COG SCI C127 Readings & Lecture Questions
Explain what you know about the "what" and "where" pathways. (Wandell reading 09/18)
"what" - ventral pathway specialized in object identification. "where" - dorsal pathway specialized in spatial transformations.
Why is fMRI a particularly powerful method for creating functional maps? (Lec. 5: Measurement, MRI, and fMRI)
- A neuron can fire very fast (several times per second) and the hemodynamic process is much slower - We are measuring these things periphery, not directly - We don't measure blood levels in capillaries right next to neurons because they are too small. We measure blood level in larger capillaries that are the size of voxels - The more protons spinning (the larger the magnet), the smaller the resolution (which is better) - BOLD signals reflect tuning in astrocytes - Astrocytes are vacuum cleaners that are around neurons (they pick up NT's and clean them up and give them back to the neurons so when a neuron fires, and astrocyte "fires" as well) and that's cellular metabolic activity (even though astrocytes don't actually compute anything, they are tuned in the same way as the neurons round them) - These astrocytes do not soak up all of the NT in the synaptic cleft but don't have same binding efficiency (less sensitive to NT as to allow some NT to travel to post synaptic cell) so astrocytes don't jolt into action until NT is highly concentrated (so neuron has to fire several times) -So, the astrocytes are actually MORE tuned to certain things that the neurons themselves because they need higher concentration of something to fire
Explain the basic concept of attention. Offer and explain two reasons that we might have evolved mechanisms for attention. (Lec. 7: Attention)
- Attention happens at all levels, because they all have some sort of top-down influence optimized to help the system perform a task - Attention does not have to operate the same way everywhere - Classical View of Attention: Broadbent's Early vs Late Selection Theory --Somethings that you can attend to are really simple and some really complicated --Some things not processed and some things really highly processed --Attend to language: pitch (low) or word (high) - Triesman's Feature Integration Theory --You have feature maps and each map/visual areas represented one visual feature (but we know that's not true now) - Itti and Koch's Salience Map Model --You need to find interesting stuff in each image (what's different from everything else) -- That's what will be most informative and important -- You need to combine/compare a bunch of features and come up with aggregate statistic for interestingness ("salience") -- You tend to (move your eyes to) the salient things -- This is bottom-up attention (what in the world attracts your attention) -Bottom up and top-down attention are constantly competing for you to process THEIR information
What does it mean to think of a neuron as a filter? (Lec. 3: Filters, Circuits and Maps)
- Because neurons cannot pass all signals they are receiving, they are filters. - What differs between neurons is what they are filtering. What determines the difference? How they are connected to other neurons. -- Ex. Transducers in retina are filters -- Retinal receptors and ganglion cells are specialized filters -Many neuronal filters taken together form a filter bank. Pooling across the bank reconstructs some of the signal. -Multiple dimensions represented by neurons -Spatial position vs orientation (neurons are tuned for both) -Filtering is inevitable if channel has smaller bandwidth than the sensory organ and some information is lost with each filtering - To fix that, we aggregate across a variety of neurons to get the full picture
What is the point-null hypothesis testing framework? (Lec. 4: Doing Brain Science)
- Beginning from some prior model, theory or hunch (MTH), identify a specific prediction of the MTH that can be tested empirically using the PNHT framework. - Design an experiment in which a simple stimulus or task parameter is manipulated across conditions, and which, if the MTH prediction is true, should produce different brain responses across those conditions. - Collect brain data from multiple subjects. - Perform statistical tests to determine whether the brain responses differ across the tested conditions more than would be expected if the hypothesis was false and any observed differences were due to random fluctuations (i.e., noise). - If differences in brain responses across conditions are greater than expected by chance, then reject the null hypothesis. If not, fail to reject the null hypothesis.
Thought q2: How would we determine whether an estimated functional map is the **best** map? (Lec. 3: Filters, Circuits and Maps)
- Best resolution - Most replicability from experiments -Makes the most sense with theories about how it may work
How does separating data into fit and test sets make results more likely to be replicable? (Lec. 4: Doing Brain Science)
- Ensures the generalizability of the experiment, increasing external validity - Avoids fitting a model too closely to the peculiarities of a single dataset -If you only have a fit (or training) data set and no test set, you might achieve a good model fit, but it might be predicting the noise in your sample (as opposed to the underlying general pattern you'd be hoping to predict). - You need a test set to know if your model holds up in predicting data that wasn't used to generate the model in the first place. And then, ideally, you'd have another out-of-sample test set to see how generalizable your model is. - so, basically to prevent over-fitting of your model, and also to be able to make inferences about its generalizability.
