All over the human body, there are receptors that help to alert the brain to potential harm. For example, intense heat on the skin elicits a signal that travels to the brain and activates many parts of the brain. Some of the same brain regions that are switched on by signals of potential bodily harm also help the brain to form expectations about events. A person’s expectations may have a strong influence on how they experience pain. For example, if a person expects that taking a pill will reduce their pain, they may feel less pain even if the pill is a fake.

Exactly how the brain processes pain signals and expectations remains unclear. Does the brain activity simply reflect how intense the heat is? Some scientists think there may be two separate processes going on: one that predicts what will happen and another that calculates the difference between the prediction and what the receptors actually detect. This difference is called a prediction error. If every unpredicted sensory signal elicits a calculation of the prediction error that would help improve the brain’s future predictions.

Now, Geuter et al. show that the predictions are a key part of how the brain perceives pain induced by heat. In the experiments, 28 people had heat applied to skin on their forearm at temperatures that were either noticeable but not painful or painful. Their brain activity was recorded using functional magnetic resonance imaging (fMRI), and measurements were taken of the pupils in their eyes and their skin’s response to heat. The fMRI scans showed that activity in the back part of a brain region called the insular cortex reflects the intensity of the heat that is applied to the person’s arm, while the front part of the same region signals pain predictions and the prediction error.

This suggests that scientists are correct that pain predictions and prediction error calculations are an integral part of the pain response. More studies are needed to determine if these brain processes might contribute to chronic pain and whether a similar process occurs in response to other types of unpleasant experiences.

Classic bottom-up views construe perception as a feedforward stream of sensory information that is passed along the neural hierarchy from receptors to high-level brain regions (Hubel and Wiesel, 1959). Accordingly, neurons are thought of as feature detectors and cortical responses to sensory stimuli are expected to scale with the presence of stimulus features, that is, activity in pain processing brain regions should reflect the activation level of nociceptors. This basic account has been extended by a wealth of findings demonstrating that top-down expectations play an important role in modulating both the pain experience and the activity in pain processing brain regions (Sawamoto et al., 2000; Koyama et al., 2005; Lorenz et al., 2005; Brown et al., 2008; Atlas et al., 2010; Bingel et al., 2011; Wiech et al., 2014b). Other neuroimaging studies have shown that stimulus-response functions differ between brain regions (Davis et al., 1998; Coghill et al., 1999; Apkarian et al., 2001; Bornhövd et al., 2002; Davis et al., 2002; Porro et al., 2004) and that brain activation is modulated by concurrent task demands (Bantick et al., 2002; Valet et al., 2004; Wiech et al., 2005; Seminowicz and Davis, 2007; Villemure and Bushnell, 2009).

However, these theories cannot explain the reduction in sensory cortical activity for expected compared to unexpected stimuli (Alink et al., 2010; Egner et al., 2010; Todorovic et al., 2011; Kok et al., 2012). In contrast, theories of Bayesian perceptual decision making, as formalized in predictive coding models (Knill and Pouget, 2004; Friston, 2005; Summerfield and de Lange, 2014), can explain such expectation suppression effects. Their proposal is that perception arises from the integration of sensory input with predictions about upcoming stimuli continuously generated by an internal model. More formally, the percept is determined by the posterior probability as computed by Bayes’ theorem from the predictions (prior) and the sensory input (likelihood of a given stimulus). Within this framework, measurements of brain activity are composed of the activity of two distinct neuronal populations – one population encoding the expected stimulus based on an internal model of the world (prediction) and one population encoding the mismatch between sensory input and the prediction (prediction error; PE) (Rao and Ballard, 1999; Friston, 2005).

A direct hypothesis derived from this framework is that sensory brain responses should be reduced when the brain’s prediction was accurate. In this situation, the resulting PE is small and regional brain activation is lower for accurate than for inaccurate predictions. This has been observed for primary visual cortex (Alink et al., 2010; Kok et al., 2012), early auditory electrophysiological responses (Todorovic et al., 2011), and the fusiform face area (Summerfield et al., 2008; den Ouden et al., 2010; Egner et al., 2010). The organization of cortical pain processing differs from other sensory modalities in that many cortical pain processing areas receive direct thalamic input and thus avoid a clear hierarchical organization (Craig, 2002; Dum et al., 2009). It is therefore unclear whether the same computational principles apply to pain as well. If pain processing is also based on predictive coding principles, this framework would offer an elegant and general computational mechanisms of perception across modalities (Wiech, 2016) and could help explain several expectation-related effects, including placebo effects (Petrovic et al., 2010; Büchel et al., 2014; Tabor et al., 2017).

In order to arbitrate between possible mechanisms underlying pain perception, we used a probabilistic heat pain task to formally compare a predictive coding model against a stimulus intensity model and a stimulus plus expectation model, respectively (Figure 1A–C). Three different visual cues manipulated expectations about an upcoming cutaneous heat stimulus (Figure 1D). Each cue was associated with a different probability of receiving painful or non-painful heat on the forearm (25, 50, or 75% chance of receiving pain and referred to as low, medium, and high cue, respectively; Figure 1E). Using functional magnetic resonance imaging (fMRI) in combination with model-based analyses in this task, we quantified evidence for all models in skin conductance responses (SCR), pupil diameter, and across the brain.

Figure 1

Download asset Open asset

According to the predictive coding model (Egner et al., 2010; Büchel et al., 2014), responses to cutaneous heat are the weighted sum of the prediction and the PE (Figure 1C). Under this framework, pain-processing regions will increase their activity with increasing probability of pain. Furthermore, they will also signal a PE if the stimulation is more painful than expected, but are not expected to show a PE for warm stimulation (Figure 1C). This pain PE is motivated by previous work (Egner et al., 2010; Büchel et al., 2014; Summerfield and de Lange, 2014), by the observation that PE for warm stimuli have topographies distinct from PE for pain (Ploghaus et al., 2000; Zeidan et al., 2015), and findings suggesting that reward and aversive PE are encoded by different neuronal populations (Yacubian et al., 2006; Belova et al., 2007; Seymour et al., 2007; Fiorillo, 2013). In addition to the pain PE, we later consider models using absolute and signed PEs, respectively (see Comparing different PE types).

Alternatively, pain-processing regions could simply encode stimulus intensity (Figure 1A) or an additive combination of intensity and expectation (Figure 1B). Each model makes different predictions about the measured response profiles within the present paradigm, which we tested both in an analysis of variance (ANOVA) framework and using formal Bayesian model comparison.

Behavioral results

Before comparing the different models against each other, we verified that the two stimuli were clearly distinguishable. Pain ratings obtained after each run showed that the 28 participants distinguished between the two stimulus intensities (t(27) = 20.9; p<0.001), that intensity ratings were close to the calibrated intensities of 30 and 75, respectively (mean warm: 29.0 ± 9.1 std.; mean pain: 75.0 ± 10.3 std.), and that warm stimuli were not perceived as painful (Figure 2A). Trial-by-trial ratings classifying stimuli as either painful or non-painful matched the stimulus intensity with 94.3% accuracy, further supporting the qualitative difference between the two stimuli. Target reaction times to color changes of the fixation dot did also not differ between two stimulus intensities (warm 520.3 ± 94 ms; pain 516.1 ± 104.8 ms; t(27)=0.51; p=0.61; Figure 2B), suggesting a similar attention allocation for both stimulation intensities.

Figure 2

Download asset Open asset

Region of interest results

We next computed ANOVA’s on brain activity extracted from anatomically defined a priori regions of interest (ROI). Among those ROIs, bilateral anterior insula (left: F(2,58)=5.5; p=0.007; right: F(2,58)= 7.5; p=0.001) and right amygdala (F(2,58)=5.4; p=0.007) showed the expected cue $\times$ stimulus interaction (Figure 3A, Table 1). Importantly, the response pattern matched the pattern expected by the predictive coding model, that is, responses in anterior insula and amygdala increased with pain expectation for warm stimuli and decreased for pain stimuli. Furthermore, all regions except for the postcentral gyrus, amygdala and PAG showed a significant main effect of stimulus (Table 1), but no ROI showed a significant main effect of cue.

