Sleep Science Podcast

Episode 1: Bob Stickgold - What do sleep and dreams do for our minds and our memories?

Penelope Lewis Season 1 Episode 1

In this episode, we speak with Professor Bob Stickgold, one of the most prominent researchers in the field of sleep science. Bob tells us about the early days of sleep research and how he and other scientists struggled to convince the world that sleep really is important for memory.   He also talks about methodological subtleties to investigate the role of sleep on cognitive functions, the purpose of dreaming, and offline default mode processing in general, and what he sees as the most interesting future research directions of sleep.

Produced by:  Eniko Simo

See Professor Stickgold's faculty profile here.  His new book 'When Brains Dream' is here.

If you'd like to read more about the topics Professor Stickgold mentioned you can find some of his related publications here:

1) Karni and Sagi's 1994 Science paper on how sleep boosts visual discrimination
2) Bob Stickgold's 2000 Nature Neuroscience paper on sleep and the visual discrimination task
3) Matt Walker's 2000 Neuron paper on how sleep boosts performance on the finger tapping task
4) a sample of Erin Wamsley's work on dreams
5) a 2000 review, in Science, of Bob's early thoughts on sleep, memory, and dreams.

Glossary:
p-value = a statistical measure to test if the findings of an experiment are significant. 0.01 means a robust paradigm. 





This recording is property of the sleep science podcast and not for resale

Bob: The brain has to decide whether it's going to just try to stabilise and strengthen a memory or whether it wants to forget the unimportant parts or whether it wants to just extract and keep the gist and forget all of the detail, how are those choices are made. 

*Music*   

Introduction:

Penny: Hello and welcome to the Sleep Science Podcast. I’m Penny Lewis, a sleep scientist, and the presenter of this show. In this programme, we talk about all things related to sleep from dreaming and sleepwalking to what sleep does for the brain and the body and how we can get more out of our sleep. The sleep science podcast is a production of the NAPS sleep lab at Cardiff University. You can find more information about the guests, the podcast, and our sleep research at Cardiff on the podcast webpage. Today's guest is Bob Stickgold. Bob is a highly eminent professor at Harvard university's Beth Israel medical centre. He's been investigating the ways in which sleep impacts on our memories for over 20 years. Bob has always been a thought leader in research on sleep and has played a core role in moving the study of sleep and memory from a fringe field to one of the hottest and fastest growing areas in cognitive neuroscience today. In this episode Bob and I will talk a bit about the history of our field, discussing some of the early experiments that helped to confirm that memories really can be strengthened across sleep and also to help to bring the study of sleep and memory into the mainstream.

[00:01:44] 

Penny: What were some of the early experiments that clued you in that sleep was important for memory and made you go down this line of research?

[00:01:52]

