1 00:00:01,640 --> 00:00:04,040 The following content is provided under a Creative 2 00:00:04,040 --> 00:00:05,580 Commons license. 3 00:00:05,580 --> 00:00:07,880 Your support will help MIT OpenCourseWare 4 00:00:07,880 --> 00:00:12,270 continue to offer high-quality educational resources for free. 5 00:00:12,270 --> 00:00:14,870 To make a donation or view additional materials 6 00:00:14,870 --> 00:00:18,830 from hundreds of MIT courses, visit MIT OpenCourseWare 7 00:00:18,830 --> 00:00:22,160 at ocw.mit.edu. 8 00:00:22,160 --> 00:00:24,020 What I'm going to be talking about 9 00:00:24,020 --> 00:00:26,330 is some of the fundamental work that 10 00:00:26,330 --> 00:00:29,030 has been done trying to understand 11 00:00:29,030 --> 00:00:35,240 neural basis for memory and spatial perception 12 00:00:35,240 --> 00:00:41,720 and cognition that comes largely from behavioral and 13 00:00:41,720 --> 00:00:45,680 electrophysiological studies in the brain system that's 14 00:00:45,680 --> 00:00:47,670 shown here-- the hippocampus. 15 00:00:47,670 --> 00:00:51,206 And as I was mentioning before we started, 16 00:00:51,206 --> 00:00:52,580 this year's Nobel Prize was given 17 00:00:52,580 --> 00:00:55,580 to John O'Keefe, who discovered the properties 18 00:00:55,580 --> 00:00:58,010 of individual neurons in the hippocampus 19 00:00:58,010 --> 00:01:02,840 by applying these methods for recording 20 00:01:02,840 --> 00:01:05,690 the discharge of single cells by putting 21 00:01:05,690 --> 00:01:08,100 little wires into the brain-- so-called extracellular 22 00:01:08,100 --> 00:01:08,600 recordings. 23 00:01:08,600 --> 00:01:09,350 You put in a wire. 24 00:01:09,350 --> 00:01:12,180 The tip of the electrode can record electrical discharge 25 00:01:12,180 --> 00:01:14,114 of cells. 26 00:01:14,114 --> 00:01:15,530 You measure the discharge of cells 27 00:01:15,530 --> 00:01:18,230 as animals move around in space. 28 00:01:18,230 --> 00:01:22,100 And you try to figure out what this part of the brain does. 29 00:01:22,100 --> 00:01:24,320 Now where the observation part of science 30 00:01:24,320 --> 00:01:26,000 comes in in this case is interesting, 31 00:01:26,000 --> 00:01:33,140 because it relates to another aspect of neuroscience. 32 00:01:33,140 --> 00:01:37,910 And that is the relationship between the neurobiology 33 00:01:37,910 --> 00:01:39,060 and behavior. 34 00:01:39,060 --> 00:01:41,710 And this is an approach known as a neuroethological approach. 35 00:01:41,710 --> 00:01:43,760 And that's studying the brain systems 36 00:01:43,760 --> 00:01:51,010 in the context in which they actually evolve and were used. 37 00:01:51,010 --> 00:01:56,540 And so that could apply to the use of song 38 00:01:56,540 --> 00:01:59,100 as a mechanism of communication in birds. 39 00:01:59,100 --> 00:02:03,920 It can come from the finding prey 40 00:02:03,920 --> 00:02:07,190 in the dark, when it comes sound-localization owls. 41 00:02:07,190 --> 00:02:14,030 In the case of rodents, O'Keefe appreciated the fact 42 00:02:14,030 --> 00:02:17,120 that the hippocampus in rodents really 43 00:02:17,120 --> 00:02:21,170 has a primary role in spatial navigation-- 44 00:02:21,170 --> 00:02:23,120 that animals that have damaged the hippocampus 45 00:02:23,120 --> 00:02:25,310 have problems with spatial navigation. 46 00:02:25,310 --> 00:02:27,770 And the hippocampus had been studied up 47 00:02:27,770 --> 00:02:31,130 to that point using animals that were head-fixed. 48 00:02:31,130 --> 00:02:33,580 So you take a little rat. 49 00:02:33,580 --> 00:02:34,370 You fix his head. 50 00:02:34,370 --> 00:02:35,786 And the reason for fixing its head 51 00:02:35,786 --> 00:02:38,480 is for convenience, largely, so that when 52 00:02:38,480 --> 00:02:40,850 you place these little electrodes into the brain, 53 00:02:40,850 --> 00:02:42,014 the animal doesn't move. 54 00:02:42,014 --> 00:02:42,680 They don't move. 55 00:02:42,680 --> 00:02:44,638 Electrode moves, you can't get good recordings. 56 00:02:44,638 --> 00:02:47,190 So you need to keep the preparation fixed. 57 00:02:47,190 --> 00:02:48,530 So you fix the animal's head. 58 00:02:48,530 --> 00:02:50,030 Then you figure out ways to actually 59 00:02:50,030 --> 00:02:54,380 look, to study the system given the constraints 60 00:02:54,380 --> 00:02:55,370 of the methodology. 61 00:02:55,370 --> 00:03:00,800 And that involved, largely, before John O'Keefe in 1970, 62 00:03:00,800 --> 00:03:02,640 using methods of classical conditioning. 63 00:03:02,640 --> 00:03:06,170 So you might be familiar with basic learning theory. 64 00:03:06,170 --> 00:03:09,320 And how do we learn how to do things? 65 00:03:09,320 --> 00:03:12,800 Well, it's basically chaining together 66 00:03:12,800 --> 00:03:15,290 stimulus response associations. 67 00:03:15,290 --> 00:03:16,950 You see something, you do something, 68 00:03:16,950 --> 00:03:18,830 you get rewarded for doing that. 69 00:03:18,830 --> 00:03:21,510 You're more likely to do that again in the future. 70 00:03:21,510 --> 00:03:24,350 This is a basic Pavlovian conditioning. 71 00:03:24,350 --> 00:03:26,090 I ring a bell, you get food. 72 00:03:26,090 --> 00:03:27,980 You associate bell with the food. 73 00:03:27,980 --> 00:03:30,710 And so that was the thinking-- that all of cognition 74 00:03:30,710 --> 00:03:34,190 can be built up from basic stimulus-response associations. 75 00:03:34,190 --> 00:03:36,050 But there was a movement around the time 76 00:03:36,050 --> 00:03:39,380 that O'Keefe was doing this work that 77 00:03:39,380 --> 00:03:41,900 proposed that that was insufficient-- 78 00:03:41,900 --> 00:03:44,547 that simple behavioral simple stimulus response learning was 79 00:03:44,547 --> 00:03:46,130 insufficient, that there was some kind 80 00:03:46,130 --> 00:03:52,400 of internal foundation upon which learning was applied. 81 00:03:52,400 --> 00:03:55,740 And this was the so-called cognitive theory 82 00:03:55,740 --> 00:03:58,070 of learning and memory. 83 00:03:58,070 --> 00:04:01,310 And the hippocampus was posited beta-site 84 00:04:01,310 --> 00:04:05,930 of one property of this cognitive learning 85 00:04:05,930 --> 00:04:07,370 as applied to space. 86 00:04:07,370 --> 00:04:10,740 And the observation was actually a fairly simple one. 87 00:04:10,740 --> 00:04:12,080 If you take a rat-- 88 00:04:12,080 --> 00:04:13,700 I see you want to learn something. 89 00:04:13,700 --> 00:04:20,839 That is, if you want to learn to associate a cue with food 90 00:04:20,839 --> 00:04:24,920 or you want to train it to go over and press a lever for food 91 00:04:24,920 --> 00:04:26,616 or-- you want it to learn something, 92 00:04:26,616 --> 00:04:27,990 and you do this in an environment 93 00:04:27,990 --> 00:04:28,966 and put it in a box. 94 00:04:28,966 --> 00:04:31,090 Well, if you take a rat and you put it in that box, 95 00:04:31,090 --> 00:04:33,674 let's say, the day before, and just let it wander around. 96 00:04:33,674 --> 00:04:34,340 It does nothing. 97 00:04:34,340 --> 00:04:35,710 It just explores space. 98 00:04:35,710 --> 00:04:36,620 You take it out. 99 00:04:36,620 --> 00:04:39,230 And now we take two rats-- one that's been in the box before 100 00:04:39,230 --> 00:04:40,550 and one that has not been in the box before-- 101 00:04:40,550 --> 00:04:42,050 and the rat that has been in the box 102 00:04:42,050 --> 00:04:45,174 before will learn faster than the animal has not 103 00:04:45,174 --> 00:04:45,840 been in the box. 104 00:04:45,840 --> 00:04:49,300 And you say, why would just passive exposure to that box 105 00:04:49,300 --> 00:04:50,900 enhance it's learning? 106 00:04:50,900 --> 00:04:53,420 And this phenomena is known as latent learning. 107 00:04:53,420 --> 00:04:55,550 The animal had learned something that it could then 108 00:04:55,550 --> 00:05:02,390 apply to this new learning even though it was not instructive. 109 00:05:02,390 --> 00:05:05,340 It wasn't rewarded for doing anything, it just explored. 110 00:05:05,340 --> 00:05:08,150 And so this idea that there was some sort of latent capacity 111 00:05:08,150 --> 00:05:14,000 to enhance learning was motivated 112 00:05:14,000 --> 00:05:17,480 the study of the hippocampus in the context 113 00:05:17,480 --> 00:05:21,140 of non-head-fixed recording. 114 00:05:21,140 --> 00:05:23,440 And so it keeps real inside. 115 00:05:23,440 --> 00:05:26,930 In fact, the paper that was cited for the Nobel Prize 116 00:05:26,930 --> 00:05:29,939 was the paper in which he first recorded from these cells-- 117 00:05:29,939 --> 00:05:32,480 and all he did was just take the animals out of the ear bars. 118 00:05:32,480 --> 00:05:34,188 So he took the animal out of the ear bars 119 00:05:34,188 --> 00:05:36,750 and just let it run around on a tabletop, just like this, 120 00:05:36,750 --> 00:05:37,250 actually. 121 00:05:37,250 --> 00:05:39,650 It was a table at the University College 122 00:05:39,650 --> 00:05:42,571 London about this size where the rat just kind of wandered 123 00:05:42,571 --> 00:05:43,070 around. 124 00:05:43,070 --> 00:05:46,040 And he made observations-- oh, here's a cell that fires when 125 00:05:46,040 --> 00:05:49,820 the animal goes over to the left hand side-- 126 00:05:49,820 --> 00:05:52,250 very descriptive, but the key insight 127 00:05:52,250 --> 00:05:54,680 was to study the hippocampus when rats are 128 00:05:54,680 --> 00:05:58,670 doing what rats normally do-- 129 00:05:58,670 --> 00:05:59,510 explore in space. 130 00:05:59,510 --> 00:06:01,340 So it was this ethological approach. 131 00:06:01,340 --> 00:06:08,034 And what he discovered was that cells in the hippocampus.-- 132 00:06:08,034 --> 00:06:09,200 and this is a cross-section. 133 00:06:09,200 --> 00:06:10,769 The hippocampus is found here. 134 00:06:10,769 --> 00:06:12,185 It's in the medial temporal lobes. 135 00:06:15,329 --> 00:06:17,620 The name hippocampus comes from the Greek for seahorse, 136 00:06:17,620 --> 00:06:21,380 so in humans, it sort of looks like a seahorse. 137 00:06:21,380 --> 00:06:25,220 The regions here-- the terminology CA, 138 00:06:25,220 --> 00:06:27,950 these different fields of the hippocampus, 139 00:06:27,950 --> 00:06:29,770 comes from cornu ammonis, or Amman's horn. 140 00:06:29,770 --> 00:06:31,520 That's because it looks like a ram's horn. 141 00:06:31,520 --> 00:06:33,940 So if you think about it, it's like little ram's horns 142 00:06:33,940 --> 00:06:35,690 right here, put them in the temporal lobes 143 00:06:35,690 --> 00:06:36,500 and you move them in there, that's 144 00:06:36,500 --> 00:06:37,583 where your hippocampus is. 145 00:06:37,583 --> 00:06:39,620 Just like this. 146 00:06:39,620 --> 00:06:43,287 And if you make a cross-section, slice through that, 147 00:06:43,287 --> 00:06:44,120 this is the circuit. 