The Relation Between Psychology and Neuroscience

welcome to the afternoon event see everybody's mood is elevated maybe that's because of the change in laws here not to say that the mood is high but people seem to be doing well and that so you come here normally to see cool things cool new results new techniques maybe meet new people definitely hear new gossip I know that's why you're here but what we don't do so often is examine what we actually do is examine the foundations of cognitive neuroscience and so we thought we'd do that a little bit today and so I assume that you're here at the cns meeting because you're actually interested in the relationship between cognition or psychology and neuroscience if you're not then I don't know why you're here you should probably leave now and but you could go to the cognitive science Society or ApS or psycho nah mcc's you could go to the Society for Neuroscience or cosine or something like that and there's pretty unlikely that it's gonna be a session that asks what's the relationship between neuroscience and neuroscience just it's just not going to be that way but in cognitive neuroscience we have to face the music we can't pretend they're not serious issues epistemic cuts or difficulties in solving the problems we're interested in now the conceptual foundations of our field are I assume more than just having correlations right we're not just interested in having a concept from the cognitive sciences of psychology and then relate it in some way to some neural measurement we make now that's that's what we do and there's nothing wrong with correlation correlation is a beautiful thing but and I assume that our our yearning is higher our aims are more ambitious and although correlation is great and let's say regression which is like the most flavor of the day and data analysis is a wonderful thing it can't be that that's all we want we must want more and so there's a few issues that you know can bug us and we're gonna we're gonna unpack them this afternoon so let me you see the title is the relationship between or the relation between psychology and neuroscience and let's just briefly review why there's a problem so I'm gonna try this now yeah this is clicker which looks like it was made with glue stick and you did say press it firmly so I pressed it firmly oh okay so here's two very briefly I will trust me at once I want a hug with stage-4 Delambre here's a quote that I rather like I'm from a book but philosophy which I actually can't read from this angle I make these guys here read it out loud okay Evan you read it out loud okay let me try cognitive neuroscience operates across the boundary between two fields neurophysiology and psychology their respective concepts of which are categorical EDA similar and I think that's where the underline should be the things we talk about our categorical EDA similar the logical or conceptual relations between the physiological and the psychological are problematic the relations between the mind and brain are bewildering and if you don't think that's true then you also are in the wrong room if you think that the things we are working on are easy and straightforward you're just mistaken so just to briefly illustrate the nature of the problem Oh a trick I just click very lightly so here's just a list of a few a few this is like the steak at lunch here's a list of a go for this side of a few kind of foundational concepts that you're familiar with from from neuroscience that are completely not bewildering because this is the kind of stuff the stuff we study stuff like a channel or a dendrite or ensemble of neurons or a process like LTP or Ltd or oscillation and there's no there this you know you can maybe debate about the minutiae of this but you're not going to have fundamental disagreements about their relevance now here possibly me yes is a similar list for some list of psychological concepts that we may use to talk about cognition behavior psychology things like attention or aspect volition memory language concepts or processes like prediction Association discrimination generalization and so on now again that's what you learn in psychology and you're not going to you know put up a big fight that those aren't relevant concepts but the problem is that we haven't even the vaguest idea about how to relate those and if you think otherwise let me know because then I can be rich and famous but the fact of the matter is we really don't beyond correlative information between these domains likewise we can say well look we all read or we we're supposed to read for class and this book by Marr which distinguishes a computational and algorithmic and an implementational level again extremely useful way to partition complicated things but to formally find relationships between these is actually quite astonished and the difficult thing and here for instance tomorrow we have a symposium on mesoscale versus macro scale and microscopic clicking behavior there yeah forceful well step on it so again I mean are you doing work at the microscopic level which is you know feels real because we're kind of duelists right you measure stuff and you can touch it so it must be real at the meso scale which are served mystical things that we think probably matter for neural coding or the macroscopic scale which is the stuff we measure and then correlate with everything so no matter how you carve the problem up and no matter what aspect of psychology or the cognitive sciences you study the problem is really quite more difficult when you start to look into where the cuts lie between the stuff that we are made of and the stuff of thought this is a difficult problem now we have four people who are solving it for us today which is great you should take excellent notes and so Leila debauchee is at the far end who's a professor at Columbia University who you all know this already and whose expertise I picked and for under each very brief description of what each person works on one of my favorite papers from each person so you know what you know run and go read it so Leila is of course very famous for work on memory and episodic memory and a bent structure on medial temporal lobe systems and so on next to Leila is Jennifer Gro Jenny is a expert in an eye movement control and coordinate transformations and neural coding so it works much more at the system's neuroscience and computational neuroscience level kate hartley is next to Jenny as a professor of psychology and works on decision-making on motivation and on development and Sharon Thompson chill is a psychology professor who works on concepts on language on semantics and those are all papers I particularly like but there's what much more stuff of them that you ought to read and we're going to forcefully get through the section it'd be since this is the last time I'll have the stage I wanted to quickly get off of my chest but I wanted to recommend do you read if you're a graduate student or a postdoc or just a human being and you can read you should read these things this is good stuff read Conrad Logan's the foundations of ethology or read Nikolaas Tinbergen the study of instinct read David Mars book vision read the essays of Howard Patti read Bennett and hackers book on the philosophy in neuroscience these are cool things and you should take your time study it and enjoy it and then go out and you know enjoy the new laws so now each participant will spend a few minutes outlining sort of what they think are the big issues and then we'll have a discussion together for a little while and then open it up to questions as soon as we feel it's the right moment which could be very soon or very late in the game depends on how this discussion goes so and Leila is the first person to come and hopefully the clicker forceful clicking either so I'll just start by saying you know a couple of minutes and I'm not gonna tell you anything that's surprising I have no answers but I really was thankful to David into the conversations I had with everyone sitting up here to just get a chance to sit with the questions about how do we think these two fields are interacting what's the value of the work that we do it's so easy to get bogged down in the experiments and the data we love the data that's why we're here right but it's also I think important to stop and think and until David had asked me to serve on this panel I realized it had been a long long time since I had sat down and started to you know give myself a chance to think about these ideas so I apologize if what I'm gonna share with you seems rather simplistic that's where my thinking is and we can continue the conversation but I thought it might be useful to just first the answer depends on what your goals are right what is the goal of what we're all doing and there are several goals we have a psychologist in cognitive neuroscientist and one is to describe behavior and one way to think about that is what happens when it happens you know how can you measure it and another level is to predict behavior what I can't even read that what varies with what so what you can move around one factor and show that behavior is moving around in another way right describing behavior but ultimately we want to explain behavior right so the idea here is you can understand more causally what kinds of manipulations lead to different kinds of effects and ideally you maybe want an explanatory you want an explanation for behavior so that you can change behavior right you can modify it in some way and a lot of people care about changing behavior but then it dawned on me that one of the questions we have to ask ourselves is what is the nature of an explanation what is an explanation and it's not a question to take for granted but you get there are many questions that people ask these are the kinds of questions that my children have asked me and maybe even ask so what's an explanation for how a bird flies what's an explanation for how an engine works why are peas green or maybe why do you not like Mondays mommy that's one that I get what do these questions all have in common what they have in common is that they describe causal relationships the answer would describe relationships between phenomena at different levels and the the important thing about that is the levels of analysis so to understand any of these sort of simple questions you have to give an answer that describes the relationship between elements its constituent elements at one level to understand the emergent properties that are there at another level so what we're doing here is here are two levels imagine cognate cognition are thinking and information processing and neuroscience and there's a lot of work you can do at each individual level to