Bob: So in retrospect, it’s striking that we didn't already know that sleep was important for memory. We talk about sleeping on a problem all the time, in all cultures and it's this implication that while you sleep your brain is continuing to process information to solve problems, to analyse information that is stored and extract solutions to problems. And this is a cultural belief that we all have, so in some level it's always fun to look back and say “Oh Gee we thought we discovered what we already knew”.  The earliest studies on this go back to the 1970s and Carlisle Smith who did truly elegant research in rats, showing that if you interfere with their sleep after they were trained on an avoidance task that they didn't remember having learned it, and that in fact it required REM sleep at particular time windows after training for that memory to become consolidated and maintained. And for whatever the personal and sleep cultural research reasons that research became sidelined and never received I think the attention it should have. And only reopened when Avi Karni and Dov Sagi from Israel did the first study using this visual discrimination task, which they got published in Science, showing that sleep could facilitate offline enhancement of performance on this task, after learning. And this is the first time that had moved into humans in sort of a cognitive domain; there had been earlier studies with French language learning showing that those who dreamt about it showed more improvement, although causality with questionable there. And this study by Karni and Sagi was the first one that really argued that there was processing going on during sleep of memory and that got a buzz but not much of an impact on the field itself. I say this because I tracked the number of publications on sleep and memory, and they didn't really take off until after a subsequent article in 2000 that was distributed free at the Neuroscience Society meeting that year and so it got big spread. And that's really when it started to take off and I think was our studies that extended Karni and Sagi’s findings on the visual discrimination tasks with tonnes of really rigorous controls to eliminate other explanations. As we sort of pushed that paradigm it became clear that in fact time awake gives you no improvement at all. So, we ran separate groups that we trained and tested either 3,6, 9 or 12 hours later or trained in the evening and tested the next morning. And showed that it was only the ones who were allowed to sleep who showed improvement at retest. We were also able to show that the amount of improvement was strongly correlated both with the amount of slow wave sleep subjects got early in the night and with the amount of REM sleep they got late in the night. And so, that allowed us to pull together a model that argued that early night slow wave sleep is necessary to initially consolidate and stabilise that memory and that REM sleep later in the night which would lead to actual enhancement. It was Sara Mednick who was a graduate student of mine at the time who did nap studies that really showed that it was the slow wave sleep that stabilises the memory and the REM sleep that enhances it. And that was really the start of it, that was what finally convinced the sleep research community that this was real. There had sort of been this attitude that ‘oh people had looked at it in the 70s and it didn't work’, that it wasn't there. It was really hard to get a change in that attitude, but it became a slow migration overtime. Matt Walker was a postdoc of mine at the time, came up with the motor sequence task. Interestingly Karni actually is the one who first published the finger tapping motor sequence task but not in relationship to sleep. He didn't study it for sleep and I remember Matt coming into my office one day, waving Karni’s paper and saying “Bob this one's gonna be sleep dependent”. That paradigm which is simply typing 41324 over and over for a dozen 30 second trials and getting faster and faster and then plateauing and staying at plateau until you've had sleep. When we first implemented that we just had people typing in Microsoft Word where we used a funky font that they couldn't see what they were typing and just told him to type 41324. And Matt had a stopwatch and he'd say start, then say stop. And when they stop, they hit a carriage return and then they'll be on the next line and then after 30 seconds he'd say start and then 30 seconds later say stop again and do the training and the testing that way. I think after 10 subjects who came in one day, trained, came back the next day and we tested, he had the P level down to 0.001.

Penny: Oh wow!

Bob: Yeah, half of those we’re same day, no sleep and half of those were overnight actually. And it was just five in each group, he was better than 0.01. It is a phenomenally robust paradigm.

[00:08:04]

Penny: What was it about that paradigm do you think made Matt think this will be sleep dependent?

[00:08:11]

Bob: I think Karni showed that people plateaued but later they did better. It smelled right, it just had this feeling of the kind of task.

[00:08:23]

Penny: So, there was some kind of hint in the data it sounds like. I mean maybe the fact that the visual discrimination task was a kind of a procedural perceptual task and this was a motor procedural task because we're already on the scent that these procedural talks were things that could improve across the sleep.

[00:08:40]

Bob: That’s right. It’s just like when we got to the dream field, when we figured out it was sleep onset is the time you want to look at. It’s one of those intuitions that make us personally feel like successful scientists when we say “this is the one”. 

[00:08:53]

Penny: We had a slightly similar intuition on a related thing. You know we've been working on the serial reaction time task which is similar. It's kind of typing essentially numbers in a sequence and getting faster at it but not realising that there's a sequence and I think we tested it out on one of my colleagues, you know quite a senior person, and he woke up in the morning and he said “I'm aware of what the sequence was”. The learning of explicitly being aware of this is sleep dependent. 

[00:09:28]

Bob: Occurred at night.

[00:08:29]

Penny: From what you've said so far, it's probably fair to say that the origins of our studies and our understanding of this effect came from looking at these procedural tasks but actually, now over the last 10-20 years we've gone much further than just looking at procedural skills. And you mentioned earlier that Sara Mednick has done some work that helps us to separate out the potential roles of slow wave and REM sleep. Could you talk a little bit about that?