How is the visual system organized at a broad scale? (Lec. 6: Retinotopy, other topies and vision)
- Entirely hierarchical, continuously gets more complex as you get higher in the system - Organized as a deep network -Distinct anatomical and functional areas arranged in a hierarchical, parallel network with substantial recurrent feedback - Nonlinear transformations between areas -Each area represents a different feature space projected onto the cortical surface Each area differentially affected by bottom-up and top-down information Bottom-up information: info comes from the outside and goes into the brain Top-down information: info comes from the brain and affects other parts of the brain
Thought q1: How would we determine whether an area has a map ? (Lec. 3: Filters, Circuits and Maps)
- Need to see distinct anatomical areas (different things or neurons) - Then need a way to measure some sort of change and make a theory to test - May need high enough resolution to view things to measure/make theories on - Theory à experiment à look at data from control vs. manipulated condition - Look at activity in the brain paired with stimuli given to participant - Or through date driven method: use regression to test multiple hypotheses. Get all the data then mine through the data to see what sticks -All neurons need to be tuned to same things: systematic change in dimensions represented in area that go along some spatial are in the cerebral cortex, module = group f neurons that all code in the same dimension
How can the brain represent more than two dimensions in a functional map ? (Lec. 3: Filters, Circuits and Maps)
- Neurons have the capability of responding into multiple dimensions - Some neurons are receptive in all stimuli while others are only selective in certain stimuli
What is statistical significance and what does it tell us? (Lec. 4: Doing Brain Science)
- Statistical significance is necessary but not sufficient for creating strong and useful field of science - Statistical significance was originally intended as a metric to be used preliminary to full data analysis, but over the past decades it has become a target to shoot for. -Statistical significance merely indicates whether the observed differences are likely to be larger than we would expect if the treatments had no effect at all. It says NOTHING about the size of the treatment effect. -Overemphasis on statistical significance has caused researchers to optimize their experiments to achieve significance at the cost of other more valuable metrics (e.g., effect size or prediction accuracy). This has lead to all manner of intentional and unintentional "p-hacking". -In most cases, obtaining a significant result provides no information that can be easily used to optimize strategy for the subsequent experiment or to estimate model class or model parameters. -PNHT provides no framework for testing predictions or generalization.
How does the data-driven voxelwise modeling approach work? (Lec. 6: Retinotopy, other topies and vision)
- Treats the whole problem like giant regression problem -Fit data set to estimate model -Use all random, natural stimuli (instead of faces vs. objects) -Get data and see what stimuli correlates with what brain activity to se what part of the brain represents what parts of these images -Calculate stimulus features (extract stuff in images) such as color, orientation, semantic category, etc. and see what lights up when certain features come up -Data-driven model: 1. Collect a large sample of brain data from several individuals under naturalistic conditions sampled from the distribution of conditions relevant to the domain of inquiry. 2. Divide the data up into separate fit and test data sets. 3. Using the fit data set, leverage the general linear model (GLM, often in the form of ridge regression) to map stimulus and/or task features onto measured brain activity. 4. Using the test data, estimate statistical significance and prediction accuracy for each feature and for each individual subject. 5. Use clustering methods to interpret the fraction of data that is well predicted by the fit models. 6. Use separate statistical models to aggregate data across the group, and perform significance testing, prediction accuracy assessment and interpretation separately at the group level.-
How are neurons organized within the brain ? (Lec. 2: Brain and Function)
- White matter = no synapses (just long axons), grey matter = synapses -Neurons in cortex organized into 6 anatomically distinct layers --Layers 1-3 are called "superficial" or "supragranular" levels --Layers 5-6 are called "deep" or "infragranular" levels -Input to the cortex enters at level 4, cortico-cortical connections are through layer 1, cortico-thalamus connections are through layer 6 -Local circuits are organized into separate functional areas or subdivisions --Projections from superficial layers go to cortex --Projections from deep layers go to the thalamus
Why does resolution affect our ability to identify functional maps ? (Lec. 3: Filters, Circuits and Maps)
- increases clarity and makes notable areas more distinct - increases the amount of information obtained from the map (info can also be increased by increasing data channels)
Give 3 examples of common scientific practices that will tend to produce published results that cannot be replicated. (Lec. 4: Doing Brain Science)
-Adoption of a low alpha level (5%) puts reasonable bounds on the rate at which errors can enter the published literature --This imposes no limit on the rate which errors may arise in the literature -Conceptual replication (more test validity for the phenomenon being tested) --Could Interact insidiously with publication bias, opening the door to literature that appear to confirm the reality of phenomena that in fact do not exist. -Errors will be pruned out if we just stay patient --No plausible concrete scenarios to back up this forecast -Flexible designs and flexible definitions for constructs --Not holding all other variables constant and not maintaining strict research protocol
How does fMRI work? What are some advantages and disadvantages? (Lec. 5: Measurement, MRI, and fMRI)
-Based on the idea that oxygen-rich blood and oxygen-poor blood have different magnetic resonance -Radio waves aimed at protons of hydrogen atoms within areas being studied -As magnetic field hits protons, they line up. Radio frequency knocks the protons out of alignment -After radio burst ends, protons align again, releasing signals that the MRI can pick up -Protons in areas of oxygenated blood produce the most signals -Mapped in squares called voxels — each represents a thousand nerve cells - Deoxy-haemoglobin is paramagnetic and has a low MR signal, oxy-haemoglobin is diamagnetic and has a normal MR signal -Advantages --Good spatial resolution --Inherently 3d --In continuous development -Disadvantages --Very expensive and complicated --Poor temporal resolution --Does not measure neural activity
What is the data driven framework? (Lec. 4: Doing Brain Science)
-Collect a large sample of brain data from several individuals under naturalistic conditions sampled from the distribution of conditions relevant to the domain of inquiry. -Divide the data up into separate fit and test data sets. Using the fit data set, leverage the general linear model (GLM, often in the form of ridge regression) to map stimulus and/or task features onto measured brain activity. -Using the test data, estimate statistical significance and prediction accuracy for each feature and for each individual subject. Use clustering methods to interpret the fraction of data that is well predicted by the fit models. - Use separate statistical models to aggregate data across the group, and perform significance testing, prediction accuracy assessment and interpretation separately at the group level.