understand the elements what are the elements of cognition how do they work what are the elements of neuroscience but what we're doing here is trying to build rules are you know understand the computations that bridge across the two levels and I think Elizabeth started the session this evening by talking about bridges in San Francisco and that's really why what we're doing is so complicated because unpacking the bridging laws is above and beyond understanding the constituent elements owe an explanation cannot is not good enough to define the love these sort of elements that are there in the brain the neurons for example even the networks or the firing it's not enough to describe the elements you have to understand how they interact and give rise to the emergent properties of memory and attention and decision-making and so it's extremely complicated but there is hope because there are emerging properties all over the place since you're a mutt couple of my favorite example is a flock of birds and when you watch hundreds of birds fly together you can't explain the behavior of the flock by anything that's in an individual bird it turns out though that there are computations in an individual bird like the distance you maintain from fellow birds and not to run into them and to follow them in some way and also to avoid predators that when put together can explain the emergent behavior of the flock but it's not enough to know what a single bird is doing the same thing with water water you know no there's nothing more simple than three molecules h2o but understanding if you put let's say a jillion molecules together you get water and water has emergent properties that are not there in the molecules that make up water one for example is liquidity like what you see on the screen that does not occur in the molecules when you freeze it it turns to ice so there are things that these are emergent properties and I think that cognition consciousness attention in memory those are emergent properties that come from the interactions the complicated interactions of the constituent elements of of the brain so I should have done more clicking here we go so entirely new properties emerge that are not explainable by the constituent parts and just to sort of break it down so you might have behavior and brain and one of the things we might be asking ourselves in our own work is well what is the right level and an important point that I want to make today is that I feel as you know a scientist that's been around for a long time and I've studied systems neuroscience and now I'm studying what I'm studying and I feel like there are these paradigms of science where there seems to be this chauvinism about what the right level of analysis is and that's all wrong every level is equally likely to contribute to the answer but you have to pick the right level based on your question so just if you just break down behavior there is the individual which is what we all work on we study cognition we want to understand how the mind works so typically you're working in with one person but then there are diets what happens to cognition when you're talking to someone what about group behavior group behavior is another emergent property that comes out of our own cognitive systems and then there are micro scales of behavior like micro behaviors micro sort of mini eye movements there are things you can measure implicit memory for example that could be a kind of micro behavior and then in the brain of course we have areas that we measure typically with bold there are neurons there are networks and then there are oscillations and all of these are other levels of analysis and I and again I just want to reiterate that there is I don't think any right level and what's really important is to think about what your questions are and one of the things I'd like to tell my students and I like to tell myself often is if I get stuck thinking about data or a question I need to take a walk and really think more clearly about what am I trying to ask here what is my question and that should help you to identify what the right level for your analysis is and then what kind of bridging laws you want to attempt to come up with to kind of bridge across those levels and this is just another example of the bridging law so you have sort of abilities and cognition computation and neuroscience and what we do is happening in the middle the edges between these areas there's a lot to do at any one of these levels but all the hard work that were engaging in I think is happening on the in the vertices on these edges here and these little lines that you might overlook that just connect the triangle that's where all of the hard work is that we're doing thank you [Applause] alright I'm gonna try pressing this firmly I understand that's the way to make it work okay so when David emailed me and asked me to be part of this panel I wrote back right away and if you've ever emailed me before you know that's not typical behavior so and and the reason is because I've been thinking a lot at sort of a meta level about you know what are we trying to do and are we on the right track and in part because of my day job affiliations I'm sort of never quite clear where the boundaries between different disciplines are I have a appointments in a neurobiology department as well as a department that it's called both psychology and neuroscience so you know figuring out where the you know what is what is neuroscience what is psychology is sometimes not clear to me so the way I conceive of the of the questions though are sort of are you working more at the electrode level or more at the scanner level and so I think I think the a thing that emerges based on these two different kind of levels of inquiry is that that we're making different assumptions about what is functionally significant and so if you can see I can't see this picture but from what I can remember I put up here there is a person sort of facing forward that I think of a sort of cognitive neuroscience or scanner scanner approaches and then there's a person walking away and to me that feels a little like what has been happening over the last ten years at the more neurobiological approaches because what I think it's been happening in neurobiology has been a focus at ever more granular levels of detail where it is it is to me not completely clear that we knew what to do with the level of granular till we had so so it's been very hard I think to to connect that with what's been happening in more of the scanner level cognitive neuroscience which has been emphasizing and I think rightly so that the brain is operating as a series of interconnected nodes that the network level of analysis is incredibly important okay sorry so another thing that has been on my mind has been the differences among these different techniques in the opportunities to make discoveries of things you were not looking for so I think so this is a cartoon from the old comic strip mutton Jeff and it starts off with a guy is looking for something and the policeman for whatever reason asks him you know why are you looking here and he says because the light is better here and I think I think you know if the field is being too driven by techniques you're at risk of doing experiments because you can and not because you necessarily have a profoundly important underlying question that needs to be addressed that way so that's something that I've been watching happen from time to time in particular I think the optogenetic approaches as an example the way for those of you who are less familiar with it one way this technique is used is that you you get a virus expression of rhodopsin channelrhodopsin in one part of the brain and you shine light and another part of a brain so that you can probe a particular connection in a particular behavioral context so I think if you do an experiment like that you can only discover something about that particular connection in that behavioral context which is of course what you know the experiment is designed to do but one of the things that I have really valued not as someone who uses fMRI but as someone who sort of observes the discoveries unfolding in the field is that you all who use this technique can discover things you were not looking for and I think those kinds of discoveries have really moved the field forward in ways that perhaps no one could have anticipated let's see ok psychology one of the things that comes up in the context of psychology is to think about replication well I sort of have some Envy about replication issues in psychology I can't remember what my bullet points are and I can't read them so I'll just go with what I can remember about them so so we have a couple of different problems one is that we can't necessarily tell when we're replicating in neuroscience because in particularly the older literature is insufficiently quantitative for us to be able to even ascertain did we get the same result as someone who did an experiment a similar experiment before a second issue is that [Music] [Music] excellent so point number two sometimes sometimes we're doing too much replication in particular when a new technique is being developed it's often tested on things that we already know the answer to that is certainly a valuable way to test the technique however it isn't necessarily something I want to be reading if I'm not interested in using that technique but then I think probably the most important point to make is that when we do fail to replicate when you know maybe that might be too strong of a way to say it but when there's some reason to doubt an established finding we are not doing the extra work that is needed to make sure that we sort of move those findings of the sort of primary list the Canon of things that get taught to new students entering the field and so the the fact that we have read checked something tested to see whether or not it replicates is kind of lost because people could just keep believing the old stuff okay final point is we cannot test everything we need a plan it is not going to be you know a understanding of how the brain works how it leads to behavior is probably not going to come from testing every single condition that you can imagine testing so I don't really have an answer to that I think that we want to start thinking about how we can focus on general principles people can think about how exactly we can generalize from the experiments that we do do and there are other disciplines that have thought about this a little bit for example in statistics there's a concept of you're going to sample some of the population you're not going to ask every single you're not going to query every single member of a population