share interactive transcript request transcript/captions live captions
download
|
MyPlaylist
SPEAKER 1: This is a production of Cornell University.
SPEAKER 2: So this is exciting because this is the first time that NB&B, to my knowledge, is at a joint seminar with Human Development. And we're in a brand new space, both in the sense that it's architecturally new and conceptually new that we're all here together around I think a really great speaker who really unifies a lot of interests around-- dispersed around the university, Daniela Schiller from Mount Sinai.
So she, I just learned, was a psychology and philosophy major as an undergrad, and I think often, people's undergraduate interests manage to resonate through their careers. And I think you can see in her work a philosophical foundation that spans both her work with animals, clinical orientation of schizophrenic rats. We can have a philosophical conversation about that. And then onto her work in social navigation in humans.
So for her graduate work at Tel Aviv University, she worked with Ina Weiner on these schizophrenic rat models, and I just learned that latent inhibition which they express is actually observed in human schizophrenics and can be treated with some antipsychotics. So she did really basic science animal model work for her PhD.
And then I think very bravely and maybe smartly, went to join dual-postdoc with Joe LeDoux and Liz Phelps at NYU where she-- I think this is-- did some of the work for which she's best known, looked at reconsolidation mechanisms, the vulnerability of memory at those brief periods around the time when you recall something. And that got a lot of press, and even, in some sense, along with your life history focus in a really interesting New Yorker article that I all recommend you read about Daniela.
And so now in the last few years, she's been really forging a new path at Mount Sinai, taking on more social questions, and also but still grounded in these basic science methods and considerations. So I'm really excited to have her here, and I think I can speak for all of us and give her a brief round of applause from everybody.
[APPLAUSE]
And it's a tradition of the Department of Neurobiology and Behavior to grant our speakers with this mug, which has a lot of the animal models used in Mudd Hall. And I looked for a human, I don't see one. But there's a human brain, right? Human brain on here, which is really what binds all of our interests. All right.
DANIELA SCHILLER: Thank you. I love my new mug. OK. Can you hear me well? All right. So thank you very much for the invitation. I had a great morning. I also marveled the bus ride over here with the coffee and all the treats and I'm very much enjoying this day.
I'm going to present a fairly new data that I didn't have a chance to present as much. I chose not to discuss the reconsolidation work basically because I saw that a lot of people here are interested in social psychology and social neuroscience. But I do plan in the future to maybe merge the two lines of research of the social representations and the reconsolidation, and the discussions I had this morning were very insightful in this sense.
So this is an example of how humans are really exceptional-- have exceptional social skills. It's running. So people seem exceptional compared to other animals, and we feel that we're very much developed in our social interactions. And this intuition led to the rapid growth of social-- sorry, it's sensitive here. To the rapid growth of the field of social neuroscience and the search for the so-called social brain.
And the idea that maybe there is a set of dedicated brain regions that evolved probably recently in evolution and maybe are unique to humans for the sole purpose of processing our social interactions and social information. But it turns out that social skills refuse to neatly map onto anatomically defined regions, and the computations that the social brain is performing are still obscure, so the search is ongoing.
Take, for example, the situation of the speed dater. So Andy here has just a few minutes to meet and evaluate and assess these women and decide if he wants to further date one of them. So see, for example, let's say Amanda. He gets a lot of information in a very short period. So for example, she's a social activist, she smokes. She's kind of selfish, seems she likes babies.
So he-- Andy has to integrate all this information. How would he do that? Or how would Andy's brain process these types of information? So imagine you're a participant. In your speed dating situation, you would see a face, and you will get these different bits of information.
For example, he told the student that he just wasn't smart enough. And then he said that everyone else had missed a main point or he was distracted during the presentation and so forth. So this was a set of three sentences that depict negative information, which are separated from additional three sentences that depict positive information.
For example, he excused the new team member's error, explored the local woods instead of going to the mall, and picked up his roommate's package on his way from work. And then you form an impression and you give your evaluation how much you like that person.
So the structure of the test is as follows. You have these three negative bits and three positive bits that are presented either first for negative or positive, or first positive-negative, it's counterbalanced, and then you give your evaluation.
So if your evaluation is positive and you like that person, then we take the positive information and assign it to a condition that we create retroactively that we call evaluation-relevant, whereas the other-- the negative information is assigned to a condition that we call evaluation-irrelevant. So now we have two conditions.
So you can see that if each one of you saw this information, some of you had a positive impression, some of you had a negative impression within the evaluation-relevant condition, we would have both positive and negative information depending on your decision, so it's subjective.
We capture the subjective preference of each person and we put it in-- we compile it into this one condition that we term evaluation-relevant versus irrelevant.
And then we want to see how the brain is processing this type of information that is subsequently going to affect or shape your evaluation. And this is similar to the DM effect, the famous DM effect in memory that you have a different set encoding due to subsequent remembering. Items that are better remembered are encoded differently than items that are later forgotten.