Figure 3

Download asset Open asset

Table 1

		Stimulus		Cue		Cue X stimulus
Region		F(1,27)	P	F(2,54)	P	F(2,54)	P
ACC	L	13.93	0.0009*	1.21	0.3053	1.23	0.3017
	R	15.99	0.0004*	0.95	0.3923	1.11	0.3372
anterior insula	L	8.3	0.0077*	0.41	0.6651	5.46	0.0069*
	R	9.69	0.0043*	1.58	0.2155	7.48	0.0014*
posterior insula	L	15.73	0.0005*	0.28	0.7538	1.58	0.2145
	R	12.12	0.0017*	0.15	0.8637	0.09	0.9111
parietal operculum	L	18.3	0.0002*	1.2	0.3089	0.02	0.9779
	R	23.35	<0.0001*	0.17	0.8408	0.72	0.4918
post central gyrus	L	6.14	0.0198	1.1	0.3409	0.18	0.839
	R	2.57	0.1206	0.18	0.8387	0.41	0.6675
amygdala	L	0.1	0.7506	0.1	0.9046	4.94	0.0107
	R	0.83	0.369	0.51	0.6033	5.39	0.0074*
thalamus	L	8	0.0087*	0.39	0.6761	1.4	0.2545
	R	7.6	0.0104*	1.32	0.275	2.62	0.0823
PAG		0.02	0.9027	1.02	0.3675	4.34	0.0178
NPS		47.73	<0.0001*	0.14	0.8708	2.18	0.1228

1. ACC: anterior cingulate cortex, PAG: periaqueductal gray, NPS: neurological pain signature.

2. *FDR q<0.05.

Although the above ROIs are associated with pain processing, a recently developed multivariate pattern, termed neurological pain signature (NPS; Wager et al., 2013), provides a more specific and sensitive estimate of heat pain intensity (Wager et al., 2013; Krishnan et al., 2016). We thus computed an ANOVA on the pattern expression values as indicators of overall pain intensity for the NPS (Figure 3B). Stimulus intensity had an effect on NPS expression (F(1,27)=47.7; p<0.001), but neither cue nor the interaction were significant (both p>0.12; Table 1). Since NPS responses are strongly correlated with experimental heat pain reports (Wager et al., 2013; Krishnan et al., 2016), they can potentially serve as an indicator of trial-by-trial pain reports in this context to test for effects of correct predictions on pain reports. A post-hoc t-test revealed that unexpected pain tended to elicit stronger responses than expected pain (t(27)=2.2; p=0.036).

Voxel-wise statistical maps

In order to test for the proposed effects in brain regions outside of the a priori defined ROIs, we computed a whole brain analysis for the effects of stimulus (pain > warm), effects of cue (cue high > cue low), and for the interaction contrast ((cue high, warm) > (cue low, warm)) > ((cue high, pain) > (cue low, pain)). The stimulus intensity contrast revealed activations in classical pain processing areas, including posterior and mid-insula, secondary somatosensory cortex, parietal operculum, and midcingulate cortex (Figure 4A). A cue $\times$ stimulus interaction was observed again in left (peak MNI coordinates: x=−30, y = 24, z=-4) and right (x = 46, y = 20, z=−8) anterior insula (Figure 4B). No other brain region showed the interaction effect at a family wise error rate of p<0.05. Testing for the main effect of cue did not reveal any significant voxels.

Figure 4

Download asset Open asset

Formal model comparisons

After observing that the response profiles of SCR, pupil, bilateral anterior insula, and right amygdala were as expected by a concurrent representation of predictions and PE, we conducted formal model comparisons using Bayes factors (BF) (Kass and Raftery, 1995; Rouder and Morey, 2012) to identify the best explanatory model. Bayes factors are computed as the ratio of marginal likelihoods of the data under each of two models and thus quantify the evidence for one model over the other given the data. This metric thus allows the identification of the best model while implicitly controlling for the number of free parameters. Following Kass and Raftery (1995), we consider log-BF >3 as strong evidence for the predictive coding model and values of log-BF < −3 as strong evidence for the alternative model.

Comparing the predictive coding model against the stimulus intensity model revealed strong evidence in favor of the predictive coding model for both SCR and pupil responses (log-BF_SCR = 6.65; log-BF_pupil = 6.75; Figure 5A). Comparing it against the stimulus plus expectation model revealed similarly decisive evidence in favor of the predictive coding model (log-BF_SCR = 8.05; log-BF_pupil = 6.09; Figure 5B). These log-BF values indicate that the predictive coding model was at least 400 times more likely than each of the two alternatives.

Figure 5

Download asset Open asset

In contrast, NPS expression was better explained by the stimulus intensity model compared to the predictive coding model (log-BF = −3.98), mirroring the previously observed main effect of stimulus (Figure 5A). Computing log-BF’s for the individual ROIs confirmed the results of the ANOVA interaction tests in that the anterior insula showed strong evidence for the predictive coding model compared to the two alternative models (Figure 5). Generally, the right hemisphere yielded a clearer picture in terms of model evidence, potentially because of stronger signal in the hemisphere contralateral to the heat stimulation on the left arm. For example, evidence for the predictive coding model against the stimulus intensity model in the left anterior insula ROI was below threshold, while the evidence was above threshold for the right anterior insula (Figure 5A,B). Interestingly, the right parietal operculum ROI showed strong evidence for the stimulus intensity model (log-BF = −3.02; Figure 5A).

Comparing the stimulus intensity against the stimulus plus expectation model did not reveal decisive evidence for one over the other model on the ROI level. Although no comparison reached the threshold of |log-BF|>3, all ROIs and physiological measures weakly supported the simpler, stimulus intensity model (log-BF range: 0.12–2.16).

In order to obtain a spatially more detailed picture of the computational processes of pain processing across the brain, we computed voxel-wise log-BF’s comparing the predictive coding model against the stimulus intensity model and the stimulus plus expectation model, respectively. Again, responses in bilateral anterior insula strongly supported the predictive coding model (Figure 5C,D). Within the posterior insula and parietal operculum, this more fine-grained analyses revealed bilateral evidence for the simpler, stimulus-intensity model, which was less evident on the ROI level. Similar results were obtained when comparing the predictive coding model against the stimulus plus expectation model (Figure 5D), but evidence for the stimulus plus expectation model was weaker. Directly comparing the stimulus intensity model against the stimulus plus expectation model revealed modest support for the stimulus intensity model (log-BF >2) in midcingulate cortex, posterior insula, and parietal operculum.

A surface projection of the non-thresholded, voxel-wise log-BF maps comparing the predictive coding model against the two alternative models within the insula, demonstrated a gradual change in evidence from anterior to posterior insula (Figure 5C,D). This gradient is also evident when the average log-BF from the insula is plotted over the anterior-posterior dimension (Figure 6A). Importantly, and in line with anatomical considerations, the gradient is also steeper in the right hemisphere (contralateral to stimulation) based on the more decisive evidence in both anterior and posterior insula.

Figure 6

Download asset Open asset

Since we expect both predictions and PEs to contribute positively to the measured brain signal, we extracted the weight parameters for predictions and PEs from the left and right anterior insula regions in which log-BF >3. Firstly, the weights for both predictions and PEs were positive as postulated by the model. Interestingly, the weight for the PE was approximately two-times as strong as the prediction (left anterior insula: 1:2.1; right: 1:2.2), which is very similar to the ratio of 1:2 reported in a previous study in the fusiform face area (Egner et al., 2010). A stronger weighting of the PE results in reduced responses for expected compared to unexpected painful stimuli as illustrated in Figure 1C and observed here in SCR (Figure 2C), pupil dilation (Figure 2D), and anterior insula and amygdala activation (Figure 3A) as well as in other studies (Alink et al., 2010; Todorovic et al., 2011).

Comparing different PE types

While neuronal coding of reward prediction errors is well understood (Schultz et al., 2015), the specifics of aversive PE coding are currently debated (Belova et al., 2007; Seymour et al., 2007; Boll et al., 2013; Fiorillo, 2013; Klavir et al., 2013; Roy et al., 2014; Matsumoto et al., 2016). We therefore compared three versions of the predictive coding model that differed in their PE specification. The original model, presented above, builds on a pain PE in which a warm stimulus does not elicit a PE within pain regions (Figures 1A and 7A). Alternative models incorporated an absolute PE (i.e., the absolute difference between stimulus and prediction) and a signed PE (i.e., the difference between stimulus intensity and prediction), respectively (Figure 7A). All models share the same prediction term.