computer science brings us a number of concepts that I think are useful such as concepts of efficient search if you know something is happening somewhere between here and here a good way to look for where it is happening is to keep splitting the different the distance in half computer science also brings us other concepts of sort of how do we how do we abstract to general principles we don't think of computing as being based on getting your transistors from a particular manufacturer there's a concept of software and algorithm that that is at play okay so some advice for young neuroscientists yeah I know but it didn't go so I guess I would say try to focus on issues of functional and computational significance and have some kind of plan for generalizing and seek ways to be lucky seek ways to put yourself in a position where you can discover something you weren't necessarily looking for and I think that comes from sort of just paying really close attention and looking carefully at your data looking at it in ways you didn't plan to necessarily okay hi so like both Layla and Jenny when I was asked by David to do this I really had to give a great deal of thought to the question we're all really in the business of studying the intersection between cognitive processes behavior and the underlying neural implementation of these constructs and one of the kind of central questions is what are the right kind sorry what are the right kind of linking hypotheses that that we should be forming that really tie these different constructs together and I think Layla made an important point that the right kind of linking hypotheses really depends on what your goal is I think many of us have a goal of understanding how cognitive processes are giving rise to behavior and I'll give you sort of what I think of as a thought experiment like what is what is the right the right way in which we can achieve this goal so increasingly we have a large amounts of neural data there technological innovations that give us the ability to record from many many neurons at once this is really an enterprise that has been sweeping through neuroscience as well in the in this scanner world we obtained lots of data from the whole brain and I think increasingly with the widespread adoption of machine learning techniques there's been a focus on using these massive amounts of data that we have to try and predict behavior and of course this is a really important enterprise if we understand something then surely we should be able to predict how neural computations are giving rise to behavior but would this be sufficient if we had a sort of black box a predictive system and here I don't mean the cognitive process of prediction but any kind of mapping between these neural measurements and the behavioral output if we're able to predict that is that sufficient does that achieve the goal of understanding behavior and I think from a psychological perspective the answer is clearly no and I think Lailah made this point as well that our goal is not simply so if we are able to predict behavior but we have no idea how that black box is functioning then I don't think we've accomplished our goal of understanding so understanding I think requires a level of explanation and I think we're all in the enterprise of either we believe that that although we are undertaking a really complex effort to identify some mapping between brain and behavior that if we are cognitive neuroscientists we believe that there that it's not a few a futile enterprise to try and derive an understanding and explanation of what that transformation function is so what would constitute an explanation so we are tasked with specifying what are the latent processes that go in this black box these the nature of these processes of course are the representations that are important for generating our behavior and and the computations that are being performed on those those representations of of our world so are there successful examples of psychological sort of theories of formulating specifying the nature of the representations and computations that are being performed that give us insight into the neural implementation so I had to think about sort of success stories of good linking hypotheses that have given us insights into into underlying neural implementation and one is try trichromatic theory so in really beautiful work in the early 19th century young and Helmholtz did a series of psychophysics experiments really having individuals match the perceived color of a single wavelength of light with one or two or three knobs that they could turn to sort of calibrate the wavelengths that they felt generated a perceived color that matched and they inferred from the behavioral data from this psychological experimentation that three knobs that would that emit a light at short medium and long wavelengths appeared to be sufficient to produce the the perceived colors and this of course predated by a hundred and fifty years I believe the discovery of three types of photoreceptor cells that are sensitive to short long and medium medium and long wavelengths so again an inference about the way the computations the brain here the eye needs to carry out in order to approximate the behavior that's being observed okay so another one and this is very much a different a different kind of theory so cognitive maps so the work of Tolman in the mid twentieth century he was really challenging a pervasive view in animal learning that behavior could be understood through sort of the formation of stimulus-response associations and showed through behavioral experimentation that it was that that current theories could not count for the types of representations that animals were displaying they had that the theory was wrong and that one needed a theory that encompassed representations of a space and place and path this and indeed in subsequent work by O'Keefe and Nadel and by the Mosers it was revealed that that in the brain there are representations that enable us to carry out these kinds of computations so again an inference that could be drawn from from behavior that was able to inform the search for what kinds of representations need to exist in the brain okay and a third one so you know this this doesn't really reflect a theory that came solely from from the domain of psychology I think there were many fields including different fields of computer science and engineering that we're looking at similar types of processes but the work of vers Courtland Wagner in which observing animal behavior and trying to formulate what are what are sort of that what are the types of computations that can give rise to the dynamics of behavior that are observed in these sort of simple reinforcement learning experiments and they proposed that what might what appears to be going on is something like an error driven learning process where the animal is sort of doing sequential updating between updating expectations and using um experienced outcomes to drive this process and then many years later or in this case not so many years later it was discovered that the functioning of dopamine and open dopaminergic neurons appear to be carrying out a computation that is reasonably well approximated of course we now know that there's much more complexity to the computations they're carrying out but again sort of specifying what what a potential computation that could drive this sort of learning might look like and indeed finding that that sort of computation is carried out in the brain so how do we go about if if the enterprise that were that we're tasked with is generating good linking sieze how do we go about doing this you know these are maybe examples of places where we've succeeded but but what do we do as a field and these are you know again these are things that these are not novel ideas these are things that we know theory building so we all do experiments in our labs we generate data we test hypotheses and we produce sort of you know what what I am calling here atomic data point sort of single studies about how some manipulation pushes around behavior in a way and I think you know the really challenging thing is to draw kind of larger inferences from these atomic data points to gather together the observations that you've made with the other observations that are present in the field and really make inferences from these data that allow allow you to account for a sort of broader set of phenomenon and ideally to to take what seemed to be sort of diverse or disparate phenomenon and account for them with one sort of unifying theory and this may be sort of well I think one way in which we we can attempt to do this is to try or one useful way of going about this enterprises to try and construct generative models to construct models that specify a process for producing a behavior or a neural measurement when we understand a phenomenon well enough then hopefully we can specify this as a formal model and I think you know this is a somewhat controversial perspective but by that I mean a sort of mathematical specification that makes quantifiable predictions and that is not because these formulas are the way the system works I put oh sorry I'm forgetting to keep you in line here you go so the quote at the bottom is when I'm fond of all the models are wrong but some models are useful so I think these kinds of normal models are that that we have in our field are wrong but they can be useful in that they make quantitative predictions we can measure how far off they are we can compare one model with another to see whether the introduction of some simulation of a cognitive process improves our predictive ability or impairs it does it seem to capture the data in a in a quantitative way so so these features of generative models are useful and this last point sort of pertains to what is the right what is what is really the so revisiting prediction which in the you know in the beginning I sort of said it can't be all that we're trying to do but it's a really important part of understanding so when when you really understand the way a system works you should be able to make predictions about how a measurement or a manipulation will influence behavior and I think as a field this really relates to the point about replicability that was raised earlier as a field we have observed that many findings that that have been demonstrated in one laboratory may not be replicated in another I think you know one questions of scientific integrity aside one thing that we have to think about is when we fit models to data there's always and this is this applies to any sort of model a regression model there's always the potential that we've overfit the model that we've introduced a number to