148 00:06:44,120 --> 00:06:48,530 So information comes in from across the brain, 149 00:06:48,530 --> 00:06:53,670 converges in the primary input to the hippocampus-- 150 00:06:53,670 --> 00:06:59,000 that's called entorhinal cortex, which is what the other Nobel 151 00:06:59,000 --> 00:07:02,540 Prize was awarded to-- 152 00:07:02,540 --> 00:07:07,310 the husband and wife team that had discovered and identified 153 00:07:07,310 --> 00:07:10,040 the properties of cells in the entorhinal cortex, 154 00:07:10,040 --> 00:07:13,130 these so-called grid cells that seem to carry information 155 00:07:13,130 --> 00:07:18,500 about actual, it seemed like, Cartesian-like spatial 156 00:07:18,500 --> 00:07:20,559 information conveyed in the hippocampus. 157 00:07:20,559 --> 00:07:22,850 So stuff comes in from the rest of the brain, converges 158 00:07:22,850 --> 00:07:24,980 in the entorhinal cortex, and then goes around 159 00:07:24,980 --> 00:07:28,490 this little loop, from these three primary subfields-- 160 00:07:28,490 --> 00:07:34,520 the dentate gyrus, CA cornu ammonis 3, CA1, subiculum, 161 00:07:34,520 --> 00:07:35,270 and back on again. 162 00:07:35,270 --> 00:07:37,820 So it goes through this loop. 163 00:07:37,820 --> 00:07:39,320 Classically, it had been referred to 164 00:07:39,320 --> 00:07:41,420 as the trisynaptic loop. 165 00:07:41,420 --> 00:07:46,620 One synapse, dentate, CA3, CA1, and then back out. 166 00:07:46,620 --> 00:07:50,150 And again, recordings in this part of the hippocampus 167 00:07:50,150 --> 00:07:52,700 reveal the properties which I'll mention, 168 00:07:52,700 --> 00:07:57,290 and that is that the cells respond to locations in space. 169 00:07:57,290 --> 00:08:00,250 But prior to that-- 170 00:08:00,250 --> 00:08:02,780 this electrophysiological work in rodents-- 171 00:08:02,780 --> 00:08:05,920 there had been human neuropsychological work, 172 00:08:05,920 --> 00:08:09,680 in particular the seminal case of the patient HM who 173 00:08:09,680 --> 00:08:12,230 had been studied here for many decades 174 00:08:12,230 --> 00:08:15,610 before he just recently passed away. 175 00:08:15,610 --> 00:08:25,010 Known as HM, or Henry Molaison, as he is sort of recently 176 00:08:25,010 --> 00:08:32,939 revealed to be a patient who had undergone bilateral resection 177 00:08:32,939 --> 00:08:33,980 of medial temporal lobes. 178 00:08:33,980 --> 00:08:37,731 Cut out parts of his hippocampus and other 179 00:08:37,731 --> 00:08:39,480 associated medial temporal lobe structures 180 00:08:39,480 --> 00:08:41,834 to treat intractable epilepsy. 181 00:08:41,834 --> 00:08:44,250 He subsequently lost the ability to form any new memories. 182 00:08:44,250 --> 00:08:48,860 So had permanent anterograde- couldn't 183 00:08:48,860 --> 00:08:51,530 form any new memories going forward in the future and lost 184 00:08:51,530 --> 00:08:53,940 some of his older memories. 185 00:08:53,940 --> 00:08:57,470 So you lose memories in humans, rodents 186 00:08:57,470 --> 00:09:00,690 can't navigate in space, also humans have a spatial deficit, 187 00:09:00,690 --> 00:09:03,440 so there's some connection between space and memory. 188 00:09:03,440 --> 00:09:05,480 And the question is, what is that? 189 00:09:05,480 --> 00:09:09,300 What would link spatial navigation and memory? 190 00:09:09,300 --> 00:09:10,967 But not just any kind of memory-- 191 00:09:10,967 --> 00:09:12,800 what we refer to as episodic memory-- memory 192 00:09:12,800 --> 00:09:14,610 for events or experiences. 193 00:09:14,610 --> 00:09:17,420 And so the working hypothesis is that these two things 194 00:09:17,420 --> 00:09:20,570 are really connected by a need, a computation imperative, 195 00:09:20,570 --> 00:09:23,800 to maintain information about time order. 196 00:09:23,800 --> 00:09:26,390 And that is, if you're going to use experience 197 00:09:26,390 --> 00:09:27,950 to guide future behavior, you need 198 00:09:27,950 --> 00:09:30,320 to figure out what the causal relationships are 199 00:09:30,320 --> 00:09:34,020 between events in the world. 200 00:09:34,020 --> 00:09:36,560 If I see A and B, what I really want to figure out-- 201 00:09:36,560 --> 00:09:38,060 I don't just want to record the fact 202 00:09:38,060 --> 00:09:39,710 that A and B happen together. 203 00:09:39,710 --> 00:09:41,990 Ultimately, I want to try to understand 204 00:09:41,990 --> 00:09:45,170 the relationship between A and B. Did A cause B, 205 00:09:45,170 --> 00:09:49,720 or how might I actually predict B given A? 206 00:09:49,720 --> 00:09:51,560 And the way I would do that would 207 00:09:51,560 --> 00:09:54,720 be to construct some kind of simple internal model, 208 00:09:54,720 --> 00:09:57,470 a predictive model, that's based on experience. 209 00:09:57,470 --> 00:10:01,202 And so the idea is, hippocampus captures experience, 210 00:10:01,202 --> 00:10:03,160 and then sort through some process, which we'll 211 00:10:03,160 --> 00:10:05,320 refer to as the process of consolidation, 212 00:10:05,320 --> 00:10:09,370 translates experience into some working model that 213 00:10:09,370 --> 00:10:13,180 can predict events and can be used to guide behavior 214 00:10:13,180 --> 00:10:14,410 in decision making. 215 00:10:14,410 --> 00:10:16,450 And critical to that is just the idea 216 00:10:16,450 --> 00:10:21,080 of time and a particular time order. 217 00:10:21,080 --> 00:10:26,740 So as I mentioned, the use of very simple technology, 218 00:10:26,740 --> 00:10:31,030 in this case extracellular neurophysiological recording-- 219 00:10:31,030 --> 00:10:35,530 taking a tiny wire-- this wire is actually four small wires, 220 00:10:35,530 --> 00:10:38,699 each one about 10 microns across. 221 00:10:38,699 --> 00:10:40,490 You twist them together in a little bundle. 222 00:10:40,490 --> 00:10:42,430 The bundle is about 35 microns. 223 00:10:42,430 --> 00:10:45,470 Human hair is on the order of typically about 50 microns, 224 00:10:45,470 --> 00:10:49,420 so these are wires, or multi contact electrodes, 225 00:10:49,420 --> 00:10:52,280 about as large as a human hair. 226 00:10:52,280 --> 00:10:55,630 We thread these things through the little oil rig 227 00:10:55,630 --> 00:10:58,210 drilling device here, which is-- 228 00:10:58,210 --> 00:11:02,950 these little micromanipulators allow this wire 229 00:11:02,950 --> 00:11:06,220 to be driven down through very small guide cannula, 230 00:11:06,220 --> 00:11:08,680 and so out the bottom will be a number 231 00:11:08,680 --> 00:11:11,290 of these independent individual electrodes, 232 00:11:11,290 --> 00:11:15,010 each controlled by their own micromanipulator. 233 00:11:15,010 --> 00:11:16,930 So we can send these electrodes down-- 234 00:11:16,930 --> 00:11:19,720 this entire device, in this case, 235 00:11:19,720 --> 00:11:24,310 weighs anywhere from 12 to 20 grams, 236 00:11:24,310 --> 00:11:27,130 and this can be placed on an animal's head 237 00:11:27,130 --> 00:11:30,430 permanently or chronically, and that is that once it goes on, 238 00:11:30,430 --> 00:11:31,570 it doesn't come off. 239 00:11:31,570 --> 00:11:36,370 So they will have this little helmet like thing 240 00:11:36,370 --> 00:11:40,540 on their head, a small opening is made in the skull, 241 00:11:40,540 --> 00:11:44,980 the wires are sent through, and the whole surgical procedure 242 00:11:44,980 --> 00:11:48,190 takes maybe 30 to 45 minutes. 243 00:11:48,190 --> 00:11:49,670 It's like an outpatient thing. 244 00:11:49,670 --> 00:11:51,211 It would probably take longer to have 245 00:11:51,211 --> 00:11:55,060 your wisdom teeth pulled than to have a brain implant installed. 246 00:11:55,060 --> 00:11:57,340 But then once this is installed, these electrodes 247 00:11:57,340 --> 00:12:01,300 can be driven down and placed permanently. 248 00:12:01,300 --> 00:12:03,790 It gives you the ability to monitor pattern activity 249 00:12:03,790 --> 00:12:09,850 across large populations over long periods of time-- 250 00:12:09,850 --> 00:12:11,710 days, weeks, months. 251 00:12:11,710 --> 00:12:17,590 So you have tapped into activity in this brain area, 252 00:12:17,590 --> 00:12:24,060 and you can monitor as animals experience, learn, and recall. 253 00:12:24,060 --> 00:12:25,860 This is what the raw data looks like. 254 00:12:25,860 --> 00:12:28,450 This is a cartoon of the electrode with four contacts. 255 00:12:28,450 --> 00:12:30,010 The idea with the four contacts is 256 00:12:30,010 --> 00:12:32,880 it gives you the ability to triangulate location, 257 00:12:32,880 --> 00:12:37,760 much as you stereoscopically can determine depth. 258 00:12:37,760 --> 00:12:39,280 These four contacts will give you 259 00:12:39,280 --> 00:12:41,590 the ability to triangulate things 260 00:12:41,590 --> 00:12:43,330 in three dimensional space. 261 00:12:43,330 --> 00:12:46,240 One of the properties of this hippocampal circuit, and that 262 00:12:46,240 --> 00:12:48,010 is-- so this little cross-section-- you 263 00:12:48,010 --> 00:12:49,060 imagine a neuron here. 264 00:12:49,060 --> 00:12:50,560 And so there are going to be, again, 265 00:12:50,560 --> 00:12:54,970 neurons distributed across the hippocampus. 266 00:12:54,970 --> 00:12:58,960 In the rat, there are on the order of maybe 200,000 267 00:12:58,960 --> 00:13:01,750 cells in area CA1. 268 00:13:01,750 --> 00:13:05,230 In the entire hippocampus, there are on the order of one 269 00:13:05,230 --> 00:13:07,370 to two million cells. 270 00:13:07,370 --> 00:13:10,720 And if you drop an electrode any place into the hippocampus, 271 00:13:10,720 --> 00:13:13,870 you will find cells that have this kind of spatial property. 272 00:13:13,870 --> 00:13:17,650 And so what that says is that spatial information 273 00:13:17,650 --> 00:13:21,480 is distributed not in a topographic way. 274 00:13:21,480 --> 00:13:25,210 In other words, it's not one location in space 275 00:13:25,210 --> 00:13:29,590 is mapped into one location in the hippocampus. 