More evidence for mixed-selectivity in the cortex.  This time with voxels in the human brain.

Jackson, J., & Woolgar, A. (2018). Adaptive coding in the human brain: Distinct object features are encoded by overlapping voxels in frontoparietal cortex. Cortex.

Read more about mixed selectivity:
Rigotti, M., Barak, O., Warden, M.R., Wang, X., Daw, N.D., Miller, E.K., & Fusi, S. (2013) The importance of mixed selectivity in complex cognitive tasks. Nature, 497, 585-590, doi:10.1038/nature12160. View PDF »        

Fusi, S., Miller, E.K., and Rigotti, M. (2016) Why neurons mix: High dimensionality for higher cognition.  Current Opinion in Neurobiology. 37:66-74  doi:10.1016/j.conb.2016.01.010. View PDF »

Still think that single neurons with specific functions rule the brain?  Let us persuade you otherwise.  We argue that cognitive control stems from dynamic, context-dependent population coding.

Stokes, M., Buschman, T.J., and Miller, E.K. (2017) Dynamic coding for flexible cognitive control.  The Wiley Handbook of Cognitive Control, The Wiley Handbook of Cognitive Control, Edited by Tobias Egner, John Wiley & Sons, 2017(Chichester, West Sussex, UK). View PDF

Randoph Helfrich and Robert Knight review evidence that the infrastructure of cognitive control is rhythmic.  The general idea is that the prefrontal cortex controls large-scale oscillatory dynamics in the cortex and subcortex.  But there is much more.  Do yourself a favor: Read it.

Helfrich, R. F., & Knight, R. T. (2016). Oscillatory Dynamics of Prefrontal Cognitive Control. Trends in Cognitive Sciences.

A very nice experiment from Matt Chafee et al (as usual).  They show that neurons in the prefrontal cortex don’t have fixed properties.  Instead, they show “mixed selectivity” that changes with behavioral context and is biased toward stimuli that inhibit prepotent responses.  Sounds like cognitive control to me.

Blackman, Rachael K., et al. “Monkey prefrontal neurons reflect logical operations for cognitive control in a variant of the AX continuous performance task (AX-CPT).” The Journal of Neuroscience 36.14 (2016): 4067-4079.

This study shows the role of alpha and beta oscillations in the prefrontal cortex and frontal eye fields in a classic test of cognitive control: anti-saccades.  It also shows how these oscillatory patterns develop with adulthood.

Hwang, Kai, et al. “Frontal preparatory neural oscillations associated with cognitive control: A developmental study comparing young adults and adolescents.NeuroImage (2016).

Erez and Duncan elegantly show that the prefrontal cortex only cares about behavioral (goal) relevance.  Human subjects detected whether images from one of two visual categories were present in a scene.  The prefrontal cortex did not distinguish between the two categories but did distinguish whether an image was one the two categories (i.e., a target) or not (a non-target).

Erez, Y. and Duncan, J. Discrimination of Visual Categories Based on Behavioral Relevance in Widespread Regions of Frontoparietal Cortex.  The Journal of Neuroscience, 9 September 2015, 35(36): 12383-12393; doi: 10.1523/JNEUROSCI.1134-15.2015

Abstract context representations are not just in the prefrontal cortex, they are also in the amygdala.  The authors also report that errors were associated with reduced context encoding.  Cool.

Saez, A., et al. “Abstract Context Representations in Primate Amygdala and Prefrontal Cortex.Neuron 87.4 (2015): 869-881.

Preview by Cohen and Paz:
Cohen, Yarden, and Rony Paz. “It All Depends on the Context, but Also on the Amygdala.” Neuron 87.4 (2015): 678-680.

Voytek et al provide more evidence that oscillatory dynamics play a critical role in neural communication and cognitive control.  As humans performed tasks that required greater abstraction, there was an increase in theta synchrony between anterior and posterior frontal cortex.  This may allow more anterior frontal cortex is communicate the higher level goals to motor cortex.