many predictors and that now we have a model that describes very well the data that we have in hand but may not generalize to behavior that has not been seen by that specific model yet and so I think as a field we need to adopt practices like and this doesn't mean do your experiment twice every time this can mean something as simple as using cross-validation techniques and collecting a slightly larger sample and keeping some held-out data this is this is the first step I think to creating models that will be able to explain behavior in subsequent observations and I think you know this is one step towards the the prediction component of the goal okay thank you [Applause] given the issues with the clicker and not being able to see the screen I'm really glad that I made the decision not to bring slides I'll just give you this it really has anyone ever seen something that looks like this before this century so unlike my esteemed panelists when David asked me to be on this panel to discuss this topic I said no because I had to be in New Jersey today so but I made it here two hours ago I just got off an airplane how many of you flew in here from somewhere show of hands all right and and on your flight if someone asks you what what you were doing going to San Francisco and I you know you said you were you know a scientist how many how many of you said you were neuroscientist show of hands okay and how many of you said you were a psychologist okay and how many of you never took your headphones off so you want to have a conversation like this okay smart so I I always answer the question neuroscientist because that stops the conversation because because people I think the assumption is wow that's really hard and I can't talk to that person and I'm just gonna go back to watching his stars born which actually it really is what I was doing in the plane but and when you say psychologists like oh I now can talk to this person because they do something I can relate to and it's easy and I think the fact that I make that distinction in my mind that there's not a transparency between the two of those that there's something different in this case I'm saying one is hard and one is easy is why this question is so important right it's not transparent what the relationship is but it can't be simple otherwise I wouldn't have to think how to answer that question because the answer wouldn't matter so I say neuroscientist because it's hard but when I was thinking about the relation between these it's really the opposite right it's much easier to measure something physical than to measure something mental and when I think about the relationship between neuroscience and psychology I think that neuroscience is giving us an easier way to do something really hard and again that's not to say neuroscience it's easy but I think it's easier and so when I struggle with what I'm doing and what I'm trying to understand it's always the psychological questions that are hanging me up and when I get a grasp on that then the neuroscience piece seems like it's gonna help out like it's gonna be a little it'd be a little easier so the next time someone asks you what you're doing on a plane just slide your head one time and but if but if I had hadn't started talking to this person I actually probably would have made some slides and then right now I'd be sitting here going like this and trying to see them and that I'll turn it back over to you [Applause] now to a different clicker yes okay thank you everybody so other than these pithy funny engaging thoughts there were a lot of isms and I thought and the isms that stuck out for me that maybe we can draw out some more about the specific question about the relationship between psychology are so and Leila talked about level chauvinism which was a kind of what we all do namely if you've worked on v1 that you think that the neurons of v1 explain everything in perception and Jani talked about a kind of methodological imperialism as if you're a single unit person or an fMRI person you see all problems through the lens of the data that you have access to and maybe that ends up Kenna lysing us into a certain way of asking questions that is you know artificially restricted and both Leila and Kay talked about a lot of you know the idea of bridging laws and linking hypotheses there was a lot of linking and bridge ISM and I think that's actually the one where we want to go because that's the that's the actually hard part and then of course there's the McCrone the thing that bothers me the most and I talked I already told Kate this at lunch predictive ISM it's not all about prediction damage I know you want to predict everything in your mouth but I just but so um if we're if we're to have successful bridging laws and linking hypotheses where do we start I mean do we start by throwing out and our narrow focus on the and the methods were used to or on the level of abstraction that we're used to I mean how do we go about what you want which is an explanatory model that's beyond just prediction as you pointed out this one I don't have a I don't have a full answer to your question I'm not sure anybody does but one thing that one clue that I feel means you're on the right track to something productive happening is if you have more than one hypothesis that you're testing so that you can you can you know frame your question as is it more like this or more like that and you know this and that don't necessarily have to in the end at the end of the day turn out to be the right thing but that comparison is more fruitful than saying is it this but do you think that students who are learning neuroscience now or psychology actually are familiar with the concept hypothesis I mean it's so that's so quaint I mean is that actually a thing I recommend that you're asking a tongue-in-cheek question but I think it's actually an important one because I think you know there I think there are people you could have asked to be on the stage today who would say that it's perfectly reasonable to do science without a hypothesis that you can take data and discover sort of reliable relationships between data and behave this is your predictive ism right you can discover reliable replicable relationships between neural measurements and behavior and have no hypothesis or no explanation about what underlies the mapping and I think this is you know increasingly common as we have very high dimensional data sets and is that where does that belong in your kind of scientific taxonomy is that is is that informing I mean this is a question for everyone is that informing our understanding of behavior I think is a really important question must all science be hypothesis driven I think hypotheses are really important for for pinning down explanation but I think there is a perspective that there can be discovery without explanation yeah I mean I agree I wasn't gonna respond to that but I think that's a great we all want discovery by accident and we all have hypotheses and I actually believe that we have hypotheses more than we let on from reading papers and part of what's gone wrong with the system and I'm feeling very rogue and rebellious about I'm getting to an age where you can reflect back and get angry about things but the what's being rewarded in reading paper you know getting a paper published and getting it out there and convincing people as a story and a story is very important because you have to know how to place the data that's in the paper into a narrative but then but the narrative always fits the data and hypotheses are meant to be to be broken are meant to not be mad and that's okay too and that's really something that we can benefit from and I don't I just don't mean null results I mean that option we have hypotheses about how behavior should move or how it should be related to representations in the brain and we get something that's a little bit different and then we're stuck with a dilemma about how do you write the paper do you write the paper so that it feels like the data fit in with your hypotheses or some hypotheses somewhere or do you write the paper true to your process and the history of what you thought and I always find it more compelling to think about why we didn't get that's the discovery maybe why you didn't get what you thought you might get but there is a little bit of nugget of the the process that the world that we're in we're all in a world and we buy into the things we need to do and and it's important to talk about where we think we are being best served as scientists and where we think are our papers and our story maybe not reflecting our hypotheses so I actually would like to assert that we have more hypotheses and we let on or tell other people about so I think the the kind of answer to your frustration is really where you started which is it's actually not enough to just have one hypothesis when you're doing science you you really need at least two because if your hypothesis is wrong it shouldn't just be a failure it should be support for something else right and so the idea that you frame something as I'm gonna do an experiment that's going to decide between two hypotheses as opposed to I'm gonna write a paper where it looks like there was one foregone hypothesis and no alternative I mean that's the answer right is that you have so having one hypothesis is really easy having two hypotheses that are both plausible is hard but what's really exciting is when not only do you have two hypotheses but where you could have two scientists identified with each one that could do adversarial science together right where so it's not a it's not a hypothetical scientist who holds this hypothesis where there actually is someone you can point to who would agree on the predictions right and so and I think that's you know sort of where you started so I would say no you shouldn't just have one hypothesis so a former colleague of mine Kevin Dunbar has done some really interesting work where he he studied the the discussions that unfolded at lab meetings in eight molecular biology labs over a two year period and then he did a linguistic analysis of all the discussions and one of the really interesting things that he found there was that there were there are a few different ways that people dealt with unexpected findings so clearly they had a hypothesis they got something unexpected and then what did they do and and there were a variety of different things they could do they could repeat the experiment the same way you know just pigheaded do it the same way again I must have done something wrong but don't have to figure out what it was just try it over again there were they or they sometimes you know reevaluated their techniques and made some