Figure 7

Download asset Open asset

The original, pain PE model provided a better fit of the pupil, NPS, and ROI responses when compared against the absolute PE model (Figure 7B). Only the PAG and bilateral amygdala tended to favor the absolute PE model, but the model evidence did not pass the threshold of log-BF < −3. The pain PE model also provided a better fit than the signed PE model for SCR and pupil diameter (Figure 7C). For the ROIs, the pain PE also provided better fits, while the signed PE tended to fit the NPS response slightly better. But again, none of the ROI comparisons revealed above threshold evidence. In summary, neither the absolute nor the signed PE model provided compelling evidence for a better fit than the pain PE model. In fact, the pain model explained the responses better than the alternatives in most of the ROIs.

Combining a probabilistic heat pain paradigm and Bayesian model comparison, we identified a functional dissociation between anterior and posterior insula suggesting that these regions implement different computations supporting pain perception. While posterior insula and parietal operculum employ stimulus intensity coding, activity in the anterior insula reflects the summation of pain expectation and prediction errors resulting from unexpected pain, thus conforming to a predictive coding account (Egner et al., 2010; Büchel et al., 2014). This functional dissociation was evident at a ROI level as well as on a voxel-wise analysis. The combination of expectation and prediction error was also by far the best model explaining SCR and pupil diameter responses to painful and non-painful heat stimuli. By contrast, the response profile of a multivariate brain pattern predictive of pain ratings (NPS; Wager et al., 2013) reflected only stimulus intensity.

The insula as a whole is involved in a multitude of different processes – at least 20 according to one review (Nieuwenhuys, 2012) – from somatosensory to emotional and conflict processing (Kurth et al., 2010; Chang et al., 2013; Wiech et al., 2014a). Interestingly, the cytoarchitectonic organization of the insula and its functional connectivity with other brain regions vary smoothly along an anterior-posterior gradient (Cerliani et al., 2012; Nieuwenhuys, 2012). A similar functional gradient was evident in the present study, thereby linking the underlying cytoarchitectonic and connectivity gradients to distinct pain processing modes.

The posterior insula is strongly involved in somatosensory and pain perception (Mazzola et al., 2012; Nieuwenhuys, 2012), it receives direct spinothalamic input (Craig, 2002; Dum et al., 2009), and it is functionally and structurally connected to somatosensory cortices (Wiech et al., 2014a). This connectivity pattern is well suited to support a stimulus intensity coding role of the posterior insula as observed here using phasic pain stimuli and as also indicated by a more tonic model of pain (Segerdahl et al., 2015). It should be noted that the sensory role of the posterior insula includes non-painful stimuli, too (Davis et al., 1998).

By contrast anterior insula activity represents a combination of pain prediction and PE. Its predictive function fits nicely with previous reports that pre-stimulus activity in the anterior insula modulates the perception of subsequent stimuli (Ploner et al., 2010; Wiech et al., 2010). Furthermore, the anterior insula flexibly connects to emotional and attentional brain regions (Taylor et al., 2009; Ploner et al., 2010) and integrates information from a diverse set of prefrontal and limbic brain regions (Critchley, 2005; Seminowicz and Davis, 2007; Kurth et al., 2010; Cerliani et al., 2012; Wiech et al., 2014a). In addition to its predictive function, the anterior insula also encodes mismatches between predictions and aversive outcomes during reinforcement learning (Seymour et al., 2004; Pessiglione et al., 2006; Boll et al., 2013) and other concurrent task demands (Seminowicz and Davis, 2007). The strong connections between anterior insula and prefrontal regions involved in contextual processing as well as its ‘hub-like’, evaluative function (Baliki et al., 2009; Chang et al., 2013; Uddin et al., 2014) render the anterior insula particularly suitable for the evaluation of predictions against sensory input. The anterior insula could thus represent a mediator between somatosensory signals in posterior insula and contextual representations in prefrontral cortex, integrating those representations for perceptual decisions and behavioral responses (Kong et al., 2006; Seminowicz and Davis, 2007; Baliki et al., 2009). Integration of multiple information streams in this brain region could thus be crucial for the construction of pain experiences that are shaped by learning and external feedback (Wiech, 2016; Geuter et al., 2017).

In addition to the anterior insula, both SCR and pupil diameter in this study – and reaction times in a similar paradigm (Wiech et al., 2014b) – showed a pattern predicted by our model. Expectations of certain visual stimuli can also sharpen their cortical representation (Kok et al., 2012), but it is unknown how this would translate to pain reports — whether predicted pain would be more or less intense. A recent study did observe no difference in pain ratings between correctly and incorrectly cued pain stimuli (Zeidan et al., 2015). Here, we opted for a time-efficient design and did not collect trial-by-trial pain ratings to address this question. However, investigating potential indicators for perceived pain, responses of the NPS (Wager et al., 2013; Krishnan et al., 2016) and autonomic measures (Geuter et al., 2014), can offer insights. All three measures – NPS, SCR, and pupil – showed stronger responses to unexpected compared to expected pain, hinting at a potential enhancement of unexpected pain stimuli that needs to be investigated more thoroughly in futures studies. Taken together, the results show that principles of predictive coding are relevant for behavioral responses in the context of pain.

Other stimulus attributes than painfulness, e.g. general aversiveness, salience, or motivational demands, co-vary with stimulus intensity. Previous studies have correlated brain activity with pain reports and stimulus intensity in order to dissociate the two (Coghill et al., 1999; Büchel et al., 2002; Davis et al., 2002, 2004; Porro et al., 2004; Baliki et al., 2009). Interestingly, anterior insula activity correlated with perceived heat even in the absence of heat stimulation (Davis et al., 2004). Studies by Downar and colleagues (Downar et al., 2000, 2003) also found that anterior insula and anterior cingulate cortex responded to unexpected or novel stimuli. Within a predictive coding framework, the overall response is decomposed into two distinct functional components – a prediction term and PE term – that are key components of learning theory. Interestingly, the anterior insula also showed a prediction error response in the present study in line with previous work (Downar et al., 2000, 2002; Seymour et al., 2004; Boll et al., 2013). From a psychological perspective, the decomposition is important because directing attention towards expected aversive events (high probability of pain, prediction term) and towards unexpected events (PE) is adaptive. The prediction error is assumed to drive learning (Rescorla and Wagner, 1972) and is thus critical for adaption to the environment. By contrast, salience – understood as the difference to preceding sensory events (Mouraux et al., 2011) – emerges after stimuli have been processed and the elicited surprise or PE has been computed. Salience can thus be considered a secondary stimulus property resulting from a high PE that in turn can modulate subsequent updating of beliefs. This process is formalized in the Pearce-Hall model of reinforcement learning in which a surprising, salient outcome, affects the learning rate in the next trial (Pearce and Hall, 1980; Boll et al., 2013; Atlas et al., 2016).

Predictive coding theories offer a parsimonious computational implementation of cross-modal, Bayesian perceptual decision making (Knill and Pouget, 2004; Friston, 2005; Summerfield and de Lange, 2014). These accounts can explain several effects within a single framework including extra-classical receptive field effects in visual cortex (Rao and Ballard, 1999), repetition suppression (Summerfield et al., 2008; Todorovic et al., 2011), and have been suggested as a framework to understand placebo effects (Büchel et al., 2014). In support of a domain general integration process of expectations and PEs, the ratio of the contributions of both processes to physiological signals observed here, mirrored the ratio previously reported for the fusiform face area in visual perception (Egner et al., 2010). Interestingly, in both studies, the PE was weighted stronger than the prediction, which suggests that learning and updating of the internal model are crucial for perception. In addition, the observation that the anterior insula also processes PEs in other modalities (Downar et al., 2000; Iglesias et al., 2013) hints towards a cross-modal role of the anterior insula within a predictive coding framework.