What methods are most commonly used in cognitive neuroscience? (Lec. 5: Measurement, MRI, and fMRI)
-EEG -EcoG -PET -fMRI -MEG -TMS
Give three examples of functional areas identified using functional localizers. What was the contrast used in each of these cases? (Lec. 6: Retinotopy, other topies and vision)
-Fusiform face area (FFA) --Contrast faces — objects -Extrastriate body area (EBA) --Contrast bodies — objects -Parahippocampal place area (PPA) --Contrast scenes — objects
Why do functional localizers reveal only a small fraction of all of the distinct visual areas that likely comprise the human visual system? (Lec. 6: Retinotopy, other topies and vision)
-In the individual-subjects functional localization approach, a region or a set of regions is defined in each subject using a contrast targeting the cognitive process of interest. For example, to identify face-selective brain regions, a contrast between faces and objects is commonly used. Once a localizer task has been developed and validated, in each subsequent study every participant is scanned on the localizer task and on the "task of interest", i.e., a task designed to evaluate a particular hypothesis about the functional profile of the region(s) in question. -There are two challenges that face researchers who want to adopt this approach for studying high-level cognitive processes, such as language: --First, it is non-trivial to decide on a contrast that would target all and only regions supporting the cognitive process of interest. --Second, many high-level cognitive tasks elicit robust and distributed activations, which sometimes makes it difficult to decide (a) what counts as a "region", and (b) how parts of activations correspond across subjects. Here are the solutions we came up with.
What is hemodynamic coupling? (Lec. 5: Measurement, MRI, and fMRI)
-Relationship between underlying neural response and subsequent change in blood flow (as function of neural activity) --Glucose consumption leads to increased blood flow --Synaptic activity is related to glucose metabolism -Measuring neural activity by looking at blood flow and oxygenation -Unclear whether it is actually tracking if oxygen consumptions increases to aid in neurons firing or to provide glucose to astrocytes Give examples of linear vs non-linear relationships between neural activity & blood flow in the brain show evidence for or against hemodynamic coupling: -NEEDS TO BE FILLED OUT
What are the disadvantages of PNHT? (Lec. 4: Doing Brain Science)
-Statistical significance is necessary but not sufficient for creating strong and useful field of science -Statistical significance was originally intended as a metric to be used preliminary to full data analysis, but over the past decades it has become a target to shoot for. - Statistical significance merely indicates whether the observed differences are likely to be larger than we would expect if the treatments had no effect at all. It says NOTHING about the size of the treatment effect. - Overemphasis on statistical significance has caused researchers to optimize their experiments to achieve significance at the cost of other more valuable metrics (e.g., effect size or prediction accuracy). This has lead to all manner of intentional and unintentional "p-hacking". -In most cases, obtaining a significant result provides no information that can be easily used to optimize strategy for the subsequent experiment or to estimate model class or model parameters. -PNHT provides no framework for testing predictions or generalization.
Name three properties that change systematically across the visual hierarchy and describe how they change. (Lec. 6: Retinotopy, other topies and vision)
-Visual areas represent increasingly complicated image features --Anatomical micro-circuitry changes --Diameter of spatial receptive field increases -Tuning of single neurons for object shapes becomes more complex --V1 represents gabor functions -- V2 tuned for angles and bullseyes -Receptive field size grows --Not confined to hemifields, represents most of visual field --Retinotopic maps easy in LGN and V1, harder as you get higher in the visual hierarchy
Give two reasons that it is useful to measure visual field maps. (Wandell reading 09/18)
1) Visual neuroscience theory proposes that specific visual field maps serve certain perceptual specializations (likely that more than one map is essential for particular function). 2) provide useful info on the likely amount of cortical surface area allocated as a function of visual field.
What are the four hypotheses concerning the discrepancy between the effects of attention in V1 as measured using neurophysiology and fMRI?(Boynton 09/25)
1) local field potential - LFPs are significantly better predictors of BOLD signals 2) delayed feedback - effects of attention in V1 show up later so slower fMRI better captures this 3) voxel vs single neuron - more noise when measuring single neuron 4) differences in experimental design
What does temporal summation of BOLD signals tell us about hemodynamic coupling? (Heeger 09/16)
According to the linear transform model, it should be possible to predict the response to a long stimulus presentation by summing the responses to shorter stimuli. In cases where this doesn't hold, however, it shows that BOLD might have an inherently nonlinear dependence on flow.
List one advantage and one disadvantage of study pre-registration. (Lec. 4: Doing Brain Science)
Advantages: prevents p-hacking by means of post-hoc data mining and is more of an open-source data mindset. Disadvantages: Increased administration and extra work.
