Dr Ben Seymour, Computational and Biological Learning Lab, Trumpington Street, Cambridge CB2 1PZ


Center for Information and Neural Networks, National Institute for Information and Communications Technology (NICT), 1-3 Yamadaoka, Suita, Osaka 565-0871, Japan.

bjs49 AT / seymour AT

F1000 reviews (paywall)



Pain at SfN

Just back from SfN, here are 3 of the most interesting pain posters/talks from SfN (in no particualr order).

1. Schulz et al (Munich): Neurophysiological correlates of tonic pain. They record EEG with a tonic thermal pain stimulus that varies over time, and compare with phasic pain. They show the clearest evidence to date that tonic and phasic are fundamentally different - both from a neuranatomical and electrophysiological perspective. In particular, the prefrontal theta tracks subjective ratings of tonic pain which is dissociable from actual thermal input. It would be fascinating to see this in patients.

2. Baliki et al (Northwestern): Hub reorganization of brain functional networks: A hallmark of the chronic pain state. They take a graph theoretic approach to rsfMRI connectivity data, and show that chronic pain is charactersied by widespread hub disruption in nodes (small regions of interest) with a sparse link density. Basically, this means that there is a fundamental change in the basic functional architecture at a whole-brain level, manifest as a change in the small-worldliness, which seems to quantitatively and objectively characterise the functional chronic pain state for the first time. But what is even more remarkable is that exactly the same changes are seen in rat rsfMRI, developing slowly after injury, showing that this is a valid tranlsational pain biomarker. This will be a landmark paper whenever it is published.  

3. Atlas et al (NYU): How instructed knowledge shapes aversive learning. In this nanosymposium talk, Lauren Atlas an co. dissociate distinct mechanisms underlying reversal learning driven either by feedback, or by instruction. Interestingly, SCRs appropriately follow the reversed contingencies following instruction, but the amygdala carries on signal feedback related teaching signals (associabilities) regardless, as if it is blind to this information. Furthermore, the ventral striatal prediction error signal is reduced, suggesting that somehow this error-based feedback circuit isn't used. What this data show is that understanding the relationship between experience based and cognitive based aversive learning is critical to developing advanced models of aversive (fear) learning.



Sensation is cool.

Humans have an amazing ability to sense temperature, with some people able to detect changes as little as 0.05C. How we do this is a bit of a mystery. Cold sensation is mediated in the periphery by the TRP family of thermoceptors, but there don’t seem to be enough known members of TRP family (each having an optimal temperature tuning response function) to code the full range of detectable temperatures with sufficient accuracy. Unless there is an as yet undiscovered large new family of thermoreceptors, then this suggests either:

  • 'Hyper-acuity' is achieved by sophisticated central brain decoding processes.
  • Each TRP receptor can adaptively and optimally tune itself to the ambient temperature, and so detect very small changes.

There's also a second, related, puzzle iwhen it comes to temperature: why is the perception of temperature strongly determined by relative temperature differences, such that a tepid bowl of water will feel cold to a hot hand, but warm to a cold hand? Again, this suggests eitehr central or peripheral processes:

  • There is a central topographic ‘state-thermometer’, sensitive to the ambient environmental temperature, against which phasic perturbations in temperature are judged.
  • Peripheral TRP thermoceptors do the adapting themselves.

Clearly both puzzles could be at least in part explained by peripheral adaptivity. In a recent study in the Journal of Neuroscience, Fujita and colleagues meaured cytosolic Ca2+ influx in cultured cells, and show that the TRPM8 cold receptor response is strongly determined by the ambient temperature of the cell before a phasic change in temperature. At 40C the temperature threshold was 35C, whereas at 30C the threshold was 28C, implying that the sensitivity to phasic temperature changes is increased in the region of ambient temperature. This is the case regardless of whether the cell arrives at this temperature from slow warming or slow cooling.  This implies temperature-sensitive molecular mechanisms that operate over different timescales, and that neuronal responses could be a function of the comparison between two underlying molecular processes. They go on to identify one such candidate mechanism, phosphatidylinositol 4,5-bisphosphate (PIP2) - which is known to be an important regulator of TRPM8 function. Intracellular depletion impairs the dependency on ambient temperature, and mutations to the binding site of PIP2 on the TRPM8 receptor also abolish the effect, providing good evidence for the necessity of PIP2 - TRPM8 signaling.

These results suggest that TRPM8 receptors ‘tune’ themselves to the ambient temperature, and hence are more sensitive to phasic changes. This illustrates a potential peripheral contribution to the adaptivity and hyper-acuity problems.

