The target detection task is a standard paradigm to study the neural correlates of perceptual awareness. In the near-threshold detection task, participants are presented with stimuli at intensities close to their individual detection thresholds, resulting in detection rates of ~50%. Neural responses underlying detected and undetected targets can then be contrasted at identical physical stimulation parameters to identify the neural correlates of conscious access.

In the somatosensory domain, research on this task has identified a range of areas that correlate with target detection. These include the thalamus, primary (SI) and secondary (SII) somatosensory cortices, motor areas, the anterior insular cortex (AIC), anterior cingulate cortex (ACC), as well as posterior parietal and prefrontal regions (Auksztulewicz et al., 2012; Bastuji et al., 2016; Allen et al., 2016; Bornhövd et al., 2002; Büchel et al., 2002; de Lafuente and Romo, 2005; de Lafuente and Romo, 2006; Frey et al., 2016; Hirvonen and Palva, 2016; Jones et al., 2007; Moore et al., 2013). This diversity of findings parallels results from the visual and auditory modalities (e.g. Carmel et al., 2006; Eriksson et al., 2007; Rees et al., 2002) and underlies the idea that perceptual awareness emerges when local information is propagated from sensory cortices to higher order brain regions to elicit a reverberating network of broadcast activity (Baars, 1997; Dehaene et al., 2006).

While a complex network activation may seem a suitable candidate to explain a complex phenomenon such as perceptual awareness, the functional specificity of the identified regions remains largely unknown. One problem that complicates the interpretation is that classical target detection tasks not only probe perceptual awareness but involve a range of correlated cognitive processes that may confound classical contrastive analyses (Aru et al., 2012; de Graaf et al., 2012). In the context of the somatosensory near-threshold detection task, four aspects are particularly problematic: 1. perception of near-threshold stimuli is difficult, and resolution of associated uncertainty and introspective processes may differ between detected and undetected targets (de Lafuente and Romo, 2011). 2. Target detection is the explicit behavioural goal of the task and therefore, detected targets have higher behavioural relevance than undetected targets (Farooqui and Manly, 2018). 3. Target detection is directly mapped to overt reports that allow for assessment of participants’ trial-by-trial perception (Tsuchiya et al., 2015). 4. Overt reports are often communicated with button presses by one hand while stimulation occurs on the other hand, which may affect cortical excitability in homologue regions of the sensorimotor homunculus (Zagha et al., 2013). All these variables potentially contribute to the commonly observed network activation and dedicated experimental paradigms are warranted to scrutinise its functional specificity for target detection and accordingly, somatosensory awareness.

Here, we used functional magnetic resonance imaging (fMRI) on a novel somatosensory detection task that explicitly varies stimulus uncertainty and controls for behavioural relevance, overt reports, and motor responses. To distinguish BOLD responses reflecting target detection from those indicating concomitant processes, we fit simple behavioural models to our fMRI data, capturing trial-wise physical stimulus intensity, target detection, detection probability, expected uncertainty, and overt reports, respectively. We dissociate corresponding representations in the brain by means of Bayesian Model Selection (BMS, Stephan, Penny, Daunizeau, Moran, & Friston, 2009), which determines which model best explains the data in every voxel of the brain based on model evidence maps. Building on insights from the visual modality (Farooqui and Manly, 2018; Frässle et al., 2014; Koch et al., 2016), we hypothesise that BOLD responses associated with target detection are restricted to somatosensory regions, whereas activity in the frontoparietal network reflects cognitive processes that follow from task requirements.

To scrutinise the neural processes underlying somatosensory target detection in humans, we employed an experimental paradigm that explicitly dissociates target detection from stimulus uncertainty, behavioural relevance, overt reports, and motor responses. Using Bayesian Model Selection on the acquired fMRI data, we observe a transformation from physical to perceptual representations as the target is propagated through the somatosensory hierarchy. This transformation primarily occurred in SI and SII, whereas expected uncertainty was represented in insular and cingulate regions and overt reports were processed in supplementary motor cortex. Our analysis reveals large overlap with the previously identified target detection network but assigns functional specificity to the involved regions.

SI is the first cortical region to receive somatosensory input from the contralateral body side. It is subdivided into four somatotopic maps, areas 3a/b, 1, and 2, that are organised along an anterior-posterior axis on the postcentral gyrus and increase in their hierarchical level (Delhaye et al., 2018; Felleman and Van Essen, 1991). Our analysis suggests that during somatosensory target detection, contralateral areas 3b, 1, and anterior parts of area 2process the physical stimulus intensity, whereas more posterior parts of area 2, extending into adjacent SPL, represent the probability to detect a target. Previous research has been ambiguous regarding the role of SI for target detection. Some studies found no difference in SI responses for detected and undetected targets (de Lafuente and Romo, 2005; Schubert et al., 2006; Wühle et al., 2010; Wühle et al., 2011) and are in line with a representation of physical stimulus properties. Others have reported stronger SI evoked magneto-/electroencephalography responses for detected compared to missed stimuli (Hirvonen and Palva, 2016; Jones et al., 2007; Palva et al., 2005). Interestingly, most studies reporting early detection-related effects in SI used the classical near-threshold detection task, whereas those that did not find effects in SI typically used alternative approaches (varying stimulus intensities: de Lafuente and Romo, 2005; de Lafuente and Romo, 2006; backward masking: Schubert et al., 2006; paired-pulse paradigms: Wühle et al., 2010, Wühle et al., 2011). As target detection at perceptual threshold changes with fluctuations in cortical excitability (Boly et al., 2007; Frey et al., 2016; Moore et al., 2013; Schubert et al., 2009; Weisz et al., 2014), this dissociation raises the question, whether differentiable SI responses for detected and undetected near-threshold stimuli are in fact markers of stimulus awareness or rather the result of background processes that may facilitate or attenuate target detection depending on pre-stimulus brain states (Schubert et al., 2006). In our data, we do observe a transformation of stimulus representations in SI as the stimulus is propagated up the local hierarchy. However, although the detection probability model captures some perceptual properties of the stimulus and may constitute a first step towards perceptual readout, it does not explicitly differentiate between detected and undetected trials and importantly, predicts the same level of activation for threshold stimuli, regardless of whether they were perceived or not. Therefore, our results do not support detection-sensitivity in SI.

Responses in bilateral SII mirror the transformation from stimulus intensity to detection probability in SI and importantly, they also show an effect of binary target detection in adjacent regions. The distribution of these effects did not align with the cytoarchitectonic subdivisions of the parietal operculum, whose distinct functional roles remain largely elusive. The less pronounced somatotopy and large interindividual variability in SII (Eickhoff et al., 2006a; Sanchez Panchuelo et al., 2018) may have concealed the exact organisation of the observed effects and further studies are needed to ascertain the specific functional topology. Nonetheless, the relevance of region SII for somatosensory target detection is well documented in the literature. It exhibits stronger somatosensory evoked potentials (Auksztulewicz et al., 2012; Wühle et al., 2010; Wühle et al., 2011), increased spike rates (de Lafuente and Romo, 2006), stronger BOLD-signal changes (Moore et al., 2013), and increased network integration (Weisz et al., 2014) when stimuli are detected compared to when they are missed. Our findings are in line with intracortical recordings in monkeys (de Lafuente and Romo, 2006) showing that SII neurons are the first to be predictive of perceptual decisions. Previous studies have further shown that SII integrates information from different body parts (Goldin et al., 2018), indexes unexpected somatosensory stimuli (Chen et al., 2008) and stimulus omissions (Andersen and Lundqvist, 2019), and can adapt to task requirements (Romo et al., 2002). In combination with the observed shift from physical to perceptual representations in our study, evidence suggests that SII might be a central relay point at which behaviourally relevant stimuli are transformed into perceptual outcomes.