But at the same token, we just took the same principle-- the DE effect, we call it, which is would see a difference in the encoding of information that is subsequently going to be relevant to your evaluation. And we asked which are the brain regions that are going to differentiate these two types of information?
So those of you who are familiar with the social neuroscience literature, there's not a lot of brain regions that you could guess would be involved. One of them is primarily the dorsal medial prefrontal cortex. So luckily, we didn't find that region, and that was the good news.
When we searched in the entire brain to see which regions are dissociating these two types of conditions, we saw the posterior cingulate cortex very robustly, and also, the amygdala dissociating these two. So these are actually regions that weren't traditionally or typically involved in social processing, but actually in valuation processing or associative learning and learning to evaluate important information in the environment and so forth.
This speaks to this trendy discussion of talking about domain-specific versus domain general in social processes, and the idea that actually maybe there isn't a social brain, but rather, social information is processed by regions that are processing-- doing basic computations in many other domains.
AUDIENCE: Is it a function [? of ?] [INAUDIBLE]?
DANIELA SCHILLER: Yes. So we also find nice correlations that the level of your evaluation correlated with the extent of the intensity of the brain signal that we find. So the more you dislike a person, the stronger the signal is in this region during the negative beat. And the more you like a person, the stronger the signal in these regions, the more you like a person.
So this is just a quick demonstration of a relatively old study from 2009 where we showed that you have a representation that is specific to the significance or the evaluation that you're going to have about that person based on the information that you're getting. So while you're processing it. And this is a formation of a first impression.
And then we ask, what happens later? Now you start to interact with that person and it's the happily ever after life. Life continues and you start to sort your social environment based on interactions-- active interactions. Here, what you see is just a passive evaluation.
But in real life, for example, you join this new department. Through interactions, you get acquainted with the people, some people get closer to you, some people remain far away, some people gain power over you, some people give you power. And over time, you start to shape your social environment and it gets shaped into a structure that we maybe even represent spatially. So people above you, people below you, close to you, and further from you. Look something like that.
And we wanted to capture how the brain is tracking this navigation through social space. So to the study, that we first had to identify the main factors that influence social relationships. What are these factors? We went back to social theory, and already in the '70s, the theory by Wiggins and colleagues of the interpersonal circle, this is a conceptualization of interpersonal variables on two dimensions of dominance and what they call love.
And vectors represent the relationships that-- or your assessment on these dimensions or on these variables. So the magnitude of the vector will be how much you're deviant in a particular interpersonal variable, and the orientation is where you are located.
So this gave rise to additional theories that captured the same idea. For example, the stereotype content model of Susan Fiske from Princeton that described the allocation of people to four quadrants that are created by the interaction between the two dimensions of warmth and competence. So warmth is the intentions of others and competence is their ability to act upon these intentions.
And the combination of these-- or the interaction between these dimensions allow you to quickly assign people into one of these four quadrants, and this facilitates stereotypic thinking.
Similarly, Alex Todorov from Princeton also developed the face evaluation model suggesting a similar two axis of power/dominance or value/trustworthiness that we quickly infer from facial features. And our inference of these dimensions from the faces can be as significant as determining the results of political elections.
So overall, we can see the two main factors that influence social relationships can be clustered under these two variables of power and affiliation. And it's a cluster because many different words or definitions could go into that. But overall, these are the two main dimensions.
So because these are so fundamental, it is possible that the brain is representing people really as locations in these two-dimensional space of power and affiliation. And within that space, we actually encode the orientation and the extent of the distances of people from us.
And we can see that also across species, for example, the social defeat model. Of course, that power and-- the power is an important factor across species for survival and affiliation for reproduction, and you can see that in primates and also in the famous Stanford prison experiment. And the brain is equipped with mechanisms to represent space-- or to represent maps and the relationship similarity and distance between dimensions.
This is a famous finding of place cells in the hippocampus. These are neurons that encode a particular location that the animal is placed in, and also, grid cells that are representing also the grid of the environment.
And the last bit of theoretical information that is relevant is the construal level theory that Yaacov Trope and Nira Liberman developed suggesting that we have different types of psychological distance in many different domains, but they all share a common construct of psychological distance that is egocentric in nature.
So we took all these bits of information and developed this idea that maybe our social map or social environment is represented on this two-dimensional map that is framed by power and affiliation. And then the specific metric that the brain is tracking could be the polar coordinates of this environment.
So in a two-dimensional space, we can use polar coordinates, the orientation, and length of a vector that would tell us the location of a particular point, or in this case, a particular person. And we were looking for the representation of these variables in the brain.
So more specifically, the study hypothesis were that the brain is tracking social relationships using these two-dimensional spatial-- or spatial representation. That this geometric representation engages neural systems that are tracking or involving spatial cognition. And the extent to which the brain is performing that is related to psychological well-being.
