beep, F2B1) fashion. Instantaneous frequency estimation performed over parieto-occipital sensors revealed slower alpha rhythms right after stimulus onset in the F2B1 condition, as compared to the F2, a pattern found to correlate with the difference between modality-specific ISIs (F2B1-F2). Of note, peristimulus alpha frequency differed also between 1 vs 2 flashes reports, although in the visual modality only (i.e., faster alpha oscillations in 2 flash percept vs 1 flash). This pattern of results was reinvigorated in a causal manner via occipital tACS, which was capable of, respectively, narrowing down vs enlarging the temporal binding window of individuals undergoing 13 Hz vs 8 Hz stimulation in the F2 modality alone. To elucidate what the oscillatory signatures of crossmodal integration might be, the authors further focused on the phase of posterior alpha rhythms. Accordingly, the Phase Opposition Sum proved to significantly differ between modalities (F2B1 vs F2) during the prestimulus time window, suggesting that audiovisual signals undergo finer processing based on the ongoing phase of occipital alpha oscillations, rather than the speed at which these rhythms cycle. As a last bit of information, a computational model factoring in the electrophysiological assumptions of both the discrete sampling hypothesis and auditory-induced phase-resetting was devised. Analyses run on such synthetic data were partially able to reproduce the patterns witnessed in the empirical dataset. While faster frequency rates broadly provide a higher probability to detect 2 flashes instead of 1, the occurrence of a concurrent auditory signal in cross-modal trials should cause a transient elongation (i.e. slower frequency rate) of the ongoing alpha cycle due to phase-reset dynamics (as revealed via inter-trial phase clustering), prompting larger ISIs during F2B1 trials. Conversely, the model provides that alpha oscillatory phase might predict how well an observer dissociates sensory information from noise (i.e., perceptual clarity), with the second flash clearly perceived as such as long as it falls within specific phase windows along the alpha cycle.

Strengths:

The authors leveraged complementary approaches (EEG, tACS, and computational modelling), the results thereof not only integrate, but depict an overarching mechanistic scenario elegantly framing phase-resetting dynamics into the broader theoretical architecture posited by the discrete sampling hypothesis. Analyses on brain oscillations (either via frequency sliding and phase opposition sum) mostly appear to be methodologically sound, and very-well supported by tACS results. Under this perspective, the modelling approach serves as a convenient tool to reconcile and shed more light on the pieces of evidence gathered on empirical data, returning an appealing account on how cross-modal stimuli interplay with ongoing alpha rhythms and differentially affect multisensory processing in humans.

Weaknesses:

Some information relative to the task and the analyses is missing. For instance, it is not entirely clear from the text what the number of flashes actually displayed in explicit short trials is (1 or 2?). We believe it is always two, but it should be explicitly stated.

Moreover, the sample size might be an issue. As highlighted by a recent meta-analysis on the matter (Samaha & Romei, 2024), an underpowered sample size may very well drive null-findings relative to tACS data in F2B1 trials, in interplay with broad and un-individualized frequency targets.

Some criticality arises regarding the actual "bistability" of bistable trials, as the statistics relative to the main task (i.e., the actual means and SEMs are missing) broadly point toward a higher proclivity to report 2 instead of 1 flash in both F2B1 and F2 trials. This makes sense to some extent, given that 2 flashes have always been displayed (at least in bistable trials), yet tells about something botched during the pretest titration procedure.