276 00:13:29,590 --> 00:13:33,100 Unlike some of the sensory areas that have this kind 277 00:13:33,100 --> 00:13:36,490 of topography, for instance, if you record in the visual 278 00:13:36,490 --> 00:13:39,850 cortex, where there was there would be some correspondence 279 00:13:39,850 --> 00:13:43,090 between location and visual field, 280 00:13:43,090 --> 00:13:48,210 and the location of cells in the cortex that respond to that 281 00:13:48,210 --> 00:13:50,050 so-called retinotopic mapping-- 282 00:13:50,050 --> 00:13:51,880 visual field mapped onto the anatomy. 283 00:13:51,880 --> 00:13:54,400 Same thing with the somatosensory system. 284 00:13:54,400 --> 00:13:56,840 As I move and I touch different parts of the body, 285 00:13:56,840 --> 00:14:00,520 the cells that respond to that will be mapped out 286 00:14:00,520 --> 00:14:04,510 in a largely one-to-one correspondence 287 00:14:04,510 --> 00:14:10,370 between the adjacency of stimuli in the input space 288 00:14:10,370 --> 00:14:12,610 and adjacency of the representation 289 00:14:12,610 --> 00:14:15,770 in the neural space that does not occur in the hippocampus. 290 00:14:15,770 --> 00:14:18,730 Two cells that are right next to one another 291 00:14:18,730 --> 00:14:23,710 are no more likely to respond to nearby locations in space 292 00:14:23,710 --> 00:14:26,570 than two cells that are distant from one another. 293 00:14:26,570 --> 00:14:29,890 So, as we'll discuss, the principle 294 00:14:29,890 --> 00:14:33,250 of information representation in the hippocampus 295 00:14:33,250 --> 00:14:38,410 appears to be one of kind of sparse distributed patterning, 296 00:14:38,410 --> 00:14:44,620 that you have lots of cells that will respond 297 00:14:44,620 --> 00:14:48,220 to different environments and individual cells 298 00:14:48,220 --> 00:14:50,355 don't have a unique relationship to locations. 299 00:14:50,355 --> 00:14:52,480 It's really the pattern across cells that gives you 300 00:14:52,480 --> 00:14:55,504 a unique signature of code. 301 00:14:55,504 --> 00:14:56,920 We're taking advantage of the fact 302 00:14:56,920 --> 00:15:00,490 that you don't necessarily have to place the electrodes 303 00:15:00,490 --> 00:15:05,060 at the proper place in order to get responses 304 00:15:05,060 --> 00:15:06,680 in some location in space. 305 00:15:06,680 --> 00:15:09,470 Anywhere you put these electrodes in the hippocampus, 306 00:15:09,470 --> 00:15:12,100 you're going to get a certain fraction-- 307 00:15:12,100 --> 00:15:14,570 generally about 30% of the cells will respond 308 00:15:14,570 --> 00:15:16,770 in the given environment. 309 00:15:16,770 --> 00:15:18,860 And this is what those responses look like. 310 00:15:18,860 --> 00:15:23,870 I won't go through the technical details, but, needless to say, 311 00:15:23,870 --> 00:15:27,050 this just shows how events, action 312 00:15:27,050 --> 00:15:29,050 potentials that are detected-- 313 00:15:29,050 --> 00:15:31,700 in here, you see a voltage trace, you pick out 314 00:15:31,700 --> 00:15:40,070 the amplitude of these little voltage transients generated 315 00:15:40,070 --> 00:15:42,570 by action potentials in the cells, 316 00:15:42,570 --> 00:15:45,552 you plot the amplitude across these different channels. 317 00:15:45,552 --> 00:15:47,510 In this way, each point is an action potential, 318 00:15:47,510 --> 00:15:51,230 and what you see is that the amplitudes will cluster 319 00:15:51,230 --> 00:15:53,480 in a way that reflect the relative position 320 00:15:53,480 --> 00:15:55,480 or location of cells-- 321 00:15:55,480 --> 00:15:59,060 the sources relative to these wires, 322 00:15:59,060 --> 00:16:03,441 using the basic principle that if you're closer to a wire, 323 00:16:03,441 --> 00:16:04,940 the signal is going to get stronger. 324 00:16:04,940 --> 00:16:11,660 So amplitude is essentially inversely related to distance. 325 00:16:11,660 --> 00:16:15,380 So here, this is the amplitude of an action potential plot 326 00:16:15,380 --> 00:16:17,180 across two channels. 327 00:16:17,180 --> 00:16:19,520 These points are larger on channel one, 328 00:16:19,520 --> 00:16:20,900 small on channel 2. 329 00:16:20,900 --> 00:16:24,739 That means it's close to channel 1, far from channel 2. 330 00:16:24,739 --> 00:16:27,030 And then different cells will have different relations. 331 00:16:27,030 --> 00:16:29,655 This will be large on channel 2, small on channel 1, et cetera. 332 00:16:29,655 --> 00:16:32,720 So you can figure out where these cells are in space, 333 00:16:32,720 --> 00:16:37,250 means you can pick out lots of cells from, in this case, 334 00:16:37,250 --> 00:16:40,850 maybe 12 to 18 electrodes can give you 50 335 00:16:40,850 --> 00:16:42,972 to 100 or more cells. 336 00:16:42,972 --> 00:16:44,930 And then looking at the activity of those cells 337 00:16:44,930 --> 00:16:45,890 in a simple box-- 338 00:16:45,890 --> 00:16:49,310 this is just a little box, one wall removed, little ceiling 339 00:16:49,310 --> 00:16:53,390 of cubes, simple architectural design, 340 00:16:53,390 --> 00:16:56,400 nice, clean and simple-- 341 00:16:56,400 --> 00:17:01,910 what you get is clean and simple mapping of spatial locations 342 00:17:01,910 --> 00:17:03,390 into these neural responses. 343 00:17:03,390 --> 00:17:08,810 So each one of these panels represents the activity 344 00:17:08,810 --> 00:17:09,740 of individual cells. 345 00:17:09,740 --> 00:17:13,010 So this is about 80 simultaneously recorded cells. 346 00:17:13,010 --> 00:17:18,530 The color of the heat map indicates the firing rate 347 00:17:18,530 --> 00:17:20,369 of these individual cells. 348 00:17:20,369 --> 00:17:24,510 Red indicates high firing rate, blue indicates no firing. 349 00:17:24,510 --> 00:17:27,036 So this is a top down view of that box. 350 00:17:27,036 --> 00:17:29,660 So this cell, for instance, when the animal is wandering around 351 00:17:29,660 --> 00:17:32,290 in the box, the cell is silent in all the blue areas, 352 00:17:32,290 --> 00:17:34,290 and when it goes in the lower right hand corner, 353 00:17:34,290 --> 00:17:35,630 the cell fires vigorously. 354 00:17:35,630 --> 00:17:38,930 So silent, fires vigorously. 355 00:17:38,930 --> 00:17:40,610 This one fires on the left hand wall. 356 00:17:40,610 --> 00:17:42,750 This one also fires on the lower right hand side. 357 00:17:42,750 --> 00:17:46,640 So this points out the combinatorial nature 358 00:17:46,640 --> 00:17:49,010 of this spatial representation, and that 359 00:17:49,010 --> 00:17:51,290 is that if the animal is in the lower right hand 360 00:17:51,290 --> 00:17:53,810 corner of this particular box, we'll 361 00:17:53,810 --> 00:17:55,250 get these two cells to fire. 362 00:17:55,250 --> 00:17:58,580 If I take the animal and I put it into a different box, 363 00:17:58,580 --> 00:18:00,994 all of these responses will be scrambled up. 364 00:18:00,994 --> 00:18:02,660 There's nothing that says this cell will 365 00:18:02,660 --> 00:18:05,482 fire in the lower right hand corner of another box. 366 00:18:05,482 --> 00:18:07,940 And certainly, even if it does fire in the lower right hand 367 00:18:07,940 --> 00:18:12,420 corner, this other cell isn't going to fire along with it. 368 00:18:12,420 --> 00:18:15,680 And at any given location, there are roughly 1% to 5% 369 00:18:15,680 --> 00:18:17,010 of the cells that are firing. 370 00:18:17,010 --> 00:18:22,062 So at any given location, there are about maybe 5,000 cells 371 00:18:22,062 --> 00:18:23,520 in the hippocampus that are active. 372 00:18:23,520 --> 00:18:29,180 So the unique location and unique context and environments 373 00:18:29,180 --> 00:18:33,980 can be conveyed across a unique pattern of about 5,000 cells 374 00:18:33,980 --> 00:18:36,800 out of 100,000 or 200,000 cells. 375 00:18:36,800 --> 00:18:38,300 So there's large combinatorial space 376 00:18:38,300 --> 00:18:44,270 in which one can represent unique locations or potentially 377 00:18:44,270 --> 00:18:48,470 even unique experiences within different locations. 378 00:18:48,470 --> 00:18:50,510 You see that many cells are silent. 379 00:18:50,510 --> 00:18:52,160 On average, about 30% of the cells 380 00:18:52,160 --> 00:18:55,130 respond in any given environment. 381 00:18:55,130 --> 00:18:56,930 The silent cells, as we'll see, they 382 00:18:56,930 --> 00:18:59,672 can be detected when the animals are not in a running 383 00:18:59,672 --> 00:19:01,880 around or experiencing space, but when they're asleep 384 00:19:01,880 --> 00:19:04,310 or in these other offline states, 385 00:19:04,310 --> 00:19:09,500 when the hippocampus is actually thinking about other stuff. 386 00:19:09,500 --> 00:19:12,320 We'll talk about that a bit. 387 00:19:12,320 --> 00:19:15,890 And then here, you can also see a small number of these cells. 388 00:19:15,890 --> 00:19:24,230 In this case, about 5% to 10% of these cells 389 00:19:24,230 --> 00:19:28,210 seem to have elevated firing rate across the entire space. 390 00:19:28,210 --> 00:19:30,210 These are actually a different class of neurons. 391 00:19:30,210 --> 00:19:31,530 These are excitatory neurons. 392 00:19:31,530 --> 00:19:32,760 These are inhibitory neurons. 393 00:19:32,760 --> 00:19:34,430 The inhibition, the idea that you 394 00:19:34,430 --> 00:19:37,880 have circuits that can both excite and inhibit, 395 00:19:37,880 --> 00:19:42,120 and that this is used as a circuit property 396 00:19:42,120 --> 00:19:47,480 to sculpt computation is something 397 00:19:47,480 --> 00:19:48,620 that we'll also discuss. 398 00:19:48,620 --> 00:19:51,080 So there's balance between excitation as kind of a circuit 399 00:19:51,080 --> 00:19:52,400 principle that's used. 400 00:19:52,400 --> 00:19:55,080 Inhibitory cells fire all over the place. 401 00:19:55,080 --> 00:19:57,170 They're not really communicating information. 402 00:19:57,170 --> 00:20:01,740 They're really modulating the processing of information. 403 00:20:01,740 --> 00:20:04,540 And another property of these cells 404 00:20:04,540 --> 00:20:07,410 that O'Keefe termed place cells is that when 405 00:20:07,410 --> 00:20:11,970 animals are constrained to move along these limited paths-- 406 00:20:11,970 --> 00:20:14,972 in this case, this is a linear track, as we refer to it. 407 00:20:14,972 --> 00:20:16,430 It's like a little corridor, and it 408 00:20:16,430 --> 00:20:18,090 was moving down this corridor. 409 00:20:18,090 --> 00:20:20,280 When they move down a corridor, cells don't only 410 00:20:20,280 --> 00:20:22,470 fire where the animal is, but also in the direction that it's 411 00:20:22,470 --> 00:20:22,970 going. 