"Overview of Recent Conceptual Advances (and Puzzles) in Cognitive Neuroscience"
Neuphi
March 1, 2007, 5-7pm
Note-taker: J.T. , E.I.
Last Updated: March 7, 2007
When it comes to two different disciplines working together, the question of how these disciplines will interact becomes important. The way I see philosophy and neuroscience coming together here is in cooperation rather than the old model on which scientist produce facts, and philosophers come onto the science after the fact. One model—mind can assist analysis of such models at the computational level as well as the conceptual level. For example, David Marr’s study of the visual system distinguished different levels/aspects of the visual system. Computational analogy, like studying mathematical algorithms, coming close to software and the counterparts, study of missing organs. This is a good model because philosophers and philosophical logicians can make interesting suggestions at the computational level, and perhaps contribute to science.
These puzzles may not be philosophical, but are significant. Neuroscience studies the mind through the function of healthy, human brains. Very successful and popular, but in its popularity certain conceptual puzzles arise with the use of fMRI technology to understand the mind.
1st puzzle: What constitutes the same brain region from individual to individual.
1) Why do we care about isolating brain regions?
WhyWeCare 1): Allows us to “triangulate” mind brain associations. To make it possible to speak about one study in relation to other studies. Single studies can never pin down function of brain. Particularly important when we try to draw an inference from two independent studies.
Example from my field of research
Two separate studies seemed to suggest that: The “Right Posterior STS” responds robustly to biological motion (relative to other kinds of motion/nonhuman objects) AND to belief attribution. The inference was: There must be something common between the belief attribution and biological motion.
However, upon closer investigation, they were actually different brain regions.
WhyWeCare 2) Allows us to make reverse inferences from neuroimaging results
Example context: When unexpected recruitment in the brain from imaging, we want to come up with hypothesis to explain the other
Example context: Get a brain region we know well enough so that we can use to ID how people think about that task.
e.g. Jessica’s story (Jessica broke her neighbor’s plates), manipulate subjects’ own experience with Jessica (she’s either nice or unfair). Subjects will make different inferences AND this will manifest differently in brain function.
Cannot make reverse attributions unless region can be isolated.
2) Why is it difficult to ID brain regions?
--How to define same-region-ness? There are three prominent different methods of detecting sameness-in-regions of the brain.
(A) Normalize to 3D coordinates by transforming the subject’s brain to the standard one. [most common way]
(B) Consider the curvature of the actual subjects’ brain and locate based on prominent sulci/gyri.
(C) Characterize brain regions via cytoarchitecture (cell type, brain layer) etc. (But they have to be dead).
The problem is that the data from these three different methods do not agree.
3) Possible approach to address the brain region issue?
The method I advocate is the “Functional Regions of Interest” method, involving anatomy and function. Method comes from visual neuroscience.
The line between V1 and V2 is increasingly and decreasingly peripheral vision.
But of course, this is not necessarily applicable to all brain regions.
Fortunately, belief attributions work like this as well.
Experimental Procedure: “Sally-Anne test” adapted for grownups:
Task Type 1) False belief task: She parked a sports car in the driveway, but at night he substituted a minivan. Where does she think her sports car is in the morning?
Task Type 2) False photo task: Tell story of volcano erupting and island being destroyed. Show photo of intact island. What did the island look like in the photograph? Must base response on a representation and not on reality.
Result: fMRI shows that Right Posterior STS has preference for beliefs over photographs
i.e. There’s a region that is recruited in false belief stories while not in photo task.
Assumptions inherent in this inference:
1) Stable relationship btw anatomical structure and function across domains. One can meaningfully generalize from one experimental context to another.
2) Structure to function links must be general in individual, and across experimental conditions.
3) Approach is good for telling average response for a whole region to stimulus, but bad at finding finer grain spacial patterns. Brain must have a much finer resolution than what we see in fMRI.
2nd puzzle [35:00]: Which cognitive function gets ascribes “special” regions? Once they got the “Special” brain regions, how do we know what they “do”?
Foder’s Hypothesis: Special domain specific mechanisms might exist for input systems only, i.e. only for perceptually defined domains. Central mechanisms (“thought”) will only be domain general. In fact, early fMRI (8-10yrs) results gave supportive evidence, that brain domains delimited by stimuli. Only faces, places, shapes, human bodies, motion had special brain regions in vision.
However, counter-evidence seems to be posed by belief-attribution study. Data indicates that contra-Foder, perhaps there IS a brain region that is specifically recruited when people think about other people’s thinking.
False Belief but not False Photographs is consistent with three hypothesis.