[00:09:54]

Bob: Yeah, I remember after our first paper under visual discrimination task, where we got a correlation between the amounts of slow wave sleep and REM and overnight improvement that had a P value less than .0001 and I sort of said “OK that's it done”, but I remember driving home one day and saying to myself,  “I'm going to be very disappointed if it turns out that the function of sleep is to help us spot diagonal bars in our peripheral vision, it has to be about more than that”. And so over the last 15 years now as this become first accepted and then sort of a hot topic of research, the research has just exploded in terms, not only of the number of people engaged in it but in the questions they're asking. So, at this point there's still a lot of work going on looking at slow wave sleep's role in episodic memory encoding or declarative memory encoding, memorising word pairs and that's been a big field recently because people are trying to boost the amount of slow wave sleep people get with devices and thereby improve memory. I still think the most exciting and the most important findings are coming from the question of REM sleep. And its role in doing a lot more than just nailing down, consolidating, stabilising, maybe slightly enhancing, what we initially learned. And REM sleep seems to have an incredible role in taking that information and understanding it. You know I tell people that I bought an apple two plus computer with 48 kilobytes of memory, about a million times less than we routinely by nowadays. And it could memorise everything I typed on the keyboard, it could memorise everything I put into a camera, it could memorise everything that came in through on microphone, but it was laughable to think of asking that computer, “OK, so what does that mean?”.  It's a big difference between memorising something and understanding what it means and if we take computers as the analogue, it turns out we didn't need 48 K of memory, but you needed you know 50 gigabytes of memory and highly sophisticated software. Neural net learning systems to actually begin to extract meaning out of what's been memorised. That's the difference between being smart and being wise, that's the difference between just cluttering a system was a lot of information and converting it into a useful form. So, everything we encode, our brain has to make a decision about each of those things; whether it wants to keep it, stabilise it, strengthen it, just allow it to decay.  Most of what we encode in a day we forget, thankfully. The steps we want to remember it has to be tagged somehow and then processed. And the processing isn't just strengthening it. It means identifying what aspects of a memory are important, discovering how to file it, whether it gets filed under ‘today’ and ‘lunch’ or whether it gets filed under the particular ‘food’ or whether it gets filed under something else. And the brain has to figure out whether there's new information that it can create by taking what it has learned and integrating it with existing memories and looking for patterns or exceptions and coming to new understandings. And Penny some of your work, some of my work, some of Sara Mednick’s, lots of people's work is now showing us that REM sleep seems to do those things. It integrates it, it figures out where to integrate it into our existing networks, it then looks at those networks and see if the brain can extract patterns that weren't obvious beforehand and what those patterns mean. It works on emotional memory and probably because emotional memories are always encoded within a context, so if you're just learning word pairs like ‘telephone-tape measure’ you say to yourself “telephone-tape measure”, and you can remember that later. But if the word pair is “sister-rape”, then it's not just the word pair. It's an emotional word pair and it brings up a plethora of related older memories that it then has to figure out whether to integrate it with and how to integrate it with. And I think that's why the emotional memory ends up being  REM dependent because emotional memories are always learned in a context and it's always important with emotional memories to figure out what that context is you know is this the context of another jerk or a potential friend. 

[00:15:24]

Penny: You're suggesting that REM cares a lot about context and that's because of it’s role in integration and so that spills over into emotional processing or emotional memories as well, is that kind of where you’re going?

Bob: That’s right, why are emotional memories stronger? Why do we remember if we are seeing a bunch of word pairs and 5 out of 40 are emotional? Why do we remember all 5 of those but only 10 of the other 35? The memory researchers will tell you it’s about depth of encoding. That the more deeply we encode a memory, the better we remember it. The definition of depth of encoding is how many things you connect it to in your memory network. So, if I tell you to remember the pair ‘floric-shnogger’, you’re going to have a really hard time, because you don’t even have memories of those two pseudo words that you can link in to. You probably couldn’t repeat them back to me now. But if I say the pair ‘chair-tape’, I can then talk to you about something else and point out that those are words that have rich associations in your memory and ask you what those words are and you can probably recall them. So, the amount of integration that happens with all things you learn, determines how well you remember. With these rich deep networks, I think that’s when REM sleep comes into play, because by definition, almost, emotional memories are important. If they weren't they wouldn't have been emotional and that's what REM is looking at. It’s looking at what’s important and how is it important. 