Coming to the analyses on brain waves, one main concern relates to the phase-reset-induced slow-down of posterior alpha rhythms being of true oscillatory nature, rather than a mere evoked response (i.e., not sustained over time). Another question calling for some further scrutiny regards the overlooked pattern linking the temporal extent of the IAF differences between F2 and F2B1 trials with the ISIs across experimental conditions (explicit short, bistable, and explicit long). That is, the wider the ISI, the longer the temporal extent of the IAF difference between sensory modalities. Although neglected by the authors, such a trend speaks in favour of a rather nuanced scenario stemming from not only auditory-induced phase-reset alpha cycle elongation, but also some non-linear and perhaps super-additive contribution of flash-induced phase-resetting. This consideration introduces some of the issues about the computational simulation, which was modelled around the assumption of phase-resetting being triggered by acoustic stimuli alone. Given how appealing the model already is, I wonder whether the authors might refine the model accordingly and integrate the phase-resetting impact of visual stimuli upon synthetic alpha rhythms. Relatedly, I would also suggest the authors to throw in a few more simulations to explore the parameter space and assay, to which quantitative extent the model still holds (e.g. allowing alpha frequency to randomly change within a range between 8 and 13 Hz, or pivoting the phase delay around 10 or 50 ms). As a last remark, I would avoid, or at least tone down, concluding that the results hereby presented might reconcile and/or explain the null effects in Buergers & Noppeney, 2022; as the relationship between IAFs and audiovisual abilities still holds when examining other cross-modal paradigms such as the Sound-Induced Flash-Illusion (Noguchi, 2022), and the aforementioned patterns might be due to other factors, such as a too small sample size (Samaha & Romei, 2024).

Reviewer #2 (Public review):

Summary:

The authors used a visual flash discrimination task in which two flashes are presented one after another with different inter-stimulus intervals. Participants either perceive one flash or two flashes. The authors show that the simultaneous presence of an auditory input extends the temporal window of integration, meaning that two flashes presented shortly after one another are more likely to be perceived as a single flash. Auditory inputs are accompanied by a reduction in alpha frequency over visual areas. Prestimulus alpha frequency predicts perceptual outcomes in the absence of auditory stimuli, whereas prestimulus alpha phase becomes the dominant predictor when auditory input is present. A computational model based on phase-resetting theory supports these findings. Additionally, a transcranial stimulation experiment confirms the causal role of alpha frequency in unimodal visual perception but not in cross-modal contexts.

Strengths:

The authors elegantly combined several approaches-from behavior to computational modeling and EEG-to provide a comprehensive overview of the mechanisms involved in visual integration in the presence or absence of auditory input. The methods used are state-of-the-art, and the authors attempted to address possible pitfalls.

Weaknesses:

The use of Bayesian statistics could further strengthen the paper, especially given that a few p-values are close to the significance threshold (lines 162 & 258), but they are interpreted differently in different cases (absence of effect vs. trend).

Overall, these results provide new insights into the role of alpha oscillations in visual processing and offer an interesting perspective on the current debate regarding the roles of alpha phase and frequency in visual perception. More generally, they contribute to our understanding of the neural dynamics of multisensory integration.

Reviewer #3 (Public review):

Summary:

The authors investigated the impact of an auditory stimulus on visual integration at the behavioral, electrophysiological, and mechanistic levels. Although the role of alpha brain oscillations on visual perception has been widely studied, how the brain dynamics in the visual cortices are influenced by a cross-modal stimulus remains ill-defined. The authors demonstrated that auditory stimulation systematically induced a drop in visual alpha frequency, increasing the time window for audio-visual integration, while in the unimodal condition, visual integration was modulated by small variations within the alpha frequency range. In addition, they only found a role of the phase of alpha brain oscillations on visual perception in the cross-modal condition. Based on the perceptual cycles' theory framework, the authors developed a model allowing them to describe their results according to a phase resetting induced by the auditory stimulation. These results showed that the influence of well-known brain dynamics on one modality can be disrupted by another modality. They provided insights into the importance of investigating cross-modal brain dynamics, and an interesting model that extends the perceptual cycle framework.

Strengths:

The results are supported by a combination of various, established experimental and analysis approaches (e.g., two-flash fusion task, psychometric curves, phase opposition), ensuring strong methodological bases and allowing direct comparisons with related findings in the literature.

The model the authors proposed is an extension and an improvement of the perceptual cycle's framework. Interestingly, this model could then be tested in other experimental approaches.

Weaknesses:

There is an increasing number of studies in cognitive neuroscience showing the importance of considering inter-individual variability. The individual alpha frequency (IAF) varied from 8 to 13 Hz with a huge variability across participants, and studies have shown that the IAF influenced visual perception. Investigating inter-individual variations of the IAF in the reported results would be of great interest, especially for the model.