changes but tried it again but the third way which was really sort of how he found that science was advancing was hey that unexpected finding led to a new hypothesis and interestingly there was a gender bias in those new hypotheses all of the new hypotheses were generated by women postdocs but so in the study of psychology turn me on it goes the I can I'm certainly not an anti hypothesis person but the there is a do you agree with the idea of just wildly measuring things in psychology there's social observatories you know sticking a camera there recording all these nice people sitting here quantifying away correlating with everything with everything is that likely to yield something in psychology that we can then bring to neuroscience or did you actually need a well-constructed hypothesis that canta that followed from some set of observations or phenomena or a theory god forbid I mean I so that we bring implicit hypotheses to the task I think that's true and it's even better to have more than one is also true had to have alternative is kind of the name of the game but the way our funding is going and the way people think about psychology is increasingly about naturalistic things right so naturalistic scenarios like huge observations of things is that gonna work for psychology or is that actually going against what we want to do I mean I think a real advantage of that approach is that if our goal is to understand behavior you know the the challenge that we always face in the lab is one of what are we measuring are we measuring something that has real ecological validity and then you know as we make discoveries we have to we have to really establish that the thing that's happening in the lab has construct validity for something that's happening in the real world so I think you know something that's very valuable about studying naturalistic behavior and I think you can study naturalistic behavior in a way that doesn't necessarily toss out sort of hypothesis driven science as well you can have hypotheses about the way cognitive processes or information processing or neural process sieze may sort of function in those ecological settings but you know I think there's real value i I think we don't do it as often we don't do it often because it's hard because it's hard to get measurements because it's hard to control and you know I think the the sort of increased movement in that direction is not intrinsically problematic I think it's I think it really has the potential to enable us to understand behavior that is more like what we sort of cognitive processes and behaviors unfold in our in our everyday life yeah I mean I think it's a little bit dramatic to think that it has to start somewhere and then somewhere because that's just like that's just about arguing and then we might as well be philosophers but so it's instead of a straight line and like you could start with like observations or hypotheses but it's really a circle right it's really that we come into it with some hypotheses based on observations and data and then we create new hypotheses and they change ever so slightly or dramatically depending on what the nature of the new data are and so this is the cycle and I think that it's healthy to be in the cycle and then there are some times where you want to go off and you become really interested in digging down and a result where maybe you have the same hypothesis for ten years that could happen too but I think that there's a lot to be said especially for you know early career researchers and I think there's a lot to be said for realizing that there's so many whether you call it levels or layers or questions that haven't even begun to be answered and that being inspired by what is interesting to you can often come from introspection and observation and I don't know how alive that is in the community whether it's pooh-poohed or like whether we have to be like you know big data or a computational model like what's the right you can just still think about a question just think about how men works for you what's interesting does it think about whatever it is and you might discover you're not only your interest and what's what do you what have you observed whether it's in a crowd or in your own mind but you might very well find that the way that you approach it hasn't actually been approached before there's so many holes that need to be filled and you and I know that when I started doing you know I was in a neurobiology department and I started doing my work I felt like I was in the lab with all of these people who knew more than I did because I was the junior graduate student with like 10,000 postdocs that's what it felt like and I felt like I know nothing in this group what do I know I just have to learn and then all it took was going to the coffee shop to look around and then I thought oh wait I know something here and what do I know and there's for me that was my process that doesn't have to be yours but realizing that there's so many questions that haven't even no one has thought of before and hasn't been addressed I think should be invigorating for for future study I mean Jenny had also said is that what you mean with get lucky or be lucky or find your luck or prepare yourself for luck or something like that was as allow yourself to introspect you do it you know I didn't really I I threw it out there you know I hope people will find a way that works for them but I think it's I think it's important to be doing experiments that go beyond the hypotheses that we can currently formulate I assume that all of your students and postdocs as least are allowed to just read and introspect and think and not be in the lab therefore they do that in the lab for sure so what are the recommend the let's get back to the hard parts here that we're trying to find you know bridging laws or linking hypotheses or something to help us understand the relationships between brain and psychology is the way to go introspections you got to give me something I mean how that's simple David I like that simple well it's just you know I don't think there's a formula I think you know introspection maybe the beginning and then Extro spectrum like reading and understanding what's there is really important too and and I think that we talked about this I think that the scholarship and knowing what has been done already cannot be overvalued and it's really difficult because there's so much work that comes out in all of our micro fields that you can barely keep up with what's happening in the last five years I remember when I started a lab I made a deal with my graduate students that I would tell them everything I knew up until 1994 because I knew all of the literature and that they had to cover 1994 until now I made dating myself but and I that was not successful so I I think you raise an interesting point which is that I think it's particularly challenging when training in cognitive neuroscience in this day and age because they're sort of an increasingly large skill set that you're expected to acquire so almost everyone sort of learns to program and learns new sort of methodological and analytical techniques and these take a long time to master and these are really important skill sets to to master in at that stage of one's career but at the same time time is finite and everyone needs to get eight hours of sleep so you know it is a zero-sum Enterprise and you know time one only if one wants to sort of ask questions in an area that hasn't been probed already well how would you know if it has if you haven't read very broadly or deeply you know many of my scientific insights have come from wandering through literature's outside of my discipline I really like reading behavioural ecology which is sort of like my fun reading area when I'm when I'm sort of not focusing on the on the primary work but you know when does one have time to do this and is this really a part of scholarly training now I think it's a really important part to be able to spend time kind of wandering I'm having an idea wondering you know who's thought this before who's shown has anyone shown this before you know and really diving down the rabbit hole of previous findings and seeing that citation and going and looking that up and it's you know this is this takes time and one as a graduate student or postdoc or you know throughout a career should spend many many many hours doing this but is this something that you know that that students have the freedom to do now I think it's I think it's an important question so one thing that we talked about at lunch then I would just wanted to include in the discussion now is that you know reading the literature for me isn't a matter of starting at the beginning of a paper and reading it straight through I am typically reading the literature to find out some specific thing and to find out whether it's and offered to find out whether or not anybody has ever done it so looking you're you you sort of transition to from like memorizing a whole bunch of stuff to trying to find the holes and it's a different way of reading a paper and you might not read every word you might just look at a couple of figures and be like okay yeah this paper does not have the information I was looking for move on to the next one I try to not read anything and remain unshackled from the literature that allows me to I want to pick up something that Sharon said that I agree with but I think it has all kinds of dangerous consequences which is that um the relationship between psychology in the brain is obviously not simple but you said tongue-in-cheek II but I think you're right that neuroscience is easy and psychology is hard and so we end up picking the problems is we use Neuros neuroscience techniques of various sorts to try to address questions that may or may not actually be interesting parts of psychology so are we artificially let's say catalyzing ourselves in terms of the kind of psychology we can do because we do Merl sciency stuff just because we can I mean the interesting thing the interesting parts of psychology I mean it may not be visible to us because we just say well we're gonna do the neuroscience parts that's much easier I mean are we missing something in psychology altogether and is there a psychology gonna be left in a few years actually there might there might be no more as I called you anyway I told you not to let this get depressing this is the start of the meeting don't bring us all down no I mean I think this is the point of that cartoon of looking where the looking I couldn't see the cartoon but the version I know is you're looking for the keys not where you lost them but where the lamppost is I don't know if that was the cartoon so I mean yeah obviously