Another feature of predictive coding models is their hierarchical organization: At each level of the neural hierarchy, predictions and PEs will be computed for the specific features encoded in this region (Rao and Ballard, 1999; Friston, 2005). For example, early visual and auditory areas process multimodal stimuli under the assumption of independent physical sources and only higher areas form joint representations using Bayesian inference (Rohe and Noppeney, 2015). Although anterior insula activity matched the response pattern proposed by the predictive coding model, other brain regions and the NPS followed a stimulus intensity model. This discrepancy could be either due to inherently distinct computations implemented in those regions or due to the fact that the nature of predictions and PEs changes across regions (Iglesias et al., 2013). Because pain is an inherently multi-faceted experience that includes sensory-discriminative, emotional, and motivational components, the present predictive coding model could capture certain aspects of this multi-faceted experience better than others. The computational difference observed between anterior and posterior insula could, at least in part, reflect such functional differences. In fact, studies investigating visual processing within a predictive coding framework also observed regionally restricted effects based on the manipulated stimulus features. For example, activity of the fusiform face area and parahippocampal place area is well described by a predictive coding model, but each region responds selectively to their respective preferred stimuli, i.e. faces and houses (den Ouden et al., 2010; Egner et al., 2010; Jiang et al., 2013). Similarly, expectations of certain low-level visual features such as grating orientation, selectively attenuate primary visual cortex activity, but not activity in higher visual areas (Alink et al., 2010; Kok et al., 2012). Based on these results, the observed computational differentiation between anterior and posterior insula indicates that both regions process distinct features of painful stimuli and these could be related to different psychological and behavioral outcomes in healthy and patient populations.

The representation of aversive prediction errors in the brain is still not fully understood. Important open questions include whether aversive PE are represented on a continuous dimension along with reward prediction errors and whether particular brain regions represent absolute, signed, or pain PE. Activity reflecting absolute aversive PE has been observed in the amygdala (Boll et al., 2013; McHugh et al., 2014), while signed aversive PE have been observed in the striatum and PAG (Seymour et al., 2005, 2007; Roy et al., 2014; Zhang et al., 2016). Within sensory cortices, unexpected omissions of visual or auditory stimuli lead to enhanced activity in auditory or visual areas, reminiscent of absolute prediction errors (den Ouden et al., 2009, 2010; Todorovic et al., 2011; Todorovic and de Lange, 2012). Comparing models incorporating different PE specifications, showed that the model based on an asymmetric, pain PE explained brain responses in the present study best. Our results thus suggest that PE encoding in the anterior insula differs between situations when the outcome is more pain than expected compared to unexpected pain omissions. The differentiation of PEs observed here is similar to a distinction observed in visual processing: in two studies, activity in the fusiform face area, a face-selective brain region, reflected prediction errors for face stimuli, but not for house stimuli (den Ouden et al., 2010; Egner et al., 2010).

In summary, the observed responses in SCR, pupil dilation, and anterior insula activation demonstrate that at least part of the pain experience can be explained by a domain general predictive coding framework. The parallels observed between pain and visual processing (Egner et al., 2010) hint towards a general processing principle based on internal predictions and PE. An interesting question for future research is how the contributions of predictions and PE shift in states of altered or chronic pain conditions that are also related to altered learning processes (Vlaeyen, 2015). As anterior insula structure and function changes profoundly in chronic pain conditions (Bushnell et al., 2013; Ceko et al., 2013; Hong et al., 2014; Flodin et al., 2015), it is possible that the precision or influence of the prediction is strongly enhanced in chronic pain conditions or that PEs are incorrectly computed (Edwards et al., 2012). If the underlying computations are domain general, this would also explain the hyper-sensitivity observed in certain chronic pain populations to non-painful tactile and visual stimuli (López-Solà et al., 2014). This framework could hence open up new ways to investigate pain processing in clinical populations.

1. 1. Alink A
   2. Schwiedrzik CM
   3. Kohler A
   4. Singer W
   5. Muckli L

   (2010) Stimulus predictability reduces responses in primary visual cortex

   Journal of Neuroscience 30:2960–2966.

   https://doi.org/10.1523/JNEUROSCI.3730-10.2010

   • PubMed
   • Google Scholar
2. 1. Apkarian AV
   2. Krauss BR
   3. Fredrickson BE
   4. Szeverenyi NM

   (2001) Imaging the pain of low back pain: functional magnetic resonance imaging in combination with monitoring subjective pain perception allows the study of clinical pain states

   Neuroscience Letters 299:57–60.

   https://doi.org/10.1016/S0304-3940(01)01504-X

   • PubMed
   • Google Scholar
3. 1. Ashburner J

   (2007) A fast diffeomorphic image registration algorithm

   NeuroImage 38:95–113.

   https://doi.org/10.1016/j.neuroimage.2007.07.007

   • PubMed
   • Google Scholar
4. 1. Atlas LY
   2. Bolger N
   3. Lindquist MA
   4. Wager TD

   (2010) Brain mediators of predictive cue effects on perceived pain

   Journal of Neuroscience 30:12964–12977.

   https://doi.org/10.1523/JNEUROSCI.0057-10.2010

   • PubMed
   • Google Scholar
5. 1. Atlas LY
   2. Doll BB
   3. Li J
   4. Daw ND
   5. Phelps EA

   (2016) Instructed knowledge shapes feedback-driven aversive learning in striatum and orbitofrontal cortex, but not the amygdala

   eLife 5:e15192.

   https://doi.org/10.7554/eLife.15192

   • PubMed
   • Google Scholar
6. 1. Baliki MN
   2. Geha PY
   3. Apkarian AV

   (2009) Parsing pain perception between nociceptive representation and magnitude estimation

   Journal of Neurophysiology 101:875–887.

   https://doi.org/10.1152/jn.91100.2008

   • PubMed
   • Google Scholar
7. 1. Bantick SJ
   2. Wise RG
   3. Ploghaus A
   4. Clare S
   5. Smith SM
   6. Tracey I

   (2002) Imaging how attention modulates pain in humans using Functional MRI

   Brain 125:310–319.

   https://doi.org/10.1093/brain/awf022

   • PubMed
   • Google Scholar
8. 1. Belova MA
   2. Paton JJ
   3. Morrison SE
   4. Salzman CD

   (2007) Expectation modulates neural responses to pleasant and aversive stimuli in primate amygdala

   Neuron 55:970–984.

   https://doi.org/10.1016/j.neuron.2007.08.004

   • PubMed
   • Google Scholar
9. 1. Bingel U
   2. Wanigasekera V
   3. Wiech K
   4. Ni Mhuircheartaigh R
   5. Lee MC
   6. Ploner M
   7. Tracey I

   (2011) The effect of treatment expectation on drug efficacy: imaging the analgesic benefit of the opioid remifentanil

   Science Translational Medicine 3:70ra14.

   https://doi.org/10.1126/scitranslmed.3001244

   • PubMed
   • Google Scholar
10. 1. Boll S
    2. Gamer M
    3. Gluth S
    4. Finsterbusch J
    5. Büchel C

    (2013) Separate amygdala subregions signal surprise and predictiveness during associative fear learning in humans

    European Journal of Neuroscience 37:758–767.

    https://doi.org/10.1111/ejn.12094

    • 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. Boucsein W
    2. Fowles DC
    3. Grimnes S
    4. Ben-Shakhar G
    5. roth WT
    6. Dawson ME
    7. Filion DL
    8. Society for Psychophysiological Research Ad Hoc Committee on Electrodermal Measures

    (2012) Publication recommendations for electrodermal measurements

    Psychophysiology 49:1017–1034.

    https://doi.org/10.1111/j.1469-8986.2012.01384.x

    • PubMed
    • Google Scholar
13. 1. Brown CA
    2. Seymour B
    3. Boyle Y
    4. El-Deredy W
    5. Jones AK

    (2008) Modulation of pain ratings by expectation and uncertainty: behavioral characteristics and anticipatory neural correlates

    Pain 135:240–250.

    https://doi.org/10.1016/j.pain.2007.05.022

    • PubMed
    • Google Scholar
14. 1. Bushnell MC
    2. Ceko M
    3. Low LA

    (2013) Cognitive and emotional control of pain and its disruption in chronic pain

    Nature Reviews Neuroscience 14:502–511.

    https://doi.org/10.1038/nrn3516

    • PubMed
    • Google Scholar
15. 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

    Journal of Neuroscience 22:970–976.