Although the use of non-invasive brain stimulation to infer causality is a method of great interest, the use of tACS in the presented work is not optimal. Instead of inducing alpha brain oscillations in visual cortices, the use of tACS to activate the auditory cortex instead of the actual auditory stimulation would have presented more interest.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

Is peristimulus alpha (8-14 Hz) frequency and/or phase involved in shaping the length of visual and audiovisual temporal binding windows, as posited by the discrete sampling hypothesis? If so, to what extent and perceptual scenario are they functionally relevant? The authors addressed such questions by collecting EEG data during the completion of the widely-known 2-flash fusion paradigm, administered both in a standard (i.e., visual only, F2) and audiovisual (i.e., 2 flashes and 1 beep, F2B1) fashion. Instantaneous frequency estimation performed over parieto-occipital sensors revealed slower alpha rhythms right after stimulus onset in the F2B1 condition, as compared to the F2, a pattern found to correlate with the difference between modality-specific ISIs (F2B1-F2). Of note, peristimulus alpha frequency differed also between 1 vs 2 flashes reports, although in the visual modality only (i.e., faster alpha oscillations in 2 flash percept vs 1 flash). This pattern of results was reinvigorated in a causal manner via occipital tACS, which was capable of, respectively, narrowing down vs enlarging the temporal binding window of individuals undergoing 13 Hz vs 8 Hz stimulation in the F2 modality alone. To elucidate what the oscillatory signatures of crossmodal integration might be, the authors further focused on the phase of posterior alpha rhythms. Accordingly, the Phase Opposition Sum proved to significantly differ between modalities (F2B1 vs F2) during the prestimulus time window, suggesting that audiovisual signals undergo finer processing based on the ongoing phase of occipital alpha oscillations, rather than the speed at which these rhythms cycle. As a last bit of information, a computational model factoring in the electrophysiological assumptions of both the discrete sampling hypothesis and auditory-induced phase-resetting was devised. Analyses run on such synthetic data were partially able to reproduce the patterns witnessed in the empirical dataset. While faster frequency rates broadly provide a higher probability to detect 2 flashes instead of 1, the occurrence of a concurrent auditory signal in cross-modal trials should cause a transient elongation (i.e. slower frequency rate) of the ongoing alpha cycle due to phase-reset dynamics (as revealed via inter-trial phase clustering), prompting larger ISIs during F2B1 trials. Conversely, the model provides that alpha oscillatory phase might predict how well an observer dissociates sensory information from noise (i.e., perceptual clarity), with the second flash clearly perceived as such as long as it falls within specific phase windows along the alpha cycle.

Strengths:

The authors leveraged complementary approaches (EEG, tACS, and computational modelling), the results thereof not only integrate, but depict an overarching mechanistic scenario elegantly framing phase-resetting dynamics into the broader theoretical architecture posited by the discrete sampling hypothesis. Analyses on brain oscillations (either via frequency sliding and phase opposition sum) mostly appear to be methodologically sound, and very-well supported by tACS results. Under this perspective, the modelling approach serves as a convenient tool to reconcile and shed more light on the pieces of evidence gathered on empirical data, returning an appealing account on how cross-modal stimuli interplay with ongoing alpha rhythms and differentially affect multisensory processing in humans.

Weaknesses:

Some information relative to the task and the analyses is missing. For instance, it is not entirely clear from the text what the number of flashes actually displayed in explicit short trials is (1 or 2?). We believe it is always two, but it should be explicitly stated.

We thank the reviewer for highlighting this important point. In our study, all explicit trials consistently presented two flashes. We will clearly state this detail in the Methods section to avoid any further confusion.

Moreover, the sample size might be an issue. As highlighted by a recent meta-analysis on the matter (Samaha & Romei, 2024), an underpowered sample size may very well drive null-findings relative to tACS data in F2B1 trials, in interplay with broad and un-individualized frequency targets.

We thank the reviewer for raising this point. First, we would like to clarify that our results do not suggest that the frequency effect is absent in the F2B1 condition; rather, it is relatively attenuated compared to the F2 condition. If the sample size were the primary issue, we would expect to observe a null effect in both conditions. Instead, the stronger frequency modulation in F2 confirms that the sound-induced modulation is present, albeit reduced in the audiovisual context. In our revised manuscript, we will explicitly note that our claim is not that there is no frequency effect in F2B1 but that the effect is weaker relative to F2, and we will also acknowledge the potential limitations associated with sample size and the lack of individualized frequency targeting.