Expected uncertainty, here defined as susceptibility to perceptual fluctuations, best explained the data in SMG/ACC and AIC. These regions are commonly referred to as the salience network, which is involved in conflict monitoring, reorienting of attention, and introspective processes (Menon and Uddin, 2010) and, interestingly, was found to be sensitive to perceptual ambiguity (Lamichhane et al., 2016). Importantly, both ACC and insular cortex have previously been implied in somatosensory target detection (Bastuji et al., 2016; Boly et al., 2007; Büchel et al., 2002; Hirvonen and Palva, 2016; Moore et al., 2013). The insula in particular has been assigned a pivotal role in interoceptive inference (Seth, 2013) and awareness (Craig, 2009; Critchley et al., 2004), although recent research suggests that activity in AIC might reflect interoceptive sensitivity and learning rather than subjective experience per se (Canales-Johnson et al., 2015). In light of our results and given the fact that interoceptive signals, such as the heartbeat, are often faint signals, future research may benefit from combining measures of perceptual uncertainty (e.g. Garfinkel et al., 2015) with neuroimaging techniques to reconcile interoceptive processing with insular response properties. With regard to its role in somatosensory perception, the AIC has further been demonstrated to show strengthened backward connectivity with somatosensory cortex upon mismatch detection (Allen et al., 2016). Considering that in our study AIC particularly activated for hard-to-detect stimuli, these findings suggest that – although it does not generally reflect target detection – AIC may exert top-down modulation on somatosensory cortex to facilitate somatosensory processing.

The primary objective of the target detection task is to uncover neural processes underlying the emergence of perceptual awareness. Our results identify responses in somatosensory cortex as the best correlates of perceptual awareness and are in line with previous research emphasising the role of dedicated sensory regions for both visual (Boehler et al., 2008; Hurme et al., 2017; Pascual-Leone and Walsh, 2001; Ress et al., 2000) and somatosensory awareness (Auksztulewicz et al., 2012; Jones et al., 2007). However, this view is not unopposed and especially the role of prefrontal cortex (PFC) remains a topic of intense debate. While opponents argue that activity in PFC is an artefact of report requirements and behavioural relevance (Boly et al., 2017; Brascamp et al., 2015; Farooqui and Manly, 2018; Frässle et al., 2014; Koch et al., 2016; Pitts et al., 2014; Tsuchiya et al., 2015), proponents uphold that PFC has been demonstrated to code perceptual content even in passive paradigms (Panagiotaropoulos et al., 2012) and that null-findings in PFC may be the result of insufficient sensitivity in the methods (Odegaard et al., 2017). Ours is the first study to explicitly dissociate stimulus awareness from overt reports and behavioural relevance in the somatosensory domain and while we can only draw conclusions to the extent of the stimuli and methods used in this study, our findings do not support reliable representations of somatosensory awareness in PFC.

Importantly, we did not find strictly detection-related responses in posterior parietal cortex either, a result that conflicts with the ‘posterior hot zone’ theory of conscious perception (Koch et al., 2016) but is in line with an earlier finding suggesting that posterior parietal activation is related to goal completion but not stimulus awareness (Farooqui and Manly, 2018). Given that we controlled for goal completion effects in our study by equating behavioural relevance across detected and undetected targets, we may conclude that the correlates of target detection are largely restricted to dedicated sensory processing sites when controlled for common task requirements.

The aim of the current study was to experimentally dissociate target detection from four potentially confounding processes and hence, preclude these processes as root causes of detection-related neural activity. 1. Stimulus uncertainty: By varying stimulus intensities across trials, we implicitly tested for specific response profiles that occur in regions reflecting target detection. Specifically, these regions were expected to show low activity for low, subthreshold stimulus intensities and high activity for high, suprathreshold stimulus intensities. Since uncertainty, no matter if subjective or objective uncertainty measures are used, decreases for high-intensity stimuli, we can exclude uncertainty related processes as the driving force of activity in regions showing such a response profile. 2. Behavioural relevance: In order to perform the task, participants had to report a match or mismatch between their target perception and a visual matching cue. Since detected and undetected targets were found to occur equally often for match and mismatch reports, the two perceptual outcomes were equally relevant for the task and thus, differences in behavioural relevance are unlikely to have caused the observed detection-related activity. 3. Overt reports: As mentioned in point 2, overt reports were independent of target detection and are therefore unlikely to have caused detection-related effects. However, participants might still form a covert but explicit decision regarding the presence or absence of the target prior to making the match/mismatch decision, which could potentially confound the observed detection effects. Two reasons speak against this interpretation: First, the task imposed a time constraint that required participants to form their decisions fast. Therefore, the task could best be solved by directly comparing the two input modalities without engaging intermediate steps, a strategy, that was verbally confirmed by a majority of participants after the experiment. Second, the only areas that showed reliable and systematic covariation with target detection were regions in bilateral SII. An explicit internal decision regarding the absence or presence of a target stimulus would be expected to involve supramodal processes (because the same decision can be made on stimuli from different modalities) and therefore recruit higher order association cortices. Since this was not the case, we consider it unlikely that the detection effects observed in SII were induced by explicit perceptual decisions. Nonetheless, rapid low-level perceptual decision processes cannot be completely excluded and alternative approaches, such as no-report paradigms (Tsuchiya et al., 2015), may aid in disentangling their impact. 4. Motor responses: Participants gave their reports by making saccades to peripheral response cues. None of the observed detection effects overlap with regions commonly observed during saccadic eye movements (Ettinger et al., 2008; Kimmig et al., 2001) and effects in SI were localised to hand areas and far removed from eye representations in the somatosensory homunculus. Changes in cortical excitability due to motor responses are therefore unlikely to have affected the results. In conclusion, we consider the detection-related responses observed in bilateral SII to be largely free of the addressed experimental confounds.

A theoretical framework that accommodates awareness-related responses in sensory cortices is the recurrent processing hypothesis of conscious perception, which argues that perceptual awareness emerges when feedforward signals from early sensory cortex are consolidated by re-entrant feedback (Lamme, 2006). Such a mechanism might account for the early transformation from stimulus intensity to detection probability that we observed in SI and in fact, previous research on somatosensory target detection suggests that stimulus awareness is best captured in recurrent interactions between SI and SII (Auksztulewicz et al., 2012; Cauller and Kulics, 1991; Kwon et al., 2016; Yang et al., 2016). However, these earlier studies have used simpler versions of the detection task and studies employing methods with high temporal resolution in combination with rigorous experimental control are warranted to scrutinise the functional specificity of recurrent processing for somatosensory awareness.

Another open question concerns the relationship between target detection and subjective experience. Clearly, subjective experience of a detected target at near-threshold intensities is not identical to that of a stimulus at much higher intensities and thus, it is not well modelled by the detection model. Likewise, subjective experience can vary for stimuli at identical physical stimulus intensities and thus, it is not well modelled by the intensity or detection probability models. To address the subjective dimensions of our task, we would require trial-by-trial awareness ratings (and likewise, confidence ratings to evaluate metacognitive aspects of uncertainty). In the current investigation, we did not acquire such ratings because, although they would endow us with a better model of subjective experience, they would also reintroduce the requirement for explicit reports on the stimuli and result in higher relevance of detected compared to undetected targets. However, building on the results of the current study and the opportunities offered by the Bayesian analysis approach, incorporating such subjective reports and probing underlying neural responses may be a promising avenue for future research. In fact, it has been shown that subjective somatosensory awareness is parametrically encoded in connectivity patterns between somatosensory and higher order regions (Auksztulewicz and Blankenburg, 2013). This line of research could potentially consolidate experimental findings arguing for local versus global perspectives.

In summary, our findings dissect the functional contribution of different regions in the target detection network and advocate more complex experimental paradigms to dissociate neural responses reflecting conscious access from those supporting collateral functions. Note, however, that this is not to say that frontal and posterior parietal areas do not contribute to the full phenomenal experience of the task. We should keep in mind that although the somatosensory detection task allows for an operationalisation of basic stimulus awareness, detection of simple electrical targets is a simplistic reduction of what constitutes the richness of our everyday conscious perception. Certainly, cognitive processes such as attention and introspection considerably influence our experience and may even alter perceptual contents and likewise, detection of more complex stimuli, such as tactile motion or objects may require additional cognitive resources further downstream the somatosensory hierarchy. Having said that, when it comes to the perceptual integration of somatosensory stimuli as simple as electrical pulses to the wrist, our data suggest that activity in secondary somatosensory cortex is the best correlate of perceptual success.

In accordance with EU's General Data Protection Regulation we are unable to share raw fMRI data. However, single subject log evidence maps and group posterior probability maps have been uploaded to figshare (https://doi.org/10.6084/m9.figshare.7347167.v1) and posterior probability maps can be directly inspected on Neurovault (https://neurovault.org/collections/4496/). Analysis code is available on GitHub (https://github.com/PiaSchroeder/SomatosensoryTargetDetection_fMRI; copy archived at https://github.com/elifesciences-publications/SomatosensoryTargetDetection_fMRI).