412 00:20:22,970 --> 00:20:25,320 And so this is the first indication 413 00:20:25,320 --> 00:20:27,890 these cells are not just-- again, it's not just a GPS. 414 00:20:27,890 --> 00:20:30,360 It's not telling you where you are. 415 00:20:30,360 --> 00:20:32,010 It's at least telling you where you're 416 00:20:32,010 --> 00:20:34,650 going, maybe what you're doing. 417 00:20:34,650 --> 00:20:36,607 So here, if I look at this yellow cell, 418 00:20:36,607 --> 00:20:38,940 I can tell that the animal is not only in this location, 419 00:20:38,940 --> 00:20:44,120 but it's moving down this linear track in this direction. 420 00:20:44,120 --> 00:20:47,940 So there's going to be a unique sequence when the animal walks 421 00:20:47,940 --> 00:20:49,020 along this path-- 422 00:20:49,020 --> 00:20:50,940 yellow cell, red cell, green cell. 423 00:20:50,940 --> 00:20:53,280 So the different cells are color coded here. 424 00:20:53,280 --> 00:20:56,070 So if you look over time, you'll see 425 00:20:56,070 --> 00:20:59,850 that there will be a unique sequence of activity 426 00:20:59,850 --> 00:21:02,780 in the hippocampus that reflects the animal's actual behavior 427 00:21:02,780 --> 00:21:04,640 experienced in that space. 428 00:21:04,640 --> 00:21:07,755 So here, this is just a little movie that shows raw-- 429 00:21:07,755 --> 00:21:09,630 this is what you would actually see if you're 430 00:21:09,630 --> 00:21:10,710 running the experiment. 431 00:21:10,710 --> 00:21:12,001 This is a little top down view. 432 00:21:12,001 --> 00:21:16,200 The green circle just highlights where the rat actually is. 433 00:21:16,200 --> 00:21:17,910 The color coding-- these are the cells 434 00:21:17,910 --> 00:21:19,420 that we're picking up over here. 435 00:21:19,420 --> 00:21:22,160 This is data as it would be coming out of a set up. 436 00:21:22,160 --> 00:21:24,480 You see, this light blue cell fires here, 437 00:21:24,480 --> 00:21:26,757 the dark blue cell fires here. 438 00:21:26,757 --> 00:21:28,090 Dark blue cell, light blue cell. 439 00:21:28,090 --> 00:21:30,690 So this is the spatial firing property, place cells. 440 00:21:35,645 --> 00:21:36,976 So the animal has stopped. 441 00:21:36,976 --> 00:21:39,350 The one thing you will notice-- you saw lots of activity, 442 00:21:39,350 --> 00:21:41,460 and animal stopped there, there was this big burst of activity. 443 00:21:41,460 --> 00:21:42,959 He'll stop again a little bit later. 444 00:21:48,814 --> 00:21:49,830 Now he's moving. 445 00:21:56,700 --> 00:21:58,880 Now when he's moving , if you listen carefully, 446 00:21:58,880 --> 00:22:01,550 you'll hear that there's this background rhythm. 447 00:22:01,550 --> 00:22:04,550 There's a modulation that's going ch-ch-ch-ch. 448 00:22:04,550 --> 00:22:07,400 Background modulation activity which 449 00:22:07,400 --> 00:22:09,470 is associated with locomotion. 450 00:22:09,470 --> 00:22:11,510 So there are really two modes-- 451 00:22:11,510 --> 00:22:14,250 you're actively engaged or when you're taking information in, 452 00:22:14,250 --> 00:22:15,940 you get this rhythm. 453 00:22:15,940 --> 00:22:18,440 When you're inattentive, not engaged, not taking information 454 00:22:18,440 --> 00:22:20,106 in, but internally evaluated, the rhythm 455 00:22:20,106 --> 00:22:24,620 goes away and is replaced by these bursts of activity. 456 00:22:24,620 --> 00:22:29,480 And so these two modes, active attentive-- 457 00:22:29,480 --> 00:22:32,350 you're processing information coming in, inattentive-- 458 00:22:32,350 --> 00:22:37,010 you're evaluating information, you're thinking about stuff-- 459 00:22:37,010 --> 00:22:40,880 can be reflected in these two characteristic modes, which 460 00:22:40,880 --> 00:22:42,420 you can literally hear. 461 00:22:42,420 --> 00:22:44,277 You can hear the difference between the two. 462 00:22:44,277 --> 00:22:46,860 Obviously, I've been listening to these things for a long time 463 00:22:46,860 --> 00:22:48,545 so it's very easy for me to pick up. 464 00:22:48,545 --> 00:22:52,720 It might be harder for you but, you listen to it a little bit 465 00:22:52,720 --> 00:22:54,710 and it's very easy to distinguish these two 466 00:22:54,710 --> 00:22:55,720 different brain routes. 467 00:22:55,720 --> 00:22:57,761 So this is going to be another view of that data. 468 00:22:57,761 --> 00:23:00,380 So same data, only now, instead of showing you the raw data, 469 00:23:00,380 --> 00:23:02,210 where is the firing. 470 00:23:02,210 --> 00:23:06,080 And that's one thing about this correlate that's so compelling. 471 00:23:06,080 --> 00:23:08,660 That was literally raw data as it was coming out. 472 00:23:08,660 --> 00:23:11,480 There's no processing at all and you could see the correlate, 473 00:23:11,480 --> 00:23:14,750 you can see the spatial correlate, which tells you, 474 00:23:14,750 --> 00:23:16,880 this is not something that requires 475 00:23:16,880 --> 00:23:22,730 I have to do multiple regressions, 476 00:23:22,730 --> 00:23:25,170 and show you all of the-- 477 00:23:25,170 --> 00:23:30,260 give you some sort of statistical confidence 478 00:23:30,260 --> 00:23:32,180 that this is what the hippocampus is doing. 479 00:23:32,180 --> 00:23:34,790 You can literally see it in individual cells. 480 00:23:34,790 --> 00:23:38,345 The spatial correlate is extremely robust, compelling, 481 00:23:38,345 --> 00:23:39,780 and consistent. 482 00:23:39,780 --> 00:23:41,390 Those cells weren't selected. 483 00:23:41,390 --> 00:23:43,670 If you record any of the cells in the hippocampus, 484 00:23:43,670 --> 00:23:44,930 they're all doing that. 485 00:23:44,930 --> 00:23:47,630 The estimate is that over 95% of the cells that you 486 00:23:47,630 --> 00:23:49,350 can record in rodent hippocampus will 487 00:23:49,350 --> 00:23:50,600 have these spatial properties. 488 00:23:50,600 --> 00:23:54,410 So it's really a fundamental property of this memory system. 489 00:23:54,410 --> 00:23:56,090 So this is the same data, except now, 490 00:23:56,090 --> 00:23:57,780 instead of showing the spiking, we're 491 00:23:57,780 --> 00:24:01,250 going to use this simple Bayesian Estimation 492 00:24:01,250 --> 00:24:03,350 Algorithm to ask the question. 493 00:24:03,350 --> 00:24:08,090 "OK, if we know which cells are firing, 494 00:24:08,090 --> 00:24:11,980 can we guess the location that the animal would likely be?" 495 00:24:11,980 --> 00:24:16,970 That is, if we know the firing as a function of location, 496 00:24:16,970 --> 00:24:20,640 we can use a Bayesian inference to guess the probability 497 00:24:20,640 --> 00:24:23,176 of location, given firing. 498 00:24:23,176 --> 00:24:24,800 And that's what we're going to do here. 499 00:24:24,800 --> 00:24:27,590 Just asking, if we know which cells are firing-- 500 00:24:27,590 --> 00:24:29,420 for instance, if the blue cell is firing, 501 00:24:29,420 --> 00:24:31,820 you would say, "Oh, the animal's probably over here." 502 00:24:31,820 --> 00:24:35,150 So recording across, looking at pattern across many cells, 503 00:24:35,150 --> 00:24:39,770 and then we're coming up every 100 milliseconds, 504 00:24:39,770 --> 00:24:41,990 we're coming up with a probability 505 00:24:41,990 --> 00:24:44,600 that that pattern could have occurred if the animal was 506 00:24:44,600 --> 00:24:45,939 at any location on this track. 507 00:24:45,939 --> 00:24:48,230 And the probability is going to be shown by a triangle. 508 00:24:48,230 --> 00:24:50,040 Big triangle means high probability. 509 00:24:50,040 --> 00:24:55,580 And what you'll see is, when we do this estimation, 510 00:24:55,580 --> 00:24:59,470 when the animal's moving, you see the triangle. 511 00:24:59,470 --> 00:25:04,370 The triangles are only highly probable 512 00:25:04,370 --> 00:25:07,310 when the animal is actually moving, 513 00:25:07,310 --> 00:25:09,410 and only at the location that the animal is. 514 00:25:09,410 --> 00:25:11,130 So the hippocampus is centrally-- 515 00:25:11,130 --> 00:25:12,505 the hippocampus representation is 516 00:25:12,505 --> 00:25:15,540 tracking the location the animal and also when the animal stops. 517 00:25:15,540 --> 00:25:20,570 Now, if you listen, you heard the burst 518 00:25:20,570 --> 00:25:22,760 and you look at the triangle, when the animal stops, 519 00:25:22,760 --> 00:25:24,275 the triangle no longer corresponds 520 00:25:24,275 --> 00:25:25,290 to where the animal is. 521 00:25:25,290 --> 00:25:26,790 In fact, when you get the burst, you 522 00:25:26,790 --> 00:25:28,580 see the triangle jumping around the track. 523 00:25:28,580 --> 00:25:30,350 So again, there's these two modes 524 00:25:30,350 --> 00:25:34,010 and the representations have these two properties. 525 00:25:34,010 --> 00:25:38,240 Moving, oscillation, track current location. 526 00:25:38,240 --> 00:25:42,832 Stopping, oscillation goes away, representation now 527 00:25:42,832 --> 00:25:43,790 jumps across the track. 528 00:25:46,550 --> 00:25:50,780 And interestingly, these little bursts, like that, 529 00:25:50,780 --> 00:25:52,906 also occur, not just when the animal stops briefly, 530 00:25:52,906 --> 00:25:54,363 but when it actually goes to sleep. 531 00:25:54,363 --> 00:25:56,224 So here on the inset, this is the animal now 532 00:25:56,224 --> 00:25:57,890 has been taken off the track altogether. 533 00:25:57,890 --> 00:25:59,931 It's just sitting in a little box somewhere else, 534 00:25:59,931 --> 00:26:02,600 it's curled up, and it's asleep. 535 00:26:02,600 --> 00:26:05,240 You get these same bursts and when you decode activity, 536 00:26:05,240 --> 00:26:08,930 you find you can decode the position on the track 537 00:26:08,930 --> 00:26:11,840 and, if you look carefully, you see that the position actually 538 00:26:11,840 --> 00:26:16,320 follows a sequence of trajectory along the track. 539 00:26:16,320 --> 00:26:20,010 And we'll go into that in a little bit more detail. 540 00:26:20,010 --> 00:26:22,520 So you think of there being these two states, the offline 541 00:26:22,520 --> 00:26:23,210 and the online. 542 00:26:23,210 --> 00:26:26,960 And the online, when the animal's moving, 543 00:26:26,960 --> 00:26:30,770 the characteristic mode of activity in the hippocampus 544 00:26:30,770 --> 00:26:34,250 is an oscillatory mode, described as the theta rhythm, 545 00:26:34,250 --> 00:26:36,900 which is this 10 hertz oscillation. 