[00:17:06]

Penny: So, what do you think REM is actually doing with either these emotional memories or with the integration?

[00:17:17]

Bob: I'm debating whether to go into dreaming at this point. I think that REM sleep and dreaming which probably evolved at about the same time, and relatively late, in the evolution of sleep and the evolution of memory. I think REM sleep evolved in part to provide a neurophysiological substrate for, I want to say, deep neural net exploration for asking questions instead of answering questions. So, with a usual memory task, all you're supposed to do is remember what you were told. “I'm gonna tell you some words, I’m gonna show you some pictures, I'm going to play you some sounds”, and then later on, I’m either going to ask you to repeat them back to me or to recognise them out of a string of some new and some old words or pictures. Originally, memory was just about saving past experiences so you could bring them to mind again or reactivate them non consciously when similar situations arose for an organism. I think REM sleep was about figuring things out. I think of REM sleep as being about divergent thinking, it's not about remembering that you already learned, but really but exploring and discovering new information. When you do your pattern recognitions and their statistical learning, it's about figuring out that it is statistical learning. It's about figuring out what those rules are. I keep going back to sleeping on a problem. I mean the classic sleeping on the problem is you get offered a really good job in a lousy location and you can’t decide how to balance those two against each other.  And if you're sort of compulsive, you make one of those lists, right, you know take the jobs pluses and minuses. And you can make that list and it never helps anybody, as best I can tell, it’s not in the right format. And then you go to sleep and you wake up the next morning and you say to your partner you wake up and you roll over and you say, “You know what I can't take that job”, and they say, “why not?”, and the answer you give is “I don't know, it's just wrong”.  So, while you're sleeping all those things, your compulsive lists of pluses and minuses, are somehow being explored and weighted of either yes or no and you lose all of the basis for that, all of the bottom-up calculations that the brain was doing outside of awareness. It does all that then it just pops up the answer to you.  And I think that's what REM sleep is really about. Whether it's about a weather prediction task that Ina Djonlagic did with me, whether it’s about your statistical learning, whether it's about deciding who to vote for in an election. 

[00:20:20]

Penny: So, these are all decisions that require the integration of lots of different types and pieces of information to make the decision. 

Bob: Yes, and too many to write them all down and come up with weighting factors and add them up and come up with an answer. That mechanism doesn't work for complicated questions like these. Almost all of this pattern extraction seems to be REM dependent, but I believe you have one that seems not to be right? 

[00:20:58]
 
 Penny: We have some abstraction task that seem to be slow wave dependent. So, we are thinking hard about what's the difference between tasks that are slow wave dependent and REM dependent, you know, in terms of abstraction and pulling out these commonalities. And actually, my current hunch on this, and this is just a hunch; is that actually it's things where there's a temporal dependency between items. So, you're trying to integrate across time, like a sequence of tones or button presses, those things seem to be more slow wave dependent but things like the weather prediction task, you know, you have images, and the timing is not so important, the sequence is not important, these seem to be REM dependent. 

[00:21:40]

Bob: I mean I point out to people I say all this stuff about REM sleep, and you know what Carlisle Smith referred to as complex cognitive procedural learning, things like the Tower of Hanoi task, improvement is REM dependent. That was one of his early studies but then I say having said all that, remember that the visual discrimination task, a learning that appears to occur in V1 itself is REM dependent and it's not time dependent whereas this finger tapping task which turns out to be stage 2 NREM dependent N2, is a temporal sequence task. So maybe that's a big piece of it. 

