Discrete working memory Discrete working memory Delay Discrete

Discrete working memory

Discrete working memory Delay

Discrete working memory

Discrete working memory

Discrete working memory Green cue Delay Red cue 27 27 Spike rate (Hz) 0 -15 Delay 0 Time (sec) 15 30 0 -15 Data from inferotemporal cortex Fuster and Jervey, Science (1981) 0 15 Time (sec) 30

Spatial working memory

Spatial working memory Direction of monkey's gaze

Spatial working memory

Spatial working memory Delay

Spatial working memory

Spatial working memory

Data from Funahashi et al. (1989) J. Neurophysiol. 61: 331

Parametric Working Memory and Sequential Discrimination Experiments by group of R. Romo et al. , UNAM Nature 399: 470 (1999), Cereb. Cort. 13: 1196 (2003)

Choose f 1 > f 2 f 1 f 2

or f 2 > f 1 f 2

base delay f 1(Hz) 10 14 18 22 26 30 34 Rastergram: Trial-averaged firing rate Firing rate (Hz) 30 Tuning curve of memory activity Firing rate (Hz) 0 0. 5 Time (sec) 18 5 3. 5 Romo et al. Nature 1999 (from Miller et al. Cerebral Cortex 2003) 10 Stimulus, f 1 (Hz) 34

Delay activity in PFC ties the task together Firing rates Primary somatosensory cortex: Number of tuned neurons Secondary somatosensory cortex: Premotor cortex: Prefrontal cortex: Romo et al. Philos Trans Roy Soc: Biol, 2002

Network model: firing rate curve (mean-field theory)

Network model: firing rates with weak feedback Firing rate curve Feedback current Rate=f(I) I Rate I=f(Rate)

Network model: firing rates with weak feedback Firing rate curve Feedback current I(app) Rate=f(I) I Rate I=f(Rate)

Network model: firing rates with strong feedback Firing rate curve Feedback current

Network model: Bistability from strong feedback

Network model: firing rates with strong feedback Firing rate curve Total current I(app) Rate=f(I) I Rate I=f(Rate)

Network model: Bistability from strong feedback

Network model: recurrent excitation = pool of tens to hundreds of self-exciting neurons

Network model: bistability from recurrent excitation Input spikes here Memory activity

Stability increases exponentially with number of neurons in pool Miller and Wang, Chaos 2006 cf Miller et al, PLOS Biol. 2005

Network model: Continuous attractor with moderate feedback Firing rate curve Feedback current W Rate=f(I) I Rate I=f(Rate)

Network model: Continuous attractor with moderate feedback Firing rate curve Feedback current I(app) Rate=f(I) I Rate I=f(Rate)

Network model: Continuous attractor with moderate feedback Firing rate curve Feedback current I(app) Rate=f(I) I Rate I=f(Rate)