546 00:26:36,900 --> 00:26:38,600 When the animal stops and becomes 547 00:26:38,600 --> 00:26:41,550 quiet and immobile, within about half a second, 548 00:26:41,550 --> 00:26:44,750 oscillation goes away and is replaced 549 00:26:44,750 --> 00:26:48,730 by these large transient aperiodic events, 550 00:26:48,730 --> 00:26:51,620 these so-called sharp waves, because 551 00:26:51,620 --> 00:26:54,530 in the extracellular electrical field 552 00:26:54,530 --> 00:26:57,847 that you measure, you see these large deflections. 553 00:26:57,847 --> 00:26:58,930 And then, if you zoom in-- 554 00:26:58,930 --> 00:27:02,310 I'll also show you shortly-- 555 00:27:02,310 --> 00:27:05,090 you can actually see that these very high frequency 556 00:27:05,090 --> 00:27:08,630 oscillations, about 100 to 200 hertz, riding on top of that, 557 00:27:08,630 --> 00:27:10,520 these are referred to as, ripples. 558 00:27:10,520 --> 00:27:16,280 And this is the term that Gyorgy Buzsaki came up with. 559 00:27:16,280 --> 00:27:18,800 He described this sharp wave ripple activity 560 00:27:18,800 --> 00:27:24,140 as corresponding to this offline state, quiet wakeful, 561 00:27:24,140 --> 00:27:28,040 and some sleep states. 562 00:27:28,040 --> 00:27:32,150 So the first thing is to look at this oscillatory state. 563 00:27:32,150 --> 00:27:33,779 So an animal's actually moving. 564 00:27:33,779 --> 00:27:36,320 And I've already shown you the spatial correlate, these place 565 00:27:36,320 --> 00:27:36,830 cells. 566 00:27:36,830 --> 00:27:39,740 But there's another property of these cells 567 00:27:39,740 --> 00:27:42,020 that O'Keefe also discovered. 568 00:27:42,020 --> 00:27:45,777 But this was in early 1990s, around 1991. 569 00:27:45,777 --> 00:27:47,360 And that is, he noted, if you actually 570 00:27:47,360 --> 00:27:50,870 look at the discharge of single spikes with respect 571 00:27:50,870 --> 00:27:53,180 to this oscillation, first of all, 572 00:27:53,180 --> 00:27:55,060 the spikes are actually phase locked. 573 00:27:55,060 --> 00:27:57,230 In other words, here spikes fire when 574 00:27:57,230 --> 00:28:01,160 the oscillation, this theta rhythm, is at its peak. 575 00:28:01,160 --> 00:28:03,170 So the idea is, this oscillation really 576 00:28:03,170 --> 00:28:05,930 reflects time varying excitability, 577 00:28:05,930 --> 00:28:08,780 sometimes where cells likely to fire, 578 00:28:08,780 --> 00:28:09,910 other times when it's not. 579 00:28:09,910 --> 00:28:14,360 And you can think of this oscillation in excitability 580 00:28:14,360 --> 00:28:19,040 as being an oscillation in relative excitability, 581 00:28:19,040 --> 00:28:19,820 or inhibitions. 582 00:28:19,820 --> 00:28:22,490 So inhibition, elevation is high, it's low, 583 00:28:22,490 --> 00:28:23,990 its high, it's low. 584 00:28:23,990 --> 00:28:25,850 When inhibition is low, cell fires. 585 00:28:25,850 --> 00:28:27,860 When it's high, cell doesn't fire. 586 00:28:27,860 --> 00:28:30,030 So this is modulation excitability. 587 00:28:30,030 --> 00:28:30,780 You see this here. 588 00:28:30,780 --> 00:28:32,280 But he noted another thing, and that 589 00:28:32,280 --> 00:28:34,610 is, if you look at the precise phase-- 590 00:28:34,610 --> 00:28:37,760 so here, yes, cells tend to fire at the peak, 591 00:28:37,760 --> 00:28:41,330 but here, this is now over time, but animal 592 00:28:41,330 --> 00:28:44,210 moving at constant velocity, space, and time, 593 00:28:44,210 --> 00:28:45,840 are interchangeable. 594 00:28:45,840 --> 00:28:48,890 So as the animal's moving through it's place field, 595 00:28:48,890 --> 00:28:53,690 the spikes start to fire earlier and earlier, as this phase 596 00:28:53,690 --> 00:28:56,250 code, spatial location. 597 00:28:56,250 --> 00:28:59,520 So you can tell, based upon the relative phase, 598 00:28:59,520 --> 00:29:03,050 whether the animal is just entering the field, spikes fire 599 00:29:03,050 --> 00:29:06,650 late, here close to the peak, versus further 600 00:29:06,650 --> 00:29:09,735 into the field, where they fire a little bit earlier. 601 00:29:09,735 --> 00:29:11,360 There's a relationship between distance 602 00:29:11,360 --> 00:29:14,480 into the field and relative phase. 603 00:29:14,480 --> 00:29:16,280 What he termed, phase precession. 604 00:29:16,280 --> 00:29:17,840 And this is a representation of that. 605 00:29:17,840 --> 00:29:19,126 This is actual data. 606 00:29:19,126 --> 00:29:20,750 This is just a cartoon that illustrates 607 00:29:20,750 --> 00:29:22,070 the basic principle. 608 00:29:22,070 --> 00:29:24,507 And so the idea that, if I have a place cell, 609 00:29:24,507 --> 00:29:26,090 if I look at the marginal distribution 610 00:29:26,090 --> 00:29:28,680 as a function of location, this would be the place field. 611 00:29:28,680 --> 00:29:30,990 Otherwise, it would just collapse all of this. 612 00:29:30,990 --> 00:29:34,700 This is the spiking now, as a function of position and phase. 613 00:29:34,700 --> 00:29:37,190 If I just look at spiking as a function of position, 614 00:29:37,190 --> 00:29:40,430 as just the density of firing, what I would get 615 00:29:40,430 --> 00:29:42,140 is a place field. 616 00:29:42,140 --> 00:29:45,500 Not many spikes here as the animal enters the field, lots 617 00:29:45,500 --> 00:29:47,210 of spikes here as you get into-- this 618 00:29:47,210 --> 00:29:50,244 would be the classic spatial receptive field. 619 00:29:50,244 --> 00:29:52,410 But now, if I introduce phase, as well, now you say, 620 00:29:52,410 --> 00:29:54,576 "Oh, wait, there's a systematic relationship as well 621 00:29:54,576 --> 00:29:57,930 between phase and location." 622 00:29:57,930 --> 00:30:02,600 In fact, phase is a better predictor of relative location 623 00:30:02,600 --> 00:30:04,265 than firing rate. 624 00:30:04,265 --> 00:30:07,220 If I asked, when did the spike occur? 625 00:30:07,220 --> 00:30:11,150 If it occurred here, late in phase, 626 00:30:11,150 --> 00:30:13,570 then I know the animal is just entering the field. 627 00:30:13,570 --> 00:30:15,850 If it occurs early in phase, I know 628 00:30:15,850 --> 00:30:18,200 it's at the end of its field. 629 00:30:18,200 --> 00:30:21,490 And that's interesting. 630 00:30:21,490 --> 00:30:24,320 And this is a simple model that says, well, 631 00:30:24,320 --> 00:30:26,510 an easy way of explaining that is, if you just 632 00:30:26,510 --> 00:30:27,925 have this excitability model-- 633 00:30:27,925 --> 00:30:29,300 and this was one of the questions 634 00:30:29,300 --> 00:30:31,880 that I guess you asked about this sweeping inhibition. 635 00:30:31,880 --> 00:30:34,940 So if we just imagine that you have an input, 636 00:30:34,940 --> 00:30:39,050 excitatory input, shown in blue, and then an inhibitory input, 637 00:30:39,050 --> 00:30:42,080 shown in red, where the inhibition is oscillating, 638 00:30:42,080 --> 00:30:47,000 and you apply a very simple biophysical model 639 00:30:47,000 --> 00:30:49,540 that says, a spike action potential is 640 00:30:49,540 --> 00:30:51,790 going to occur when excitation and blue exceeds 641 00:30:51,790 --> 00:30:52,750 inhibition and red. 642 00:30:52,750 --> 00:30:54,920 More excitation inhibition, you get a spike. 643 00:30:54,920 --> 00:31:01,270 So here, in this oscillation, anywhere the red trace 644 00:31:01,270 --> 00:31:03,640 is higher than the blue trace, no spiking. 645 00:31:03,640 --> 00:31:07,180 So in this case, when excitation is low, 646 00:31:07,180 --> 00:31:09,370 you have to wait until inhibition drops all the way 647 00:31:09,370 --> 00:31:11,050 to here to get a spike. 648 00:31:11,050 --> 00:31:12,770 This is late-phase. 649 00:31:12,770 --> 00:31:14,440 So weak excitation means you have 650 00:31:14,440 --> 00:31:17,080 to wait until inhibition is low. 651 00:31:17,080 --> 00:31:18,847 So this is the sweeping inhibition, 652 00:31:18,847 --> 00:31:20,180 you have to wait until it's low. 653 00:31:20,180 --> 00:31:23,560 When excitation is strong, I don't have to wait so long. 654 00:31:23,560 --> 00:31:25,250 It can fire earlier. 655 00:31:25,250 --> 00:31:27,730 So the principle here is that there is a relationship 656 00:31:27,730 --> 00:31:29,260 between magnitude and latency. 657 00:31:29,260 --> 00:31:30,376 Stronger means earlier. 658 00:31:30,376 --> 00:31:31,750 That's the biophysical principle. 659 00:31:31,750 --> 00:31:32,500 Very simple. 660 00:31:32,500 --> 00:31:36,430 Stronger earlier, weaker later. 661 00:31:36,430 --> 00:31:39,510 That will give you this phase precession property. 662 00:31:39,510 --> 00:31:41,800 Phase precession, you might think 663 00:31:41,800 --> 00:31:44,110 of as this biophysical curiosity. 664 00:31:44,110 --> 00:31:45,850 But when you think about how that 665 00:31:45,850 --> 00:31:49,060 would apply when you have more than one place cell-- in fact, 666 00:31:49,060 --> 00:31:53,140 I have two place cells here, one in blue, one here in purple, 667 00:31:53,140 --> 00:31:56,270 where place cell one is to the left of place cell two. 668 00:31:56,270 --> 00:31:57,920 So as the animal's going through here, 669 00:31:57,920 --> 00:32:00,880 first the blue cell will fire, and then, the purple cell 670 00:32:00,880 --> 00:32:01,900 will fire. 671 00:32:01,900 --> 00:32:04,700 And if you look at the excitatory 672 00:32:04,700 --> 00:32:08,960 drive, shown here as a ramp in blue, ramp in purple, 673 00:32:08,960 --> 00:32:10,750 and now you ask, "When is the blue cell 674 00:32:10,750 --> 00:32:13,420 going to fire, and the purple going to fire, 675 00:32:13,420 --> 00:32:16,407 during each one of these oscillatory cycles?" 676 00:32:16,407 --> 00:32:18,490 The answer will be, well, because the blue's-- the 677 00:32:18,490 --> 00:32:21,799 excitatory drives the blue cells higher than the purple cell, 678 00:32:21,799 --> 00:32:23,590 blue is always going to fire before purple. 679 00:32:23,590 --> 00:32:25,960 In fact, in each and every cycle, 680 00:32:25,960 --> 00:32:28,360 it will be blue purple blue purple blue purple. 681 00:32:28,360 --> 00:32:32,530 So this principle of phase precession, 682 00:32:32,530 --> 00:32:35,590 or phase coding, for single cells, 683 00:32:35,590 --> 00:32:39,490 actually gives you sequential encoding across a population 684 00:32:39,490 --> 00:32:41,800 that you will actually on each and every cycle 685 00:32:41,800 --> 00:32:43,360 have a sequence. 