    • PubMed
    • Google Scholar
16. 1. Büchel C
    2. Geuter S
    3. Sprenger C
    4. Eippert F

    (2014) Placebo analgesia: a predictive coding perspective

    Neuron 81:1223–1239.

    https://doi.org/10.1016/j.neuron.2014.02.042

    • PubMed
    • Google Scholar
17. 1. Ceko M
    2. Bushnell MC
    3. Fitzcharles MA
    4. Schweinhardt P

    (2013) Fibromyalgia interacts with age to change the brain

    NeuroImage: Clinical 3:249–260.

    https://doi.org/10.1016/j.nicl.2013.08.015

    • PubMed
    • Google Scholar
18. 1. Cerliani L
    2. Thomas RM
    3. Jbabdi S
    4. Siero JC
    5. Nanetti L
    6. Crippa A
    7. Gazzola V
    8. D'Arceuil H
    9. Keysers C

    (2012) Probabilistic tractography recovers a rostrocaudal trajectory of connectivity variability in the human insular cortex

    Human Brain Mapping 33:2005–2034.

    https://doi.org/10.1002/hbm.21338

    • PubMed
    • Google Scholar
19. 1. Chang LJ
    2. Yarkoni T
    3. Khaw MW
    4. Sanfey AG

    (2013) Decoding the role of the insula in human cognition: functional parcellation and large-scale reverse inference

    Cerebral Cortex 23:739–749.

    https://doi.org/10.1093/cercor/bhs065

    • PubMed
    • Google Scholar
20. 1. Coghill RC
    2. Sang CN
    3. Maisog JM
    4. Iadarola MJ

    (1999)

    Pain intensity processing within the human brain: a bilateral, distributed mechanism

    Journal of Neurophysiology 82:1934–1943.

    • PubMed
    • Google Scholar
21. 1. Craig AD

    (2002) How do you feel? Interoception: the sense of the physiological condition of the body

    Nature Reviews Neuroscience 3:655–666.

    https://doi.org/10.1038/nrn894

    • PubMed
    • Google Scholar
22. 1. Critchley HD

    (2005) Neural mechanisms of autonomic, affective, and cognitive integration

    The Journal of Comparative Neurology 493:154–166.

    https://doi.org/10.1002/cne.20749

    • PubMed
    • Google Scholar
23. 1. Davis KD
    2. Kwan CL
    3. Crawley AP
    4. Mikulis DJ

    (1998)

    Functional MRI study of thalamic and cortical activations evoked by cutaneous heat, cold, and tactile stimuli

    Journal of Neurophysiology 80:1533–1546.

    • PubMed
    • Google Scholar
24. 1. Davis KD
    2. Pope GE
    3. Crawley AP
    4. Mikulis DJ

    (2002) Neural correlates of prickle sensation: a percept-related fMRI study

    Nature Neuroscience 5:1121–1122.

    https://doi.org/10.1038/nn955

    • PubMed
    • Google Scholar
25. 1. Davis KD
    2. Pope GE
    3. Crawley AP
    4. Mikulis DJ

    (2004) Perceptual illusion of "paradoxical heat" engages the insular cortex

    Journal of Neurophysiology 92:1248–1251.

    https://doi.org/10.1152/jn.00084.2004

    • PubMed
    • Google Scholar
26. 1. den Ouden HE
    2. Daunizeau J
    3. Roiser J
    4. Friston KJ
    5. Stephan KE

    (2010) Striatal prediction error modulates cortical coupling

    Journal of Neuroscience 30:3210–3219.

    https://doi.org/10.1523/JNEUROSCI.4458-09.2010

    • PubMed
    • Google Scholar
27. 1. den Ouden HE
    2. Friston KJ
    3. Daw ND
    4. McIntosh AR
    5. Stephan KE

    (2009) A dual role for prediction error in associative learning

    Cerebral Cortex 19:1175–1185.

    https://doi.org/10.1093/cercor/bhn161

    • PubMed
    • Google Scholar
28. 1. Desikan RS
    2. Ségonne F
    3. Fischl B
    4. Quinn BT
    5. Dickerson BC
    6. Blacker D
    7. Buckner RL
    8. Dale AM
    9. Maguire RP
    10. Hyman BT
    11. Albert MS
    12. Killiany RJ

    (2006) An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest

    NeuroImage 31:968–980.

    https://doi.org/10.1016/j.neuroimage.2006.01.021

    • PubMed
    • Google Scholar
29. 1. Downar J
    2. Crawley AP
    3. Mikulis DJ
    4. Davis KD

    (2000) A multimodal cortical network for the detection of changes in the sensory environment

    Nature Neuroscience 3:277–283.

    https://doi.org/10.1038/72991

    • PubMed
    • Google Scholar
30. 1. Downar J
    2. Crawley AP
    3. Mikulis DJ
    4. Davis KD

    (2002) A cortical network sensitive to stimulus salience in a neutral behavioral context across multiple sensory modalities

    Journal of Neurophysiology 87:615–620.

    https://doi.org/10.1152/jn.00636.2001

    • PubMed
    • Google Scholar
31. 1. Downar J
    2. Mikulis DJ
    3. Davis KD

    (2003) Neural correlates of the prolonged salience of painful stimulation

    NeuroImage 20:1540–1551.

    https://doi.org/10.1016/S1053-8119(03)00407-5

    • PubMed
    • Google Scholar
32. 1. Dum RP
    2. Levinthal DJ
    3. Strick PL

    (2009) The spinothalamic system targets motor and sensory areas in the cerebral cortex of monkeys

    Journal of Neuroscience 29:14223–14235.

    https://doi.org/10.1523/JNEUROSCI.3398-09.2009

    • PubMed
    • Google Scholar
33. 1. Edwards MJ
    2. Adams RA
    3. Brown H
    4. Pareés I
    5. Friston KJ

    (2012) A bayesian account of 'hysteria'

    Brain 135:3495–3512.

    https://doi.org/10.1093/brain/aws129

    • PubMed
    • Google Scholar
34. 1. Egner T
    2. Monti JM
    3. Summerfield C

    (2010) Expectation and surprise determine neural population responses in the ventral visual stream

    Journal of Neuroscience 30:16601–16608.

    https://doi.org/10.1523/JNEUROSCI.2770-10.2010

    • PubMed
    • Google Scholar
35. 1. Einhäuser W
    2. Stout J
    3. Koch C
    4. Carter O

    (2008) Pupil dilation reflects perceptual selection and predicts subsequent stability in perceptual rivalry

    PNAS 105:1704–1709.

    https://doi.org/10.1073/pnas.0707727105

    • PubMed
    • Google Scholar
36. 1. Eklund A
    2. Nichols TE
    3. Knutsson H

    (2016) Cluster failure: why fMRI inferences for spatial extent have inflated false-positive rates

    PNAS 113:7900–7905.

    https://doi.org/10.1073/pnas.1602413113

    • PubMed
    • Google Scholar
37. 1. Fiorillo CD

    (2013) Two dimensions of value: dopamine neurons represent reward but not aversiveness

    Science 341:546–549.

    https://doi.org/10.1126/science.1238699

    • PubMed
    • Google Scholar
38. 1. Flodin P
    2. Martinsen S
    3. Mannerkorpi K
    4. Löfgren M
    5. Bileviciute-Ljungar I
    6. Kosek E
    7. Fransson P

    (2015) Normalization of aberrant resting state functional connectivity in Fibromyalgia patients following a three month physical exercise therapy

    NeuroImage: Clinical 9:134–139.

    https://doi.org/10.1016/j.nicl.2015.08.004

    • PubMed
    • Google Scholar
39. 1. Friston K

    (2005) A theory of cortical responses

    Philosophical Transactions of the Royal Society B: Biological Sciences 360:815–836.

    https://doi.org/10.1098/rstb.2005.1622

    • PubMed
    • Google Scholar
40. 1. Geuter S
    2. Gamer M
    3. Onat S
    4. Büchel C

    (2014) Parametric trial-by-trial prediction of pain by easily available physiological measures

    Pain 155:994–1001.

    https://doi.org/10.1016/j.pain.2014.02.005

    • PubMed
    • Google Scholar
41. 1. Geuter S
    2. Koban L
    3. Wager TD

    (2017) The cognitive neuroscience of Placebo Effects: concepts, predictions, and Physiology