I mean you know thinking about the slide with the levels of analysis of behavior I mean we because most not all but most neuroscience methods are limited to individuals but huge facets of interesting behavior is not at the individual level it's the diet or the group or the societal level right so there's an example right there of something that you're missing when your unit of analysis is just the individual that's not a criticism just of neuroscience I mean that's you could criticism of a lot of psychological methods too but yeah I think that's you know this comes from again we're focused on learning all these tools learning all these methods and then we get anchored to them and we ask the questions that those methods are good for and we don't have time to stop back and say what are the questions that I think are interesting and should I go do something else right so I'm agreeing with you even though I told you not to let this get depressing so but I think that's again where this discussion like this I mean I I don't know if you guys are having fun or you're just politely waiting for the reception to start at 6:30 but you know I mean I think that's what was fun about this is cuz we don't have time to stop because we're all constantly doing things too and you know just stop and you know think about these things that say okay let me go back to why I got into this in the first place what do I really want to understand and and then ask the question what's the relationship of neuroscience and psychology around that question and maybe the answer is no for for some of these questions yeah I mean I think I didn't want to sound chauvinistic either by arguing that understanding only comes from across level analysis because maybe your goal isn't understanding maybe your goal is predicting or knowing what works so for example you don't need to know anything about what CBT is doing to know that it's helpful to two patients when they when they come out of a session you there's so many examples of that and do if you want to understand behavior in the mind or you want to understand more about what correlations exist or how to move around there's so many laws that are out there in memory for example we know that there are ways to enhance memory and those are phenomena we don't understand them in terms of levels of analysis like the trichromatic the one example Kate was able to come up with which I was impressed with so for example the spacing effect levels of processing the contiguity effect now has maybe maybe that's one champion that we can talk about but you can you can learn a lot about cognition by understanding the internal workings how to nudge it around how it works like the spacing effect works all the time right almost but we don't understand why and there are theories and we haven't gotten I would argue we haven't really gotten further in our understanding so maybe and it's okay if your goal isn't to understand it at a lower level but I think the idea the trope is that if you can understand a higher level at a lower level of instantiation that you're gonna constrain and have a better understanding of the higher level that's the goal and that's what we started with but that could take a lifetime or two right and that's there's a long so the question is how much time is too much to ask a question I don't know these are interesting thoughts but if the arc is so long are you really asking a useful question anymore and then useful depends on for whom and I think you're making you're sort of advancing an argument that that I know David has advanced which is sort of that that neuroscience needs both behavior and theories that in order to sort of serve the function of [Music] explanation that it one can't simply one can't simply use neural measures to predict outcomes what you know that that being able to sort of test I have a hypothesis or arbitrate between two theories by testing two hypotheses that this is sort of taking a well-understood construct and probing how it may or may not be implemented is doing neuroscience in a way that fosters understanding yeah so to make reference to the sleep talk just before this session I mean I think that's a really good example of something that isn't really fundamentally explainable by anything we know from a strictly psychological domain why would we have to be unconscious for eight hours a day that can't be good evolutionarily speaking bad things could happen to us during that period of time so there must be some fundamental biological need and we sort of we see this in the connection to all the it'll ill consequences from a health perspective and you know I've been really struck by new findings suggesting that during sleep the lymph fluid can flow through the brain and so you know it's possible that there needs to be some sort of cleaning process that happens to catch us up homeostatic Lee to you know to get rid of all the stuff that you know we had to use during the day to think and and maybe there just it wasn't possible for evolution to come up with an alternative to this so I think that you know that I don't think this is established yet but that's an example of where there's the possibility that that at a neurobiological level at a biological level there will be an understanding of what the constraints that evolution is operating under that can't come from just looking at the behavior I think we're willing to take questions but while we're while you guys are formulating your questions which are likely to be very easy and pleasant and polite what would you want your students to to tell you right now and the basis of our discussion at lunchtime discussion now if you're a student of mine you've been here is now what I think you should do read only philosophy and engineering don't bother with neuroscience and don't bother with psychology learn the toolbox of engineering and learn how to think critically and read a lot of philosophy and the other stuff will be like falling off a truck maybe that may be wildly optimistic but that's what I think what what should they say so we have a lot of people lined up for questions so maybe we start on the left and then we'll alternate go ahead sir yes my chauvinism is for cognitive functional models based on behavior and I see a real asymmetry between the cognitive psychology and the neuroscience levels neuroscientists seem to heavily rely on guidance from cognitive psychology particularly the work in fMRI is an example going the other way I I'd like to learn about the benefit of being able to adjudicate between competing cognitive models on the basis of neuroscience evidence I think there's very little or no progress that in that direction and I'll just close with two examples that I've been interested in are there there are functional models of two varieties with regard to how you read English aloud do you use two routes or one another research topic again of dis personal interest is well whether or not speaking two languages enhances domain-general cognitive functioning now I think there's been a great deal of neuroscience work on both of these questions but I don't think it's helped us at all in deciding which cognitive models are on the right track I was just gonna say didn't you hear the part where we were supposed to get easy questions well I don't I don't know the area but I think the the one could argue that the existing if the existing neural data better predicted the behavior then that is a way that it could constrain which model of behavior is better so that's how what we try to do in the lab is try to always use behavior as a factor to use the neural data and try to understand whether the neural data is predicting the behavior of different theories so that's one it's an ideal possibility I don't know in the area of of linguistics and language learning whether that applies one thing that I thought of when you were saying that that I forgot to say is that I do feel like and this could be because of my lineage and where I came from in my history that I and hence my students and the people that I interact with here at the meeting have been very inspired by animal models and neuroscience over the decades and a lot of the work that we do builds in the knowledge of what single units are doing in that brain area that we think is relevant and so that's a part of how we think and what I one piece that I think is really missing and I think you touched on in your first question is the neuroscience community doesn't do the same right so they I think in and this is broad strokes there aren't a lot of neuroscience papers that are looking at single units and brain regions or oscillations that are inspired by findings in psychology right and so there's there's a bit of a like a bottleneck and that we know a lot about how the translation translational that's a thing NIH loves what's translational like can you do an animal experiment and a human why is that translational why not understanding human behavior at an implementation level in the animal I think there's something missing and it's just not working like I think it really to make progress there has to be a union across the fields where we're exchanging ideas learning from what the other field is learning and that's not happening yet I agree that I was thought that human speech was a good model for birdsong at the risk of make surprising Sharon I actually think there I'm actually optimistic about that there are in fact cases where neurobiological data adjudicate between cognitive science models in particular in psycholinguistics I mean this is not the moment to you know advertise that kind of stuff but I have colleagues and friends who actually do so I think I think there is that opportunity to have carefully specified cognitive science models that are then decided where the neural data can be decisive so I think I'm more optimistic on that we need to talk yeah on this side I have a question with reference to a metaphor for the mind brain relationship and and the metaphor is this it's the whether the relationship between the atmosphere and weather as compared to the brain the mind because the the weather we experience around the globe is an emergent property of the atmosphere that we have you're on earth if you go to different planets you have different weather based on the physics the energy eating of the Sun things like this and I'm wondering if that metaphor is reasonable for the mind brain relationship and I was gonna ask that to anybody on the panel that would want to comment on that metaphor you don't like the birds or the water well you know the the atmosphere no one no one directs what the weather is gonna be on the planet no one we've got billions of neurons and they create