What is the AS? Point to it
Angular Sulcus
What is the AG? Point to it
Angular gyrus
What are the three ways that attention can affect neural responses? (Lec. 7: Attention)
Baseline response rate - (additive) Shift in spontaneous firing rate of the neuron (literally, baseline elevates or decreases) Change in response gain - (multiplicative) Turns down filters not responsive to intended target -> more neural response Shift in tuning curve - of other neurons to that orientation - change in filtering properties
What are some other important brain structures outside of the cerebral cortex ? (Lec. 2: Brain and Function)
Cerebellum: regulation of motor movements Thalamus: relay motor and sensory signals to the cerebral cortex
What is the CgS? Point to it
Cingulate Sulcus
What is the COS? Point to it
Collateral Sulcus
Give three lines of evidence suggesting that human area MT+ is selective for motion. (Grill-Spector 09/23)
Comparison of coherent vs incoherent light spots activatesMT+ linearly as coherence of motion increases MT+ adapts to patterened motion (global motion) as opposed to lower visual areas that adapt to component motion (local motion) Lesions near MT+ lead to akinetopsia
What are Marr's three levels of analysis? (Poeppel reading 08/28)
Computational: general problem Algorithmic: how problems can be solved Implementational: physical medium in which computation is performed
Why do cognitive neuroscientists make functional maps? What is the point? (Poeppel reading 08/28)
Create local topographical organization. Understand the spatial organization of systems in the brain. It's worked for functional systems such as motor and visual, so cognitive scientists and neuroscientists want to create it for cognitive functions as well
What is the coordinate system that neuroanatomists use to talk about the brain ? (Lec. 2: Brain and Function)
Dorsal: Top of the brain Ventral: Bottom of the human brain Rostral: Front of the human brain (above the eyes) Caudal: Back of the brain when viewed on the horizontal plane Horizontal plane: across the middle Sagittal plane: across the corpus callosum Coronal plane: halfway across the brain
How are neurons, columns and areas connected to one another ? (Lec. 2: Brain and Function)
Each column has 6 layers, it's important to remember that -The input to the cortex occurs at layer 4 -Cortico-cortical connections are through layer 1 -Cortico-thalamic connections are through layer 6
What is the difference between FFA and PPA? (Grill-Spector 09/23)
FFA - fusiform face area: face-selective, manifests foveal bias PPA - parahippocampal place area: place-selective, manifests peripheral field bias
What is generalization and why is it important? (Lec. 4: Doing Brain Science)
Generalization is how closely the study results will apply to the real world. I.e High generalizability -> very relevant/replicable in the real world
What new information was provided by studies of H.M. that was not available earlier? (Squire 10/02)
H.M. provided four principles about how memory is organized in the brain.
What is Poeppel's proposed solution to the mapping problem? Do you think that it could work? (Poeppel reading 08/28)
He proposes we rethink the current primitive parts list for neuroscience and psychology, which may call for a 'radical decomposition' of previously agreed upon primitives.
What is the HS? Point to it
Heschel's Sulcus
Explain the hierarchical view of human visual cortex organization proposed by the authors of this paper. (Grill-Spector 09/23)
Hierarchical processing proposes that visual perception is achieved via a gradual stagewise process in which information is first represented in a localized and simple form and is transformed into more abstract/holistic/multidimensional representations as you go to higher visual fields.
Why might functional maps that we recover from the human cerebral cortex depend on the spatial resolution of our measurement device? (Op de Beeck 09/09)
Higher spatial resolution allows for greater detection of minute circuitry, thus allowing function to be specialized into smaller, more specific locations for a map. Thus, increasing map accuracy.
Based on the arguments in this review, do you think that attention and working memory are the same thing or different things? Explain and justify your answer using evidence from the review. (CHUN reading 09/30)
IDK if you can say they are the same but rather significantly correlated (sustained visual attention increases working memory's capacity to retain info). What makes it unclear is whether attending representations in the absence of sensory input is still defined as attention or if it crosses the boundary to working memory.
What is the IFS? Point to it
Inferior Frontal Sulcus
What is the ITS ? Point to it
Inferior Temporal Sulcus
What is the ISJ? Point to it
Intermediate Sulcus of Jenson
Under what conditions does fMRI favor measuring signals in large blood vessels and under what conditions does it favor measuring signals in small vessels? (Heeger 09/16)
Intravascular signals are suppressed at higher magnetic field strengths because of the inhomogeneity in the magnetic field within larger vessels. The acquisition can be modified to de-emphasize the BOLD signals from larger vessels, by suppressing signals that are associated with higher flow velocities.
What is invariance and why is it important for understanding vision?(Grill-Spector 09/23)
Invariance refers to the lack of change in response to stimulus. This is important because higher-level areas show more invariant tuning to contrast sensitivity and high sensitivity to parafoveal stimuli, implying larger receptive fields in higher-level areas.