Many interesting questions emerge: what are the response properties from lower temperatures (such as 20C)? How does it interact with TRPM8 ligands such as menthol? To what extent these processes underlie behavioural and perceptual aspects of cold perception? Notwithstanding this, the results provide an important new finding in our understanding of thermosensory processing and perception.

Fujita F, Uchida K, Takaishi M, Sokabe T, Tominaga M. Ambient temperature affects the temperaature threshold for TRPM8 activation through interaction of phosphatatidylinositol 4,5- bisphosphate. J Neurosci. 2013 Apr 3; 33(14):6154-9

Based on a F1000 review with Hiro Mano 


Relief ≠ reward

It’s tempting to assume that since relief of something bad (such as pain) is pleasant, then relief can be considered the same as other rewards, including when it comes to brain processes. This would be very convenient, since it would mean we could apply everything we know about reward processing to pain relief processing. But is it true?


Well, there is obvious parity in terms of valence: as long as we can show that relief motivates behaviour, then it is by a psychological definition a reward. But this relationship turns out to have a more nuanced history in theories of motivation and opponency - the approach-withdrawal yin-yang of reward and punishment systems. In Konorski’s classic account, relief and disappointment are inhibitory states that signify omission of otherwise expected punishment or reward, respectively. In Solomon and Corbit’s opponency model, interruption or terminatation of a punishment or reward is considered in a similar manner. Furthermore, that relief can motivate behaviour lay at the heart of (Mowrer’s) classic two-factor theory of avoidance: when one learns that a certain action reduces punishment, the outcome state that signals its relief becomes ‘rewarding’ (as a conditioned inhibitor), and then reinforces avoidance actions in the conventional manner (as a conditioned reinforcer). So far so good.


But flies can be seen circling the ointment when one notices the conspicuous paucity of good evidence that conditioned inhibitors can easily act as conditioned reinforcers. Furthermore, conditioned inhibitors sometimes seem to have different properties – they are remarkably resistant to extinction for instance, unlike reward excitators (i.e. normal reward-predicting cues). This raises an awkward possibility: perhaps relief might actually be a fundamentally different process


Two recent papers hint at some neurophysiological support to this hypothesis. Sangha and colleagues compare responses in rat amygdala neurons in a Palvovian learning task involving different cues associated with shock, safety and reward. They show neurons in the amygdala show selective cue responses discriminating the predictions of each. Although some neurons responded similarly to reward and safety, many neurons responded differentially.  In another paper, Fernando and colleagues compared behavioural and pharmacological properties of appetitive conditioned excitators (normal Pavlovian reward predictors) and aversive conditioned inhibitors (safety signals), and showed that whereas reward predictors support the acquisition of a new response, safety signals did not. Furthermore, systemic d-amphetamine boosted only reward-predicting instrumental behaviour, suggesting a selective role for dopamine in reward behaviour and not safety signal behaviour. Taken together, these results provide behavioural, electrophysiological and pharmacological evidence against the motivational equivalence of relief and reward.


This suggests that pain neuroscientists should cautious about equating relief and reward, and hopefully this should stimulate more research aimed at understanding the neural basis for of relief, including the possible differences between different types of relief (omission of expected pain, and termination of ongoing pain). It’s also important for understanding conditions such as Obsessive Compulsive Disorder, which is thought to be an inflation of avoidance ‘habits’ – over-activity of the system that supports maintenance of avoidance.


Safety encoding in the basal amygdala. Sangha S, Chadick JZ, Janak PH. J Neurosci. 2013 Feb 27;33(9):3744-51. PMID: 23447586 

Comparison of the conditioned reinforcing properties of a safety signal and appetitive stimulus: effects of d-amphetamine and anxiolytics. Fernando AB, Urcelay GP, Mar AC, Dickinson A, Robbins T. Psychopharmacology (Berl). 2013 May;227(2):195-208. PMID: 23299096


This blog is based on a recently published F1000 review with Hiro Mano.


Itchy neurons... 

ITCH is one of the most curious of sensations - it's hard to know how you would describe it if we didn't have the word 'itch'. On one hand it's aversive, and indeed for patients with severe itch due to skin or neurological disorders, it is nothing short of pure torment. But for most of us, it is also enormously satisfying to scratch - indeed arguably much more satisfying than the original itch was unpleasant (compare it with pain and pain relief for instance). Itch also has the remarkable capacity to be induced by seeing someone else scratching. Actually just reading the word itch will make you itch (if you've read this far, you've almost certainly scratched at least once).