1. 1. Schröder P

   (2018) Neurovault

   Somatosensory target detection.

   https://neurovault.org/collections/4496/
2. 1. Schröder P

   (2018) Figshare

   Neural basis of somatosensory target detection.

   https://doi.org/10.6084/m9.figshare.7347167.v1

1. 1. Allen M
   2. Fardo F
   3. Dietz MJ
   4. Hillebrandt H
   5. Friston KJ
   6. Rees G
   7. Roepstorff A

   (2016) Anterior insula coordinates hierarchical processing of tactile mismatch responses

   NeuroImage 127:34–43.

   https://doi.org/10.1016/j.neuroimage.2015.11.030

   • PubMed
   • Google Scholar
2. 1. Andersen LM
   2. Lundqvist D

   (2019) Somatosensory responses to nothing: an MEG study of expectations during omission of tactile stimulations

   NeuroImage 184:78–89.

   https://doi.org/10.1016/j.neuroimage.2018.09.014

   • Google Scholar
3. 1. Aru J
   2. Bachmann T
   3. Singer W
   4. Melloni L

   (2012) Distilling the neural correlates of consciousness

   Neuroscience & Biobehavioral Reviews 36:737–746.

   https://doi.org/10.1016/j.neubiorev.2011.12.003

   • PubMed
   • Google Scholar
4. 1. Auksztulewicz R
   2. Spitzer B
   3. Blankenburg F

   (2012) Recurrent neural processing and somatosensory awareness

   Journal of Neuroscience 32:799–805.

   https://doi.org/10.1523/JNEUROSCI.3974-11.2012

   • PubMed
   • Google Scholar
5. 1. Auksztulewicz R
   2. Blankenburg F

   (2013) Subjective rating of weak tactile stimuli is parametrically encoded in event-related potentials

   Journal of Neuroscience 33:11878–11887.

   https://doi.org/10.1523/JNEUROSCI.4243-12.2013

   • PubMed
   • Google Scholar
6. 1. Baars BJ

   (1997)

   In the theater of consciousness - Global workspace theory, A rigorous scientific theory of consciousness

   Journal of Consciousness Studies 4:.

   • Google Scholar
7. 1. Bastuji H
   2. Frot M
   3. Perchet C
   4. Magnin M
   5. Garcia-Larrea L

   (2016) Pain networks from the inside: spatiotemporal analysis of brain responses leading from nociception to conscious perception

   Human Brain Mapping 37:4301–4315.

   https://doi.org/10.1002/hbm.23310

   • PubMed
   • Google Scholar
8. 1. Boehler CN
   2. Schoenfeld MA
   3. Heinze HJ
   4. Hopf JM

   (2008) Rapid recurrent processing gates awareness in primary visual cortex

   PNAS 105:8742–8747.

   https://doi.org/10.1073/pnas.0801999105

   • PubMed
   • Google Scholar
9. 1. Boly M
   2. Balteau E
   3. Schnakers C
   4. Degueldre C
   5. Moonen G
   6. Luxen A
   7. Phillips C
   8. Peigneux P
   9. Maquet P
   10. Laureys S

   (2007) Baseline brain activity fluctuations predict somatosensory perception in humans

   PNAS 104:12187–12192.

   https://doi.org/10.1073/pnas.0611404104

   • PubMed
   • Google Scholar
10. 1. Boly M
    2. Massimini M
    3. Tsuchiya N
    4. Postle BR
    5. Koch C
    6. Tononi G

    (2017) Are the neural correlates of consciousness in the front or in the back of the cerebral cortex? clinical and neuroimaging evidence

    The Journal of Neuroscience 37:9603–9613.

    https://doi.org/10.1523/JNEUROSCI.3218-16.2017

    • PubMed
    • Google Scholar
11. 1. Bornhövd K
    2. Quante M
    3. Glauche V
    4. Bromm B
    5. Weiller C
    6. Büchel C

    (2002) Painful stimuli evoke different stimulus-response functions in the Amygdala, prefrontal, insula and somatosensory cortex: a single-trial fMRI study

    Brain 125:1326–1336.

    https://doi.org/10.1093/brain/awf137

    • PubMed
    • Google Scholar
12. 1. Brainard DH

    (1997) The psychophysics toolbox

    Spatial Vision 10:433–436.

    https://doi.org/10.1163/156856897X00357

    • PubMed
    • Google Scholar
13. 1. Brascamp J
    2. Blake R
    3. Knapen T

    (2015) Negligible fronto-parietal BOLD activity accompanying unreportable switches in bistable perception

    Nature Neuroscience 18:1672–1678.

    https://doi.org/10.1038/nn.4130

    • PubMed
    • Google Scholar
14. 1. Büchel C
    2. Bornhovd K
    3. Quante M
    4. Glauche V
    5. Bromm B
    6. Weiller C

    (2002) Dissociable neural responses related to pain intensity, stimulus intensity, and stimulus awareness within the anterior cingulate cortex: a parametric single-trial laser functional magnetic resonance imaging study

    The Journal of Neuroscience 22:970–976.

    https://doi.org/10.1523/JNEUROSCI.22-03-00970.2002

    • PubMed
    • Google Scholar
15. 1. Canales-Johnson A
    2. Silva C
    3. Huepe D
    4. Rivera-Rei Á
    5. Noreika V
    6. Garcia MC
    7. Silva W
    8. Ciraolo C
    9. Vaucheret E
    10. Sedeño L
    11. Couto B
    12. Kargieman L
    13. Baglivo F
    14. Sigman M
    15. Chennu S
    16. Ibáñez A
    17. Rodríguez E
    18. Bekinschtein TA

    (2015) Auditory feedback differentially modulates behavioral and neural markers of objective and subjective performance when tapping to your heartbeat

    Cerebral Cortex 25:4490–4503.

    https://doi.org/10.1093/cercor/bhv076

    • PubMed
    • Google Scholar
16. 1. Carmel D
    2. Lavie N
    3. Rees G

    (2006) Conscious awareness of flicker in humans involves frontal and parietal cortex

    Current Biology 16:907–911.

    https://doi.org/10.1016/j.cub.2006.03.055

    • PubMed
    • Google Scholar
17. 1. Cauller LJ
    2. Kulics AT

    (1991) The neural basis of the behaviorally relevant N1 component of the somatosensory-evoked potential in SI cortex of awake monkeys: evidence that backward cortical projections signal conscious touch sensation

    Experimental Brain Research 84:607–619.

    https://doi.org/10.1007/BF00230973

    • PubMed
    • Google Scholar
18. 1. Chen TL
    2. Babiloni C
    3. Ferretti A
    4. Perrucci MG
    5. Romani GL
    6. Rossini PM
    7. Tartaro A
    8. Del Gratta C

    (2008) Human secondary somatosensory cortex is involved in the processing of somatosensory rare stimuli: an fMRI study

    NeuroImage 40:1765–1771.

    https://doi.org/10.1016/j.neuroimage.2008.01.020

    • PubMed
    • Google Scholar
19. 1. Cornelissen FW
    2. Peters EM
    3. Palmer J

    (2002) The eyelink toolbox: eye tracking with MATLAB and the psychophysics toolbox

    Behavior Research Methods, Instruments, & Computers 34:613–617.

    https://doi.org/10.3758/BF03195489

    • Google Scholar
20. 1. Craig AD

    (2009) How do you feel--now? the anterior insula and human awareness

    Nature Reviews Neuroscience 10:59–70.

    https://doi.org/10.1038/nrn2555

    • PubMed
    • Google Scholar
21. 1. Critchley HD
    2. Wiens S
    3. Rotshtein P
    4. Ohman A
    5. Dolan RJ

    (2004) Neural systems supporting interoceptive awareness

    Nature Neuroscience 7:189–195.

    https://doi.org/10.1038/nn1176

    • PubMed
    • Google Scholar
22. 1. de Graaf TA
    2. Hsieh PJ
    3. Sack AT

    (2012) The 'correlates' in neural correlates of consciousness

    Neuroscience & Biobehavioral Reviews 36:191–197.

    https://doi.org/10.1016/j.neubiorev.2011.05.012

    • PubMed
    • Google Scholar
23. 1. de Lafuente V
    2. Romo R

    (2005) Neuronal correlates of subjective sensory experience

    Nature Neuroscience 8:1698–1703.

    https://doi.org/10.1038/nn1587

    • PubMed
    • Google Scholar
24. 1. de Lafuente V
    2. Romo R

    (2006) Neural correlate of subjective sensory experience gradually builds up across cortical areas

    PNAS 103:14266–14271.

    https://doi.org/10.1073/pnas.0605826103

    • PubMed
    • Google Scholar
25. 1. de Lafuente V
    2. Romo R

    (2011) Dopamine neurons code subjective sensory experience and uncertainty of perceptual decisions

    PNAS 108:19767–19771.

    https://doi.org/10.1073/pnas.1117636108

    • PubMed
    • Google Scholar
26. 1. Dehaene S
    2. Changeux JP
    3. Naccache L
    4. Sackur J
    5. Sergent C

    (2006) Conscious, Preconscious, and subliminal processing: a testable taxonomy

    Trends in Cognitive Sciences 10:204–211.

    https://doi.org/10.1016/j.tics.2006.03.007

    • PubMed
    • Google Scholar
27. 1. Delhaye BP
    2. Long KH
    3. Bensmaia SJ

    (2018)

    Comprehensive Physiology

    1575–1602, Neural Basis of Touch and Proprioception in Primate Cortex, Comprehensive Physiology, Hoboken, John Wiley & Sons, Inc, 10.1002/cphy.c170033.