[00:22:22]

Penny: I think it's an interesting question of how much the temporal sequence matters and it's something that we should study 

[00:22:29]

Bob: I'm writing a book with Tony Zadra who is a dream researcher up in Montreal, and it has been a wonderful journey for me to learn a lot of this stuff about dreaming that I didn't know, that Tony is an expert on but also to sort of try to wrap my mind around the functionality of dreaming. Dreaming is primarily about exploring networks, looking for interesting new ideas, if you will. It's about doing what we wish as scientists we had more time to do, which is to just kick back and our chair and think about stuff without any particular goal in mind. And I think that's really what dreaming is about. It's about finding weakly associated memories in our networks, combining them into narrative stories and then observing our emotional responses within the dreams to those narratives. That it is in essence saying do these go together in an interesting way? Is there something to be gained by linking these, two initially only weakly associated memories, linking them more strongly? We look at a lot of our dream reports after we train people on tasks and we’re struck by the fact both; that they're obviously related to the task and 2nd that the obviously useless, if we think that the purpose is to enhance memory or to enhance performance. So, we have people trying to learn a visual maze on the computer, they’re wandering through the maze under their own control, trying to learn the layout of the maze and before they really learned it well, we stop them. Then we let them sleep and we test them again afterwards and yes in fact as with so many other studies they performed better after sleeping as opposed to an equal amount of time awake, but Erin Wamsley did these wonderful studies where she also woke them up and collected dream reports. And so, you get these dream reports about someone saying “Oh, I dreamt that I was lost in a forest, or I dreamt that I was in a bat cave, and I was thinking about how bat caves are sort of like mazes”. Well, that's nice you know, it's not going to help them perform the task better, not going to enhance their memory for the layout of the maze but that's not what it's trying to do. It's trying to discover oh you know what, next time, you go down into a cave you might think about this maze study and think about how you tried to memorise and what worked trying to memorise the layout. Or next time you're trying to learn a maze, you might think about your experiences in the bat cave and how you remember your path in the bat cave. It’s trying to find novel associations that might be useful in the future and the brain when we’re dreaming is very happy to fail. After a study we did using Alpine racer, which is an arcade game where you do downhill skiing, where you’re actually standing on a platform with ski poles and seeing a screen of an avatar skiing down a slope. Some report a dream where they are moving swiftly through a forest without their legs moving at all. They perfectly describe themselves with the posture that you hold when you're skiing and then they “It was like I was on a conveyor belt”.  And so, it's putting together these different associations, you know, rushing through a forest and they were doing a slalom course on the screen earlier where there's skiing between wooden poles and so now, they're rushing between wooden trees with their legs not moving at all. It suggested trying to figure out something else that might be relevant, but someone else reports seeing a squirrel skiing that warns of loser right, but the brain is sort of taking the role of a venture capitalist when it dreams. It's happy to have high failure rates in its search because it's looking for the unexpected and quite valuable association.

[00:26:59]

Penny: Is that true for all dreams? Or is that specifically REM dreams? Do you make a distinction between dreams and Non-REM and REM sleep ?

[00:27:12]