So how did we do that? We created a roleplaying game we basically invented. We wrote something. So imagine you're a participant, you're going into the scanner, and these are the instructions. You move to a new city, Greenville, and you're going to meet characters that are going to speak to you, and you'll have a chance to interact with them. And you have a task. Your task is to find a job and a place to live by interacting with the townspeople.
So in level 1, your first goal is to get a. Job. So first, we have a narrative that we wrote. It's morning, you're heading out for the first time. It's a nice day. And then suddenly, someone approaches you. This first character is Oliva, and Oliva speaks to you. And she reminds you that you were actually together in high school.
So this is an example of an affiliation trial. Oliva goes for a hug. And now you have two choices that you can, with a button press, indicate. So either you hug her for a long moment or you just give her a brief pat on the back.
So this is how the narrative develops and you meet characters. There's Peter there. And it turns out that they had a bad breakup, and then you're put in this awkward situation that you have to choose between them. So you're pretty much forced to get close to one and not to the other.
You have characters that exert more power. So for instance, you have Mrs. Newcomb and she's very powerful in the city, but she's also good friends with your parents that she knew since you were a little kid. So it's a very intricate narrative that we had fun with.
But it's very well-measured in the sense that you have a specific number of interactions where you can either exert or take power, and specific number of interaction where you can get closer or away. So we had-- I'll give you the parameters, but fixed number of power interactions and affiliation interactions that were pre-tested to make sure that this is what they reflect.
That's just an example of a power interaction where a character essentially begs you to do whatever you want. So he said, I'll do anything for you, and then you can choose. You choose either to take advantage of that or you can start manipulating the person. So that's an example of a power interaction.
And then Mr. Hayworth, he's like the Donald Trump of the city. He kind of owns everything, but people really like him. And then you can decide. So how badly do you need the job? And then you can say, "I really need it, boss," or you're not that desperate. So then you've succeeded.
So overall, the task is kind of fun and people are very much engaged and they're very disappointed that the game is over and they want to know if they won.
[LAUGHTER]
So there's a lot of potential there. So it is kind of very naturalistic, but it was very well-quantified. So we have a narrative section of two to 10 seconds, and then we have the options trials. And within the options, we measure when they did the choice, and then the rest of it becomes blank, which served as our baseline condition.
And we hypothesized that the parameters that we're looking for the location in the social space are within the options trial. When you make the option and pretty much-- it's like a gameboard where you move a character. So you put it in the new location.
So we had 18 subjects and total time of 26 minutes. Six characters. One of them was neutral type of character that you just talk about the weather with or neutral stuff. Six affiliations and six power interactions. And we had counterbalance scripts such that the big boss was sometimes a woman. And the administrative assistant-- we counterbalanced the gender of the characters.
A specific definition was that power was follow or give instructions. These are sentences where you use imperatives. And affiliation is either you engage or engage in personal conversation. There was also a component of physical touch and affection.
We also tested for memory to see that the activation that we're tracking is not related to just the encoding of memory. We had an implicit explicit sociality task, which I will describe, and a series of questionnaires, which-- we had an hypothesis that it will be related to psychological well-being. So we took the main assays that capture the main personality traits and social function. So the social anxiety scale has an avoidance and a fear component in it.
How did we model social relationship? So we devised this geometric modeling where we had the y-axis was the power and the x-axis was affiliation. And then this is why I mentioned the construal level theory that suggested that we represent psychological distance across domains from an egocentric point of view.
So we hypothesized that we represent our social environment from an egocentric perspective. So the vector or the distance would be between ourselves and another person as opposed to viewing them two-dimensionally separated from us.
So during the game, when you make a choice, the care-- if you choose to have Peter drive you home, then you took power from Peter or reduce the power of Peter. So he goes down in power, then you get closer in the next trial, then further closer. Then in the end of the game, Peter ends up in a certain coordinate on the space.
So I just want to emphasize that the participant didn't see that. This was just their own calculation or modeling of the social space. All the characters seen was just the bubbles and the text and their choices. So then Oliva, also, you have the interaction through the narrative, and she has a different trajectory. So she ends up-- it's just slow. So she ends up somewhere else.
And then for each character, we-- in each trial, we measure the length of the vector and the angle, the orientation of the vector. So overall, we have the choices that we record in the task. We calculate, we extract these parameters for each and every trial, and then we create a parametric regressor. So our analysis of brain function is, which region in the brain is tracking this particular value throughout the task in each and every trial? This is a parametric regressor.
What-- we try to identify the brain activation that varies in the same way as if it is tracking these particular values. So these are two separate analysis. One is the length of the vector and the other is the orientation of the vector. These are the two parameters that you require both of them to represent the location of a person in the social environment.