What is a functional map? (Lec. 3: Filters, Circuits and Maps)
It is a region in which stimulus or task selectivity changes smoothly and systematically across the map. Not all functional areas contain smooth functional maps, and not all stimulus or task dimensions represented within an area are represented as smoothly changing maps. To create one we need a way to measure (MRI, etc.) it and identify what it's doing (function)
Bonus question: Why do you think the human area called MT+ when the corresponding area in non-human primates is called MT? (Grill-Spector 09/23)
It is responsible for more functionality?
List one advantage and one disadvantage of publishing null results. (Lec. 4: Doing Brain Science)
It provides a historical record of all failed experiments and can inform future studies. Pollutes literature with the failures that occur the most often, looks "bad" for researchers (although not really)
What is the relationship between BOLD and LFPs (Local Field Potentials)? (Heeger 09/16)
LFP reflects the superposition of synchronized dendritic currents, averaged over a larger volume of tissue and has been shown to be a good predictor of BOLD signals.
Why does collecting more data often improve the reliability and predictive power of an experiment? (Yarkoni 09/11)
Large data sets provide natural guard against overfitting. The larger the sample, the more representative it is of the population
What is the LOS? Point to it
Lateral Occipital Sulcus
What is a retinotopic map? (Lec. 6: Retinotopy, other topies and vision)
Mapping of the visual input from the retina to the neurons within the visual stream (primary visual cortex)
What is the mFus ? Point to it
Medial Fusiform Sulcus
What are the four principles of memory organization suggest by studies of H.M.? (Squire 10/02)
Memory is not one single thing Memory is not involved in intellectual and perceptual functions (as evidenced by damaged structures) Medial temporal lobe structures are not needed for immediate memory or for the rehearsal and maintenance of material in working memory Medial temporal lobe cannot be the ultimate storage site for long term memory. Permanent memory must be stored elsewhere (presumably the neocortex)
Name three general properties that change systematically across the human visual hierarchy from more peripheral to more central visual areas, and describe how they change. (Grill-Spector 09/23)
Micro-circuitry increases Diameter of spatial receptive fields increases Tuning of single neurons for objects becomes more complex
What general kinds of information is represented in relatively more central visual areas? (Lec. 6: Retinotopy, other topies and vision)
More central areas represent more complex features (later areas)
What general kinds of information is represented in relatively more peripheral visual areas? (Lec. 6: Retinotopy, other topies and vision)
More peripheral areas represent simpler features (earlier areas)
Explain the "matched filter" view of attention. (Lec. 7: Attention)
NEEDS TO BE FILLED OUT
Explain the "routing system" view of attention. (Lec. 7: Attention)
NEEDS TO BE FILLED OUT
Summarize the Cukur et al. study of semantic attention. (Lec. 7: Attention)
NEEDS TO BE FILLED OUT
What is the main difference in neural responses found with spatial versus feature-based attention? (Lec. 7: Attention)
NEEDS TO BE FILLED OUT
What is the local coding principle ? (Lec. 3: Filters, Circuits and Maps)
Neurons are usually sensitive only to a narrow range of values along a few dimensions.
What is the pCgs? Point to it
Para-cingulate fissure
What is the POS? Point to it
Parieto-Occipital Sulcus
What is the PoCes? Point to it
Post-Central Sulcus
What is the difference between anterograde and retrograde amnesia? What brain structures are affected in these two situations? (Squire 10/02)
Retrograde: pre-existing memories are lost. Hippocampus, parahippocampal gyrus (this led to longer episode of amnesia) Anterograde: ability to memorize new things is impaired.
What is the difference between spatial attention and feature-based attention? (Lec. 7: Attention)
Spatial attention - response in V1 from voxels associated with attended peripheral stimuli compared to unattended stimuli placed in the opposite visual hemifield. (Basically, setting/location you inhabit) --fMRI responses in human V1 are now known to modulate with spatial attention even in the absence of a physical stimulus. The effect of attention on the fMRI contrast response function in V1 and other early visual areas is additive. [Additive - basis of linearity -For simple stimulus experiments -Brain is not linear] 0Feature-based - such as a direction of motion or orientation, enhances the response to visual neurons selective to that feature and suppressed response to neurons tuned away. This feature-based effect has been shown to operate on neurons with receptive fields well outside the spatial focus of attention. --Feature-based effects have been found in macaque areas MT and V4 but so far not in area V1. fMRI responses in V1 have been shown to be strongly modulated by feature-based attention.
What is the difference between significance and importance? (Lec. 4: Doing Brain Science)
Statistical significance is a measure of certainty as it relates to random variation. While importance is a measure of the effects of the event.