Some criticality arises regarding the actual "bistability" of bistable trials, as the statistics relative to the main task (i.e., the actual means and SEMs are missing) broadly point toward a higher proclivity to report 2 instead of 1 flash in both F2B1 and F2 trials. This makes sense to some extent, given that 2 flashes have always been displayed (at least in bistable trials), yet tells about something botched during the pretest titration procedure.

We thank the reviewer for pointing out the potential bias toward reporting “two flashes” in the bistable trials. Because our experimental design involves presenting two flashes in both explicit and bistable trials, a slight tendency to report two flashes may naturally arise, especially at threshold levels determined during pretesting. We believe, however, that this bias does not undermine our primary findings. Our psychophysical procedure is designed to align the inter-stimulus interval with each participant’s fusion threshold, aiming for a near 50/50 split between “one-flash” and “two-flash” reports. However, given that two flashes are always presented, participants may be predisposed to report two flashes when uncertain. This reflects a plausible perceptual bias inherent in the bistable design, rather than a systematic flaw. Importantly, this tendency appears at comparable levels in both the F2 and F2B1 conditions, indicating that it does not selectively affect any particular condition. In the revised manuscript, we will include additional descriptive statistics, such as means and standard deviations, to demonstrate that the observed bias remains within an acceptable range and does not compromise our core conclusions regarding the modulatory effect of auditory input on visual integration.

Coming to the analyses on brain waves, one main concern relates to the phase-reset-induced slow-down of posterior alpha rhythms being of true oscillatory nature, rather than a mere evoked response (i.e., not sustained over time).

We appreciate the reviewer’s concern regarding this issue. First, the sustained decrease in posterior alpha frequency observed in our study—persisting for approximately 280 ms—substantially exceeds the typical duration of an auditory evoked potential (generally 50–200 ms) (Näätänen and Picton, 1987). This extended period of modulation suggests that it is not merely a transient evoked response.

Second, our analysis of alpha power further supports this interpretation. A purely evoked response is usually accompanied by a corresponding increase in signal power; however, our results show no such power increase when comparing the F2B1 condition with the F2 condition.

Moreover, the observed increase in alpha phase resetting—as measured by inter-trial phase coherence (ITC)—does not significantly correlate with changes in alpha power. This dissociation further indicates that the auditory-induced effects are unlikely to be driven solely by evoked potentials, but are more consistent with a reorganization of the intrinsic neural oscillatory activity.

Together, these lines of evidence strongly support the view that the auditory-induced decrease in alpha frequency reflects true changes in ongoing oscillatory dynamics, rather than being merely a transient evoked response.

Another question calling for some further scrutiny regards the overlooked pattern linking the temporal extent of the IAF differences between F2 and F2B1 trials with the ISIs across experimental conditions (explicit short, bistable, and explicit long). That is, the wider the ISI, the longer the temporal extent of the IAF difference between sensory modalities. Although neglected by the authors, such a trend speaks in favour of a rather nuanced scenario stemming from not only auditory-induced phase-reset alpha cycle elongation, but also some non-linear and perhaps super-additive contribution of flash-induced phase-resetting. This consideration introduces some of the issues about the computational simulation, which was modelled around the assumption of phase-resetting being triggered by acoustic stimuli alone. Given how appealing the model already is, I wonder whether the authors might refine the model accordingly and integrate the phase-resetting impact of visual stimuli upon synthetic alpha rhythms.

We appreciate the reviewer’s insightful comment regarding the potential influence of flash-induced phase resetting on the temporal extent of the IAF differences. We acknowledge that the observation—that wider ISIs are associated with a longer period of IAF differences—hints at a non-linear or even super-additive interaction between auditory- and flash-induced phase resetting mechanisms.

However, the primary focus of our current study is on how auditory stimuli affect alpha oscillatory dynamics. Our experimental design and computational model were specifically optimized to capture auditory-induced phase resetting. Incorporating the additional influence of flash-induced effects would require a significantly more refined experimental framework and a more complex modeling approach. This added complexity could obscure the interpretation of our main findings, which are centered on auditory influences.