AUDIENCE: So is that-- is each individual trajectory predetermined? [INAUDIBLE] any way on the choices that [INAUDIBLE]? Or is that all just part of the narrative from the get-go?
DANIELA SCHILLER: It's completely dependent on your choices. So it's-- the principle of it is like a Choose Your Own Adventure game.
AUDIENCE: Oh, OK.
DANIELA SCHILLER: The narrative is fixed, but because you make different choices, you end up with pretty much a different narrative and different location. And I actually-- so I can show you a result. So each--
AUDIENCE: Just a question. In your regression model, did you co-vary out angle when you looked at distance? Or did you orthogonalize them with respect to each other?
DANIELA SCHILLER: Yeah. Because they were somewhat correlated, we test them just in different models. They weren't in the same model.
AUDIENCE: But you didn't control for one to examine the uniqueness of activity for the other?
DANIELA SCHILLER: No, they shared too much-- they correlated, so we just did separate analysis or separate models. So this is just a schematic representation of how it would end up just with three characters, but you would construct this in two-dimensional space, you can track the trajectories of the characters in the entire game.
And these are real examples from six participants. So you can almost see a personality trend. So for instance, this person in this narrow space, that each color represents a different character. So they didn't give-- they weren't very hierarchical, but very much been spread. And some were much more hierarchical. Some didn't let anyone close. Some let-- so you can vaguely see how it captures the different personalities or different tendencies. It's very subjective.
This is a similar representation only on three dimensions. So you can see the development of the trajectories over time from three different participants, and each color is a different character. So again, each one is spreading more, this one just got closer to one, this one didn't get closer to anybody, and so forth. So it captures your individual tendencies.
These are the final values of the angle and the length. And this is the average across participants. So this is just to get a sense of how the data looks. And the differences between the characters reflect the narrative. So there was something stereotypic. So you can see how Mr. Hayworth, slash, Trump was very different from Oliva, your high school friend. So across all participants, you did see differences between the characters, but you also see the large variance that reflects the different choices in the game.
We also had this implicit attitude where we asked them to assign houses to the characters. So a lot of people just give Mr. Hayworth the biggest house, although it doesn't make-- it's not necessary and for themselves. They took the small house. But some people just took the big house and put themselves in the center, or put themselves like way up and everybody were downwards.
And then, this is a critical task. After the scan, that we asked them to position the characters, and this is the first time they did see these two-dimensional space. And we just asked them to put the dots that represent the characters. So you can see here, Mr. Hayworth. You can see the different just tendencies.
So what we did, we tested the ecological validity. What we did is took these locations and compare them to the locations that we measured from their choices in the game. And we calculated the distance. So for example, if we had a location from the choices of Mr. and Mrs. Hayworth here, then we calculated the distance and averaged that.
And we found that the subjective and the game-computed locations were significantly closer to each other compared to the game and randomly-assigned dots. So this validated for us that the choices that we captured in the game actually do reflect their subjective perception of their social space. What did we find in the imaging results? So just to remind you, we created these--
AUDIENCE: When you randomly assigned the dots, did you use the variance that-- such as-- so did you just spread them across the space or you just took the variance of the subjects, actually the characteristics of the subject choices?
DANIELA SCHILLER: I'm not sure I understand the--
AUDIENCE: So when you randomly-- so the randomly-assigned dots. It seems that some people are much closer and use just much less of the space than other people.
DANIELA SCHILLER: Right.
AUDIENCE: When you randomly assign the dots, do you take the space that subjects use? Or do you just space them out?
DANIELA SCHILLER: I use all the available space, which was available to them. So if there was no relationship, there shouldn't be a difference, but if we do capture what they are using, then it should be related.
AUDIENCE: I have a question. So was the magnitude of the movement of these characters in space something you're trying to capture? Because-- so one decision is one unit in space, right? But I imagine that there are some things that-- like getting a job is different from just getting a ride on your way home.
And so you just approximated by collapsing all of these into a change of one unit? Or did you capture the magnitude of the action in some way?
DANIELA SCHILLER: Yeah, so I completely agree. In real life, probably different decisions have different weights. But in this case, just because this was a first and a proof of principle, we treated it as discrete steps.
And we assumed that it would still capture it, but we have to design another experiment. I mean, now that we have this evidence, we can design an experiment that will really parametrically vary these types of decisions. Because in this type of task, we don't really have a way to assess what is the weight of each-- of the decision.
It's also possible that with time-- or depending on the location of the person in space, then the steps become smaller and smaller. For instance, there's an asymptotic. So we have to incorporate that into the model.
So first, the parametric regressor that tracked the angle of the vector. So in a whole-brain analysis, corrected for multiple comparisons. We found the hippocampus is an additional three regions, which I will just describe. But the hippocampus was the main, and you see immediately why.