    Annual Review of Neuroscience 40:167–188.

    https://doi.org/10.1146/annurev-neuro-072116-031132

    • PubMed
    • Google Scholar
42. 1. Holmes AP
    2. Blair RC
    3. Watson JD
    4. Ford I

    (1996) Nonparametric analysis of statistic images from functional mapping experiments

    Journal of Cerebral Blood Flow and Metabolism : Official Journal of the International Society of Cerebral Blood Flow and Metabolism 16:7–22.

    https://doi.org/10.1097/00004647-199601000-00002

    • PubMed
    • Google Scholar
43. 1. Hong JY
    2. Kilpatrick LA
    3. Labus JS
    4. Gupta A
    5. Katibian D
    6. Ashe-McNalley C
    7. Stains J
    8. Heendeniya N
    9. Smith SR
    10. Tillisch K
    11. Naliboff B
    12. Mayer EA

    (2014) Sex and disease-related alterations of anterior insula functional connectivity in chronic abdominal pain

    Journal of Neuroscience 34:14252–14259.

    https://doi.org/10.1523/JNEUROSCI.1683-14.2014

    • PubMed
    • Google Scholar
44. 1. Hubel DH
    2. Wiesel TN

    (1959) Receptive fields of single neurones in the cat's striate cortex

    The Journal of Physiology 148:574–591.

    https://doi.org/10.1113/jphysiol.1959.sp006308

    • PubMed
    • Google Scholar
45. 1. Iglesias S
    2. Mathys C
    3. Brodersen KH
    4. Kasper L
    5. Piccirelli M
    6. den Ouden HE
    7. Stephan KE

    (2013) Hierarchical prediction errors in midbrain and basal forebrain during sensory learning

    Neuron 80:519–530.

    https://doi.org/10.1016/j.neuron.2013.09.009

    • PubMed
    • Google Scholar
46. Book

    1. Jeffreys H

    (1961)

    Theory of Probability

    Oxford, UK: Oxford University Press.

    • Google Scholar
47. 1. Jiang J
    2. Summerfield C
    3. Egner T

    (2013) Attention sharpens the distinction between expected and unexpected percepts in the visual brain

    Journal of Neuroscience 33:18438–18447.

    https://doi.org/10.1523/JNEUROSCI.3308-13.2013

    • PubMed
    • Google Scholar
48. 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
49. 1. Kietzmann TC
    2. Geuter S
    3. König P

    (2011) Overt visual attention as a causal factor of perceptual awareness

    PLoS One 6:e22614.

    https://doi.org/10.1371/journal.pone.0022614

    • PubMed
    • Google Scholar
50. 1. Klavir O
    2. Genud-Gabai R
    3. Paz R

    (2013) Functional connectivity between amygdala and cingulate cortex for adaptive aversive learning

    Neuron 80:1290–1300.

    https://doi.org/10.1016/j.neuron.2013.09.035

    • PubMed
    • Google Scholar
51. 1. Knill DC
    2. Pouget A

    (2004) The bayesian brain: the role of uncertainty in neural coding and computation

    Trends in Neurosciences 27:712–719.

    https://doi.org/10.1016/j.tins.2004.10.007

    • PubMed
    • Google Scholar
52. 1. Kok P
    2. Failing MF
    3. de Lange FP

    (2014) Prior expectations evoke stimulus templates in the primary visual cortex

    Journal of Cognitive Neuroscience 26:1546–1554.

    https://doi.org/10.1162/jocn_a_00562

    • PubMed
    • Google Scholar
53. 1. Kok P
    2. Jehee JF
    3. de Lange FP

    (2012) Less is more: expectation sharpens representations in the primary visual cortex

    Neuron 75:265–270.

    https://doi.org/10.1016/j.neuron.2012.04.034

    • PubMed
    • Google Scholar
54. 1. Kong J
    2. White NS
    3. Kwong KK
    4. Vangel MG
    5. Rosman IS
    6. Gracely RH
    7. Gollub RL

    (2006) Using fMRI to dissociate sensory encoding from cognitive evaluation of heat pain intensity

    Human Brain Mapping 27:715–721.

    https://doi.org/10.1002/hbm.20213

    • PubMed
    • Google Scholar
55. 1. Koyama T
    2. McHaffie JG
    3. Laurienti PJ
    4. Coghill RC

    (2005) The subjective experience of pain: where expectations become reality

    PNAS 102:12950–12955.

    https://doi.org/10.1073/pnas.0408576102

    • PubMed
    • Google Scholar
56. 1. Krishnan A
    2. Woo CW
    3. Chang LJ
    4. Ruzic L
    5. Gu X
    6. López-Solà M
    7. Jackson PL
    8. Pujol J
    9. Fan J
    10. Wager TD

    (2016) Somatic and vicarious pain are represented by dissociable multivariate brain patterns

    eLife 5:e15166.

    https://doi.org/10.7554/eLife.15166

    • PubMed
    • Google Scholar
57. 1. Kurth F
    2. Zilles K
    3. Fox PT
    4. Laird AR
    5. Eickhoff SB

    (2010) A link between the systems: functional differentiation and integration within the human insula revealed by meta-analysis

    Brain Structure and Function 214:519–534.

    https://doi.org/10.1007/s00429-010-0255-z

    • PubMed
    • Google Scholar
58. Book

    1. Lee MD
    2. Wagenmakers E-J

    (2014)

    Bayesian Cognitive Modeling: A Practical Course

    Cambridge: Cambridge University Press.

    • Google Scholar
59. 1. Lodewyckx T
    2. Kim W
    3. Lee MD
    4. Tuerlinckx F
    5. Kuppens P
    6. Wagenmakers E-J

    (2011) A tutorial on Bayes factor estimation with the product space method

    Journal of Mathematical Psychology 55:331–347.

    https://doi.org/10.1016/j.jmp.2011.06.001

    • Google Scholar
60. 1. Lorenz J
    2. Hauck M
    3. Paur RC
    4. Nakamura Y
    5. Zimmermann R
    6. Bromm B
    7. Engel AK

    (2005) Cortical correlates of false expectations during pain intensity judgments--a possible manifestation of placebo/nocebo cognitions

    Brain, Behavior, and Immunity 19:283–295.

    https://doi.org/10.1016/j.bbi.2005.03.010

    • PubMed
    • Google Scholar
61. 1. López-Solà M
    2. Pujol J
    3. Wager TD
    4. Garcia-Fontanals A
    5. Blanco-Hinojo L
    6. Garcia-Blanco S
    7. Poca-Dias V
    8. Harrison BJ
    9. Contreras-Rodríguez O
    10. Monfort J
    11. Garcia-Fructuoso F
    12. Deus J

    (2014) Altered functional magnetic resonance imaging responses to nonpainful sensory stimulation in Fibromyalgia patients

    Arthritis & Rheumatology 66:3200–3209.

    https://doi.org/10.1002/art.38781

    • PubMed
    • Google Scholar
62. 1. Matsumoto H
    2. Tian J
    3. Uchida N
    4. Watabe-Uchida M

    (2016) Midbrain dopamine neurons signal aversion in a reward-context-dependent manner

    eLife 5:e17328.

    https://doi.org/10.7554/eLife.17328

    • PubMed
    • Google Scholar
63. 1. Mazzola L
    2. Isnard J
    3. Peyron R
    4. Mauguière F

    (2012) Stimulation of the human cortex and the experience of pain: wilder Penfield's observations revisited

    Brain 135:631–640.

    https://doi.org/10.1093/brain/awr265

    • PubMed
    • Google Scholar
64. 1. McHugh SB
    2. Barkus C
    3. Huber A
    4. Capitão L
    5. Lima J
    6. Lowry JP
    7. Bannerman DM

    (2014) Aversive prediction error signals in the amygdala

    Journal of Neuroscience 34:9024–9033.

    https://doi.org/10.1523/JNEUROSCI.4465-13.2014

    • PubMed
    • Google Scholar
65. 1. Mouraux A
    2. Diukova A
    3. Lee MC
    4. Wise RG
    5. Iannetti GD

    (2011) A multisensory investigation of the functional significance of the "pain matrix"

    NeuroImage 54:2237–2249.

    https://doi.org/10.1016/j.neuroimage.2010.09.084

    • PubMed
    • Google Scholar
66. Book

    1. Naidich T
    2. Duvernoy H
    3. Delman B
    4. Sorensen AG
    5. Kollias S
    6. Haake E

    (2009)

    Duvernoy’s Atlas of the Human Brain Stem and Cerebellum

    New York: Springer.