    • Google Scholar
28. 1. Eickhoff SB
    2. Stephan KE
    3. Mohlberg H
    4. Grefkes C
    5. Fink GR
    6. Amunts K
    7. Zilles K

    (2005) A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data

    NeuroImage 25:1325–1335.

    https://doi.org/10.1016/j.neuroimage.2004.12.034

    • PubMed
    • Google Scholar
29. 1. Eickhoff SB
    2. Amunts K
    3. Mohlberg H
    4. Zilles K

    (2006a) The human parietal operculum. II. stereotaxic maps and correlation with functional imaging results

    Cerebral Cortex 16:268–279.

    https://doi.org/10.1093/cercor/bhi106

    • PubMed
    • Google Scholar
30. 1. Eickhoff SB
    2. Schleicher A
    3. Zilles K
    4. Amunts K

    (2006b) The human parietal operculum. I. cytoarchitectonic mapping of subdivisions

    Cerebral Cortex 16:254–267.

    https://doi.org/10.1093/cercor/bhi105

    • PubMed
    • Google Scholar
31. 1. Eriksson J
    2. Larsson A
    3. Ahlström KR
    4. Nyberg L

    (2007) Similar frontal and distinct posterior cortical regions mediate visual and auditory perceptual awareness

    Cerebral Cortex 17:760–765.

    https://doi.org/10.1093/cercor/bhk029

    • PubMed
    • Google Scholar
32. 1. Ettinger U
    2. Ffytche DH
    3. Kumari V
    4. Kathmann N
    5. Reuter B
    6. Zelaya F
    7. Williams SC

    (2008) Decomposing the neural correlates of antisaccade eye movements using event-related FMRI

    Cerebral Cortex 18:1148–1159.

    https://doi.org/10.1093/cercor/bhm147

    • PubMed
    • Google Scholar
33. 1. Farooqui AA
    2. Manly T

    (2018) When attended and conscious perception deactivates fronto-parietal regions

    Cortex 107:166–179.

    https://doi.org/10.1016/j.cortex.2017.09.004

    • Google Scholar
34. 1. Felleman DJ
    2. Van Essen DC

    (1991) Distributed hierarchical processing in the primate cerebral cortex

    Cerebral Cortex 1:1–47.

    https://doi.org/10.1093/cercor/1.1.1

    • Google Scholar
35. 1. Frässle S
    2. Sommer J
    3. Jansen A
    4. Naber M
    5. Einhäuser W

    (2014) Binocular rivalry: frontal activity relates to introspection and action but not to perception

    Journal of Neuroscience 34:1738–1747.

    https://doi.org/10.1523/JNEUROSCI.4403-13.2014

    • PubMed
    • Google Scholar
36. 1. Frey JN
    2. Ruhnau P
    3. Leske S
    4. Siegel M
    5. Braun C
    6. Weisz N

    (2016) The tactile window to consciousness is characterized by Frequency-Specific integration and segregation of the primary somatosensory cortex

    Scientific Reports 6:20805.

    https://doi.org/10.1038/srep20805

    • PubMed
    • Google Scholar
37. 1. Garfinkel SN
    2. Seth AK
    3. Barrett AB
    4. Suzuki K
    5. Critchley HD

    (2015) Knowing your own heart: distinguishing interoceptive accuracy from interoceptive awareness

    Biological Psychology 104:65–74.

    https://doi.org/10.1016/j.biopsycho.2014.11.004

    • PubMed
    • Google Scholar
38. 1. Geyer S
    2. Schleicher A
    3. Zilles K

    (1999) Areas 3a, 3b, and 1 of human primary somatosensory cortex: 1. microstructural organization and interindividual variability

    NeuroImage 10:63–83.

    https://doi.org/10.1006/nimg.1999.0440

    • Google Scholar
39. 1. Goldin MA
    2. Harrell ER
    3. Estebanez L
    4. Shulz DE

    (2018) Rich spatio-temporal stimulus dynamics unveil sensory specialization in cortical area S2

    Nature Communications 9:4053.

    https://doi.org/10.1038/s41467-018-06585-4

    • PubMed
    • Google Scholar
40. 1. Grefkes C
    2. Geyer S
    3. Schormann T
    4. Roland P
    5. Zilles K

    (2001) Human somatosensory area 2: observer-independent cytoarchitectonic mapping, interindividual variability, and population map

    NeuroImage 14:617–631.

    https://doi.org/10.1006/nimg.2001.0858

    • Google Scholar
41. 1. Hirvonen J
    2. Palva S

    (2016) Cortical localization of phase and amplitude dynamics predicting access to somatosensory awareness

    Human Brain Mapping 37:311–326.

    https://doi.org/10.1002/hbm.23033

    • PubMed
    • Google Scholar
42. 1. Hoeting JA
    2. Madigan D
    3. Raftery AE
    4. Volinsky CT

    (1999)

    Bayesian model averaging: a tutorial

    Statistical Science 14:382–401.

    • Google Scholar
43. 1. Hurme M
    2. Koivisto M
    3. Revonsuo A
    4. Railo H

    (2017) Early processing in primary visual cortex is necessary for conscious and unconscious vision while late processing is necessary only for conscious vision in neurologically healthy humans

    NeuroImage 150:230–238.

    https://doi.org/10.1016/j.neuroimage.2017.02.060

    • PubMed
    • Google Scholar
44. Book

    1. Johnson VE
    2. Albert JH

    (2006)

    Ordinal Data Modeling

    Springer Science & Business Media.

    • Google Scholar
45. 1. Jones SR
    2. Pritchett DL
    3. Stufflebeam SM
    4. Hämäläinen M
    5. Moore CI

    (2007) Neural correlates of tactile detection: a combined magnetoencephalography and biophysically based computational modeling study

    Journal of Neuroscience 27:10751–10764.

    https://doi.org/10.1523/JNEUROSCI.0482-07.2007

    • PubMed
    • Google Scholar
46. 1. Kass RE
    2. Raftery AE

    (1995) Bayes factors

    Journal of the American Statistical Association 90:773–795.

    https://doi.org/10.1080/01621459.1995.10476572

    • Google Scholar
47. 1. Kimmig H
    2. Greenlee MW
    3. Gondan M
    4. Schira M
    5. Kassubek J
    6. Mergner T

    (2001) Relationship between saccadic eye movements and cortical activity as measured by fMRI: quantitative and qualitative aspects

    Experimental Brain Research 141:184–194.

    https://doi.org/10.1007/s002210100844

    • PubMed
    • Google Scholar
48. 1. Koch C
    2. Massimini M
    3. Boly M
    4. Tononi G

    (2016) Neural correlates of consciousness: progress and problems

    Nature Reviews Neuroscience 17:307–321.

    https://doi.org/10.1038/nrn.2016.22

    • PubMed
    • Google Scholar
49. Software

    1. Krekelberg B

    (2019) bayesFactor

    GitHub.

    https://www.github.com/klabhub/bayesFactor
50. 1. Kwon SE
    2. Yang H
    3. Minamisawa G
    4. O'Connor DH

    (2016) Sensory and decision-related activity propagate in a cortical feedback loop during touch perception

    Nature Neuroscience 19:1243–1249.

    https://doi.org/10.1038/nn.4356

    • PubMed
    • Google Scholar
51. 1. Lamichhane B
    2. Adhikari BM
    3. Dhamala M

    (2016) Salience network activity in perceptual decisions

    Brain Connectivity 6:558–571.

    https://doi.org/10.1089/brain.2015.0392

    • PubMed
    • Google Scholar
52. 1. Lamme VA

    (2006) Towards a true neural stance on consciousness

    Trends in Cognitive Sciences 10:494–501.

    https://doi.org/10.1016/j.tics.2006.09.001

    • PubMed
    • Google Scholar
53. 1. Menon V
    2. Uddin LQ

    (2010) Saliency, switching, attention and control: a network model of insula function

    Brain Structure and Function 214:655–667.

    https://doi.org/10.1007/s00429-010-0262-0

    • PubMed
    • Google Scholar
54. 1. Moore CI
    2. Crosier E
    3. Greve DN
    4. Savoy R
    5. Merzenich MM
    6. Dale AM

    (2013) Neocortical correlates of vibrotactile detection in humans

    Journal of Cognitive Neuroscience 25:49–61.

    https://doi.org/10.1162/jocn_a_00315

    • PubMed
    • Google Scholar
55. 1. Odegaard B
    2. Knight RT
    3. Lau H

    (2017) Should a few null findings falsify prefrontal theories of conscious perception?

