Our new paper describing a new model of working memory. Actually, not so much a new model as an update to the classic model. The classic model posited that we hold thoughts “in mind” (i.e., in working memory) via the persistent spiking of neurons. That is not wrong. It is right to a certain level of approximation. There is little doubt that spikes help maintain working memories. However, a closer examination revealed that there is much more going on than persistent spiking.
It is important to keep in mind (pun intended) that virtually all evidence for persistent spiking comes from experiments that averaged neural activity across trials. The assumption was that averaging boosts signal and decreases noise (“noise” meaning changes in activity from trial to trial). But what if that noise was not noise but real neural dynamics? We don’t always think about the same thing in the same way. Averaging assumes we do.
With this in mind, we and others have been leveraging multiple-electrode recording to examine neural activity in “real time” (on individual trials). This has revealed that working memory-related spiking occurs in sparse, coordinated bursts of activity. It also revealed oscillatory dynamics between brain waves in two frequency bands, beta and gamma. Gamma seems to act as a carrier wave that maintains the contents of working memory. Beta seems carry the top-down control signals that allow us to exert volitional control over working memory.
Miller, E.K., Lundqvist, M., and Bastos, A.M. Working Memory 2.0 Neuron DOI:https://doi.org/10.1016/j.neuron.2018.09.023 (Download a *free* copy for the next 50 days)
Here’s the abstract:
Working memory is the fundamental function by which we break free from reflexive input-output reactions to gain control over our own thoughts. It has two types of mechanisms: online maintenance of information and its volitional or executive control. Classic models proposed persistent spiking for maintenance but have not explicitly addressed executive control. We review recent theoretical and empirical studies that suggest updates and additions to the classic model. Synaptic weight changes between sparse bursts of spiking strengthen working memory maintenance. Executive control acts via interplay between network oscillations in gamma (30–100 Hz) in superficial cortical layers (layers 2 and 3) and alpha and beta (10–30 Hz) in deep cortical layers (layers 5 and 6). Deep-layer alpha and beta are associated with top-down information and inhibition. It regulates the flow of bottom-up sensory information associated with superficial layer gamma. We propose that interactions between different rhythms in distinct cortical layers underlie working memory maintenance and its volitional control.
Wasmuht, D. F., Spaak, E., Buschman, T. J., Miller, E. K., & Stokes, M. G. (2018). Intrinsic neuronal dynamics predict distinct functional roles during working memory. Nature Communications.
Abstract:
Working memory (WM) is characterized by the ability to maintain stable representations over time; however, neural activity associated with WM maintenance can be highly dynamic. We explore whether complex population coding dynamics during WM relate to the intrinsic temporal properties of single neurons in lateral prefrontal cortex (lPFC), the frontal eye fields (FEF), and lateral intraparietal cortex (LIP) of two monkeys (Macaca mulatta). We find that cells with short timescales carry memory information relatively early during memory encoding in lPFC; whereas long-timescale cells play a greater role later during processing, dominating coding in the delay period. We also observe a link between functional connectivity at rest and the intrinsic timescale in FEF and LIP. Our results indicate that individual differences in the temporal processing capacity predict complex neuronal dynamics during WM, ranging from rapid dynamic encoding of stimuli to slower, but stable, maintenance of mnemonic information.
Our dual Perspectives “debate” paper re: Does persistent spiking hold memories “in mind” (i.e., working memory):
Paper and link to opposing paper: Working Memory: Delay Activity, Yes! Persistent Activity? Maybe Not
Press release: To understand working memory, scientists must resolve this debate
Our two cents:
Surprised this became a debate. All we are saying is that if you look at delay activity more closely (on single trials) it’s bursty. Something else (synaptic weight changes) could be helping. Adding synaptic mechanisms saves energy and confers functional advantages.
Lundqvist, M., Rose, J., Herman, P., Brincat, S. L., Buschman, T. J., & Miller, E. K. (2016). Gamma and beta bursts underlie working memory. Neuron, 90(1), 152-164.
And it leaves room for network rhythms that may underlie executive control of working memory.
Bastos, A. M., Loonis, R., Kornblith, S., Lundqvist, M., & Miller, E. K. (2018). Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. Proceedings of the National Academy of Sciences, 201710323.
Lundqvist, M., Herman, P., Warden, M. R., Brincat, S. L., & Miller, E. K. (2018). Gamma and beta bursts during working memory readout suggest roles in its volitional control. Nature Communications, 9(1), 394.