Oscillatory dynamics coordinating human frontal networks in support of goal maintenance
Bradley Voytek, Andrew S Kayser, David Badre, David Fegen, Edward F Chang, Nathan E Crone, Josef Parvizi, Robert T Knight & Mark D’Esposito.  Nature Neuroscience

Sussillo reviews the use of recurrent neural networks (RNNs) to study cortical neurons.  RNNs can explain the high-dimensional, mixed-selectivity properties and oscillatory temporal dynamics of cortical neurons.  They share many features of cortical networks including feedback, nonlinearity, and parallel and distributed computing

Genovesio et al trained monkeys to judge whether red square or blue circle were farther from a reference point.  Even though information about the previous trial was irrelevant to the current trial, prefrontal cortex neurons conveyed the outcome of the previous trial and other irrelevant information about it.  Information about previous outcomes can often be helpful.  This study shows that this is automatically tracked by the prefrontal cortex even when it is not helpful.

Eiselt and Nieder trained monkeys to make greater/less than judgments to line lengths and dot numerosities.  They compared neural activity in the prefrontal cortex (PFC), anterior cingulate (AC), and premotor cortex (PMC).  The greatest proportion of greater/less than rule neurons were found in the PFC.  Further, only the PFC had neurons that were “generalists”; they signaled the greater/less than rules for both judgments.  Neurons in other areas were specialized for one judgment or the other.

This is consistent with our work showing that a large proportion of PFC neurons are multifunction, mixed selectivity neurons.  They may be key in providing the computational power for complex, flexible behavior.  For further reading see:

Rigotti, M., Barak, O., Warden, M.R., Wang, X., Daw, N.D., Miller, E.K., & Fusi, S. (2013) “The importance of mixed selectivity in complex cognitive tasks”. Nature, 497, 585-590,  doi:10.1038/nature12160. View PDF

Cromer, J.A., Roy, J.E., and Miller, E.K. (2010) Representation of multiple, independent categories in the primate prefrontal cortex. Neuron, 66: 796-807. View PDF »

Rouhinen et al provide evidence for the role of neural oscillations in the limitations of cognitive capacity.  Subjects tracked multiple objects.  Strength of oscillations were different preceding detected vs undetected objects.  Suppression of low-frequency oscillations (<20 Hz) and strengthening of high-frequency oscillations (>20 Hz) in the frontoparietal cortex was correlated with attentional load.   Load-dependent strengthening of 20-90 Hz oscillations was predictive of individual capacity.  This supports hypotheses that oscillations play major role in attention and are responsible for the limited bandwidth of cognition.

Further reading on attention, capacity, and oscillations:

  • Buschman,T.J. and Miller, E.K. (2010) Shifting the Spotlight of Attention: Evidence for Discrete Computations in Cognition. Frontiers in Human Neuroscience. 4(194): 1-9. View PDF »
  • Miller, E.K. and Buschman, T.J. (2013) Cortical circuits for the control of attention.  Current Opinion in Neurobiology.  23:216–222  View PDF »
  • Buschman, T.J. and Miller, E.K. (2009) Serial, covert, shifts of attention during visual search are reflected by the frontal eye fields and correlated with population oscillations. Neuron, 63: 386-396. View PDF »
  • Buschman, T.J. and Miller, E.K. (2007) Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science. 315: 1860-1862  The Scientist’s “Hot Paper” for October 2009. View PDF »
  • Siegel, M., Warden, M.R., and Miller, E.K. (2009) Phase-dependent neuronal coding of objects in short-term memory. Proceedings of the National Academy of Sciences, 106: 21341-21346. View PDF »

Nee and Jonides argue that short-term memory (STM) is not monolithic, but instead involves multiple processes with different characteristics.  There are frontal selection mechanisms (normally associated with attention), medial temporal binding mechanisms (associated with long-term memory) and synaptic plasticity.  As a result, STM involves a single representation that can be focused on, a set of active representations that focused can be switched to, and passive long-term memory representations with residual traces that can be easily activated.  The authors show how this model can explain many discrepancies across studies.

Our work with Stefano Fusi’s Lab makes  The Wall Street Journal.

Miller Lab alumnus Jon Wallis and crew studied two different types of cost-benefit decisions (delay vs effort).  They found that different neurons in the dorsolateral prefrontal cortex, orbitofrontal, and anterior cingulate encoded the different types of decisions.  Thus, rather than have neurons encode decisions on an abstract level, frontal cortex neurons encode stimuli based on their exact consequences.

Blogger John Borghi lists the most highly cited papers in neuroscience and has kind words for Miller and Cohen (2001).  Thanks, John!