behavior and this massive atmosphere creates weather and I was wondering if that metaphor applied to what we experience with the mind brain relationship where you have billions of neurons creating the mind as an emergent property yes I think it's as good as an emergent an example it's a great example of an emergent property and I think and you know it's one of a number of things that are in nature that and that also is the case with the mind brain relationship I think we don't know enough and hence there clearly is an emergent emergent characteristics of the mind yet going along with that then it seems to me that the cognitive neuroscience community is about where meteorology was a hundred years ago seriously in terms of coming up with understanding and formulas so just the thought thank you yeah I'm gonna go with no because it's even more difficult than that actually say 200 years ago yeah I'm sort of familiar I'm in a bit of a position that I'm sure a lot of people can relate to which is what I like to call the interdepartmental shuffle where you know I because I'm interested in both cognition and the brain that I end up between psychology and biology departments and also other departments but that's besides the point and I find that a lot of the people in my shoes who are interested in this and in this interdisciplinary zone are at the scanner level whereas those that tend to be in that more biological path tend to be the quote unquote electrode level and I'm curious in terms of getting that electrode level look at things to be more to be looking more at the cognitive aspect if we should be focusing really a on getting those pure biology people to start asking questions about cognition or if people like us who are already doing the interdepartmental shuffle should be looking at that sort of methodology more or if see at this point we shouldn't even be bothering because looking at something at such a small level is hard harder to relate to cognition i'm with sharon and saying i think it might be easier than euro sighs maybe easier I would go with trying to get your colleagues to get excited about questions at a different level because otherwise you're just adding yet another task to your day and you're shuffling already between a bunch of departments well I guess not necessarily for myself but should people like us just deciding to go into the other types of methodology as their focus or rather getting the people already in that methodology to shift their mindset a bit I I think this is a complicated enough enterprise that were we sort of broadly psychology and neuroscience are sort of good and cognitive neuroscience specifically that we are embarking upon that sort of work at every level of investigation is important for something you know even neuroscience that is not at all interested in explaining cognitive function or even behavior an understanding of how you know if neuronal signaling is going to be important for for building upon that level you know I think I think every level of investigation ultimately is going to contribute valuably to the enterprise of understanding how brain gives gives rise to mental phenomena but of course everyone has to find the level of investigation to work at and I think the right level is where your questions are yeah I was gonna say what's your question your real your deep question yeah and and then you should you should take it upon yourself and your training to do all that you can do to put together the pieces that you will want to do in the future thank you I have one more comment to make on this about I think it's so helpful to be welcoming to people that are coming from different backgrounds and wanting to ask you know questions that are sort of touching on the domain that we work in it's scary to go outside of your own and you're you know you'll speak with an accent in the new domain and if people judge you for simply speaking with an accent or using the wrong you know using the terms using jargon and not the right you know socially approved way then that that's an impediment to people succeeding in these interdisciplinary endeavors thanks for talking about this I have a question that's not directly related to psychology and neuroscience but I think it's important you guys already brought it up about when we find things that are not consistent with our hypotheses and people often fit hypotheses to have a narrative or a paper and ideally to get in high to your journal and that is like a really valuable thing and someone's career as a young scientist to get a postdoc and get a position and I'm curious like incentive structure is built that you have to kind of tell a story that's simple and exciting to a wide range of people which is why a journal like nature of science are the only ones that you know non scientist might actually read I mean that's not exactly exactly true but you know what I'm saying and so I'm wondering what can we do as a young scientist and postdocs and grad students to you know recognize when we're wrong in a paper and still get it published and then what can you guys do as well as faculty who are deciding who gets a job based on their publications and when you're reviewing articles like I think we need to be talking about this and doing it a lot better and so what do you guys think we should do well I've actually been really excited by the recent surge of rebellious nough surround publishing and putting out results on in whatever format your blog or bio archive or whatever because I think that there is a push that's happening that will end up for good in the future the other thing I will say is I think there we I'm sort of over it I'm over the high profile Journal I don't think it tells me about an individual it tells me that they not only had a finding but that they were able to package it and get it into a high profile Journal which is good that's what it tells me but it doesn't tell me if they're a good scientist and when I'm on a panel a chair of a committee to hire which I've had several opportunities to be I spend time reading the research statement and getting a chance to think does this person think in a way that's interesting and attractive and deep and you know have all the elements that I think would make a good colleague and scientist of course that's easy for me to say you you have to publish something but I don't think you have to publish in those journals to get a job and you know there isn't prot there may be an advantage it might help you get your foot in the door maybe but it doesn't get you the job because you still have to show up and talk to people about your science and that is not about where your papers get published that's just about how you think what the results mean to you what are your theories are you scholarly have you read the literature do you ask good questions those are the things that I look for when I'm on a search committee I just want to echo the bio archive and add to this that there's stuff that that the community can be doing to to sort of diminish the dominance of the glamour mags in particular in in in my lab my students pick the papers that we are going to discuss in lab meeting and they pick them from these journals and that's not good you know you so I would urge students postdocs who are choosing the literature that your lab reads to not just pick papers that are in nature neurons sell whatever but you know can try to find something you might find I mean I think the articles that come out in Journal of Neuroscience for example end up being a lot more readable than what gets published in some of these really compact formats very brief remark if I may on the finding things that don't work out or having a hard time with the kind of thing that I think a good Maxim to follow was the the wedding vows that Lori Anderson Lou Reed Lou Reed and Lori Anderson had a this was their pact which I think applies to the young scientists or all scientists and be completely fearless don't be a rate of anyone or anything have a really good detector and be tender be nice to each other be nice to people in the lab and I think those things and you can deal with even non findings and hypothesis non conformed things but just don't be afraid because in the end of this more of a bummer for yourself I think so that's a great argument for maintaining the tenure system so you don't have to afraid you can so when I stood up in line to ask a question I didn't know the person for me was gonna talk about the tool dual route versus single reading models and connecting it to the brain my student has a poster tomorrow and that very thing v62 connecting it the models to the brain and testing them against each other anyway so I those are great Helmholtz examples Kate but the the connecting you know having a generative theory that makes predictions that you can then test I would add to that I guess the vareniki like time model even though I get you know we know it's really too simplistic it did also take brain and behavior and make a novel prediction that turned out turned out wrong though right well conduction aphasia was found to be a thing but yeah I mean beyond that sure there's plenty of details you could argue about but two to David's reading list I guess I would add William James principles of psychology because what's cool is Russ pulled right gives a similar talk about connecting brain and behavior and he says well if you look at the terms William James uses and I saw a lot of them on your slide to the mental terms that they haven't changed in a hundred years but that doesn't mean haven't learned anything and you can read his stuff to see that well with memory anyway I I'm rambling and I'm gonna stop now but um great discussion thank you hi so what are the major issues that's inherent to psychology is that there is no agreed-upon definition for many constructs this is especially apparent in higher-order concepts such as intelligence emotion and personality and my question is how how is neuroscience how can neuroscience help psychology of this issue or if I want to be provocative is try to find the neural substrate substrate for undefined concept in exercise in futility yes I mean I think what you're saying is is is it worth diving into the brain if you don't have a good construct of the this I think that would be crazy that's what neuroscientists do all the time but they're not looking for they're not looking for explanations of you know sort of emotional experience in humans or personality should personality actually exist that's you know so I mean I think that's the hard work that Sharon was talking about right that's exactly why it's so hard is understanding or having a construct or a description or hypothesis or theory about