Bob: I do. So, the early studies have reported that if you ask people to identify the waking sources of elements in their dreams, when you wake them from Non-REM, and this will be almost exclusively Stage 2 NREM, the sources tend to be episodic memories. “Oh, in my dream a flying saucer landed, a really small one, and these aliens got out and I put red sauce on it and put it in the oven”, and then they say, “Oh yeah, I had pizza for dinner last night I'm sure that's why I had that dream”.  That's the Non-REM explanation. When they had the same dream report in REM, they'll say “Oh yeah that's because I just really love pizza”. So, it turns out that this source, the identifiable waking sources for non-REM dreams tend to be episodic memories and tend to be recent ones. The waking sources for REM dreams tend to be semantic knowledge, I love pizza, I eat pizza all the time and often much older memories. So it almost seems like the Non-REM dreams are doing something closer to what we normally talk about as sleep dependent memory processing and trying to explore associations to that. Whereas in REM sleep, I think it's really doing a much deeper historical search of memories and looking for much weaker associations to construct those dreams and again the reports are that the REM dreams are more bizarre than the non REM dreams. And then the sleep on set dreams I think actually are largely still tagging memories. I think we have to tag memories to identify what we want our brains or what our brains want to process further during the night and I think that tagging happens at sleep onset, and I think it happens before we fall sleep also. Look at the thoughts that go tumbling through your mind as you're in the process of falling asleep, there all about things from the day or things from tomorrow that are active concerns. Things from the day that you didn't finish or things that you didn't quite understand or a thing for tomorrow that you didn't finish or that you don't know what to expect. It's as if the brain is taking that time to look at recent history, your recent history, and ask what are the unanswered questions that I need to be working on. So, I think that those sleep onset dreams which are so identifiably related to specific learning tasks for example, those are really just the start of a night's processing of those memories and that the serious processing is going to happen later in the night. This sort of surface processing will happen in non-REM sleep, in terms of slow wave sleep dependent stabilisation and enhancement of memories and dreams that are really just looking at surface associations, stronger associations. And then REM sleep both in terms of what we talked about earlier in terms of sleep dependent memory processing and, but I would talk about now in terms of dreams, is looking for that sort of deeper, less expected, more obscure, more abstract associations because that's what humans do that's so phenomenal.  It's not that they have better memories. Humans have a growth function of knowledge, that I don't know how widespread it is in the mammalian Kingdom, but it's certainly quintessentially intense in humans of abstracting and extending from what we learn to create new knowledge that is incredibly valuable to our survival.

[00:31:25]

Penny: It's reminding me, as you say that, of something that I think I remember you telling me years ago, about dreams processing information or scenarios that are somehow unresolved or not fitting into our schemas. Can you talk a little bit about that and how that relates to everything else that you've just said about dreams?

Bob: Well, if you think again about the thoughts that go through your mind as you're falling asleep, they’re almost always about incomplete cognitive or emotional processes. Things that you started doing in your mind and aren't finished with. You’re trying to remember a name and then you can't and then you know an hour later it just pops back into your mind. I have a sort of more generalised theory which I refer to as ‘snip and fit’ and fit is the functional incompleteness theorem. The functional incompleteness theorem is that the brain has evolved to take advantage of every moment whether awake or asleep that it's not actively engaged in some other task, and spend that time trying to explicate and resolve incomplete processes, incomplete thoughts, problems that haven't been solved, current concerns.  I mean, everything that has come up since I first conceptualised, and never wrote about this concept, about the default network seems to say that when we're not doing anything else the brain shifted into this mode that activates the same memory structures that we use when we're recalling episodic memories, when we're imagining future scenarios, when we're dealing with a theory of mind problems, when we're doing navigation. It's like what did she mean when she said that, where am I going with this, what did that mean that the person said, how am I gonna deal with this in the future. These are the things that one does to your mind during the day when you're not actively doing something or if I can get onto a little rant, what we used to do, now we don't. Now we pull out our phone and we look at funny pictures of cats or we put in our air buds and we listened to music at that critically large volume where it cuts thought out of your mind. I think we have a culture now that is the fault network averse. 

[00:34:09]

Penny: Yes I know what you mean

Bob: I think - I think it's true getting your car and come to a stop light then you feel like I need to check my phone. It started with the Walkman way back when people started to always have intense sensory input coming in a very explicit way to stop us from being left with their own thoughts. And I think that's disastrous for the individual, I think it's disastrous for the society, and it's disastrous for the species. And I think that that is an amazingly important time and as we become uncomfortable with it, we become more shallow human beings because we know what we learnt, we know the facts that we memorise, but we've never stopped to think about them. So, the functional incompleteness theorem says that we have evolved to spend all of our spare time doing that and so I said it was ‘fit and snip’. The snip is sustained nonconscious incompleteness processing and that just says that we spend all of our time out of conscious awareness awake or asleep processing that information. Here's where you have this situation where you're trying to remember someone's name and you can't and then an hour later it just pops into your mind, and you realise that your brain had continued to search for that name that whole time outside of conscious awareness and outside of conscious intent. It just does that all the time and dreaming then becomes a manifestation of it. 

[00:35:37]

Penny: And so those are the snips.