(1) Region reacts to anything about a person
(2) Involved in attributing any internal, subjective and visible state
(3) Involved in attributing representational mental states (e.g. thoughts/desires)
Test hypotheses by three new story types:
A. About a person but not beliefs (“He was ganglion-looking, wore baggy jeans”)
B. Subjective, internal, visible states (hungry, achy, thirsty), but not representational mental state (“Skipped breakfast. Her stomach was rumbling.”)
C. About a belief but not about false beliefs (“He knew flight was his sister’s. It was the only one that was delayed.”)
Data & Inference [41:00]:
RTPJ (Right temporo-parietal junction) responds only to story type C → Seems to support hypothesis #3, that is, RTPJ is involved specifically in attributing representational mental states, rather than reacting to anything about a person or about any kind of state of a person.
Additionally…
-Modality of presentation (audio or video), and type of story does not make a difference.
-Does not make a difference whether mental beliefs are explicitly stated (e.g Story C—esplicitly stated “He knew….”) or implied (e.g. Jessica story—implied but never stated what Jessica was thinking). As long as subjects can be provoked to infer or attribute beliefs to the characters, the RTPJ responds.
-RTPJ also responds to identical, non-verbal stimuli.
Evidence is from “Girl & Chocolate experiment”.
Prompt 1 [instructs to follow rules]: “If girl is facing box, respond where it went last. If girl is looking at box, respond where chocolate went first.”
Prompt 2 [instructs to attribute belief]: “Where does the girl think the chocolate is?”
Only prompt 2 elicits RTPJ response.
3rd Puzzle: Is mind specialization innate? How do specializations develop?
Foder: Modular mechanisms are innate.
OR
Neuroanatomical development as a process of Canalization, increasing specialization with development.
Any evidence where non-innate specialization? Ex. Words. (It appears that specialization regions develop to handle words, clearly not innate.) Left fusiform gyrus recruited in looking for words in own language/language you read than in any other language.
But what about RTPJ to belief attribution? Innate or derived/developed?
Experimental Task: Children listened to short stories with three subsections. (A. Pond is… [physical scene information] B. Mr. McFeeglebee is old. Georgie is Mr. McFeeglebee’s nephew [social but non-mental-state information]. C. Georgie likes…, knows…, decides… [mental state information])
Response of RTPJ in children (6-11yrs old) to different subsections of the story was monitored and compare to adults.
Result: As kids get older, the less the RTPJ responds to B. For 6 year olds, the RTPJ responds non-selectively to stories B and C. Specificity increases as the subjects become older. 12 year olds’ and adults’ RTPJ have no response to B.
[54:00]
Q: You are interested in the “region”. Naively, we understand the region as spacialy defined, but in addition it may (you say) be architecturally or functionally defined—this seems to be a problem. How do we know that there is a 1-on-1 correspondence between region and function? Maybe the same region will appear in different functionalities.
A: We believe that the brain functions are correlated with specific behavior of neurons, and it happens empirically that these cells group together systemically into a region. I don’t think a region is an ontological unit, certainly not a natural kind. In fact, we can’t even make a claim as to the causal efficiency of regions.
Q: Was the order of the chocolate box test fixed?
A: No, random in order, but the subjects saw all combos.
Q: How different can brains be? How large a transformation is necessary to normalize them?
A: Very different. That makes such procedures highly inaccurate.
Q: Are there any data about how the specialization of belief attribution brain region develops in autistic people?
A: Not yet, tests are starting.
Q: Are regions involved in justification of beliefs?
A: We don’t yet know whether the regions involved cares about, e.g. the justification of a belief.
Q: Clinical studies of patients with damage in that region?
A: The best studies of such patients agree with these results, though they tend also to be impaired in other ways (esp representational cognition, maps and signs)
Q: The “regions of the brain” story is fairly plausible, but it is surprising. Have you given any thought to why these regions develop?
A: I have no good answer, but I think it has a similar reason as the fact that we have a special region of the brain for face recognition—instead of simply relying upon general shape recognition function. There would seem to be a mechanical or computational advantage.
Q: Do you make use of the pragmatics of speech vs logical inference?
A: I am very interested in this. Because of the high cost of belief attribution, it would seem impossible for us to attribute belief every time we use the word “the”. Wer are currently exploring the differences that show up with propositions like “It is raining” and “She believes it is raining”.
Q: Why is being hungry not a representational state?
A: This analysis comes out of a theory meant to explain the development of children’s conceptualization of minds. The trouble is to explain why children are fluent with certain concepts of mind but ignorant of other aspects (e.g. the difference btw ignorance and being wrong). It turns out that (attribution of) representation is one of these aspects. This indicates that hunger is not conceptualized as representational.
Download mp3: http://www.neuphi.com/images/mp3s/070301 Neuphi Saxe 1.mp3