686 00:32:43,360 --> 00:32:45,520 The code in the hippocampus is not a location, 687 00:32:45,520 --> 00:32:49,360 it's actually a sequence, a trajectory. 688 00:32:49,360 --> 00:32:50,860 This is just raw data, but I'm going 689 00:32:50,860 --> 00:32:52,240 to quickly go-- this is just talking about what 690 00:32:52,240 --> 00:32:53,330 the spiking looks like. 691 00:32:53,330 --> 00:32:54,788 And you can see, if you look at it, 692 00:32:54,788 --> 00:32:57,490 you see these spike sequences. 693 00:32:57,490 --> 00:33:00,100 But the property's more clearly demonstrated 694 00:33:00,100 --> 00:33:01,780 when I do this decoding. 695 00:33:01,780 --> 00:33:03,650 So what I'm showing you here-- 696 00:33:03,650 --> 00:33:04,720 this is, again, raw data. 697 00:33:04,720 --> 00:33:06,386 But now, instead of showing the spiking, 698 00:33:06,386 --> 00:33:08,290 I'm showing you that the result of doing 699 00:33:08,290 --> 00:33:11,530 this decoding, this Bayesian decoding, where there's 700 00:33:11,530 --> 00:33:15,550 a probability estimate that given spiking activity 701 00:33:15,550 --> 00:33:18,190 in each one of a set of successive 20 702 00:33:18,190 --> 00:33:20,560 millisecond windows-- 703 00:33:20,560 --> 00:33:22,900 so I'm walking along, now I'm firing 704 00:33:22,900 --> 00:33:25,630 at a finer temporal resolution every 20 milliseconds. 705 00:33:25,630 --> 00:33:27,580 I'm doing the same decoding, probability 706 00:33:27,580 --> 00:33:30,562 is indicated in grayscale. 707 00:33:30,562 --> 00:33:32,020 So you can see here the dark areas, 708 00:33:32,020 --> 00:33:34,147 this is high probability that this pattern-- 709 00:33:34,147 --> 00:33:35,980 the animal would have been in this location, 710 00:33:35,980 --> 00:33:37,780 or to get this pattern of activity 711 00:33:37,780 --> 00:33:39,240 in this 20 millisecond bin. 712 00:33:39,240 --> 00:33:42,560 And so, what you can see here is that the probability, 713 00:33:42,560 --> 00:33:45,610 these Bayesian decoded probabilities, 714 00:33:45,610 --> 00:33:49,630 form short sequences every single theta cycle, 715 00:33:49,630 --> 00:33:51,460 what we refer to as theta sequences. 716 00:33:51,460 --> 00:33:54,130 And the theta sequences actually move from just 717 00:33:54,130 --> 00:33:55,267 behind the animal-- 718 00:33:55,267 --> 00:33:57,100 dotted line is where the animal actually is. 719 00:33:57,100 --> 00:33:59,200 The estimate goes from just behind the animal to just 720 00:33:59,200 --> 00:34:00,158 in front of the animal. 721 00:34:00,158 --> 00:34:02,710 So 10 times a second the hippocampus 722 00:34:02,710 --> 00:34:05,530 is actually expressing a representation 723 00:34:05,530 --> 00:34:08,707 of spatial sequence that goes from behind 724 00:34:08,707 --> 00:34:09,790 to in front of the animal. 725 00:34:09,790 --> 00:34:11,560 You can think of behind and in front 726 00:34:11,560 --> 00:34:16,070 as also reflecting recent past and near future. 727 00:34:16,070 --> 00:34:24,190 So there is this relative predictive differentiation 728 00:34:24,190 --> 00:34:28,610 of response as a function of oscillatory phase. 729 00:34:28,610 --> 00:34:31,540 So if I want to look at, "Gee, where did I just come from?" 730 00:34:31,540 --> 00:34:34,540 I just have to look at activity here, 731 00:34:34,540 --> 00:34:36,909 at this slightly earlier phase. 732 00:34:36,909 --> 00:34:39,730 If I want to ask the question, "Gee, where am I likely to go 733 00:34:39,730 --> 00:34:41,040 to?" 734 00:34:41,040 --> 00:34:43,449 I simply have to shift the phase, the channel that I 735 00:34:43,449 --> 00:34:45,520 look at here, and I can see where 736 00:34:45,520 --> 00:34:46,989 the likely future location is. 737 00:34:46,989 --> 00:34:50,260 So you can think that there is actually a code, 738 00:34:50,260 --> 00:34:55,929 not just of location, but of the relative causal relationship 739 00:34:55,929 --> 00:34:58,794 between locations mapped into phase. 740 00:34:58,794 --> 00:35:00,460 And we'll see that we can experimentally 741 00:35:00,460 --> 00:35:03,400 test this idea, that that's just not 742 00:35:03,400 --> 00:35:07,420 an artifact of our decoding, but the animals are actually 743 00:35:07,420 --> 00:35:11,680 using information of these different phases 744 00:35:11,680 --> 00:35:16,810 to drive spatial navigation in particular ways 745 00:35:16,810 --> 00:35:20,680 by using some of the tools for manipulating activity 746 00:35:20,680 --> 00:35:23,560 in a closed loop optogenetics. 747 00:35:23,560 --> 00:35:26,050 So we see we can manipulate activity, specifically 748 00:35:26,050 --> 00:35:28,660 manipulate activity, hippocampal activity, different phases 749 00:35:28,660 --> 00:35:30,880 and show these different phases actually 750 00:35:30,880 --> 00:35:33,380 carry different functional consequences. 751 00:35:33,380 --> 00:35:35,120 So we've got these sequences. 752 00:35:35,120 --> 00:35:36,730 So we've got these sequences captured 753 00:35:36,730 --> 00:35:38,560 during these oscillations. 754 00:35:38,560 --> 00:35:40,210 Are these things actually meaningful? 755 00:35:40,210 --> 00:35:42,160 Well, some of the indications that, 756 00:35:42,160 --> 00:35:44,840 A, the oscillations in the sequences are meaningful 757 00:35:44,840 --> 00:35:48,050 first come from the observation that, as I mentioned, 758 00:35:48,050 --> 00:35:49,582 actually successful, using memory-- 759 00:35:49,582 --> 00:35:52,040 information that the hippocampus requires that you actually 760 00:35:52,040 --> 00:35:54,530 communicate this to executive structures 761 00:35:54,530 --> 00:35:56,662 that can guide behavior, make decisions. 762 00:35:56,662 --> 00:35:58,370 That would include the prefrontal cortex. 763 00:35:58,370 --> 00:36:01,620 The portion of the prefrontal cortex, 764 00:36:01,620 --> 00:36:03,640 which form part of this limbic circuit, 765 00:36:03,640 --> 00:36:07,770 referred to as the limbic prefrontal cortex, 766 00:36:07,770 --> 00:36:12,240 which have direct connections from the hippocampus. 767 00:36:12,240 --> 00:36:13,990 These structures are-- in the hippocampus, 768 00:36:13,990 --> 00:36:16,940 you have deficits in spatial learning memory. 769 00:36:16,940 --> 00:36:22,370 Prefrontal cortex-- you think of this as deficits 770 00:36:22,370 --> 00:36:24,770 in working memory and retrieval, executive control, 771 00:36:24,770 --> 00:36:27,020 decision-making. 772 00:36:27,020 --> 00:36:32,270 But you can think of these two things as working together. 773 00:36:32,270 --> 00:36:36,770 Hippocampus providing information 774 00:36:36,770 --> 00:36:40,520 that the prefrontal cortex can use in order to direct behavior 775 00:36:40,520 --> 00:36:41,425 and decision-making. 776 00:36:41,425 --> 00:36:43,550 And you damage either one of these things, animals, 777 00:36:43,550 --> 00:36:45,341 rats can't find their way around the space. 778 00:36:45,341 --> 00:36:47,490 You damage either one of these in humans, 779 00:36:47,490 --> 00:36:50,840 you're going to get memory deficits. 780 00:36:50,840 --> 00:36:53,870 Dementia is-- when you think of dementia as being problems 781 00:36:53,870 --> 00:36:57,410 of cognition and memory, you can get them temporal lobe 782 00:36:57,410 --> 00:37:00,320 dementias, frontal lobe dementias, you can have-- 783 00:37:03,240 --> 00:37:07,830 they're really very closely related. 784 00:37:07,830 --> 00:37:10,810 And so we can look at a simple task. 785 00:37:10,810 --> 00:37:13,420 This is a simple task, testing a so-called working memory. 786 00:37:13,420 --> 00:37:15,060 Just remember, what did you do last? 787 00:37:15,060 --> 00:37:17,630 And this task, it's just a simple alternation, 788 00:37:17,630 --> 00:37:18,950 where you turn left-- 789 00:37:18,950 --> 00:37:21,140 first time you turn left, next time to go right. 790 00:37:21,140 --> 00:37:22,970 Ethologically, it's like, "Look, I just 791 00:37:22,970 --> 00:37:26,290 got food over here, why don't I check out the other place? 792 00:37:26,290 --> 00:37:29,790 I'll check out places where I didn't get food." 793 00:37:29,790 --> 00:37:31,880 It's a so-called win-shift strategy. 794 00:37:31,880 --> 00:37:35,150 I got something here this time, I'm going to go someplace else. 795 00:37:35,150 --> 00:37:38,510 As opposed to-- a reference memory strategy 796 00:37:38,510 --> 00:37:40,880 would be referred to as a win-stay. 797 00:37:40,880 --> 00:37:42,000 That's really a good spot. 798 00:37:42,000 --> 00:37:43,310 Home. 799 00:37:43,310 --> 00:37:47,942 I love going home because I've got my TV, I got my microwave, 800 00:37:47,942 --> 00:37:49,400 I'm going to stick with that place. 801 00:37:49,400 --> 00:37:52,564 Win-stay, you keep going back to the place that rewards you. 802 00:37:52,564 --> 00:37:54,980 And so you can think of both-- again, both of these things 803 00:37:54,980 --> 00:37:56,450 have ethological value. 804 00:37:56,450 --> 00:37:58,460 Home is a good place, that's a reference place. 805 00:37:58,460 --> 00:38:01,430 If you're foraging, win-shift, don't go back 806 00:38:01,430 --> 00:38:02,990 to the place I just looted, right? 807 00:38:07,190 --> 00:38:10,250 Again, these two structures-- 808 00:38:13,490 --> 00:38:16,330 classically, you think of-- in the prefrontal cortex we 809 00:38:16,330 --> 00:38:17,990 think of working memory cells. 810 00:38:17,990 --> 00:38:20,460 And this is the short term, the idea of short term memory. 811 00:38:20,460 --> 00:38:21,920 So prefrontal cortex typically has 812 00:38:21,920 --> 00:38:23,960 been thought of as subservient. 813 00:38:23,960 --> 00:38:25,790 Working memory, where working memory 814 00:38:25,790 --> 00:38:29,540 is about holding information over short delays. 815 00:38:29,540 --> 00:38:34,006 It's an overlap in terminology, where 816 00:38:34,006 --> 00:38:35,630 working memory in the prefrontal cortex 817 00:38:35,630 --> 00:38:37,610 is really thinking about time. 818 00:38:37,610 --> 00:38:39,290 Working memory in hippocampus is really 819 00:38:39,290 --> 00:38:43,460 thinking about the context in which it's used, 820 00:38:43,460 --> 00:38:47,600 relative, session specific information, trial 821 00:38:47,600 --> 00:38:48,933 specific information. 822 00:38:52,180 --> 00:38:54,100 So I'm not going to go through the details. 823 00:38:54,100 --> 00:38:55,225 You have these two systems. 