Yuri Buzsaki pointed out that his lab reported that in PFC working memories are maintained by internally generated cell assembly sequences. The few persistently firing neurons were interneurons.
Fujisawa S, Amarasingham A, Harrison MT, Buzsáki G.
Nat Neurosci. 2008.
Also, the idea that synaptic weight changes help maintain working memories is not altogether new. Goldman-Rakic suggested such a mechanism. Her lab found that sparse firing in the PFC produces temporary changes in synaptic weights. Importantly, if neurons firing too fast, inhibitory mechanisms kick in and you don’t get the weight changes. See:
Wang, Y., Markram, H., Goodman, P.H., Berger, T.K., Ma, J., and Goldman-Rakic, P.S. (2006). Heterogeneity in the pyramidal network of the medial prefrontal cortex. Nat. Neurosci. 9, 534–542.
But, hey, don’t take our word for it: Look at memory delay activity on single trials and tell us what *you* see.
Nice result from Buschman Lab. Error-correcting dynamics introduce bias into working memory while reducing noise.
Error-correcting dynamics in visual working memory
Matthew F Panichello, Brian DePasquale, Jonathan W Pillow, Timothy Buschman
doi: https://doi.org/10.1101/319103
Press release for our new paper:
A heavy working memory load may sink brainwave ‘synch’
The paper:
Pinotsis, D.A., Buschman, T.J. and Miller, E.K. (2018) Working Memory Load Modulates Neuronal Coupling. Cerebral Cortex. https://doi.org/10.1093/cercor/bhy065 View PDF
Pinotsis, D.A., Buschman, T.J. and Miller, E.K. (2018) Working Memory Load Modulates Neuronal Coupling. Cerebral Cortex, 2018 https://doi.org/10.1093/cercor/bhy065
Abstract: There is a severe limitation in the number of items that can be held in working memory. However, the neurophysiological limits remain unknown. We asked whether the capacity limit might be explained by differences in neuronal coupling. We developed a theoretical model based on Predictive Coding and used it to analyze Cross Spectral Density data from the prefrontal cortex (PFC), frontal eye fields (FEF), and lateral intraparietal area (LIP). Monkeys performed a change detection task. The number of objects that had to be remembered (memory load) was varied (1–3 objects in the same visual hemifield). Changes in memory load changed the connectivity in the PFC–FEF–LIP network. Feedback (top-down) coupling broke down when the number of objects exceeded cognitive capacity. Thus, impaired behavioral performance coincided with a break-down of Prediction signals. This provides new insights into the neuronal underpinnings of cognitive capacity and how coupling in a distributed working memory network is affected by memory load.
An article in Science News about new ideas on the role of brain waves. It also discuss three new papers from the Miller Lab.
Brain waves may focus attention and keep information flowing Science New March 13, 2018
Here are the papers that are discussed:
Lundqvist, M., Herman, P. Warden, M.R., Brincat, S.L., and Miller, E.K. (2018) Gamma and beta bursts during working memory read-out suggest roles in its volitional control. Nature Communications. 9, 394 View PDF
Wutz, A., Loonis, R., Roy, J.E., Donoghue, J.A., and Miller, E.K. (2018) Different levels of category abstraction by different dynamics in different prefrontal areas. Neuron 97: 1-11. View PDF
Bastos, A.M., Loonis, R., Kornblith, S., Lundqvist, M., and Miller, E.K. (2018) Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. Proceedings of the National Academy of Sciences. View PDF
Lundqvist, M., Herman, P. Warden, M.R., Brincat, S.L., and Miller, E.K. (2018) Gamma and beta bursts during working memory read-out suggest roles in its volitional control. Nature Communications. 9, Article number: 394 doi:10.1038/s41467-017-02791-8
Abstract:
Working memory (WM) activity is not as stationary or sustained as previously thought. There are brief bursts of gamma (~50–120 Hz) and beta (~20–35 Hz) oscillations, the former linked to stimulus information in spiking. We examined these dynamics in relation to readout and control mechanisms of WM. Monkeys held sequences of two objects in WM to match to subsequent sequences. Changes in beta and gamma bursting suggested their distinct roles. In anticipation of having to use an object for the match decision, there was an increase in gamma and spiking information about that object and reduced beta bursting. This readout signal was only seen before relevant test objects, and was related to premotor activity. When the objects were no longer needed, beta increased and gamma decreased together with object spiking information. Deviations from these dynamics predicted behavioral errors. Thus, beta could regulate gamma and the information in WM.