[00:35:39]

Bob: In my family snipping is now a verb because it turns out it’s not just nonconscious processing but it's non-intentional so that’s when you realise you've been obsessing about something and weren't even aware of it. We don't ever actually replay the episodic memories, we don't use that episodic memory in the dream, we dream about it. You're almost in a car accident and you dream about being on bumper cars in the amusement park, you dream about it. You have a fight with someone at work and you come home, and you have a dream about having a fight with your spouse, that's got the same tone exactly as the one at work. So, we don't use episodic memories, we don't apparently even need them to know that this is important to dream about and to dream about it. 

[00:36:33]

Penny: So, our dreams are not just built from episodic memories, I guess that's part of the point, right?

[00:36:36]

Bob: They are absolutely not built from episodic memories. And I told people after the amnesia study, I said don’t tell anybody, but it turns out that dreams are the royal path to the unconscious, which of course is the famous Freudian statement.  

[00:36:54]

Penny: Yes

[00:36:55]

Bob: We're seeing unconscious memories in these amnesiacs being replayed in their dreams. So dreaming is again not so much about exactly what happened; it's about other things related to it and it's found through semantic and emotional networks in the cortex and not through the hippocampus.

[00:37:22]

Penny: Which links back to everything that we were saying earlier about REM dreaming. So, you've spent the last 20 or more years working in this field, what do you see as the biggest remaining questions in the field that we should be trying to address?

[00:37:39]

Bob: We still don't know what the range of memory processes are that can take place during sleep. At first all we were seeing is stabilising and strengthening of procedural memory then we got stabilising of declarative memories. Now we've moved into all of these more associative problem-solving creativity domains. How widespread is that domain? At this point it's starting to feel like it's everything we can think of that’s memory related, which would be a very profound statement if we could reach the point that we feel comfortable saying it, that’s one question. A second question is how many of those are done almost exclusively during sleep? So I'm thinking about the fact that we don't get better on the visual discrimination task, we don't get better on the motor sequence task across periods of wake. And then I think we don't have any useful understanding at the cellular network or physiological level of how those processes are initiated, how the memories are chosen for it and then we talk about the memories getting tagged but nobody knows what it tag looks like. The brain has to decide whether it's going to just try to stabilise and strengthen a memory or whether it wants to forget the unimportant parts or whether it wants to just extract and keep the gist and forget all of the details, how are those choices made? Part of the problem that we face is that we don't have answers to these questions during wake. The trick was getting it from being a pain in the butt to be really exciting. 

[00:39:29]

Penny: I think now we're in a place where it is really exciting 

[00:39:32]

Bob: Yes but I will tell you almost all of that came from graduate students joining labs and saying I joined your memory research lab because I want to study sleep in memory. And I would get these phone calls from really sort of annoyed big name memory people saying Bob I have this graduate student who is insisting she wants to do her thesis on sleep in memory, is there anyway-

[00:39:56]

Penny: That's really great to hear. I've never heard that before- 

[00:40:00]

Bob:- it was driven by the students coming in saying I want to study this because they were open to it. They hadn't yet locked their paradigms down as, Penny, you and I have now, you know. So, when someone comes and says to you I want to study what handedness has to do with sleep dependent memory consolidation, your answer is no.

[00:40:23]

Penny: Oh well actually we're just writing a paper on that Bob! So, but I know what you mean. It was the perfect example.

[00:40:34]

Bob: It was the perfect example! I got one more thing 'cause you're asking big questions, and the other one is dreaming. We are still at a profoundly naive and psychologically and almost psychoanalytically driven models for dream function and I think we really have to move out of that, and we really have to move into a cognitive theory of dreams function which you will find in Tony and I’s book. We have to address the question of how you study it. 

[00:41:06]

Penny: You mentioned earlier about technologies for capitalising on our burgeoning understanding of how sleep influences memories and I wonder what did you have in mind and what do you think about these technologies? Do you see this as a useful kind of way forward?