824 00:38:57,570 --> 00:39:00,450 One interesting property, if you look at these two systems 825 00:39:00,450 --> 00:39:04,140 simultaneously, what you find is that the same idea 826 00:39:04,140 --> 00:39:08,515 about oscillatory phase governing the firing of cells 827 00:39:08,515 --> 00:39:10,890 in the hippocampus also applies in the prefrontal cortex. 828 00:39:10,890 --> 00:39:12,900 That is, cells in the prefrontal cortex 829 00:39:12,900 --> 00:39:15,450 like to fire at a certain phase of the hippocampal theta 830 00:39:15,450 --> 00:39:15,950 rhythm. 831 00:39:15,950 --> 00:39:19,560 They care about, their listening to, 832 00:39:19,560 --> 00:39:22,980 this rhythm in the hippocampus. 833 00:39:22,980 --> 00:39:25,060 And so, we can look at a task. 834 00:39:25,060 --> 00:39:26,775 This is basically the same kind of task, 835 00:39:26,775 --> 00:39:28,400 only now there's a back-to-back T-maze. 836 00:39:28,400 --> 00:39:30,191 So there can be two T-mazes, and we're just 837 00:39:30,191 --> 00:39:33,680 going to do a simple variation on that, 838 00:39:33,680 --> 00:39:40,961 on that working memory task, in which the animal's 839 00:39:40,961 --> 00:39:42,460 going to start from one of these two 840 00:39:42,460 --> 00:39:44,180 arms-- it's going to walk down this arm 841 00:39:44,180 --> 00:39:48,700 and then, it has to go to the arm that's on the same side 842 00:39:48,700 --> 00:39:49,760 as it started from. 843 00:39:49,760 --> 00:39:52,540 So it has to remember, "Where did I start from? 844 00:39:52,540 --> 00:39:54,150 Oh, that's the side I have to go to." 845 00:39:54,150 --> 00:39:55,650 Then it's going to turn around, it's 846 00:39:55,650 --> 00:39:57,186 going to come back, run down here, 847 00:39:57,186 --> 00:39:59,560 and then, we're going to force it into one of these arms. 848 00:39:59,560 --> 00:40:02,800 So two back-to-back T-mazes. 849 00:40:02,800 --> 00:40:06,560 In this direction, the animal chooses, in this direction, 850 00:40:06,560 --> 00:40:07,545 we choose. 851 00:40:07,545 --> 00:40:10,030 In this direction, there's a working memory demand. 852 00:40:10,030 --> 00:40:11,900 It has to remember where it came from. 853 00:40:11,900 --> 00:40:14,230 In this direction, doesn't have to remember anything, 854 00:40:14,230 --> 00:40:15,396 doesn't make any difference. 855 00:40:15,396 --> 00:40:19,060 So behaviorally, it's symmetric, but cognitively, here there's 856 00:40:19,060 --> 00:40:21,310 a working memory demand, here there's not. 857 00:40:21,310 --> 00:40:24,610 And then we're going to look at all these things, oscillations, 858 00:40:24,610 --> 00:40:31,320 spiking, and the oscillations, look at the peak time, 859 00:40:31,320 --> 00:40:34,450 look at the firing with respect to that, firing, 860 00:40:34,450 --> 00:40:39,220 spiking, as a function of this oscillation in both structures. 861 00:40:39,220 --> 00:40:42,520 The bottom line is what you find is, 862 00:40:42,520 --> 00:40:47,680 that this is a measure of relative phase locking 863 00:40:47,680 --> 00:40:55,180 and as how well do prefrontal cells actually 864 00:40:55,180 --> 00:41:00,100 lock to this data oscillation. 865 00:41:00,100 --> 00:41:02,775 And what you find is that the degree to which they locked 866 00:41:02,775 --> 00:41:06,730 the theta rhythm is a function of whether or not the animal 867 00:41:06,730 --> 00:41:08,030 has to choose. 868 00:41:08,030 --> 00:41:10,812 So the red ones are when the animal's going down the arm 869 00:41:10,812 --> 00:41:11,770 and it's got to choose. 870 00:41:11,770 --> 00:41:15,710 The gray one is when it doesn't have to choose, when we choose. 871 00:41:15,710 --> 00:41:18,910 Then in addition, the solid red is 872 00:41:18,910 --> 00:41:23,560 when the animal had to choose and it got it right. 873 00:41:23,560 --> 00:41:25,690 And the stipple red is when it had to choose 874 00:41:25,690 --> 00:41:27,190 and it got it wrong. 875 00:41:27,190 --> 00:41:30,940 So the degree to which the prefrontal cortex actually 876 00:41:30,940 --> 00:41:35,290 locks, successfully locks, to the hippocampal theta rhythm, 877 00:41:35,290 --> 00:41:37,360 predicts whether the animal actually makes 878 00:41:37,360 --> 00:41:38,530 the correct choice or not. 879 00:41:38,530 --> 00:41:41,620 So it's as though this is a channel that 880 00:41:41,620 --> 00:41:45,340 is necessary for effectively communicating information 881 00:41:45,340 --> 00:41:48,190 between the two structures. 882 00:41:48,190 --> 00:41:54,760 And, that not only is it in the blocking of the spikes, 883 00:41:54,760 --> 00:41:57,980 spikes in the prefrontal cortex, the rhythm in the hippocampus, 884 00:41:57,980 --> 00:42:00,580 it also comes in the blocking of the rhythms themselves. 885 00:42:00,580 --> 00:42:02,320 So you can think of there being-- 886 00:42:02,320 --> 00:42:06,070 this is the theta rhythm that you 887 00:42:06,070 --> 00:42:09,260 can detect in the prefrontal cortex and in the hippocampus. 888 00:42:09,260 --> 00:42:11,470 And what you can see is there are two conditions. 889 00:42:11,470 --> 00:42:14,230 One, the animal's actually choosing, is running down here, 890 00:42:14,230 --> 00:42:15,244 it's making a choice. 891 00:42:15,244 --> 00:42:17,410 And in about the half second before the animal makes 892 00:42:17,410 --> 00:42:20,050 a choice, you see these two rhythms actually 893 00:42:20,050 --> 00:42:21,820 lock, they become coherent. 894 00:42:21,820 --> 00:42:26,530 So transient coherence is predictive 895 00:42:26,530 --> 00:42:30,040 of correct choice behavior. 896 00:42:30,040 --> 00:42:33,910 The thinking is that the rhythms themselves 897 00:42:33,910 --> 00:42:36,970 can be generating coordinated through the regulation 898 00:42:36,970 --> 00:42:38,810 of these local inhibitory circuits. 899 00:42:38,810 --> 00:42:41,060 In fact, there's been a lot of interest, for instance, 900 00:42:41,060 --> 00:42:45,130 in the relationship of local inhibitory 901 00:42:45,130 --> 00:42:49,550 control and neuropsychiatric disease and disorders. 902 00:42:49,550 --> 00:42:53,170 So, for instance, disrupting local inhibitory rhythms. 903 00:42:53,170 --> 00:42:54,700 In particular, some, for instance, 904 00:42:54,700 --> 00:42:58,300 the theta rhythms in prefrontal cortex 905 00:42:58,300 --> 00:43:00,729 can be associated with disorders like schizophrenia, 906 00:43:00,729 --> 00:43:01,270 for instance. 907 00:43:01,270 --> 00:43:04,840 So the inability to effectively impose these modes 908 00:43:04,840 --> 00:43:09,820 and then synchronize these modes can introduce disruptions 909 00:43:09,820 --> 00:43:13,900 in the ability to communicate or use memory. 910 00:43:13,900 --> 00:43:16,390 Cognitive disruption coming through disruption 911 00:43:16,390 --> 00:43:21,670 of these oscillations through disruption of inhibition 912 00:43:21,670 --> 00:43:23,110 in these local circuits. 913 00:43:23,110 --> 00:43:28,900 Now how you actually coordinate these two subjective states. 914 00:43:28,900 --> 00:43:33,710 We've been looking at the role of mid-line thalamic nuclei. 915 00:43:33,710 --> 00:43:36,820 So the thalamus as being a set of structures 916 00:43:36,820 --> 00:43:39,010 that have widespread connectivity to all 917 00:43:39,010 --> 00:43:43,030 these cortical areas, that much of their connectivity 918 00:43:43,030 --> 00:43:45,970 is inhibitory, and so they have the ability to actually 919 00:43:45,970 --> 00:43:51,640 coordinate, modulate and coordinate, these oscillatory 920 00:43:51,640 --> 00:43:52,570 modes. 921 00:43:52,570 --> 00:43:57,230 We've even just recently published on individual cells 922 00:43:57,230 --> 00:43:59,410 we found in some of these mid-line thalamic nuclei 923 00:43:59,410 --> 00:44:02,050 that, for instance, will branch. 924 00:44:02,050 --> 00:44:04,390 Single cells will branch and target cells 925 00:44:04,390 --> 00:44:06,300 in the hippocampus and the prefrontal cortex. 926 00:44:06,300 --> 00:44:08,470 So they would be ideally positioned 927 00:44:08,470 --> 00:44:11,380 to introduce, to select and impose, 928 00:44:11,380 --> 00:44:13,780 this synchronization there. 929 00:44:13,780 --> 00:44:17,590 So that's the way we think about a lot 930 00:44:17,590 --> 00:44:20,980 of this dynamic connectivity being established, 931 00:44:20,980 --> 00:44:24,610 presumably through some sort of thalamic regulation. 932 00:44:24,610 --> 00:44:26,470 And then, you think about the thalamus 933 00:44:26,470 --> 00:44:30,430 as being regulated by those so-called thalamic reticular 934 00:44:30,430 --> 00:44:33,430 nucleus that regulates the thalamus 935 00:44:33,430 --> 00:44:35,164 and it sets up these modes. 936 00:44:35,164 --> 00:44:36,580 And there's also a lot of interest 937 00:44:36,580 --> 00:44:39,490 now in the thalamic reticular nucleus 938 00:44:39,490 --> 00:44:41,560 in disease and disorders. 939 00:44:41,560 --> 00:44:45,250 A lot of the genetic screening has 940 00:44:45,250 --> 00:44:49,527 identified targets in the thalamic reticular nucleus. 941 00:44:49,527 --> 00:44:50,860 So you can think about it again. 942 00:44:50,860 --> 00:44:58,930 It's like the oscillator in your computer or your radio that 943 00:44:58,930 --> 00:45:01,210 breaks down, you can't-- 944 00:45:01,210 --> 00:45:03,667 the information can be there, but you can't tune into it. 945 00:45:03,667 --> 00:45:05,000 So it's this fundamental tuning. 946 00:45:05,000 --> 00:45:06,100 You've got to have the modes and you've 947 00:45:06,100 --> 00:45:07,990 got to be able to lock to these frequencies. 948 00:45:07,990 --> 00:45:10,180 And then, beyond that, as we'll see, 949 00:45:10,180 --> 00:45:12,760 it's not just the frequencies, but it's also 950 00:45:12,760 --> 00:45:16,334 the precise phase within those oscillatory modes that 951 00:45:16,334 --> 00:45:17,500 carry different information. 952 00:45:17,500 --> 00:45:21,220 That we actually determine by taking advantage 953 00:45:21,220 --> 00:45:24,130 of these techniques, optogenetic techniques, 954 00:45:24,130 --> 00:45:25,760 for targeted manipulation. 955 00:45:25,760 --> 00:45:28,420 So you infect the inhibitory neurons 956 00:45:28,420 --> 00:45:32,980 in the hippocampus with this optogenetically 957 00:45:32,980 --> 00:45:39,880 encoded and controllable channel so that we can optically 958 00:45:39,880 --> 00:45:42,760 excite inhibitory cells. 