What is the structure of a mammalian CNS neuron and how do these neurons work ? (Lec. 2: Brain and Function)
Structure of a CNS neuron -- Dendrites -- Cell body -- Axon hillock -- Myelin sheath -- Axon -- Axon terminal How Neurons Work - Dendrites receive information which is then sent out through the axon terminals --it should be noted dendrites participate in analogue summation which is then outputted as a digital signal (1's and 0) - The synapse (All I think we really need to know if that the synapse is where NT release occurs; astrocytes suck up extra NT that doesn't bind to the synapse of the post synaptic membrane; this is one theory for the basis of hemodynamic coupling) - Action Potential: 1.) depolarization to threshold (-55mV) 2.) activation of Na+ (sodium ion) channels (+40mV) 3.) repolarization ---inactivation of Na+ channels & activation of K+ (potassium ion) channels 4.) hyperpolarization ---return to normal permeability ---dips below -70mV then returns 5.) refractory period ---another signal cannot be fired in quick succession
What are the basic anatomical subdivisions of the cerebral cortex ? (Lec. 2: Brain and Function)
Subdivisions -Frontal lobe -Parietal lobe -Occipital lobe -Temporal lobe Major Borders -Central Sulcus: Splits Frontal and Parietal -Parieto-occipital Sulcus: Splits Parietal and Occipital lobe -Preoccipital notch: Splits Occipital and temporal lobe
What is the SFS? Point to it
Superior Frontal Sulcus
What is the STS? Point to it
Superior Temporal Sulcus
What is the SMG? Point to it
Supra-marginal gyrus
What is the SyF ? Point to it
Sylvian Fissure
What is the noise ceiling and why is it important? (Lec. 4: Doing Brain Science)
The noise ceiling describes when an independent variable no longer has an effect on a dependent variable, or the level above which variance in an independent variable is no longer measurable. A model's maximal accuracy is dictated by the noise ceiling (i.e it can only be as good as the noise ceiling which exists between 0-1)
What is the difference between primate Area V4 and human Area hV4? (Wandell reading 09/18)
The retinotopic map of hV4 was debated. Monkey V4 has clear dorsal and ventral halves but hV4 only has a clear ventral half. It appears that there is also a dorsal half, but that is clipped when compared to the dorsal halves of V1/2/3 human hV4 area represents more selectivity features than the monkey area, but scientists still do not have a clear understanding of the retinotopic map of hV4.
What factors might influence the way that functional maps and modules are organized across the surface of the cerebral cortex? (Op de Beeck 09/09)
The ways that different maps might be combined, depending on the relationship in their coding for specific functional properties (e.g. uncorrelated and additive, correlated and additive, and correlated & non-linearly combined)
What is the "mapping problem" that Poeppel states is one of the main impediments to successfully linking cognitive neuroscience to psychology? (Poeppel reading 08/28)
There is no accepted relationship between the primitive units (parts list) of each respective field. Poeppel argues that this is in part responsible for the lack of progress made towards linking physical activity to cognitive behavior.
What is the replication crisis? (Lec. 4: Doing Brain Science)
This is the problem in which many reported findings are not replicable. -This stems from over-reliance on statistical significance. - How to fix: -- Make things double blind -- Pre-register -- Publish null hypotheses -- Open data -- Adopt current best practices from ML -- Purely data-driven approach -Example of data-driven fMRI experiment -- Want to know what locations in brain represent different pictures. -- Collect estimation data -- Extract features from stimuli -- Estimate model weights -- Maximize predictions -- Collect validation data -- Calculate stimulus feature -- Pass through models -- Test predictions -- Interpret results of model
Explain the difference between exogenous (bottom-up) and endogenous (top-down) attention. (Lec. 7: Attention)
Top down: comes from somewhere within the brain and affects other regions of the brain (anticipatory) Bottom - up: comes from outside world into the brain (reflexive)
How is the visual field arranged across primary visual cortex? (State your answer in terms of visual quadrants and the horizontal and vertical meridia.) (Wandell reading 09/18)
V1 is in the center of V2 and V3. V2 and V3 both contain discontinuous hemifield maps, which are divided along the horizontal meridian. This discontinuity creates two quarter field maps in V2 and V3, each of which has one long edge representing the horizontal meridian and a second representing the vertical meridian. The vertical meridian representations of V1/V2 are adjacent to one another, as are the horizontal meridian representations of V2/V3
Why might attentional modulation be so different in single neurons and in voxels? (Lec. 7: Attention)
Voxels represent a population of neurons. In fMRI, attentional effects are much more noticeable due to LFP's (activity that doesn't include actual firing, example: EPSP or IPSP). In single neurons, you will only see firing for a particular stimulus; you can't tell what it's connected to or what exactly it's firing for without a contrast in stimulus. In neurophysiology, you also get smaller spikes from attention in comparison to what fMRI picks up (captures inputs to neurons).
Could the observed differences in the effects of attention in V1 as measured using neurophysiology and fMRI reflect differences in the experimental methods used in these two fields? Explain your answer.(Boynton 09/25)
YES. for example, electrophys mainly on animals yet they are not specialized for language and speech so the results are accurate but not exactly comparable to humans.
What is the visual field cluster idea proposed by Wandell? (Wandell reading 09/18)
a cluster is a group of maps with parallel eccentricity representations, different clusters have distinct eccentricity maps.
What is a "processing stream"? (Grill-Spector 09/23)
a hierarchical sequence specialized for a particular functional task.
Compare and contrast achramotopsia versus prospagnosia. Be sure to include their implications for memory. (Squire 10/02)
achromatopsia: inability to perceive color (still able to distinguish wavelength in the form of shades of grey), if early memories depend on cortical structures for color perception, then those memories previously retrieved in color should now be black/white -> true. prosopagnosia: once-recognizable faces no longer yield a memory signal, same areas that support the perception of faces also support the long-term memory of faces.