So this activation in the hippocampus tracked the parametric regressor. This is the z-score score by their weights. And you can see that for the other conditions, the narrative part or the non-parametric options part, because the parametric regressor is occurring at the time of the options, but we also had a regressor that is non-parametric, just the fact that there was an option trial.
And you can see that this doesn't correlate with hippocampal activation, or it doesn't predict hippocampal activation it's just a parametric regressor that varies according to the participant choices is capturing the activation in the hippocampus.
Just to see the direction of this activation, we-- after the experiment, we just divided the values into high value and high-angle values and lower-angle values. This was the cosine of the angle. So values closer to 1 or closer to minus 1. And you can see that when the values are closer to 1, you have higher activation in the hippocampus. And this is the direction of the parametric variation.
Probably suggesting that the more power people have relative to affiliation or the more power they have over you, you get higher activation in the hippocampus. So this is the general location in the space.
So what was another interesting thing that we found? Is that the beta weights of the hippocampus were correlated with some social and personality traits. So for example, social avoidance, that was negatively correlated. Also from the NEO personality, the Big Five personality traits, was negatively correlated with neuroticism, but positively with conscientiousness.
AUDIENCE: What's are beta weights?
DANIELA SCHILLER: So beta weights is the extent to which-- to say it in simple terms, the hippocampus is accurately or with relatively high-fidelity tracking the parametric regressor. So how--
AUDIENCE: --coefficient or--
DANIELA SCHILLER: Yes, I suppose you can say like that now. So the extent to which the hippocampal activation matched the values throughout the task, the extent to which this was close, this is how it was correlated with the personality traits.
AUDIENCE: And high is 1?
DANIELA SCHILLER: Yes. So if you-- so what you suggested, that if you accurately track or track with high fidelity, these values you have less social avoidance and less neuroticism, but you're more conscientious. That's the direction. So basically, it goes with the idea that if your hippocampus is tracking social occasions as you interact with other people, you're better off of in terms of psychological and social well-being.
So-- and this was unique to the hippocampus. The other three regions that we found one of them was the inferior parietal lobule. That's for you. And we know that it's involved in spatial-- visual language and arithmetic skills. And also, there was an interesting study that found the same region, encoding distance from one cell in different domains. Spatial temporal and social domains. And in their case, social was familiarity. Yes, do you have a question? Or no?
AUDIENCE: [INAUDIBLE] [? about that. ?]
DANIELA SCHILLER: OK. Another region was the dorsolateral prefrontal cortex, which we know is doing a lot of things, executive function. But also involving spatial working memory, goal-directed behavior.
And then also-- so this is a dorsal medial prefrontal cortex region, but more posterior. So it's not the typical social processing region that you would normally see in impression formation, but something that is more closer to the pre-SMA and dorsal ACC related to motor decision-making and movement planning, attention, and orientation.
So the hippocampus, to the extent which it is representing this parameter of social space, is not doing that in isolation. There are other regions. But the hippocampus seems to be core because that was the only one that was correlated with the actual personality traits.
So now the other parameter of the vector length, this, in a whole brain analysis, corrected for multiple comparisons. The only brain region we found was the posterior cingulate cortex. So I'm looking at you because we just discussed, what is the posterior cingulate cortex is doing?
And we do see this region in many, many tasks. I think it's still a mystery what exactly this process-- this region is doing. But you can see that it wasn't activated or tracking the other conditions, not the narrative and not the nonparametric options trials. We are identifying-- well, as I just showed you, we identified this region in the first impression study.
And also, other studies also found this region involved in updating of impressions of social neuroscience, but it is involved in many other processes, especially value processing, reinforcement learning, autobiographical memory, in word-related thoughts, and so forth.
So in this case, what it captured, it seems, is the psychological distance or the magnitude of the distance. But one important difference is that this is a two-dimensional representation. It actually captures the interactions between the dimension-- in this case, power and affiliation. It's not a one-dimensional representation of distance in this case.
So that was the basic findings that we had, and this followed our initial intuition and the hypothesis that we devised. But of course, we had to do a lot of testing of alternative hypotheses to prove our point.
So one option is, would there be an independent representation of power and affiliation? And other studies did find the independent representations, but not in these regions. So what we did is just took independent regressors of power and affiliation, which is the values on the y-axis or the values on the x-axis. Or, if you will, the Cartesian coordinates as opposed to polar coordinates.
So for affiliation, we found nothing. For the power, we found the left middle temporal gyrus. But you can see that it was much better correlated with the other the narrative in the options condition. So it suggests that it's probably performing a more general probably language-related processing in this particular task.
We also looked for a conjunction between power and affiliation. That is, to look for a region that is representing both. So we just put them in the same design matrix. But we found nothing when we did that. We also did a simple multiplication of the two regressors, which is also a form of interaction between them. That is, instead of taking the polar coordinates, we just took the two parametric regressor and multiply them. We found nothing there, too.