    The Journal of Neuroscience 37:9593–9602.

    https://doi.org/10.1523/JNEUROSCI.3217-16.2017

    • PubMed
    • Google Scholar
56. 1. Palva S
    2. Linkenkaer-Hansen K
    3. Näätänen R
    4. Palva JM

    (2005) Early neural correlates of conscious somatosensory perception

    Journal of Neuroscience 25:5248–5258.

    https://doi.org/10.1523/JNEUROSCI.0141-05.2005

    • PubMed
    • Google Scholar
57. 1. Panagiotaropoulos TI
    2. Deco G
    3. Kapoor V
    4. Logothetis NK

    (2012) Neuronal discharges and gamma oscillations explicitly reflect visual consciousness in the lateral prefrontal cortex

    Neuron 74:924–935.

    https://doi.org/10.1016/j.neuron.2012.04.013

    • PubMed
    • Google Scholar
58. 1. Pascual-Leone A
    2. Walsh V

    (2001) Fast backprojections from the motion to the primary visual area necessary for visual awareness

    Science 292:510–512.

    https://doi.org/10.1126/science.1057099

    • PubMed
    • Google Scholar
59. 1. Penny WD
    2. Trujillo-Barreto NJ
    3. Friston KJ

    (2005) Bayesian fMRI time series analysis with spatial priors

    NeuroImage 24:350–362.

    https://doi.org/10.1016/j.neuroimage.2004.08.034

    • PubMed
    • Google Scholar
60. 1. Penny WD
    2. Flandin G
    3. Trujillo-Barreto N

    (2007) Bayesian comparison of spatially regularised general linear models

    Human Brain Mapping 28:275–293.

    https://doi.org/10.1002/hbm.20327

    • PubMed
    • Google Scholar
61. 1. Penny WD
    2. Stephan KE
    3. Daunizeau J
    4. Rosa MJ
    5. Friston KJ
    6. Schofield TM
    7. Leff AP

    (2010) Comparing families of dynamic causal models

    PLOS Computational Biology 6:e1000709.

    https://doi.org/10.1371/journal.pcbi.1000709

    • PubMed
    • Google Scholar
62. 1. Pitts MA
    2. Metzler S
    3. Hillyard SA

    (2014) Isolating neural correlates of conscious perception from neural correlates of reporting one's perception

    Frontiers in Psychology 5:1078.

    https://doi.org/10.3389/fpsyg.2014.01078

    • PubMed
    • Google Scholar
63. 1. Rees G
    2. Kreiman G
    3. Koch C

    (2002) Neural correlates of consciousness in humans

    Nature Reviews Neuroscience 3:261–270.

    https://doi.org/10.1038/nrn783

    • PubMed
    • Google Scholar
64. 1. Ress D
    2. Backus BT
    3. Heeger DJ

    (2000) Activity in primary visual cortex predicts performance in a visual detection task

    Nature Neuroscience 3:940–945.

    https://doi.org/10.1038/78856

    • PubMed
    • Google Scholar
65. 1. Romo R
    2. Hernández A
    3. Zainos A
    4. Lemus L
    5. Brody CD

    (2002) Neuronal correlates of decision-making in secondary somatosensory cortex

    Nature Neuroscience 5:1217–1225.

    https://doi.org/10.1038/nn950

    • PubMed
    • Google Scholar
66. 1. Rosa MJ
    2. Bestmann S
    3. Harrison L
    4. Penny W

    (2010) Bayesian model selection maps for group studies

    NeuroImage 49:217–224.

    https://doi.org/10.1016/j.neuroimage.2009.08.051

    • PubMed
    • Google Scholar
67. 1. Sanchez Panchuelo RM
    2. Besle J
    3. Schluppeck D
    4. Humberstone M
    5. Francis S

    (2018) Somatotopy in the human somatosensory system

    Frontiers in Human Neuroscience 12:235.

    https://doi.org/10.3389/fnhum.2018.00235

    • PubMed
    • Google Scholar
68. 1. Schubert R
    2. Blankenburg F
    3. Lemm S
    4. Villringer A
    5. Curio G

    (2006) Now you feel it--now you don't: erp correlates of somatosensory awareness

    Psychophysiology 43:31–40.

    https://doi.org/10.1111/j.1469-8986.2006.00379.x

    • PubMed
    • Google Scholar
69. 1. Schubert R
    2. Haufe S
    3. Blankenburg F
    4. Villringer A
    5. Curio G

    (2009) Now you'll feel it, now you won't: eeg rhythms predict the effectiveness of perceptual masking

    Journal of Cognitive Neuroscience 21:2407–2419.

    https://doi.org/10.1162/jocn.2008.21174

    • PubMed
    • Google Scholar
70. 1. Seth AK

    (2013) Interoceptive inference, emotion, and the embodied self

    Trends in Cognitive Sciences 17:565–573.

    https://doi.org/10.1016/j.tics.2013.09.007

    • PubMed
    • Google Scholar
71. 1. Stephan KE
    2. Penny WD
    3. Daunizeau J
    4. Moran RJ
    5. Friston KJ

    (2009) Bayesian model selection for group studies

    NeuroImage 46:1004–1017.

    https://doi.org/10.1016/j.neuroimage.2009.03.025

    • PubMed
    • Google Scholar
72. 1. Tsuchiya N
    2. Wilke M
    3. Frässle S
    4. Lamme VAF

    (2015) No-Report paradigms: extracting the true neural correlates of consciousness

    Trends in Cognitive Sciences 19:757–770.

    https://doi.org/10.1016/j.tics.2015.10.002

    • PubMed
    • Google Scholar
73. 1. Weisz N
    2. Wühle A
    3. Monittola G
    4. Demarchi G
    5. Frey J
    6. Popov T
    7. Braun C

    (2014) Prestimulus oscillatory power and connectivity patterns predispose conscious somatosensory perception

    PNAS 111:E417–E425.

    https://doi.org/10.1073/pnas.1317267111

    • PubMed
    • Google Scholar
74. 1. Wichmann FA
    2. Hill NJ

    (2001) The psychometric function: I. fitting, sampling, and goodness of fit

    Perception & Psychophysics 63:1293–1313.

    https://doi.org/10.3758/BF03194544

    • PubMed
    • Google Scholar
75. 1. Wühle A
    2. Mertiens L
    3. Rüter J
    4. Ostwald D
    5. Braun C

    (2010) Cortical processing of near-threshold tactile stimuli: an MEG study

    Psychophysiology 47:523–534.

    https://doi.org/10.1111/j.1469-8986.2010.00964.x

    • PubMed
    • Google Scholar
76. 1. Wühle A
    2. Preissl H
    3. Braun C

    (2011) Cortical processing of near-threshold tactile stimuli in a paired-stimulus paradigm--an MEG study

    European Journal of Neuroscience 34:641–651.

    https://doi.org/10.1111/j.1460-9568.2011.07770.x

    • PubMed
    • Google Scholar
77. 1. Yang H
    2. Kwon SE
    3. Severson KS
    4. O'Connor DH

    (2016) Origins of choice-related activity in mouse somatosensory cortex

    Nature Neuroscience 19:127–134.

    https://doi.org/10.1038/nn.4183

    • PubMed
    • Google Scholar
78. 1. Zagha E
    2. Casale AE
    3. Sachdev RN
    4. McGinley MJ
    5. McCormick DA