This is the first of three papers that all lead to the same general conclusion: Sensory (bottom-up) information is fed forward through cortex by gamma (>50 Hz) waves in superficial cortical layers. Executive (top-down) information is fed back through cortex by alpha/beta waves (4-22 Hz) in deep cortical layers. The beta waves in deep layers regulate superficial layer gamma in a push-pull fashion thereby allowing top-down information to control the flow of bottom-up sensory information. This allows volitional control over what we hold in mind. Stayed tuned for the other two papers. They will appear in the next few weeks.
Bastos, A.M., Loonis, R., Kornblith, S., Lundqvist, M., and Miller, E.K. (2018) Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory. Proceedings of the National Academy of Sciences. published ahead of print January 16, 2018, doi:10.1073/pnas.1710323115
Read MIT press release here.
The authors suggest a hybrid model of working memory. The current focus of attention is encoded by spiking activity. Other items held in the working memory that are not the current focus of attention are held by temporary changes in synaptic weights per the activity-silent models of Lundqvist and Stokes.
Manohar, S. G., Zokaei, N., Fallon, S. J., Vogels, T., & Husain, M. (2017). A neural model of working memory. bioRxiv, 233007.
For more on activity-silent models, see:
Lunqvist, M., Rose, J., Herman, P, Brincat, S.L, Buschman, T.J., and Miller, E.K. (2016) Gamma and beta bursts underlie working memory. Neuron, published online March 17, 2016. View PDF »
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, (Chichester, West Sussex, UK). View PDF
Wasmuht, D. F., Spaak, E., Buschman, T. J., Miller, E. K., & Stokes, M. G. (2017). Intrinsic neuronal dynamics predict distinct functional roles during working memory. bioRxiv, 233171.
Stokes, M. G. (2015). ‘Activity-silent’working memory in prefrontal cortex: a dynamic coding framework. Trends in Cognitive Sciences, 19(7), 394-405.
Cavanagh et al test the sustained vs dynamic activity models of working memory. They find that sustained activity does not maintain the memory of a cue location/response past a distractor. Instead of sustained activity, they conclude that a dynamic population-level process underlies working memory.
Cavanagh, S. E., Towers, J. P., Wallis, J. D., Hunt, L. T., & Kennerley, S. W. (2017). Reconciling persistent and dynamic hypotheses of working memory coding in prefrontal cortex. bioRxiv, 231506.
The Dean of Prefrontal Cortex, Joaquin Fuster, breaks down prefrontal function along three lines: Executive attention, working memory, and decision-making.
Fuster, J. M. (2017). Prefrontal Executive Functions Predict and Preadapt. In Executive Functions in Health and Disease(pp. 3-19).
There is a growing consensus that there may be more to working memory than simple maintenance of spiking. On a single-trial, moment-to-moment basis, memory delay spiking is sparse, not sustained. Instead, spiking may produce changes in synaptic weights and that is where the working memories are actually stored.
Trübutschek, D., Marti, S., Ojeda, A., King, J. R., Mi, Y., Tsodyks, M., & Dehaene, S. (2017). A theory of working memory without consciousness or sustained activity. Elife, 6.
For further reading see:
Lunqvist, M., Rose, J., Herman, P, Brincat, S.L, Buschman, T.J., and Miller, E.K. (2016) Gamma and beta bursts underlie working memory. Neuron, published online March 17, 2016. View PDF »
Review of the neural mechanisms behind persistent spiking activity and working memory.
Zylberberg, J., & Strowbridge, B. W. (2017). Mechanisms of persistent activity in cortical circuits: possible neural substrates for working memory. Annual Review of Neuroscience, 40.
There is little doubt that spiking during memory delays play a role in working memory. But how persistent is the activity and how are the memories actually stored? For another perspective see:
Lunqvist, M., Rose, J., Herman, P, Brincat, S.L, Buschman, T.J., and Miller, E.K. (2016) Gamma and beta bursts underlie working memory. Neuron, published online March 17, 2016. View PDF »
We show how limitations in cognitive capacity (how many thoughts you can think at the same time – very few) may be due to changes in rhythmic coupling between cortical areas. More specifically, feedback coupling breaks down when capacity is exceeded.