The yin-and-yang of averision and relief is paralelled in many of the so-called interoceptive sensations, signaling the departure and back of a variety of homeostatic states. And these are typically signalled but what has traditionally been considered the pain and temperature pathway, mediated in the peripheral mostly by a humble but fascinating group of small unmyelinated neurons: c-fibers. The c-fiber was long considered the poor mans sensory fiber, but their true sophistication and diversity is only now fully emerging. Incorporating sensations such as warmth, pain, itch, tickle and sensual touch, they convey a plethora of interoceptive sensations that support a phenomenological richness that their large myelinated cousins could only dream of. The story of their ascendancy is framed in one of the great debates of modern neuroscience - the ‘labeled line hypothesis’. In brief, do these experientially different sensations have their own dedicated neural pathways peripherally and centrally, or do they share a common set of pathways that are decoded by the brain. Or is the answer somewhere in between, involving an intermediate degree of afferent specificity.

The ‘itch-line’ hypothesis has been highlighted by a number of seminal papers: Schmelz and colleagues discovery of a specific subpopulation of histamine responsive c-fibers who’s responses correlated with the perceptual evolution of itch after histamine administration [1]; Andrew and Craig’s identification of a specific itch responsive spinothalamic tract neurons in the cat [2]; and Sun and colleagues discovery of a population of gastrin-releasing peptide receptor lamina 1 dorsal horn neurons selectively involved in itch [3].

Now, Han and colleagues report a population of MrgprA3 expressing c-fibers that seem to be specific for non-histamine involving itch. Ablation of these fibers resulted in reduction of itch to a range of itch stimuli, but importantly, not pain. Furthermore, activation of TRPV1 co-expressing MrgprA3 fibers induced itch but not pain behaviour.

Together, these findings elucidate a new population of non-histaminergic itch c-fibers, suggesting a highly selective itch-line pathway, and a new target for clinical interventions.


[1] Schmelz M, Schmidt R, Bickel A, Handwerker HO, Torebjörk HE. Specific C-receptors for itch in human skin. J Neurosci 1997; 17: 8003–8008.

[2] Andrew D, Craig AD. Spinothalamic lamina I neurons selectively sensitive to histamine: a central neural pathway for itch. Nat Neurosci 2001; 4: 72–77.

[3] Sun Y-G, Chen Z-F. A gastrin-releasing peptide receptor mediates the itch sensation in the spinal cord. Nature 2007; 448: 700–703.

Based on a recently published F1000 post-publication peer review with Hiro Mano


Phantoms and robots.

Losing a limb is a double-edged sword – not only do you lose the motor function of your arm or leg, but often the absent limb is plagued with unbearable phantom pain. Theories of phantom limb pain place central importance on the notion of aberrant reogranisation of the deafferented cortical representation of the affected limb. Support for this comes from the apparent therapeutic benefit of procedures like the famous mirror box - they temporarily support less abnormal representations through previously learned sensory associations. Thus for this and many other types of pain syndrome caused by lesions to the pain system, restorative approaches are likely to offer significant promise.

But how do you restore a missing limb? From a motor perspective, you can use brain-machine interfaces: by decoding activity from micro-electrode arrays implanted over M1, it’s possible to control quite complex movements of a robotic arm. However most robotic limbs don't usually feel. Last year, O’Doherty and colleagues published a seminal paper in Nature showing that microstimulation of primate S1 can be used to guide sensory-motor exploration of visually identified targets. They showed that monkeys could guide virtual hand and ‘feel’ objects, and then grab them with a virtual arm controlled by simultaneous motor decoding over M1. This provided the first example of how so called brain-machine-brain interfaces can be designed to yield closed loop sensory-motor systems, yielding full avatar systems.

In their latest paper, the same group take these findings one step further, by showing that variation of the statistics of S1 cortical microstimulation – by altering the periodicity and regularity of a pulse train - leads to discriminable artificial touch. This is important, because it moves beyond a crude pulse of sensory stimulation to the demonstration that potentially quite sophisticated sensory information can be encoded in the stimulation statistics. This means that in principle, complex and multidimensional information could be fed back from sensor-containing robotic prosthetic limbs in amputee patients, allowing cortical sensory representation, and a subsequent physiological embodiment of the prosthesis. And if current theories are correct, this should abolish phantom limb pain.

It also raises the intriguing possibility that more advanced sensation could be added - why stop at traditional sensations of touch, temperature and vibration, when you could add a metal detector, or a sonar, to your fingertips...


O'Doherty, J.E., Lebedev, M.A., Ifft, P.J., Zhuang, K.Z., Shokur, S. Bleuler, H. & M.A.L. Nicolelis. (2011). 
Active tactile exploration using a brain-machine-brain interface. 
Nature, 479(7372): 228-231. 

O'Doherty, J.E., Lebedev, M.A., Li, Z. & M.A.L. Nicolelis. (2012). 
Virtual active touch using randomly patterned intracortical microstimulation. 
IEEE Transactions on Neural Systems and Rehabilitation Engineering, 20(1): 85-93.