    • Google Scholar
67. 1. Nichols TE
    2. Holmes AP

    (2002) Nonparametric permutation tests for functional neuroimaging: a primer with examples

    Human Brain Mapping 15:1–25.

    https://doi.org/10.1002/hbm.1058

    • PubMed
    • Google Scholar
68. 1. Nieuwenhuys R

    (2012) The insular cortex: a review

    Progress in Brain Research 195:123–163.

    https://doi.org/10.1016/B978-0-444-53860-4.00007-6

    • PubMed
    • Google Scholar
69. 1. Pearce JM
    2. Hall G

    (1980) A model for pavlovian learning: variations in the effectiveness of conditioned but not of unconditioned stimuli

    Psychological Review 87:532.

    https://doi.org/10.1037/0033-295X.87.6.532

    • PubMed
    • Google Scholar
70. 1. Pessiglione M
    2. Seymour B
    3. Flandin G
    4. Dolan RJ
    5. Frith CD

    (2006) Dopamine-dependent prediction errors underpin reward-seeking behaviour in humans

    Nature 442:1042–1045.

    https://doi.org/10.1038/nature05051

    • PubMed
    • Google Scholar
71. 1. Petrovic P
    2. Kalso E
    3. Petersson KM
    4. Andersson J
    5. Fransson P
    6. Ingvar M

    (2010) A prefrontal non-opioid mechanism in placebo analgesia

    Pain 150:59–65.

    https://doi.org/10.1016/j.pain.2010.03.011

    • PubMed
    • Google Scholar
72. 1. Ploghaus A
    2. Tracey I
    3. Clare S
    4. Gati JS
    5. Rawlins JN
    6. Matthews PM

    (2000) Learning about pain: the neural substrate of the prediction error for aversive events

    PNAS 97:9281–9286.

    https://doi.org/10.1073/pnas.160266497

    • PubMed
    • Google Scholar
73. 1. Ploner M
    2. Lee MC
    3. Wiech K
    4. Bingel U
    5. Tracey I

    (2010) Prestimulus functional connectivity determines pain perception in humans

    PNAS 107:355–360.

    https://doi.org/10.1073/pnas.0906186106

    • PubMed
    • Google Scholar
74. 1. Ploner M
    2. Lee MC
    3. Wiech K
    4. Bingel U
    5. Tracey I

    (2011) Flexible cerebral connectivity patterns subserve contextual modulations of pain

    Cerebral Cortex 21:719–726.

    https://doi.org/10.1093/cercor/bhq146

    • PubMed
    • Google Scholar
75. 1. Porro CA
    2. Lui F
    3. Facchin P
    4. Maieron M
    5. Baraldi P

    (2004) Percept-related activity in the human somatosensory system: functional magnetic resonance imaging studies

    Magnetic Resonance Imaging 22:1539–1548.

    https://doi.org/10.1016/j.mri.2004.10.003

    • PubMed
    • Google Scholar
76. 1. Rao RP
    2. Ballard DH

    (1999) Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects

    Nature Neuroscience 2:79–87.

    https://doi.org/10.1038/4580

    • PubMed
    • Google Scholar
77. Book

    1. Rescorla RA
    2. Wagner AR

    (1972)

    A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement

    In: Black A. H, Prokasy W. F, editors. Classical Conditioning II Current Research and Theory. Appleton-Century-Crofts. pp. 64–99.

    • Google Scholar
78. 1. Rohe T
    2. Noppeney U

    (2015) Cortical hierarchies perform bayesian causal inference in multisensory perception

    PLoS Biology 13:e1002073.

    https://doi.org/10.1371/journal.pbio.1002073

    • PubMed
    • Google Scholar
79. 1. Rouder JN
    2. Morey RD

    (2012) Default Bayes factors for Model selection in regression

    Multivariate Behavioral Research 47:877–903.

    https://doi.org/10.1080/00273171.2012.734737

    • PubMed
    • Google Scholar
80. 1. Roy M
    2. Shohamy D
    3. Daw N
    4. Jepma M
    5. Wimmer GE
    6. Wager TD

    (2014) Representation of aversive prediction errors in the human periaqueductal gray

    Nature Neuroscience 17:1607–1612.

    https://doi.org/10.1038/nn.3832

    • PubMed
    • Google Scholar
81. 1. Sawamoto N
    2. Honda M
    3. Okada T
    4. Hanakawa T
    5. Kanda M
    6. Fukuyama H
    7. Konishi J
    8. Shibasaki H

    (2000)

    Expectation of pain enhances responses to nonpainful somatosensory stimulation in the anterior cingulate cortex and parietal operculum/posterior insula: an event-related functional magnetic resonance imaging study

    Journal of Neuroscience 20:7438–7445.

    • PubMed
    • Google Scholar
82. 1. Schultz W
    2. Carelli RM
    3. Wightman RM

    (2015) Phasic dopamine signals: from subjective reward value to formal economic utility

    Current Opinion in Behavioral Sciences 5:147–154.

    https://doi.org/10.1016/j.cobeha.2015.09.006

    • PubMed
    • Google Scholar
83. 1. Scott JG
    2. Berger JO

    (2006) An exploration of aspects of bayesian multiple testing

    Journal of Statistical Planning and Inference 136:2144–2162.

    https://doi.org/10.1016/j.jspi.2005.08.031

    • Google Scholar
84. 1. Segerdahl AR
    2. Mezue M
    3. Okell TW
    4. Farrar JT
    5. Tracey I

    (2015) The dorsal posterior insula subserves a fundamental role in human pain

    Nature Neuroscience 18:499–500.

    https://doi.org/10.1038/nn.3969

    • PubMed
    • Google Scholar
85. 1. Seminowicz DA
    2. Davis KD

    (2007) Interactions of pain intensity and cognitive load: the brain stays on task

    Cerebral Cortex 17:1412–1422.

    https://doi.org/10.1093/cercor/bhl052

    • PubMed
    • Google Scholar
86. 1. Seymour B
    2. Daw N
    3. Dayan P
    4. Singer T
    5. Dolan R

    (2007) Differential encoding of losses and gains in the human striatum

    Journal of Neuroscience 27:4826–4831.

    https://doi.org/10.1523/JNEUROSCI.0400-07.2007

    • PubMed
    • Google Scholar
87. 1. Seymour B
    2. O'Doherty JP
    3. Dayan P
    4. Koltzenburg M
    5. Jones AK
    6. Dolan RJ
    7. Friston KJ
    8. Frackowiak RS

    (2004) Temporal difference models describe higher-order learning in humans

    Nature 429:664–667.

    https://doi.org/10.1038/nature02581

    • PubMed
    • Google Scholar
88. 1. Seymour B
    2. O'Doherty JP
    3. Koltzenburg M
    4. Wiech K
    5. Frackowiak R
    6. Friston K
    7. Dolan R

    (2005) Opponent appetitive-aversive neural processes underlie predictive learning of pain relief

    Nature Neuroscience 8:1234–1240.

    https://doi.org/10.1038/nn1527

    • PubMed
    • Google Scholar
89. 1. Stein N
    2. Sprenger C
    3. Scholz J
    4. Wiech K
    5. Bingel U

    (2012) White matter integrity of the descending pain modulatory system is associated with interindividual differences in placebo analgesia

    Pain 153:2210–2217.

    https://doi.org/10.1016/j.pain.2012.07.010

    • PubMed
    • Google Scholar
90. 1. Stephan KE
    2. Penny WD
    3. Moran RJ
    4. den Ouden HE
    5. Daunizeau J
    6. Friston KJ

    (2010) Ten simple rules for dynamic causal modeling

    NeuroImage 49:3099–3109.

    https://doi.org/10.1016/j.neuroimage.2009.11.015

    • PubMed
    • Google Scholar
91. 1. Summerfield C
    2. de Lange FP

    (2014) Expectation in perceptual decision making: neural and computational mechanisms

    Nature Reviews Neuroscience 15:745–756.