In the revised manuscript, we will address this intriguing possibility in the Discussion section. We will acknowledge that while the data hint at a potential visual contribution, our present model deliberately isolates auditory-induced phase resetting to maintain clarity. We also propose that future research, with more precise experimental designs and enhanced modeling techniques, is necessary to fully disentangle and capture the interplay between auditory and flash-induced phase resetting mechanisms.

Relatedly, I would also suggest the authors to throw in a few more simulations to explore the parameter space and assay, to which quantitative extent the model still holds (e.g. allowing alpha frequency to randomly change within a range between 8 and 13 Hz, or pivoting the phase delay around 10 or 50 ms).

We appreciate the reviewer’s suggestion to further explore our model’s parameter space. In response, we will conduct additional simulations that incorporate variability in alpha frequency—sampling randomly between 8 and 13 Hz—and examine alternative phase delays (e.g., around 10 and 50 ms). By systematically adjusting these parameters, we can more thoroughly evaluate the model’s robustness and delineate its boundaries under a broader range of neurophysiological conditions. We will present these results in the revised manuscript and discuss how they inform our understanding of alpha-driven visual integration in cross-modal contexts.

As a last remark, I would avoid, or at least tone down, concluding that the results hereby presented might reconcile and/or explain the null effects in Buergers & Noppeney, 2022; as the relationship between IAFs and audiovisual abilities still holds when examining other cross-modal paradigms such as the Sound-Induced Flash-Illusion (Noguchi, 2022), and the aforementioned patterns might be due to other factors, such as a too small sample size (Samaha & Romei, 2024).

We appreciate the reviewer’s suggestion and will revise our claims accordingly. In the revised manuscript, we will clarify that while our study demonstrates a mechanism by which alpha oscillations influence audiovisual integration in certain paradigms, this does not mean that our findings fully reconcile all conflicting results in the literature. We will emphasize that our mechanism may help explain why alpha frequency plays a critical role in some experimental settings, but that factors such as sample size, task parameters, and experimental design differences likely contribute to the divergent results observed across studies. Accordingly, we acknowledge that further research with larger samples and more refined methodologies is necessary to fully reconcile these discrepancies. This more cautious interpretation will be clearly discussed in the revised manuscript.

Reviewer #2 (Public review):

Summary:

The authors used a visual flash discrimination task in which two flashes are presented one after another with different inter-stimulus intervals. Participants either perceive one flash or two flashes. The authors show that the simultaneous presence of an auditory input extends the temporal window of integration, meaning that two flashes presented shortly after one another are more likely to be perceived as a single flash. Auditory inputs are accompanied by a reduction in alpha frequency over visual areas. Prestimulus alpha frequency predicts perceptual outcomes in the absence of auditory stimuli, whereas prestimulus alpha phase becomes the dominant predictor when auditory input is present. A computational model based on phase-resetting theory supports these findings. Additionally, a transcranial stimulation experiment confirms the causal role of alpha frequency in unimodal visual perception but not in cross-modal contexts.

Strengths:

The authors elegantly combined several approaches-from behavior to computational modeling and EEG-to provide a comprehensive overview of the mechanisms involved in visual integration in the presence or absence of auditory input. The methods used are state-of-the-art, and the authors attempted to address possible pitfalls.

Weaknesses:

The use of Bayesian statistics could further strengthen the paper, especially given that a few p-values are close to the significance threshold (lines 162 & 258), but they are interpreted differently in different cases (absence of effect vs. trend).

We appreciate the reviewer’s suggestion regarding the use of Bayesian statistics. We agree that a Bayesian framework can offer valuable complementary insights to our analysis by helping to distinguish whether a marginal p-value represents a trend or truly indicates the absence of an effect. To enhance the robustness of our conclusions, we will incorporate supplemental Bayesian analyses in the revised manuscript.

Overall, these results provide new insights into the role of alpha oscillations in visual processing and offer an interesting perspective on the current debate regarding the roles of alpha phase and frequency in visual perception. More generally, they contribute to our understanding of the neural dynamics of multisensory integration.