Working Memory Load Modulates Neuronal Coupling Dimitris A Pinotsis, Timothy J Buschman, Earl K Miller
doi: https://doi.org/10.1101/192336
And they show the same independence between the visual hemifields that we saw in primates.
Balakhonov, D., & Rose, J. (2017). Crows Rival Monkeys in Cognitive Capacity. Scientific Reports, 7.
For further reading:
Buschman,T.J., Siegel, M., Roy, J.E. and Miller, E.K. (2011) Neural substrates of cognitive capacity limitations. Proceedings of the National Academy of Sciences. 108(27):11252-5. View PDF »
Miller, E.K. and Buschman, T.J. (2015) Working memory capacity: Limits on the bandwidth of cognition. Daedalus, Vol. 144, No. 1, Pages 112-122. View PDF »
Kornblith, S., Buschman, T.J., and Miller, E.K. (2015) Stimulus load and oscillatory activity in higher cortex. Cerebral Cortex. Published online August 18, 2015 doi: 10.1093/cercor/bhv182. View PDF »
Neurons in the prefrontal cortex keeps track of elapsed time (even though time was not explicitly relevant) via sequential firing of neurons. The overlap of sequences depended on the degree of similarity of the item being held in memory. The time-keeping showed a Weber-fraction-like decrease in precision as time passed.
Compressed timeline of recent experience in monkey lPFC
Zoran Tiganj, Jason A Cromer, Jefferson E Roy, Earl K Miller, Marc W Howard
doi: https://doi.org/10.1101/126219
Miller Lab alum Andreas Nieder and crew show how dopamine receptors in the prefrontal cortex regulate access to working memory and its protection from interference.
Jacob, Simon N., Maximilian Stalter, and Andreas Nieder. “Cell-type-specific modulation of targets and distractors by dopamine D1 receptors in primate prefrontal cortex.” Nature Communications (2016): 13218.
Nice review of past work on the neurobiology of working memory and capacity limits:
Constantinidis, Christos, and Torkel Klingberg. “The neuroscience of working memory capacity and training.” Nature Reviews Neuroscience (2016).
Although there is a caveat: More recent work suggests that the substrate of working memory is *not* sustained spiking activity. That is an artifact of cross-trial averaging. “Delay activity” is more sparse and bursty on single trials. This suggests a different memory substrate.
See:
Lunqvist, M., Rose, J., Herman, P, Brincat, S.L, Buschman, T.J., and Miller, E.K. (2016) Gamma and beta bursts underlie working memory. Neuron, published online March 17, 2016. View PDF
Stokes, Mark G. “‘Activity-silent’working memory in prefrontal cortex: a dynamic coding framework.” Trends in cognitive sciences 19.7 (2015): 394-405.
Stokes and Spaak review our recent work on single-trial analysis of working memory “delay” activity. This showed that the classic profile of sustained activity as the memory substrate is an artifact of averaging across trials. The assumption is that averaging cancels out noise. Instead, it may be covering up important details of the dynamics of neural activity.
Read more here:
The Importance of Single-Trial Analyses in Cognitive Neuroscience
Mark Stokes and Eelke Spaak
Trends in Cognitive Sciences
DOI: http://dx.doi.org/10.1016/j.tics.2016.05.008
The original paper:
Lunqvist, M., Rose, J., Herman, P, Brincat, S.L, Buschman, T.J., and Miller, E.K. (2016) Gamma and beta bursts underlie working memory. Neuron, published online March 17, 2016. View PDF »
Free access to our new paper:
Lundqvist, M., Rose, J., Herman, P., Brincat, S. L., Buschman, T. J., & Miller, E. K. (2016). Gamma and Beta Bursts Underlie Working Memory. Neuron.
Valid until May 26, 2016
Roux and Uhlhaas propose an interesting and provocative theory: Different frequencies of neural oscillations carry different information in working memory. Gamma oscillations maintain information in working memory. Alpha suppresses irrelevant information. Theta orders information. Gamma is thought to be coupled to both alpha and theta.
This is consistent with our observations for phase-coding of different working memories in gamma (Siegel et al., 2009) and alpha suppressing a dominant, but current irrelevant, neural ensemble (Buschman et al., 2012).
Buschman, T.J., Denovellis, E.L., Diogo, C., Bullock, D. and Miller, E.K. (2012) Synchronous oscillatory neural ensembles for rules in the prefrontal cortex. Neuron. 76: 838-846. 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 »