    (2013) Motor cortex feedback influences sensory processing by modulating network state

    Neuron 79:567–578.

    https://doi.org/10.1016/j.neuron.2013.06.008

    • PubMed
    • Google Scholar

Thank you for submitting your article "Neural basis of somatosensory target detection in humans." for consideration by eLife. Your article has been reviewed by two peer reviewers, including Floris de Lange as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Joshua Gold as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Tristan A Bekinschtein (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The authors examine the neural correlates of somatosensory target detection, by applying median nerve stimulation electrical pulses of varying strength to human volunteers (N=27). They go beyond earlier investigations of sensory awareness by controlling for several factors that are associated with sensory awareness, i.e. stimulus uncertainty, behavioral relevance, overt reporting and motor responding. They observe that neural responses that can be specifically linked to target detection reside in secondary somatosensory cortex (SII), arguing for a specific role of sensory systems in generating awareness (instead of a broadcasting signal in fronto-parietal cortex). This is a well motivated, well executed and well written piece on a topic of general interest. It is notoriously difficult to disentangle awareness-related activity modulations from its precursors and consequences. I believe the authors have brought this quest a step further, by controlling for several relevant 'confounding factors' and using Bayesian Model Selection in an attempt to disentangle the respective contributions of several factors to the activity profile of sensory and non-sensory regions.

Essential revisions:

1) Anatomical localization of effects

The neural locus of intensity, P(detection) and detection effects are sometimes incredibly close to each other. There is no mention of the smoothing kernel of the fMRI data (please add!), but this does raise the question of what causes the regions of shared variance (intermixed colors). Are they really sensitive to all these effects, or is the intermixing the result of spatial and/or normalization-induced smoothing?

2) Smooth transformation from sensation to perception

The authors argue that there is a smooth transformation from physical to perceptual representation in the somatosensory system. It would be nice if this could be better visualized, and more formally tested – for example by testing whether peak anatomical coordinates for the different task components have a monotonic progression on the posterior-anterior axis.

3) The title seems a bit general, it read like a review title, please consider some version more accurate that maybe reflects this experiment analysis, and conclusions?

4) Several possible confounding factors were raised that should be at least discussed further, or possibly included in new analyses:

a) Perception of near threshold stimuli is difficult, and resolution of associated uncertainty and introspective processes may differ between detected and undetected targets (de Lafuente and Romo, 2011). How might this affect the interpretation of your results?

b) Target detection is the explicit behavioural goal of the task and therefore, detected targets have higher behavioural relevance than undetected targets (Farooqui and Manly, 2018). How might this affect the interpretation of your results?

c) Target detection is directly mapped to overt reports that allow for assessment of participants' trial-by-trial perception (Tsuchiya, Wilke, Frässle, and Lamme, 2015). Can this be teased apart from a possibly covert perceptual process?

d) Overt reports are often communicated with button presses by one hand while stimulation occurs on the other hand, which may affect cortical excitability in homologue regions of the sensorimotor homunculus (Zagha, Casale, Sachdev, McGinley, and McCormick, 2013).

5) Subsection “Behaviour”, the last phrase "Chi squared tests of independence showed no association of target detection with overt reports for any of the participants (all p > .2), confirming that our task design rendered these variables independent."

In the manner I interpret it seems that the authors are assuming that a lack of significance in the test is evidence for a no association, please either rephrase to say the test did not show enough evidence for an association or that there were not significantly associated; alternatively, run an appropriate test pointing to evidence for the null effect (enough evidence of no association) like a bayes factor.

6) In the subsection “fMRI”, I am happy to see that the activity after the BMS is not enough for the authors to claim neural association, they are also demanding fit to the each model (dichotomous, sigmoideal). In this respect they say: "The lack of such a relationship may suggest that the detection-related response that drives the model fit in these regions does not constitute an all-or-nothing response to target detection but may be restricted to only a subset of trials (see Discussion)". I thought that maybe it was also restricted to a subset of participants. I was thinking in asking if they author could provide of an impression of each subject fit and variability in the fMRI data. I am curious to understand the contribution of the participants to these interesting results and one way to do it is to plot single subject parameters. Would the authors be happy to think of a manner to show the direction and strength of effect per participant for the main results? I understand this may seem like a lot of work, it shouldn't. The fitting parameters per participant are easy to extract and the so are the bold estimates. This would allow for a good understanding of the variability of the 27 and also for the contributions each make.

Uncertainty here is not formal report uncertainty (subjective measure of uncertainty) but a toy model of expected uncertainty, it would be good for the authors to comment and give some details on this. Which concept of uncertainty do they refer to, some closer to the objective aspect of detection or closer to a metacognitive evaluation? I know it is defined later in the Discussion but a short account early on would help. Thanks.

Figure 5 legend, at the end the authors refer again to a lack of effect as "confirming that overt reports were independent of stimulus intensity and target detection." I actually like the result but I am also aware of the language vs. the test. Does the test perform confirms or give evidence for a no effect? It is truly independence or it is a lack of dependence? Apologies if I start to sound like the stats pedant.

7) On the topic of "Considering that interoceptive signals, such as the heartbeat, are often faint signals, it would be interesting to see how explicit control of perceptual uncertainty might affect these interpretations." Some year ago we also investigated the role of the insula in faint signals (heartbeat, Canales-Johnson et al., 2015), and found that both in SI and SII but more importantly in insula, the performance modulated the signal, but not the -subjective- impression of improvement. This is not to ask the authors to cite this piece, it is only to show the complexity of the interoceptive argument. There are many papers tapping into the heartbeat as a faint sensory signal, this is mainly because it might be that it is in fact not interoception but a secondary feeling on the heart pounding in the chest or the blood in the neck, ears, etc.

8) Discussion, seventh paragraph, maybe it does not capture any frontal or parietal cortex because the stimuli is so simple? I know this is borderline on strawman but it might be that the simple zap on the skin creates very little demand for detection and hence we are under-powered to capture the extent of the network involve as the demands are too little? This is just food for thought, not a really strong view that I hold. It is always difficult with these faint stimuli, and with detection task in general. I have not forgotten that the experimental design here is elegant and partially controls for this possible critique… It is in fact commented by the end of the Discussion.

Essential revisions:

1) Anatomical localization of effects

The neural locus of intensity, P(detection) and detection effects are sometimes incredibly close to each other. There is no mention of the smoothing kernel of the fMRI data (please add!), but this does raise the question of what causes the regions of shared variance (intermixed colors). Are they really sensitive to all these effects, or is the intermixing the result of spatial and/or normalization-induced smoothing?

When using Bayesian Model Selection, spatial smoothing is usually done after the 1^st level analysis. This is because the Bayesian GLM estimation capitalises on spatial priors, which allow estimating the smoothness of experimental effects from the data. Spatial smoothing is then performed on the resulting evidence maps to facilitate group analyses. For this reason, we did not describe the smoothing in the general preprocessing section, but following the description of the first level analysis:

“These evidence maps were smoothed with an 8mm FWHM Gaussian filter, resampled to 2x2x2 mm³ voxel size, and subjected to voxel-wise random effects BMS resulting in one EP map per model for group-level inference (Rosa, Bestmann, Harrison, and Penny, 2010; Stephan, Penny, Daunizeau, Moran, and Friston, 2009).”

To increase visibility of this information, we have added a reference to it in the preprocessing section, where readers would likely look for it:

“Spatial smoothing was not performed prior to the 1^st level analysis because the Bayesian GLM approach estimates the smoothness of experimental effects from the data using spatial priors (Penny, Trujillo-Barreto, and Friston, 2005) (for information on smoothing of evidence maps and beta images, see below).”

In addition, we have added information on the spatial prior used in the GLM estimation, as well as information on smoothing of beta images for the extraction of SRPs, which was missing beforehand:

“As spatial prior, we used the recommended Unweighted Graph Laplacian prior, which softly constrains effects to be similar in neighbouring voxels (where the strength of the constraint for each regressor is estimated from the data).”

“The resulting beta images were smoothed at 8mm FWHM and resampled to 2x2x2 mm³.”