Which of these factors is reflected in the BOLD signal: (a) blood oxygenation, (b) blood flow rate, (c) blood flow volume. (Heeger 09/16)
all of the above
What is the aarSyF? Point to it
anterior ascending ramus of the sylvian fissure
What is the ahrSyF? Point to it
anterior horizontal ramus of the sylvian fissure
What is the alITS? Point to it
ascending limb of the Inferior Temporal Sulcus
What is the SyFar? Point to it
ascending ramus of the sylvian fissure
How can we identify the border of a functional module or map? (Op de Beeck 09/09)
by measuring clear discontinuities in activity across cortical regions (in response to a stimulus).
How might we determine whether a functionally-specific region of the cerebral cortex is part of a larger functional map? (Op de Beeck 09/09)
by measuring peak activation in response to function-specific stimulus. If shifts are gradual, this is indicates the incorporation of a larger area of the cortical layer.
What is the CeS ? Point to it
central sulcus
What is the CSI ? Point to it
circular sulcus of the insula
What is the discrete slot model of working memory and how does it differ from the continuous model? (CHUN reading 09/30)
discrete-slot (fixed resolution): working memory can store representations in a set number of slots wherein resolution doesn't decay as load increases. continuous (variable-resolution): resolution of a representation decreases as load increases.
How do retinotopic maps change across the visual hierarchy, and why? (Grill-Spector 09/23)
early areas V1, V2, V3 - high retinotopy, low motion sensitivity, low object sensitivity. Intermediate areas V3a, hV4 - lower retinotopy, stronger responses to motion and objects. high level areas - low retinotopy, high degree of specialization. Early areas associated with low-level features of stimuli and perception whereas high-level areas demonstrate more specialized tuning.
Why does cross-validation help reduce some of these problems? (Yarkoni 09/11)
estimates prediction error. A way of "testing" within the training set to have better estimate of how it will perform in the real werld.
Explain the difference between explanation and prediction in Psychology research. (Yarkoni 09/11)
explanation aims to draw correlation/causation relationships between observed behaviors whereas prediction is aimed towards behaviors that have not yet been observed and their accurate forecast.
How is the BOLD signal measured by fMRI? (Heeger 09/16)
fMRI uses a magnetic field to introduce a characteristic radio frequency to hydrogen atoms found in the water molecules of the brain causing some to spin uniformly (precession). Different blood affects the strength of this signal. deoxyhemoglobin decreases signal/image intensity and oxyhemoglobin increases it. fMRI measures the changes in inhomogeneity of magnetic field that results in blood oxygenation (influenced by influx of new blood that hasn't been spun yet).
What is the relationship between feature-based attention effects in V1 as measured using neurophysiology and as measured using fMRI?(Boynton 09/25)
fMRI, studies have shown feature-based attention effects on responses in both attended and unattended stimuli. Effect has not been prominent in neurophysiology.
What is the fTS? Point to it
first Transverse Sulcus
How does the number research teams that are pursuing a question or the "hotness" of the question affect likelihood that a published research finding is true? (Loannidis 09/02)
greater number of research teams and "hotter" the question makes findings less likely to be true.
What is the difference between primate Area MT and human Area MT+? (Wandell reading 09/18)
hMT+ is highly motion-sensitive. hMT+ has been found to represent more than just motion, but since our resolution is not yet good enough to understand what more it represents we just know that it represents more than primate area MT.
How does increasing flexibility in study design and operational definitions affect likelihood that a published research finding is true? (Loannidis 09/02)
higher flexibility makes findings less likely to be true.
How do the visual field maps in humans and visual field maps in non-human primates correspond to one another? (Wandell reading 09/18)
human visual cortex contains greater # of neurons that primates presumably for the processing of language and reading. Homologous maps but nonhomologous functional properties (e.g. V3 in humans vs V3a in primates share similar map topography but V3 strong response for motion whereas not so in V3a)
What is the bias-variance tradeoff? (Yarkoni 09/11)
increase bias reduce variance
What is the IPS ? Point to it
intra-parietal sulcus
Name three brain regions that experiments suggest represent information in working memory. (CHUN reading 09/30)
intraparietal sulcus, parahippocampal cortex, primary visual cortex
Summarize the results of Warrington and Shallice (1984) concerning the representation of semantic knowledge. (Squire 10/02)
knowledge systems may be organized by semantic categories. category-specific retrograde memory deficits reflect the loss of knowledge that was initially acquired through category-specific processing.
Name two pieces of evidence that suggest that fMRI signals are linearly related to neural signals, and two pieces of evidence suggesting that fMRI signals are not linearly related to neural signals. (Boynton 09/25)
linearly related: temporal summation, point spread function (BOLD responses pooled across spatially separate areas should equal the sum of those signals). Not linearly related: flashing light for 1ms should have 1/10th response of 1s but does not, adaptation effect (estimated response to repeated stimuli is smaller than expected)
According to the authors, what is the difference between a cortical "module" and a cortical "map"? (Op de Beeck 09/09)
map: a gradient of selectivities along the cortical sheet. module: clustering of selectivities in discrete regions with clear discontinuities at the boundaries (bigger than a column).