  • Miller, E.K. and Cohen, J.D. (2001) An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24:167-202.
    Designated a Current Classic by Thomson Scientific as among the most cited papers in Neuroscience and Behavior. View PDF »

Behavior can be fast and automatic, but inflexible (model-free) vs slower, more deliberate, and flexible (model-based) Ray Dolan and crew show that disruption of the prefrontal cortex by transcranial magnetic stimulation (TMS) pushes humans towards more inflexible, model-free, behavior.

More evidence for domain-general processing in higher-level cortex.  Federenko et al tested human subjects with seven tasks with different cognitive demands.  FMRI revealed overlapping activation zones in the frontal and parietal cortex.  This is consistent with neurophysiological studies showing that many neurons in these areas are multifunctional.  Rigotti et al recently demonstrated that these multifunctional “mixed selectivity” neurons provide the computational power needed for high-level cognition.

For further reading:

Rigotti, M., Barak, O., Warden, M.R., Wang, X., Daw, N.D., Miller, E.K., & Fusi, S. “The importance of mixed selectivity in complex cognitive tasks”. Nature, 497, 585-590, 2013 doi:10.1038/nature12160. View PDF

Miller, E.K. and Fusi, S. (2013) Limber neurons for a nimble mind. Neuron. 78:211-213. View PDF

Mirpour and Bisley recorded neural responses and local field potentials from the lateral intraparietal cortex (LIP) during visual search.  Previously fixated non-target stimuli elicited greater lower frequency (alpha and beta) oscillations.  This suggested that reduced neural responses (and attention) to previously seen stimuli results from oscillatory-based top-down influences from the frontal cortex.

John Duncan and colleagues examined dynamic allocation of attention in the prefrontal cortex.  A behaviorally relevant target and non-target were simultaneously presented in both visual hemifields.  At first, activity in each hemifield was dominated by the stimulus in the contralateral field but then all activity became dominated by the target alone.  The speed and degree of attentional reallocation depend on relative attentional weights; more experience with a target led to faster and greater allocation to the target.  Because neurons rapidly shifted their representation from an irrelevant to relevant stimulus in the opposite hemifield, these results are consistent with adaptive coding models of neural representation.
Kadohisa et al (2013) Dynamic Construction of a Coherent Attentional State in a Prefrontal Cell Population

Further reading on adaptive coding:
Miller, E.K. and Cohen, J.D. (2001) An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24:167-202.  View PDF »

Duncan, J. and Miller, E.K. (2013) Adaptive neural coding in frontal and parietal cortex. In: Stuss, D.T. and Knight, R.T. (Eds). Principles of Frontal Lobe Function: Second Edition.

Rigotti, M., Barak, O., Warden, M.R., Wang, X., Daw, N.D., Miller, E.K., & Fusi, S. “The importance of mixed selectivity in complex cognitive tasks”. Nature, 497, 585-590, 2013 doi:10.1038/nature12160. View PDF

Bea Luna and colleagues used graph theory to examine the development of functional hubs in the human brain.  The hub architecture develops earlier, but connections between the hubs and “spokes” continue to develop and change into adulthood.

Adam Gazzaley and company show, for the first time, that training on a video game results in benefits that transfer to other tests of cognition.  Training on the NeuroRacer game produced long-lasting improvements in cognitive abilities of older adults (age 65-80).  How did they do it?  Their trick was to focus on multitasking and attention.
Anguera et al (2013) Nature

The Atlantic: How To Rebuild An Attention Span

Matt Chafee and colleagues used multiple-electrode recording in the prefrontal and parietal cortices to examine the temporal dynamics of their neural activity during a categorization task.   They decoded category signals from patterns of simultaneously recorded in small bins and asked whether the resulting  information  time series in one area could predict the other.  This showed that  “executive” top-down signals flow from the prefrontal to parietal cortex.

Ranulfo Romo and crew show delta band (1-4 Hz) synchrony between frontal and parietal cortex that varies with decisions.  When there were no decisions to be made, frontal-parietal delta was reduced.

An article in MIT’s Technology Review magazine about our work on how multitasking “mixed selectivity” neurons may be key for cognition.
Do-It-All Neurons – A key to cognitive flexibility by Anne Trafton