The question whether regions of shared variance are truly responsive to more than one behavioural dimension is very interesting but also notoriously difficult to answer using fMRI. However, to address your question at least to the level of fMRI voxels (3x3x3 mm³), we repeated our analysis, once with a global shrinkage prior instead of the spatial UGL prior in the first level GLM estimation and once with a smaller smoothing kernel (4mm FWHM) on the evidence maps for second level BMS. Neither of these approaches produced considerably more distinct results (see below). In fact, using a global shrinkage prior (which does not impose any prior assumption on the smoothness of the data) resulted in more overlap, suggesting that the use of a spatial prior increased the sensitivity of our analysis. Further, we inspected BMS maps of individual participants and found that these were no less smooth than the group results. Following these observations, we carefully conclude that the regions of shared variance are genuine at least to the resolution allowed by conventional group-level fMRI analyses. However, we would like to stress that the many levels of smoothing that are inherent in fMRI analyses (spatial specificity of the BOLD response, averaging over neurons contained within a voxel, and explicit smoothing during data analysis) limit our ability to draw final conclusions regarding the observed overlap of effects.

Author response image 1

Download asset Open asset

We have added a brief description of the alternative smoothing approaches to the Materials and methods section of the manuscript.

“To assess the impact of the employed spatial prior and smoothing parameters on the overlap of effects, we repeated the analyses, once using a global shrinkage prior (which does not constrain the smoothness of the data) in the GLM estimation, and once using a 4mm FWHM smoothing kernel on the evidence maps used for BMS. Neither of these methods considerably reduced the overlap of effects and we report the results using the original parameters.”

2) Smooth transformation from sensation to perception

The authors argue that there is a smooth transformation from physical to perceptual representation in the somatosensory system. It would be nice if this could be better visualized, and more formally tested – for example by testing whether peak anatomical coordinates for the different task components have a monotonic progression on the posterior-anterior axis.

We did not mean to imply a smooth spatial gradient, please excuse this misunderstanding. We merely observe that more posterior cytoarchitectonic subregions of SI contain more voxels sensitive to detection probability compared to more anterior regions in the group level BMS. To illustrate this observation and clarify that we are referring to distinct regions, we added a supplementary figure (Figure 3—figure supplement 1), which displays the relative number of voxels best explained by the different +family models in different subregions of SI and SII as defined by the Anatomy Toolbox. In addition, we briefly discuss why we do not find a similar organisational principle in SII. Note however, that these are purely descriptive results, since the voxel counts were obtained from the second level BMS maps.

Nonetheless, the idea of a smooth gradient is very interesting, and we agree that our BMS results in SI hint to the presence of such a gradient. However, given the considerations about spatial resolution raised in question one, we do not feel confident in claiming such a gradient based on the current data. Future high-resolution fMRI studies or even invasive recordings may give more insights into this question.

”Within SI, the spatial distribution of voxels explained by the intensity and detection probability models, respectively was found to follow the known cytoarchitectonic subdivisions of the anterior parietal cortex (Figure 3—figure supplement 1). In SII on the other hand, voxels with sensitivity to intensity, detection probability, and detection did not show an apparent organisation along the cytoarchitectonic subdivisions of the parietal operculum.”

”The distribution of these effects did not align with the cytoarchitectonic subdivisions of the parietal operculum, whose distinct functional roles remain largely elusive. The less pronounced somatotopy and large interindividual variability in SII (Eickhoff, Amunts, Mohlberg, and Zilles, 2006; Sanchez Panchuelo, Besle, Schluppeck, Humberstone, and Francis, 2018) may have concealed the exact organisation of the observed effects and further studies are needed to ascertain the specific functional topology.”

“Finally, we examined the distribution of voxels best explained by the different models of the +family across the known cytoarchitectonic subregions of primary and secondary somatosensory cortex (Eickhoff, Schleicher, Zilles, and Amunts, 2006; Geyer, Schleicher, and Zilles, 1999; Grefkes, Geyer, Schormann, Roland, and Zilles, 2001). […] For each cytoarchitectonic subregion (BA 3b, 1, 2, OP1-4), we then determined the proportion of voxels labelled by the respective +family models to obtain a descriptive summary of the BMS results across regions.”

3) The title seems a bit general, it read like a review title, please consider some version more accurate that maybe reflects this experiment analysis, and conclusions?

Thank you for pointing this out. We have changed the title to be more representative of the study. The title now reads:

“Neural basis of somatosensory target detection independent of uncertainty, relevance, and reports”

4) Several possible confounding factors were raised that should be at least discussed further, or possibly included in new analyses:

a) Perception of near threshold stimuli is difficult, and resolution of associated uncertainty and introspective processes may differ between detected and undetected targets (de Lafuente and Romo, 2011). How might this affect the interpretation of your results?

b) Target detection is the explicit behavioural goal of the task and therefore, detected targets have higher behavioural relevance than undetected targets (Farooqui and Manly, 2018). How might this affect the interpretation of your results?

c) Target detection is directly mapped to overt reports that allow for assessment of participants' trial-by-trial perception (Tsuchiya, Wilke, Frässle, and Lamme, 2015). Can this be teased apart from a possibly covert perceptual process?

d) Overt reports are often communicated with button presses by one hand while stimulation occurs on the other hand, which may affect cortical excitability in homologue regions of the sensorimotor homunculus (Zagha, Casale, Sachdev, McGinley, and McCormick, 2013).

We have added a new paragraph in the Discussion that addresses each of the raised points and discusses in how far our task was successful in controlling for the different potential confounds:

“The aim of the current study was to experimentally dissociate target detection from four potentially confounding processes and hence, preclude these processes as root causes of detection related neural activity. […] In conclusion, we consider the detection-related responses observed in bilateral SII to be largely free of the addressed experimental confounds.”

5) Subsection “Behaviour”, the last phrase "Chi squared tests of independence showed no association of target detection with overt reports for any of the participants (all p > .2), confirming that our task design rendered these variables independent."

In the manner I interpret it seems that the authors are assuming that a lack of significance in the test is evidence for a no association, please either rephrase to say the test did not show enough evidence for an association or that there were not significantly associated; alternatively, run an appropriate test pointing to evidence for the null effect (enough evidence of no association) like a bayes factor.

We entirely agree with this comment and apologise for misrepresenting the test results. We have now replaced all statistical tests with Bayesian equivalents and report Bayes factors capturing evidence for deviation of mean reaction times from zero and Bayes factors for association and independence for the relationship between overt reports and target detection.

“A Bayesian equivalence of the paired-sample t-test suggests strong evidence for shorter reaction times for detected than undetected targets (detected: 352.11 ± 50.98ms, undetected: 363.88 ± 55.97ms, difference: 11.77 ± 14.84ms, BF10 = 88.01). […] In contrast, we found positive evidence for independence for all participants (BF01 > 4) except one (BF01 = 2.62).”

“Differences in reaction times between detected and undetected targets were assessed using a Bayesian equivalent of the paired t-test (Krekelberg, 2019) and the Bayes factor quantifying the evidence for a mean deviation from zero (BF10) is reported. […] Following the guidelines by Kass and Raftery, 1995, we consider 1 < BF < 3 negligible, 3 < BF < 20 positive, 20 < BF < 150 strong, and 150 < BF very strong evidence.”

6) In the subsection “fMRI”, I am happy to see that the activity after the BMS is not enough for the authors to claim neural association, they are also demanding fit to the each model (dichotomous, sigmoideal). In this respect they say: "The lack of such a relationship may suggest that the detection-related response that drives the model fit in these regions does not constitute an all-or-nothing response to target detection but may be restricted to only a subset of trials (see Discussion)". I thought that maybe it was also restricted to a subset of participants. I was thinking in asking if they author could provide of an impression of each subject fit and variability in the fMRI data. I am curious to understand the contribution of the participants to these interesting results and one way to do it is to plot single subject parameters. Would the authors be happy to think of a manner to show the direction and strength of effect per participant for the main results? I understand this may seem like a lot of work, it shouldn't. The fitting parameters per participant are easy to extract and the so are the bold estimates. This would allow for a good understanding of the variability of the 27 and also for the contributions each make.

Thank you for this excellent suggestion. Following your advice, we inspected the single subject parameter estimates in more detail and noticed that indeed there are differences between participants in the direction of the effects. In particular, in frontal and parietal regions that show good fit of the detection model but no clear reflection of the psychometric function in their SRPs, some participants show positive beta estimates for the experimental regressor whereas others show negative estimates, leading to an average estimate close to zero and accordingly, unsystematic SRPs. This finding has led us to reconsider a peculiarity of the model selection procedure, which is that it does not take the directionality of effects into account (since it is based on model evidence, which can be driven by both positive and negative effects). This means that regions with mean parameter estimates of zero but large variability across participants can be picked up by the BMS procedure. However, since we were interested in systematic covariation between the BOLD signal and our experimental regressors, we decided to introduce a second step of statistical testing, in which we computed the evidence for the hypothesis that beta estimates of the winning models’ experimental regressors systematically deviated from zero. We then only considered those regions as regions of interest that showed positive evidence for a deviation. For the detection model, this procedure excluded a subregion of right SII, as well as left SFG, left IPL, and L V3 from the regions of interest. In addition, left thalamus and right SMarG were excluded from areas showing a report effect. We added a description of the procedure and its results to the manuscript, as well as plots showing individual beta estimates for all winning models to Figures 3, 4, and 5 and the mean extracted beta estimates with corresponding Bayes factors to Table 1.