mental states and how they give rise to experience is extremely difficult but I would argue that that for me at least is a really important starter before using the neuroscience to try to you know fine-tune or inform those theories a little bit better and test some of those hypotheses but is there any way to use nearest science to help psychology of these issues that they're that psychology has been tackling for centuries like defining personality defining intelligence you know no way I mean those are terms from you know you need the psychology of believes desires you know issue I recommend some Fodor it's for a Mental Hygiene so that you know what is the likelihood that neuroscience comes up with a appropriate decomposition of those complicated concepts that then have any psychological force it just seems vanishingly unlikely because they're categorical II different retina this is the quote that I showed at the beginning they're different in kind this is why we need linking hypotheses and this is why we're stuck they're different in kind that's a problem so no that's my answer sorry yeah I mean I think at some level constructs like an emotion are terms that refer to a diversity of subjective experience and so you know going looking for neural substrates of emotion as it's probably a futile effort I think instead you have to try and isolate some component of subject's subjective experience that you're interested in studying or some function of a subjective state you know but you know those things I think are enterprises that cognitive neuroscience has you know begun to probe but I think the idea that we will discover things that reflect some neural embedding of a concept like personality or emotion agree it's it won't exist in some like clear form of implementation because it's it's just not yeah it doesn't function that way well I disagree with my colleagues up here I do I do think that neuroscience can contribute to some kinds of some of the domains that you mentioned perhaps intelligence might be a fruitful one not at the individual human level but in comparison of humans to near animal relatives I I could imagine you know a few decades hence there being some findings that would be useful for understanding what it is that makes humans have the flexible intelligence that we have in comparison to other species well that's another kind of chauvinism isn't it yeah except we haven't we haven't learned how to sleep but animals have figured that out apparently thank you hi thanks my question was relating to the different levels of analysis like behavior versus brain versus microbiome behavior versus oscillations and the dimensionality of the phenomena at each of those levels so basically I am curious if the say observing something at the behavior level we can decompose it into a low dimensional kind of explainable theory and does it become more difficult to do that when we get it to the neural level or the data that we observe are very high dimensional and is it useful to kind of I guess by high dimensional if we're doing fMRI we have like a hundred thousand voxels or so per brain slice and we're trying to decompose that into maybe some type of relevant network then we try decompose that into 17 networks and then so on and so forth to try to decompose the data into this low dimensional structure that we can explain easily and I'm wondering if that's even if just one if there's a relationship between the dimensionality of the of the data of the phenomena that we observe depending on the level of analysis that we look at and if explain ability at the level of behavior looks the same as the explained ability at the level of neural data or if predictions and like hi the easiest way to explain neural data is still with like a very high dimensional model to try to explain the brain so it seems like you're going back and forth between thinking about the dimensionality of the measurements we have and the dimensionality of the phenomena the constructs and and there's where I would take issue with with the premise of your question I I don't think the mental states are fundamentally lower lower dimensional that's psychological measurements are tapping the lower dimensional constructs but rather we tend to measure fewer things using psychological tools than we do with neural tools and that makes that makes it seem easier because then you just it looks like you have something in lower dimensional but I think this goes back to the very first question of where neuroscience data can be helpful because often times we have very blunt instruments with our behavioral measurements that are actually taking a highly multi-dimensional mental space and projecting it onto one or two dimensions and then the hope is that we can maybe get at more of that dimensionality by taking multiple measurements but I it seem like in your question there was this belief that maybe them the mental spaces were somehow had a lower dimensionality than the physical ones and that I don't think is right I think it's just a measure it's just a measurement measurement issue right three quick punchy pithy questions okay yeah thanks we should talk about this more the first thing I'm a psychologist by training and now I'm finding myself in defending neuroscience because I think we can the one thing I really think it's which is important is that we can learn about the constraints of the machinery so the brain has constraints like we have this 10 watts or something which which is the energy we have cognitive models usually do not account for these ideas so I think we actually can learn a lot from neuroscience and the second thing is for evaluation of our models I would say we have to go one or two steps further and say let's say do diagnostics and interventions with the theories we have so develop the things one step further and look at if we can diagnose the stuff we're interested in and actually train it to be sure and I don't know what you guys think about that but yeah well your first point is one that we made already so I think any levels of analysis that you're bridging will do you provide constraints and that's part of the argument for why these Brit that you know why going to the lower level is interesting and provides an explanation because it will constrain that that's not just between behavior and neuroscience that's between networks and neurons that's maybe between neurons and molecules like any higher level has these emergent properties that could be explained so I think that we're on board with the constraints and I didn't quite catch the model testing part that you the second part I want to bring up something a little controversial and it's already come up a couple of times about subjective experience I think that's kind of the elephant in the room it feels like what we are trying to bridge is something about human behavior and obviously you know the the physical level of the brain right and I think you know it came up and the response generally was no we can't we can't get down to the bottom of emotions or personality or intelligence well intelligence was maybe suggested as something that we could look at and I just wonder is there room for introspection is there room for admitting that we are all people who are thinking all the time and know a lot about thinking because we are doing it always and can we say yeah this will be hard to measure but maybe that's the thing that we need be measuring maybe that's the connection between these two more objective fields so I guess is there is less fear about talking about subjectivity maybe one of the keys I want to make sure that so I come i complete my earlier response to the question about is you know can we pin down can we operationalize emotion a construct like emotion and a sufficiently specific way to study it in the brain and I think my point was that the words the sort of lay concepts that we have for subjective experience are not there's not a sort of one-to-one mapping between the word and the sort of complexity of the subjective experience subjective experience of course should be amenable to scientific study I think I think we really all agree and I think the idea that I'm William James on someone's recommended reading list really you know a lot of a lot of a lot of his insights into the into cognitive function were really derived from from introspection about the way the way the mind works I think that's you know certainly a valid source of hypotheses and I also think subjective experience it's absolutely a valid kind of topic of scientific study just to be clear but it's challenging and I don't think the concepts that they'll a concepts that we have to talk about these things are the right kind of way of operationalizing them for experimentation I think you bring up a really good point and I would say that in my experience the world of neuroscience / neurobiology is not talking about this and I've made a little habit of mentioning the theory of embodied or grounded cognition to my neurobiological colleagues and very few of them have heard of it before so there's you know there's there's conversation that needs to happen to bring that focus into the neurosis level thank you hi my question is what's the difference between prediction and understanding and I keep wondering why predicting behaviors based on your todavía sound like machine learning blackbox doesn't feel satisfying to me and my own answer is that because as human being we like to make predictions by ourselves we like to assimilate the future by our own brain and I wonder whether you agree with my notion of understanding and we don't have we don't have the time to discuss about the next question but if you do agree with me the next question will be whether we are just not smart enough to reach that they walk understanding I completely agree with you understanding happens in our minds right so the unsatisfying thing about prediction without without that is that you know we've no idea how it works the whole enterprise of trying to figure it out fails if we can predict it but we have no idea how the system works and then your second question field right but I think you know this is the enterprise were embarking on I think we you know there's there's some belief that the optimism of the room is that you know that there's some potential for for discovery for it for explanation that but human progress will enable us to achieve but yeah it's a great way of finishing off it's a very hard question thank you and thank you panelists for discussing [Applause]

Channel: Cognitive Neuroscience Society

Views: 13,691

Keywords: neuroscience, psychology, science, brain

Id: Ix9bAx8Mwrs

Length: 100min 27sec (6027 seconds)

Published: Fri Apr 05 2019