We also shifted the parametric regressor. As I told you, we looked for the coordinate representation during the options, but it could be that you actually have the representation already in the narrative before you even make the choice. So we shifted the regressor. We found nothing there.
And we also tried to do a non-egocentric model. So as I told you, we assume that this is a egocentric, the vector is always calculated from your perspective. But what we thought is maybe it's actually not related to you. Maybe it's a different angle, as if you're watching a movie that is distant from you and it's not relative to you. It's relative to, let's say, the origin of the character. So if it started somewhere here in space and moved, then this is the angle you represent, just as an alternative geometric model.
Here, we actually found a number of regions. Some of them overlapping with the original analysis. But then you can see again that they actually have better activation or stronger relationship with the non-parametric option condition again, suggesting that they have a different type of more general processing in this task.
So we also did a series of validation of the results. We did the ecological validity, which I told you about, the comparison between the subjective and game locations. We did scrambling analysis. We basically took choices from one participant and assigned them to a different participant and mixed all the choices and then got nothing there, too.
We tested the reaction time to see if it's correlated with any of the regressor and there was no correlation. We tested the memory test and we saw that all characters were remembered above chance. So there wasn't a significant differences in memory demands across characters that could explain the pattern of results.
We also just looked separately at the narrative and options trials, more like a sanity check to see that we get the activation that we would expect. So for example, in the options trials, when it's non-parametric, it's just the occurrence of an option, we found striatal-caudal activation, which is the main region involved in decision-making and choice.
And you can see how it's strongly active during the options, but regardless of the parametric variation of it, so it doesn't reflect these particular values that we showed we found in the hippocampus. So you can see how it's strongly active during options as opposed to narrative.
So in conclusion, what we see is the hippocampus seems to contribute to social cognition in ways that we didn't expect before by tracking the outcome of social interactions through a social space, as if you're navigating through an abstract space.
So one suggestion is that the hippocampus is maybe contributing to constructing a cognitive map. And we know that currently there are two main camps. There's the spatial camp, the camp that is in the hippocampus, is doing spatial navigation, and it's dedicated it's like the GPS of the brain, and there was a Nobel Prize for that, too.
But also the other camp is suggesting that the hippocampus is involved in declarative memory and episodic memory. So what is the relationship between the two?
One suggestion is that-- our Howard Eichenbaum has been an important player in promoting this approach, is that the hippocampus is performing relational learning, and he's suggesting that we might benefit from viewing the hippocampus through the lens of Tolman's original conceptualization from 1948 of the cognitive map as a system for organizing complex information in the environment across dimensions in different domains.
And this suggests that we represent space not only in the physical sense, but also in abstract sense. So we can organize information in music processing, for instance, in language processing, and in this case, in social processing.
And then it gives us an idea of how maybe companies can contribute to abnormal social behavior. It's interesting because in borderline personality, for example, which is a dysfunction that is characterized by very severe social deficits, you do see hippocampal dysfunction, but it wasn't very much related to their actual social skills.
And also, you see it in depression. There are severe impairments in hippocampus. But the direct link between hippocampus and social dysfunction hasn't been made, in at least in psychiatric research so far, so this is where we're heading now.
So that's pretty much it. The conclusions are that maybe our metaphors that we're using climbing up the social ladder or having a tight social circle or using expressions like "he's above me," "she's beneath me," and so forth are not just metaphors, they really do reflect the way we represent space and we navigate through space.
And probably the social brain is not necessarily a recent design evolution. We actually do-- so if we want to-- I guess I can give you my take on this domain-specific, domain-general debate. That we might have regions, especially in the prefrontal cortex, that are dedicated to processing of social information.
But this is a-- it's as if-- it's like the visual cortex of the social domain. That this is a region that is dedicated for gathering and organizing the social information, but the actual computations are relying on other structures like the stratum that we now begin to see in hippocampus and amygdala and so forth.
And yeah, that's it. So we hope we can suggest a novel role for hippocampus in representation of abstract social space. This is the team. Rita Tavares is a postdoc, and she's very special because was my first postdoc, but also, she came from immunology, so she didn't do neuroscience before.
So she joined me in the first lab and I took her without experience. It was sort of-- and we did this high-risk study and it turned out pretty nicely. And Matthew Shapiro, he studies spatial navigation. He does recording in rats' hippocampus. Yaacov Trope is the one who developed the construal level theory. And a team of really great lab members helped us in the project. And Taylor Williams created the original illustrations for the study. It seems ugly at first, but they grow on you.
[LAUGHTER]
Thank you.
[APPLAUSE]
AUDIENCE: Really interesting. Now if I'm following this-- and feel free to correct me, what I think that series of hypotheses showed was that this was a configurable effect of some kind. That it wasn't additive, if I'm getting this right. So that's very intriguing.