Reviewer #3 (Public review):

Summary:

The authors investigated the impact of an auditory stimulus on visual integration at the behavioral, electrophysiological, and mechanistic levels. Although the role of alpha brain oscillations on visual perception has been widely studied, how the brain dynamics in the visual cortices are influenced by a cross-modal stimulus remains ill-defined. The authors demonstrated that auditory stimulation systematically induced a drop in visual alpha frequency, increasing the time window for audio-visual integration, while in the unimodal condition, visual integration was modulated by small variations within the alpha frequency range. In addition, they only found a role of the phase of alpha brain oscillations on visual perception in the cross-modal condition. Based on the perceptual cycles' theory framework, the authors developed a model allowing them to describe their results according to a phase resetting induced by the auditory stimulation. These results showed that the influence of well-known brain dynamics on one modality can be disrupted by another modality. They provided insights into the importance of investigating cross-modal brain dynamics, and an interesting model that extends the perceptual cycle framework.

Strengths:

The results are supported by a combination of various, established experimental and analysis approaches (e.g., two-flash fusion task, psychometric curves, phase opposition), ensuring strong methodological bases and allowing direct comparisons with related findings in the literature.

The model the authors proposed is an extension and an improvement of the perceptual cycle's framework. Interestingly, this model could then be tested in other experimental approaches.

Weaknesses:

There is an increasing number of studies in cognitive neuroscience showing the importance of considering inter-individual variability. The individual alpha frequency (IAF) varied from 8 to 13 Hz with a huge variability across participants, and studies have shown that the IAF influenced visual perception. Investigating inter-individual variations of the IAF in the reported results would be of great interest, especially for the model.

We appreciate the reviewer’s valuable feedback regarding the importance of inter-individual variability in alpha frequency. In our current study, we have already addressed participant-level variability in our neural data by performing inter-subject correlation analyses, investigating whether individual reductions in alpha frequency correlate with broader temporal integration windows at the behavioral level.

Moreover, our computational model incorporates physiologically realistic distributions for key parameters, including frequency and amplitude, which captures some degree of individual variability. Nevertheless, we acknowledge that a more targeted examination of how different IAF values specifically affect the model’s predictions would be highly valuable. In response, we will expand our simulations to systematically explore a range of IAF values and assess their impact on temporal integration windows and related measures of audiovisual processing. These additional analyses will help clarify the role of inter-individual variability in alpha frequency and further strengthen the mechanistic account offered by our model. We will detail these enhancements and discuss their implications in the revised manuscript.

Although the use of non-invasive brain stimulation to infer causality is a method of great interest, the use of tACS in the presented work is not optimal. Instead of inducing alpha brain oscillations in visual cortices, the use of tACS to activate the auditory cortex instead of the actual auditory stimulation would have presented more interest.

We appreciate the reviewer’s suggestion and acknowledge that non-invasive brain stimulation offers promising avenues for inferring causality. In our study, our primary hypothesis focused on the role of occipital alpha oscillations in defining the temporal window for visual integration, and accordingly we targeted visual cortex in our tACS protocol.

We recognize that stimulating the auditory cortex could provide additional insights into auditory contributions to phase resetting. However, accurately targeting the auditory cortex with tACS presents technical challenges. The auditory cortex is located deeper within the temporal lobe, and factors such as variable skull thickness and complex current spread make it difficult to reliably modulate its neural activity compared to the more superficial visual areas. Indeed, recent studies have demonstrated that tACS-induced electric fields in the temporal regions tend to be weaker and less focal—for example, Huang et al. (2017) and Opitz et al. (2016) highlight the limitations in achieving robust stimulation of deeper or anatomically complex brain regions using conventional tACS approaches.

Given these considerations, while we agree that future investigations could benefit from exploring auditory cortex stimulation—either as an alternative or as a complementary approach—the present study remains focused on visual alpha modulation, where our protocol is well validated and yields reliable results. In the revised manuscript, we will clearly discuss these issues and acknowledge the potential, yet technically challenging, possibility of stimulating the auditory cortex in future work to further disentangle the contributions of auditory and visual inputs to cross-modal integration.