To ensure that none of the excluded regions show a bimodal distribution of beta estimates, which could point towards subgroups in the data, we used Hartigan’s DIP statistic to test beta estimates for bimodality. None of the regions yielded a significant result, suggesting that the null hypothesis of unimodality could not be rejected at the .05 significance level (Hartigan, 1985). Further, we used a Lilliefors test as implemented in Matlab to test for normality and again, none of the regions gave a significant test result, suggesting that the null hypothesis of normality could not be rejected. We conclude that there were no subgroups present in the data and that regions without systematic beta estimates were picked up by the BMS due to large variance across participants. All test results are listed in the table below, however, we did not include these statistics in the manuscript because we consider the beta plots in Figures 3, 4, and 5 sufficiently informative.

Test results for Hartigan’s DIP statistic, testing for unimodality and the Lilliefors test, testing for normality of the extracted beta estimates. None of the tests exceeded the significance level of p < .05.

“Having established regions that were well explained by the +family models, we further examined the underlying model parameters. […] Since the model selection procedure does not account for directionality of the underlying effects (i.e. both positive and negative parameter estimates may contribute to the model evidence), we imposed this second constraint on the identified regions to detect systematic variation across participants and thus, systematic association of the behavioural regressors with the observed BOLD activity.”

“Next, we tested the beta estimates of the winning models’ experimental regressors for systematic deviation from zero. […] We then extracted individual beta estimates from these peaks using the beta images of the respective winning models that were previously obtained from the 1^st level Bayesian GLM estimation and tested their deviation from zero using the Bayesian equivalent of a one-sample t-test (Krekelberg, 2019).”

Uncertainty here is not formal report uncertainty (subjective measure of uncertainty) but a toy model of expected uncertainty, it would be good for the authors to comment and give some details on this. Which concept of uncertainty do they refer to, some closer to the objective aspect of detection or closer to a metacognitive evaluation? I know it is defined later in the Discussion but a short account early on would help. Thanks.

Thank you for pointing this out. We have added a brief definition to the first description of the task and we now refer to our uncertainty model as expected uncertainty throughout the manuscript. Moreover, we state in the Discussion that confidence ratings would be required to address metacognitive aspects of the task.

“Accordingly, stimuli presented near individual 50% detection thresholds were expected to be associated with higher uncertainty (as defined by larger trial-by-trial variability in target detection) than clearly sub- or supraliminal stimuli.”

“To address the subjective dimensions of our task, we would require trial-by-trial awareness ratings (and likewise, confidence ratings to evaluate metacognitive aspects of uncertainty).”

Figure 5 legend, at the end the authors refer again to a lack of effect as "confirming that overt reports were independent of stimulus intensity and target detection." I actually like the result but I am also aware of the language vs. the test. Does the test perform confirms or give evidence for a no effect? It is truly independence or it is a lack of dependence? Apologies if I start to sound like the stats pedant.

Again, we agree with this remark and have removed the respective statement. The paragraph now refers to the previously established behavioural result and is otherwise restricted to the descriptive nature of the SRPs.

“As expected from the lack of association between overt reports and target detection, SRPs in these areas did not reflect a systematic relationship with stimulus intensity (Figure 5B).”

7) On the topic of "Considering that interoceptive signals, such as the heartbeat, are often faint signals, it would be interesting to see how explicit control of perceptual uncertainty might affect these interpretations." Some year ago we also investigated the role of the insula in faint signals (heartbeat, Canales-Johnson et al., 2015), and found that both in SI and SII but more importantly in insula, the performance modulated the signal, but not the -subjective- impression of improvement. This is not to ask the authors to cite this piece, it is only to show the complexity of the interoceptive argument. There are many papers tapping into the heartbeat as a faint sensory signal, this is mainly because it might be that it is in fact not interoception but a secondary feeling on the heart pounding in the chest or the blood in the neck, ears, etc.

Thank you for pointing us to this interesting paper. We did not mean to imply that the link between insula and interoceptive awareness is fully established, please excuse if this was unclear. In fact, we agree with the conclusions drawn in the paper you mentioned that suggest that signals in the insula may reflect processes that are related to interoceptive processing and influence interoceptive sensitivity but do not reflect awareness. Building on our results, we suggest that incorporating perceptual uncertainty as an explicit behavioural dimension in future studies might be a fruitful approach to further understand the exact role of the insula in the interoceptive process, notwithstanding existing evidence. We have rephrased the respective paragraph in the manuscript and hope that this makes it clearer.

“The insula in particular has been assigned a pivotal role in interoceptive inference (Seth, 2013) and awareness (Craig, 2009; Critchley, Wiens, Rotshtein, Öhman, and Dolan, 2004), although recent research suggests that activity in AIC might reflect interoceptive sensitivity and learning rather than subjective experience per se (Canales-Johnson et al., 2015). In light of our results and given the fact that interoceptive signals, such as the heartbeat, are often faint signals, future research may benefit from combining measures of perceptual uncertainty (e.g. Garfinkel, Seth, Barrett, Suzuki, and Critchley, 2015) with neuroimaging techniques to reconcile interoceptive processing with insular response properties.”

8) Discussion, seventh paragraph, maybe it does not capture any frontal or parietal cortex because the stimuli is so simple? I know this is borderline on strawman but it might be that the simple zap on the skin creates very little demand for detection and hence we are under-powered to capture the extent of the network involve as the demands are too little? This is just food for thought, not a really strong view that I hold. It is always difficult with these faint stimuli, and with detection task in general. I have not forgotten that the experimental design here is elegant and partially controls for this possible critique… It is in fact commented by the end of the Discussion.

This is indeed an interesting question. It is well conceivable, that more complex stimuli would require engagement of different regions and more extended networks and we do not wish to claim that SII is the neural substrate of any kind of conscious somatosensory percept. However, the most popular frontoparietal network theories of conscious perception postulate that any conscious percept occurs with a global ignition and “intense activation spreading to the frontoparietal network” (Dehaene, Changeux, Naccache, Sackur, and Sergent, 2006). Even though in our task the stimuli were simple and relatively faint on threshold trials, participants were arguably conscious of them and definitely so on trials with high intensities. If a global ignition of the frontoparietal network did underlie these conscious percepts, we would expect to find corresponding activation in our results, especially considering that frontoparietal activity is ubiquitous in neuroimaging studies and by no means difficult to detect. The fact that we did not find reliable engagement of the frontoparietal network in target detection, at the very least, does not offer supporting evidence for the frontoparietal network as part of “the minimum neural mechanisms jointly sufficient for any one specific conscious experience” (Koch, Massimini, Boly, and Tononi, 2016). Instead, we find SII as the most likely neural correlate of conscious somatosensory perception during target detection. In the manuscript, we have added the possibility of a more extended network for more complex stimuli but believe that a more comprehensive discussion of this question is beyond the scope of the current study.

“Certainly, cognitive processes such as attention and introspection considerably influence our experience and may even alter perceptual contents and likewise, detection of more complex stimuli, such as tactile motion or objects may require additional cognitive resources further downstream the somatosensory hierarchy.”

Author details

1. Pia Schröder

   Neurocomputation and Neuroimaging Unit, Freie Universität Berlin, Berlin, Germany

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Interpretation of results

   For correspondence

   pia.schroeder@fu-berlin.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6135-0691

2. Timo Torsten Schmidt

   Neurocomputation and Neuroimaging Unit, Freie Universität Berlin, Berlin, Germany

   Contribution

   Conceptualization, Investigation, Writing—review and editing, Interpretation of results

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1612-1301

3. Felix Blankenburg

   Neurocomputation and Neuroimaging Unit, Freie Universität Berlin, Berlin, Germany

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing—review and editing, Interpretation of results

   Competing interests

   No competing interests declared

• 1,862

  Page views

• 186

  Downloads

• 16

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.