[00:41:24]

Bob: There are now a number of good research groups and a number of powerful corporate groups that are looking at this question of ways to enhance slow wave sleep and thereby enhancing memory. They started out with transcranial direct current stimulation and have largely moved into the use of tones to do this. So, they will have a participant wearing earbuds now and also having electrodes on their scalp and they will, in real time, be doing slow wave detection and then playing sounds in phase with the slow waves to increase the amplitude of those slow waves and the amount of slow waves that the brain produces. They have been successful at inducing increases in slow waves and so with amplitudes and in memory the next morning. They've done it with elderly subjects, they've done it with younger subjects and in both cases seeing improvement in word pair memory. I think it's a fascinating research tool. It might have a clinical application in the elderly and in those with memory impairments. I think as a general conceptual approach to memory I think it's a mistake, because I think it's pushing entirely of focus on these episodic declarative memories and I don't think you want to increase the amount of slow wave sleep people get. I think the brain is a brilliant machine for calculating how much slow wave sleep and how much REM sleep you need. We know that it will vary it on the night-by-night basis depending on what you learned the day before. If you've been putting a lot of time in a REM dependent task, you will have more REM sleep the following night and upping your slow wave sleep that night can be counterproductive. So, except in cases of pathology I suspect it's gonna turn out not to be a really good idea. Now pathology might be that you've got a French final tomorrow ,I think that's a pathological state and so when right after cramming dates or irregular French verbs increasing your slow wave sleep might be productive but I think as a long term enhancement of memory I don't put much stock in it. I think it'll go the way of Prozac. There was that where it was thought everybody should be on Prozac and it turns out that you don't want to just up serotonin in your brain that's not what you want to do and you don't want to just up slow wave sleep 

[00:44:20]

Penny: So what you've talked about is closed the auditory stimulation to boost slow oscillations but what about targeted memory reactivation where tones are played to kind of trigger reactivation?

[00:44:30]

Bob: Well, I think that's great if there's something that you want to be sure to remember and remember perhaps better the next day, but, you know, what is the expression? You have to rob Peter to pay Paul - it's not free. No one has done this study where they have you do one of these tasks like learning the locations of objects on the screen paired with tones and then they play the tones while you're asleep at night and you're better at remembering the location of those associated objects the next morning. No one has then asked “OK, but do you remember what you learned in chemistry class today before you went and participated in this experiment?”, “Did you process that conversation that was rather uncomfortable that you had with a good friend yesterday?”, or, “Did you manage to get your brain to only pay attention to what you wanted it to?”.  I don't think we know the answer to that, first of all. I don't know whether it materially interferes with other processing of memory during the night. Before we go gung-ho and start distributing these to 200 million Americans, I think it's important to consider unintended side effects and whether overtime you create a brain that will not reactivate memories on its own anymore. It will assume that you're going to reactivate the memories for it, just like if you start using addictive drugs your brain stops releasing as much dopamine. We don't know how the brain adapts to that kind of treatment so if it's a one-off experiment, if it's twice a semester before exams, I think I would comfortably support its use but as a general mechanism on a more consistent basis; who's smarter you or your brain? 

[00:46:33]

Penny: Definitely my brain, yeah. I see your point. So, basically you think we need to approach these new technologies with a lot of caution and do these studies and try and work out what damage or interference or disruption they are causing along with the benefit. 

[00:46:53]

Bob: That’s right. And again, as a research tool I think they're fantastic. 

[00:46:57]

Penny: I think now probably is a good time to wrap up. We covered a lot of topics. There are more, many more that we could cover I know we could probably talk for another hour. 

[00:47:07]

Bob: Oh, go out and have a beer now and play darts and keep going!

*Music*

*Outro*

Penny: You've been listening to the Sleep Science Podcast with me, Penny Lewis. My guest today was Professor Bob Stickgold of Harvard University and our producer was the Eniko Simo. If you've enjoyed the show, please consider subscribing to the podcast or liking us on Twitter. If you have questions or comments, please get in touch by a sleepsciencepodcast@gmail.com. Thanks for listening and until next time sleep well!