How does PET work? What are some advantages and disadvantages? (Lec. 5: Measurement, MRI, and fMRI)
measures decay of radioactive tracers in brain to look at oxygen/glucose/etc. -Advantages --Tracers can be very specific --Tomographic (2d slices) -Disadvantages --Poor signal quality --Poor spatial resolution --Involves ionizing radiation
How does EcoG work? What are some advantages and disadvantages? (Lec. 5: Measurement, MRI, and fMRI)
measures electricity of brain on the actual brain itself (not outside skull) + you can stimulate the brain -Advantages: -- Good signal quality -- Good temporal resolution -- Measures neural activity -- Simulation is a causal method -Disadvantage --Invasive (gotta get into the brain -- 2d -- Signal quality low (faint signal in operating room) -- Patients often do not perform well
How does EEG work ? What are some advantages and disadvantages ? (Lec. 5: Measurement, MRI, and fMRI)
measures electricity of the brain -Advantage: --Simple --Cheap --Good temporal resolution --Measures neural activity -Disadvantage --Inherently 2d --Poor spatial resolution --Measure is biased
How does MEG work? What are some advantages and disadvantages? (Lec. 5: Measurement, MRI, and fMRI)
measures magnetic activity using superconducting sensor -Advantages --Better signal quality than eeg --Good temporal resolution --Measures neural activity -Disadvantages --Expensive --Poor spatial resolution --Neural activity measure is biased --Inherently 2d
Is memory likely to be one system, two systems or multiple systems? Why? (Squire 10/02)
multiple
What is overfitting? What is p-hacking? How are these related? (Yarkoni 09/11)
overfitting: statistical models mistakingly fitting sample-specific noise as if it were signal. p-hacking: the practice of flexibly selecting analytical procedures based in part on the quality of the results they produce. both serve to maximize sample-specific results and lower test accuracy.
What is the PreCeS ? Point to it
pre-central sulcus
Explain current views of the role of the hippocampus versus the neocortex in memory. (Squire 10/02)
rather than broadly encoding and consolidating memories, like the medial temporal lobe, each neocortical region stores only specific features of an experience. hippocampus uniquely specialized to establish/maintain declarative memories, other structures support initial perception/processing important to long term storage
What is the difference between a computational operation and a representational primitive? (Poeppel reading 08/28)
representational primitive: one of the many basic units in any arbitrary . field of study used to explain other phenomena (e.g. neuron). computational operation: high level function of these primitives and their networks.
Why is spatial resolution critically important in cognitive neuroscience? (Poeppel reading 08/28)
resolution can serve to enhance or limit the precision with which cognitive neuroscientists can measure localized neural activity at an anatomical level.
How does the choice of stimuli (or tasks) affect our ability to identify cortical modules or maps? (Op de Beeck 09/09)
singular features are limiting. multidimensional analysis allows for greater findings, but have to be careful when measuring which components of a feature are being held constant.
What is the "interpretability problem" that occurs when using (some) machine learning methods to fit predictive models to data? (Yarkoni 09/11)
some approaches are seen as "black box" algorithms that are very difficult to explain/understand (i.e. neural net)
What is the difference between spatial and feature-based attention? (Boynton 09/25)
spatial: selective attention directed to a location in space. feature-based: selective attention directed to a non-spatial feature such as color or orientation
What is the relationship between spatial attention effects in V1 as measured using neurophysiology and as measured using fMRI?(Boynton 09/25)
strong spatial attention effects studied with fMRI, to date no robust effects in neurophysiology
What is the sbPS? Point to it
sub-parietal sulcus
Describe the traveling wave method. Can you think of any other ways to measure visual field maps? (Wandell reading 09/18)
subject fixates on a point. series of flashing rings measure eccentricity. series of rotating wedges measure angular response. Together, the measurements specify the most effective visual field position in polar coordinates.
What is the idea of capacity limitation? Provide an example to support your answer. (CHUN reading 09/30)
the ability to hold items in visual working memory drops considerably after a handful (i.e. reduced when multiple features in same dimension must be encoded)
Explain the physiological basis of hemodynamic coupling in the brain. (Heeger 09/16)
the relationship between blood flow / oxygenation and neural activity in the brain. Not fully understood. One approach is that active neurons require greater amounts of oxygenated blood)
How does TMS work? What are some advantages and disadvantages? (Lec. 5: Measurement, MRI, and fMRI)
using magnetic fields to stimulate electrical field in the brain -Advantages --Causal method --Can be used with normal subjects --Relatively easy and cheap -Disadvantages --Like being hit with a hammer --Poor spatial resolution --inherently 2d --Difficult to localize
What is Chun's definition of working memory? (CHUN reading 09/30)
visual working memory allows visual information to be actively extended over time in the absence of sensory input but has a limited capacity.
Is hemodynamic coupling linear? Explain and justify your answer. (Heeger 09/16)
yes, that is an underlying assumption for the hemodynamic coupling theory. An increase in oxygenated blood to the neuron is representative of an increase in activity.