1. Brain Rhythms and Short-Term Memory Earl K. Miller The Picower Institute for Learning and Memory and Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology www.ekmiller.org Adler Foundation Symposium February 2010

2. Monkey Human The prefrontal cortex Our Goal: To understand the neural basis of high-level cognition. Our Approach: Multiple-electrode recording in trained monkeys. Allows detailed comparisons of the timing of activity between neurons.

3. The ability to hold multiple items in short-term memory is critical for planning and executing goal-directed behavior. Increasing evidence for a role of oscillatory activity in short- term memory. One model suggests phase-dependent coding of memory items (e.g., Lisman and Idiart, Science, 1995). How Do We Hold Multiple Thoughts in Mind? This model attempted to explain why short- term memory has a severe limitation in capacity.

4. Cognitive capacity: How many things can you hold in mind simultaneously? Individual differences in capacity limits can explain about 25-50% of the individual differences in tests of intelligence It is linked to normal cognition and intelligence: Capacity is highest in younger adults and reduced in many neuropsychiatric disorders Schizophrenia Parkinson’s Disease Vogel et al (2001); Gold et al (2003); Cowan et al (2006); Hackley et al (2009) www.ekmiller.org

5. Q1: Is there oscillatory synchronization during short-term memory? Q2: Do spikes at particular phases of LFP oscillations carry more information about items in memory? Q3: Is there more information about different memory items in different LFP phases? How Does the Brain Hold Multiple Items in Memory?

6. Behavioral Task and Neurophysiological Recording Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. Two monkeys New stimulus set (4 objects) each day. Object identity fully balanced with order. Eight electrodes simultaneously implanted in the DL prefrontal cortex Local field potentials and multi-unit spikes from 140 recording sites

7. Sustained LFP Gamma Oscillations During Memory Delays LFP power at different frequencies Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. 1 st delay (one stimulus) 2 nd delay (two stimuli sequence) Average of all randomly selected spikes

8. LFP power at different frequencies Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. 1 st delay (one stimulus) 2 nd delay (two stimuli sequence) Average of all randomly selected spikes Q1: Is there oscillatory synchronization during short-term memory? Yes

9. Do Spikes Synchronize to LFP Oscillations? 32 Hz 3 Hz Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. Preferred spiking phase across recording sites Proportion of recording sites with significant spike- LFP phase-locking

10. 32 Hz 3 Hz Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. Preferred spiking phase across recording sites Proportion of recording sites with significant spike- LFP phase-locking Q2: Do spikes at particular phases of LFP oscillations carry more information about items in memory? Yes

11. Stimulus Information in Average Spiking Activity Information about stimulus identity (Neural variance explained by stimulus factor) Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. Average firing rate does not clearly distinguish object order

12. Information about stimulus identity from spiking activity in each LFP phase bin. Stimulus Information in Average Spiking Activity By LFP Phase Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci.

13. Stimuli were balanced by order Stimulus Information By Stimulus Order in Different LFP Phases 32 Hz Objects were balanced by order LFP Power Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci.

14. Object Information By Object Order in Different LFP Phases Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. 32 Hz 3 Hz Coding of objects in different phases was observed at 32 Hz, not 3 Hz

15. Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. 32 Hz Object phases overlap Error trials (32 Hz) (corrects + errors) Even though overall spike-phase synchrony is unchanged Object Information By Object Order in Different LFP Phases

16. Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. 32 Hz Object phases overlap Error trials (32 Hz) (corrects + errors) Even though overall spike-phase synchrony is unchanged Q3: Is there more information about different memory items in different LFP phases? Yes

17. Conclusions During short-term memory, prefrontal activity shows oscillatory synchronization in gamma and theta bands. Spikes carry more information about stimuli held in memory at particular LFP phases. The first stimulus of the memorized sequence is encoded earlier in the LFP cycle than the second stimulus for the gamma band but not the theta band. Phase-dependent coding may serve to flexibly represent sequences in memory on a generic neuronal time-scale. This may also explain why short-term memory has a capacity limitation. Siegel, Warden, and Miller (2009) Proc. Nat. Acad. Sci. Gamma band = spike-timing dependent plasticity?

18. Miller Lab Oct 2009 www.ekmiller.org