959 00:45:42,760 --> 00:45:47,660 So we infect this excitatory channel into inhibitory cells, 960 00:45:47,660 --> 00:45:50,860 and then, we can transiently, giving very brief pulses 961 00:45:50,860 --> 00:45:55,840 of laser light, we can activate drive, inhibitory cells, 962 00:45:55,840 --> 00:45:57,610 and then, because the local circuit, 963 00:45:57,610 --> 00:46:01,160 those inhibitory cells will inhibit the excitatory cells. 964 00:46:01,160 --> 00:46:05,530 And so we have about 20-25 millisecond control. 965 00:46:05,530 --> 00:46:12,580 And now, we can lock, or control, 966 00:46:12,580 --> 00:46:16,370 inhibition based upon the phase of this oscillation. 967 00:46:16,370 --> 00:46:19,570 So the idea is, we're going to selectively 968 00:46:19,570 --> 00:46:22,780 disrupt, or inactivate, the hippocampus 969 00:46:22,780 --> 00:46:24,520 at different phases. 970 00:46:24,520 --> 00:46:27,430 We're going to ask, "Do those phases, 971 00:46:27,430 --> 00:46:30,820 do they differ in terms of their contribution 972 00:46:30,820 --> 00:46:32,960 to behavior and performance?" 973 00:46:32,960 --> 00:46:35,590 So here, we're going to lock inhibition 974 00:46:35,590 --> 00:46:38,470 to either the peak of the trough of this state of oscillation. 975 00:46:38,470 --> 00:46:40,490 And so in this task-- 976 00:46:40,490 --> 00:46:43,090 and we're going to do this manipulation, that is, 977 00:46:43,090 --> 00:46:44,175 selectively inhibit-- 978 00:46:44,175 --> 00:46:45,550 you're picking hippocampal output 979 00:46:45,550 --> 00:46:47,675 at either the peak of the trough, the theta rhythm, 980 00:46:47,675 --> 00:46:49,027 at two behavioral phases. 981 00:46:49,027 --> 00:46:51,610 So in this task, animal's going to start on one of these arms, 982 00:46:51,610 --> 00:46:53,700 going to run up, and it's going to choose. 983 00:46:53,700 --> 00:46:57,360 So we're going to think about the starting arms as-- 984 00:46:57,360 --> 00:46:59,110 we'll refer to this as the encoding phase. 985 00:46:59,110 --> 00:47:01,190 This is where you have to keep track of where you are. 986 00:47:01,190 --> 00:47:02,759 And then, here in the central arm-- 987 00:47:02,759 --> 00:47:04,550 we'll refer to this as the retrieval phase. 988 00:47:04,550 --> 00:47:05,390 This is where you have to remember, 989 00:47:05,390 --> 00:47:06,680 "Oh, where did I come from?" 990 00:47:06,680 --> 00:47:10,950 And then use that to decide where you're going to go to. 991 00:47:10,950 --> 00:47:12,860 And what we found was pretty surprising. 992 00:47:12,860 --> 00:47:16,940 And that is that, you might think, well, 993 00:47:16,940 --> 00:47:19,400 if you shut off hippocampal output, turn off 994 00:47:19,400 --> 00:47:21,770 the hippocampus, you're going to get an impairment. 995 00:47:21,770 --> 00:47:23,810 It's just like in the examples that I 996 00:47:23,810 --> 00:47:26,355 gave of suggesting hippocampal's prefrontal cortex. 997 00:47:26,355 --> 00:47:27,980 Most of those come from lesion studies. 998 00:47:27,980 --> 00:47:29,300 You damage the hippocampus, the animal 999 00:47:29,300 --> 00:47:30,390 can't find its way around. 1000 00:47:30,390 --> 00:47:34,010 So if you were to optogentically lesion, or turn off 1001 00:47:34,010 --> 00:47:35,910 the hippocampus, you might imagine, 1002 00:47:35,910 --> 00:47:39,570 OK, animals won't be able to find a way around. 1003 00:47:39,570 --> 00:47:42,350 So you could ask, "This experiment 1004 00:47:42,350 --> 00:47:50,030 is going to identify which phase, which behavioral phase, 1005 00:47:50,030 --> 00:47:53,300 are most effective in disrupting performance?" 1006 00:47:53,300 --> 00:47:57,440 But what we actually found was, when you selectively 1007 00:47:57,440 --> 00:47:59,442 inhibit activity at different phases, 1008 00:47:59,442 --> 00:48:01,400 you actually get an enhancement of performance. 1009 00:48:01,400 --> 00:48:02,870 They get better. 1010 00:48:02,870 --> 00:48:05,330 And it depended-- there wasn't just 1011 00:48:05,330 --> 00:48:07,490 one phase, a good phase and a bad phase, 1012 00:48:07,490 --> 00:48:09,610 but both at the peak and the trough 1013 00:48:09,610 --> 00:48:12,920 were both effective in enhancing performance, 1014 00:48:12,920 --> 00:48:16,820 but only when applied at certain behavioral phases. 1015 00:48:16,820 --> 00:48:19,370 In fact, there was this double dissociation. 1016 00:48:19,370 --> 00:48:23,360 And that is, that trough stimulation, 1017 00:48:23,360 --> 00:48:25,040 when applied here-- 1018 00:48:25,040 --> 00:48:30,210 so when you stimulate in the trough, 1019 00:48:30,210 --> 00:48:33,410 in the retrieval segment, animals got better. 1020 00:48:33,410 --> 00:48:37,160 When you stimulate at the peak, in the encoding segment, 1021 00:48:37,160 --> 00:48:38,990 animals got better. 1022 00:48:38,990 --> 00:48:41,430 So it wasn't that the peak is good or the trough as good, 1023 00:48:41,430 --> 00:48:43,910 it's the peak. 1024 00:48:43,910 --> 00:48:48,110 The peak is good during retrieval 1025 00:48:48,110 --> 00:48:50,930 and the trough is good during encoding. 1026 00:48:50,930 --> 00:48:53,090 So what it says is, the peak and the trough 1027 00:48:53,090 --> 00:48:55,624 had two different functions. 1028 00:48:55,624 --> 00:48:58,040 And if you actually think about what I was showing before, 1029 00:48:58,040 --> 00:48:59,584 these sequences, these sequences that 1030 00:48:59,584 --> 00:49:02,000 are going from just behind to just in front of the animal. 1031 00:49:02,000 --> 00:49:04,400 What it says is, oh yeah, these different phases, peak 1032 00:49:04,400 --> 00:49:08,790 and trough, you can think of as like past and future. 1033 00:49:08,790 --> 00:49:13,100 And so if I'm sitting over here in this encoding segment, what 1034 00:49:13,100 --> 00:49:14,870 I'm really trying to do is I'm just 1035 00:49:14,870 --> 00:49:17,570 trying to keep track of where am I right now. 1036 00:49:17,570 --> 00:49:20,030 Now, I may simultaneously also be thinking about, oh, 1037 00:49:20,030 --> 00:49:21,860 where am I going to go? 1038 00:49:21,860 --> 00:49:24,472 But for this task, it's not really helpful. 1039 00:49:24,472 --> 00:49:25,430 It's not really useful. 1040 00:49:25,430 --> 00:49:29,152 At this point, I need to focus on where I am right now. 1041 00:49:29,152 --> 00:49:30,610 So you can think of simultaneously, 1042 00:49:30,610 --> 00:49:32,180 the actions of these two channels, 1043 00:49:32,180 --> 00:49:34,310 where am I, where am I going to go. 1044 00:49:34,310 --> 00:49:38,690 And I can enhance performance by shutting off or inhibiting 1045 00:49:38,690 --> 00:49:40,490 the non-relevant one. 1046 00:49:40,490 --> 00:49:44,150 So when I shut off the retrieval channel here in the encoding, 1047 00:49:44,150 --> 00:49:44,710 I get better. 1048 00:49:44,710 --> 00:49:45,940 It's like focusing attention. 1049 00:49:45,940 --> 00:49:48,290 It's like, pay attention to what you're doing now. 1050 00:49:48,290 --> 00:49:51,210 Stop thinking about stuff. 1051 00:49:51,210 --> 00:49:54,230 Similarly, when I'm here in the retrieval segment, 1052 00:49:54,230 --> 00:49:55,020 I'm also encoding. 1053 00:49:55,020 --> 00:49:57,980 I'm trying to keep track of where I am as well as, think 1054 00:49:57,980 --> 00:49:59,180 about where I'm going to. 1055 00:49:59,180 --> 00:50:02,810 The thing is, keeping track of where I am here, 1056 00:50:02,810 --> 00:50:05,180 it's not relevant for the task. 1057 00:50:05,180 --> 00:50:07,062 It might be broadly relevant for the animal, 1058 00:50:07,062 --> 00:50:08,520 but it's not relevant for the task. 1059 00:50:08,520 --> 00:50:12,400 So when I shut off, I turn off that encoding channel, 1060 00:50:12,400 --> 00:50:17,420 I'm able to enhance the retrieval information that 1061 00:50:17,420 --> 00:50:17,920 goes out. 1062 00:50:20,799 --> 00:50:22,340 You might interpret this saying, "Oh, 1063 00:50:22,340 --> 00:50:25,250 so this is how we can improve memory by selectively shutting 1064 00:50:25,250 --> 00:50:26,150 off the hippocampus." 1065 00:50:26,150 --> 00:50:30,770 Well, this is not a strategy for general cognitive enhancement. 1066 00:50:30,770 --> 00:50:33,890 Hippocampus is working much better when all these phases 1067 00:50:33,890 --> 00:50:35,062 are in operation. 1068 00:50:35,062 --> 00:50:36,770 And that's because the hippocampus is not 1069 00:50:36,770 --> 00:50:39,260 designed to solve this task. 1070 00:50:39,260 --> 00:50:41,720 Hippocampus is designed to solve the broader task. 1071 00:50:41,720 --> 00:50:45,690 It's trying to figure out how is this task relevant to all 1072 00:50:45,690 --> 00:50:47,149 the other things that I have to do? 1073 00:50:47,149 --> 00:50:49,606 In other words, you're trying to integrate this information 1074 00:50:49,606 --> 00:50:51,500 into all the other information you have. 1075 00:50:51,500 --> 00:50:55,714 And that requires connecting the encoding or retrieval. 1076 00:50:55,714 --> 00:50:58,130 You really need to have all of those pieces of information 1077 00:50:58,130 --> 00:50:59,600 available. 1078 00:50:59,600 --> 00:51:02,780 But it does point out that one can actually 1079 00:51:02,780 --> 00:51:04,790 dissociate the function of information 1080 00:51:04,790 --> 00:51:06,380 of these two different phases. 1081 00:51:06,380 --> 00:51:08,130 The phase actually matters. 1082 00:51:08,130 --> 00:51:09,950 And it matters at the level of high level 1083 00:51:09,950 --> 00:51:11,670 behavior, decision-making. 1084 00:51:11,670 --> 00:51:17,030 It's not just a idiosyncrasy or artifact of-- 1085 00:51:17,030 --> 00:51:19,730 excitability is a function of phase. 1086 00:51:19,730 --> 00:51:22,259 This is really how information is being used. 1087 00:51:22,259 --> 00:51:24,050 Sources of content the respective copyright 1088 00:51:24,050 --> 00:51:25,250 holders, all rights reserved, excluded from our Creative 1089 00:51:25,250 --> 00:51:25,916 Commons license. 1090 00:51:25,916 --> 00:51:28,534 For more information, see http://ocw.mit.edu/fairuse. 1091 00:51:28,534 --> 00:51:30,200 Sources of content used with permission: 1092 00:51:30,200 --> 00:51:31,100 Rat brain, hippocampus EEG, theta sequences, 1093 00:51:31,100 --> 00:51:32,000 bar/raster plot, courtesy of Elsevier, Inc., 1094 00:51:32,000 --> 00:51:32,300 http://www.sciencedirect.com. 1095 00:51:32,300 --> 00:51:33,200 Hippocampus electrode, courtesy of Plos. 1096 00:51:33,200 --> 00:51:35,033 Hippocampus optogenetics, courtesy of eLife. 1097 00:51:35,033 --> 00:51:35,950 License CC by 4.0.