And there's been ideas, as you know, about the hippocampus being about context as opposed to just the stimulus. So I wonder if you could expand on that. And I think this leads to a very interesting, very speculative prediction about autism and so on being non-configurable. But anyway. You don't have to deal with that last part, but just-- what do you think about the notion of interpreting this as an emergent property of these dimensions?
DANIELA SCHILLER: Well, it relates to the definition of context. I completely agree that it's a configurable representation. It's a representation of relationship between different items in context. And in this case, it's the social context. So I think it performs the same type of calculation that you see in representation of context in other domains like spatial domain.
I think it's related to autism. We-- actually, I'm going to give a talk at the Simons Foundation soon, so we'll see what they have to say. But yeah, I don't know. I--
AUDIENCE: [INAUDIBLE] autism, for example? Sort of non-configurable?
DANIELA SCHILLER: In what way?
AUDIENCE: So in other words, it sounded, when you ruled out those hypotheses, you ruled out that this was a non-configurable effect, that there weren't independent effects. But you might argue that autism, if it's more literal and so on, it's not picking up on these configurable cues, which is why the social impairments may occur and theory of mind may occur as a result of those information processing, inability to process configurable cues, and it would be more additive, possibly.
DANIELA SCHILLER: So what type of brain activation would you expect to see?
AUDIENCE: You might-- again, totally speculative, but you might see activation that was independent rather than configurable. So all of those hypotheses you ruled out would be ruled in and the configure would be absent.
DANIELA SCHILLER: Yeah.
AUDIENCE: So it's sort of an acontextual.
DANIELA SCHILLER: Yeah. Yeah, I think it's very plausible. But I hope to study that-- we'll have access to that population soon and then we can--
AUDIENCE: What you said about personality axis, what kind of sampling did you do? How many total people worked through this as subjects?
DANIELA SCHILLER: 18.
AUDIENCE: 18? OK. So I don't know how you and your team modeled personality types. I mean, you can't sample 16 different Myers-Briggs loci. But if you had 18 replicates for four personality types, I'd be really curious as to how the mapping would go, if you could independently track that as a variable.
DANIELA SCHILLER: Yeah, I think it's an interesting idea. I mean, in this case, we just measure their scores on the personality questionnaires and we corrected for the multiple comparisons of the subscales. That's what we found. I think-- I mean, I agree, we can take this in many interesting directions. So you can compare between populations.
You can-- actually, the narrative itself is a pretty powerful tool. That you can create different narratives, and then in this case, try to influence-- or speak to different types of personalities or something of that sort.
AUDIENCE: So as someone who worked on animal experiments, too, as well as human experiments, you're in a position to maybe comment on whether you can-- there's an animal model that might allow you, then, to get a more crisp picture of the actual circuit, seeing all the lower level stuff. Is that have you thought about that as a--
DANIELA SCHILLER: Yes. That's actually a nice case of transitioning from the humans to animals if possible, and we have been discussing that. So I mentioned the social defeat protocol. This is where our rat is encountering an aggressive conspecific, and then and then you test their willingness to approach that.
So you can work with that type of paradigm, but also controlling for the identity of the animal and just like manipulating their social rank, that could be and then record from the hippocampus at that time and see the separate representations. That could be one case.
I also know that there was a study in published in Nature maybe a couple of years ago, that they recorded from V2. And I think it's pretty hard for it's a tiny part of the hippocampus. And they found that activation in that region related to memory of the social rank of the conspecific of the other.
So I think it could be-- it's actually very relevant to this paradigm because the way they did it is the animal was placed with the other animal in the same arena, and then they found the activation that is representing the social rank of that animal. But if the lesion, they can remember that. But it could be that it's not really memory, it's the actual representation when they meet.
AUDIENCE: So you talked about NEO covariates. I'm wondering if there were any interests in moral cognition. So for example, if there is an individual who places higher emphasis on authority in moral cognition, if then the computations of power distance might have a differential effect versus somebody who has a lower moral cognitive weight being placed on authority.
DANIELA SCHILLER: Yeah. So we didn't test that, but I think it's a really interesting direction to try to start on vary these parameters and put different weights based on ingroup, outgroup, similarity, prior experience, expectations, and so forth. I think it has a lot of potential.
But also just by observing the differences between the participants, you can already see that some people were more sensitive to authority, and some people were very sensitive to social distance and so forth.
AUDIENCE: So do you think your results are really specific to hippocampal support of that initial, whether it's a first impression or a relationship formation in those 26 minutes, how would you see this maybe extending to the relationship maintenance? Especially for something like affiliation where you would hope that that maybe doesn't develop right away, that it develops over time. Would you see-- would you expect the same neural correlates?
DANIELA SCHILLER: That's interesting. Really, the original intuition for that whole study came from thinking about resilience and how people cope with trauma and adverse situations. So it seems to be something that if you do it and you do it well, then you have a much better maintenance of your social life and your everyday life.