    https://doi.org/10.1038/nrn3838

    • PubMed
    • Google Scholar
92. 1. Summerfield C
    2. Trittschuh EH
    3. Monti JM
    4. Mesulam MM
    5. Egner T

    (2008) Neural repetition suppression reflects fulfilled perceptual expectations

    Nature Neuroscience 11:1004–1006.

    https://doi.org/10.1038/nn.2163

    • PubMed
    • Google Scholar
93. 1. Tabor A
    2. Thacker MA
    3. Moseley GL
    4. Körding KP

    (2017) Pain: a Statistical Account

    PLOS Computational Biology 13:e1005142.

    https://doi.org/10.1371/journal.pcbi.1005142

    • PubMed
    • Google Scholar
94. 1. Taylor KS
    2. Seminowicz DA
    3. Davis KD

    (2009) Two systems of resting state connectivity between the insula and cingulate cortex

    Human Brain Mapping 30:2731–2745.

    https://doi.org/10.1002/hbm.20705

    • PubMed
    • Google Scholar
95. 1. Tobler PN
    2. O'doherty JP
    3. Dolan RJ
    4. Schultz W

    (2006) Human neural learning depends on reward prediction errors in the blocking paradigm

    Journal of Neurophysiology 95:301–310.

    https://doi.org/10.1152/jn.00762.2005

    • PubMed
    • Google Scholar
96. 1. Todorovic A
    2. de Lange FP

    (2012) Repetition suppression and expectation suppression are dissociable in time in early auditory evoked fields

    Journal of Neuroscience 32:13389–13395.

    https://doi.org/10.1523/JNEUROSCI.2227-12.2012

    • PubMed
    • Google Scholar
97. 1. Todorovic A
    2. van Ede F
    3. Maris E
    4. de Lange FP
    5. de LFP

    (2011) Prior expectation mediates neural adaptation to repeated sounds in the auditory cortex: an MEG study

    Journal of Neuroscience 31:9118–9123.

    https://doi.org/10.1523/JNEUROSCI.1425-11.2011

    • PubMed
    • Google Scholar
98. 1. Uddin LQ
    2. Kinnison J
    3. Pessoa L
    4. Anderson ML

    (2014) Beyond the tripartite cognition-emotion-interoception model of the human insular cortex

    Journal of Cognitive Neuroscience 26:16–27.

    https://doi.org/10.1162/jocn_a_00462

    • PubMed
    • Google Scholar
99. 1. Valet M
    2. Sprenger T
    3. Boecker H
    4. Willoch F
    5. Rummeny E
    6. Conrad B
    7. Erhard P
    8. Tolle TR

    (2004) Distraction modulates connectivity of the cingulo-frontal cortex and the midbrain during pain--an fMRI analysis

    Pain 109:399–408.

    https://doi.org/10.1016/j.pain.2004.02.033

    • PubMed
    • Google Scholar
100. 1. Villemure C
     2. Bushnell MC

     (2009) Mood influences supraspinal pain processing separately from attention

     Journal of Neuroscience 29:705–715.

     https://doi.org/10.1523/JNEUROSCI.3822-08.2009

     • PubMed
     • Google Scholar
101. 1. Vlaeyen JW

     (2015) Learning to predict and control harmful events: chronic pain and conditioning

     Pain 156 Suppl 1:S86–S93.

     https://doi.org/10.1097/j.pain.0000000000000107

     • PubMed
     • Google Scholar
102. 1. Wager TD
     2. Atlas LY
     3. Leotti LA
     4. Rilling JK

     (2011) Predicting individual differences in placebo analgesia: contributions of brain activity during anticipation and pain experience

     Journal of Neuroscience 31:439–452.

     https://doi.org/10.1523/JNEUROSCI.3420-10.2011

     • PubMed
     • Google Scholar
103. 1. Wager TD
     2. Atlas LY
     3. Lindquist MA
     4. Roy M
     5. Woo CW
     6. Kross E

     (2013) An fMRI-based neurologic signature of physical pain

     New England Journal of Medicine 368:1388–1397.

     https://doi.org/10.1056/NEJMoa1204471

     • PubMed
     • Google Scholar
104. 1. Wiech K
     2. Jbabdi S
     3. Lin CS
     4. Andersson J
     5. Tracey I

     (2014a) Differential structural and resting state connectivity between insular subdivisions and other pain-related brain regions

     Pain 155:2047–2055.

     https://doi.org/10.1016/j.pain.2014.07.009

     • PubMed
     • Google Scholar
105. 1. Wiech K
     2. Lin CS
     3. Brodersen KH
     4. Bingel U
     5. Ploner M
     6. Tracey I

     (2010) Anterior insula integrates information about salience into perceptual decisions about pain

     Journal of Neuroscience 30:16324–16331.

     https://doi.org/10.1523/JNEUROSCI.2087-10.2010

     • PubMed
     • Google Scholar
106. 1. Wiech K
     2. Seymour B
     3. Kalisch R
     4. Stephan KE
     5. Koltzenburg M
     6. Driver J
     7. Dolan RJ

     (2005) Modulation of pain processing in hyperalgesia by cognitive demand

     NeuroImage 27:59–69.

     https://doi.org/10.1016/j.neuroimage.2005.03.044

     • PubMed
     • Google Scholar
107. 1. Wiech K
     2. Vandekerckhove J
     3. Zaman J
     4. Tuerlinckx F
     5. Vlaeyen JW
     6. Tracey I

     (2014b) Influence of prior information on pain involves biased perceptual decision-making

     Current Biology 24:R679–R681.

     https://doi.org/10.1016/j.cub.2014.06.022

     • PubMed
     • Google Scholar
108. 1. Wiech K

     (2016) Deconstructing the sensation of pain: the influence of cognitive processes on pain perception

     Science 354:584–587.

     https://doi.org/10.1126/science.aaf8934

     • PubMed
     • Google Scholar
109. 1. Yacubian J
     2. Gläscher J
     3. Schroeder K
     4. Sommer T
     5. Braus DF
     6. Büchel C

     (2006) Dissociable systems for gain- and loss-related value predictions and errors of prediction in the human brain

     Journal of Neuroscience 26:9530–9537.

     https://doi.org/10.1523/JNEUROSCI.2915-06.2006

     • PubMed
     • Google Scholar
110. 1. Zeidan F
     2. Lobanov OV
     3. Kraft RA
     4. Coghill RC

     (2015) Brain mechanisms supporting violated expectations of pain

     PAIN 156:1772–1785.

     https://doi.org/10.1097/j.pain.0000000000000231

     • PubMed
     • Google Scholar
111. 1. Zhang S
     2. Mano H
     3. Ganesh G
     4. Robbins T
     5. Seymour B

     (2016) Dissociable Learning Processes Underlie Human Pain Conditioning

     Current Biology 26:52–58.

     https://doi.org/10.1016/j.cub.2015.10.066

     • PubMed
     • Google Scholar

Author details

1. Stephan Geuter

   1. Department of Systems Neuroscience, University Medical Center Hamburg Eppendorf, Hamburg, Germany
   2. Institute of Cognitive Science, University of Colorado Boulder, Boulder, United States
   3. Department of Psychology and Neuroscience, University of Colorado Boulder, Boulder, United States

   Contribution

   SG, Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft

   For correspondence

   stephan.geuter@colorado.edu

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-4935-5692

2. Sabrina Boll

   1. Department of Systems Neuroscience, University Medical Center Hamburg Eppendorf, Hamburg, Germany
   2. Department of General Psychiatry, University Hospital Heidelberg, Heidelberg, Germany

   Contribution

   SB, Conceptualization, Investigation, Methodology, Writing—review and editing

   Competing interests

   The authors declare that no competing interests exist.

3. Falk Eippert

   Centre for Functional Magnetic Resonance Imaging of the Brain, University of Oxford, Oxford, United Kingdom

   Contribution

   FE, Conceptualization, Methodology, Writing—review and editing

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-3986-1719

4. Christian Büchel

   Department of Systems Neuroscience, University Medical Center Hamburg Eppendorf, Hamburg, Germany

   Contribution

   CB, Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing—review and editing

   Competing interests

   The authors declare that no competing interests exist.

• 4,109

  views

• 777

  downloads

• 136

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.