We did some subsequent analysis. It's all very preliminary, but for instance, we looked at the distance of the social space from the participant so some people tended to have people closer to them, whereas other people tended to have the whole social space further from them.
And that was correlated with the better weights of the posterior cingulate cortex activation. So that's like several levels, but to put it in simple terms or possible interpretation, is that if people are over time get closer and closer to you, you actually stop tracking this or stop representing-- or you don't represent the social space. Maybe it shifts to something else, a different type of more emotional type of representation.
But if you have a large distance, if people are disparate and far from you, then you engage in these calculations. So I don't know if this speaks to the maintenance idea, but--
AUDIENCE: --really interesting, especially with respect to research on the overlap. But when you get closer, to overlap with that representation to your own--
DANIELA SCHILLER: Yeah. Over time, that definitely-- that could definitely be a difference. I think it's related to the question here, that the weights gets different, maybe the increments are much smaller. Maybe if someone is further from you-- or closer to you and then gaining power, then it has more weight and so forth, so it can get more sophisticated.
AUDIENCE: Along those lines-- so people also naturally differ. It's really neat and nicely controlled, but do people naturally differ in the-- in how large networks get formed and there is, overall, an evolutionary advantage to be in a point where you have a large network with a lot of connections.
Do you think that there might be individual differences in ability to represent social distance that could translate, then, to ability to stand and sort of densely connected?
DANIELA SCHILLER: Yeah. I think this is some of the implications of that study, that there's definitely large individual variance. Some additional analysis we just continue playing with the data and see if we get some more stuff there.
But so for example, we saw that the size of the space early on predicts your scores in social anxiety. So if you had high social-- and social efficacy also. So people with high anxiety and lower social efficacy tended to have lower space. Smaller space. And also, they tended to give less power to people early on.
And so only these parameters early on were correlated with social anxiety and social efficacy. So it seems that there is great variance in how we manage the social space, and it is related to how we function later on.
AUDIENCE: So do you think the brain representations of this distance and angle could be moved around depending on interactions between characters?
So, for example like maybe part of the story is that you have a sort of alliance with some people, but then independent of what you are doing with the character, two characters break up and you find out that the other is now out of your coalition, so you know that person's status in relation to you has changed. Do you think something like that would have a similar effect on movement in the social space or does that fit with--
DANIELA SCHILLER: Yeah, so that's an interesting question. I think we try to get at that by looking at this other angle, which is not the non-egocentric model. The way-- which is just tracking the trajectory of the characters from this random original point. And there, we didn't find correlates.
You can also compare it in terms of the hippocampal research language, which is allocentric versus egocentric representation. And it seems that in this case, it's not doing the allocentric representation, which is just the representation of relationships in the environment.
And also, other studies didn't find when they took this approach to see if the hippocampus is representing social relationships. When it wasn't related to our own movement and our own interaction, they didn't find correlates. But it's very little data, and we didn't vary parametrically. And maybe if you're part of the story-- so this observation would still be relevant to you and then we will see this type of tracking.
AUDIENCE: But for now, it seems like it's something about the interaction itself?
DANIELA SCHILLER: Yes.
AUDIENCE: OK.
DANIELA SCHILLER: Yeah. I'm very glad you mentioned that because most studies are passive, and they're just our observations and evaluations. Whereas here, the key is that you actually do navigate. It's really about the choices that you make, and we don't find that. So we don't find activation to the narrative itself or just to the fact that there are options it's really about the trajectory.
AUDIENCE: You mentioned posterior cingulate activations in a lot of things. I wonder if you could comment on how confident you are that it's really posterior cingulate and not retrosplenial because retrosplenial cortex has a lot of shared functionality with the hippocampus and involvement in spatial orientations and things like that. So I'm just wondering if you have any thoughts on-- how much of it is posterior cingulate, how much of it is retrosplenial, or--
DANIELA SCHILLER: Yeah. I couldn't say for sure. It sounds very plausible and I'm actually very glad you mentioned that because it's a very interesting possibility and actually highly likely. We are thinking of now initiating a study of 7 Tesla , on the 7 Tesla scanner. So we might be able to capture the more fine-detailed anatomy also within the hippocampus, because here, the activation was more interior, which is for also larger scales. So we might be able to capture also this differentiation within the PCC.
SPEAKER 2: OK, great. Well, thanks for-- Thank, Daniela, for your wonderful talk.
[APPLAUSE]
SPEAKER 1: This has been a production of Cornell University. On the web at cornell.edu.
Daniela Schiller, assistant professor of psychiatry and neuroscience at Friedman Brain Institute, Mount Sinai Hospital, presents current research on the role of hippocampal brain structures in the processing of information about social interactions. Recorded Sept. 14, 2015 as part of Cornell's Human Development Outreach and Extension Program.
