Readings

Question 1

What does the paper “identifying natural images from human brain activity” try to do that previous studies haven’t done?

Answer 1

• Previous studies decoded orientation, position and object category from activity in the visual cortex using simple stimuli or images from fixed categories
• This study develops a decoding method based on quantitative receptive-field models that characterise the relationship between visual stimuli and fMRI activity in early visual areas

Question 2

What do the receptive-field models used in the paper “identifying natural images from human brain activity” do?

Answer 2

-These models describe the tuning of individual voxels for space, orientation and spatial frequency and are estimated directly from responses evoked by natural images

Question 3

What goal of neuroscience is the paper “identifying natural images from human brain activity” trying to help solve?

Answer 3

A goal in neuroscience is to be able to decode mental content from brain activity

Question 4

What is the main result of the paper “identifying natural images from human brain activity”?

Answer 4

Results show that these receptive field models used make it possible to identify which specific image was seen by the observer from a large set of completely novel natural images

Question 5

What does the result of “identifying natural images from human brain activity” say about spatial frequency?

Answer 5

• Simpler receptive-field models that describe only spatial tuning yield much poorer performance
• Results indicate that both orientation and spatial frequency tuning contribute to identification performance, but spatial frequency makes the larger contribution

Question 6

What do the results of “identifying natural images from human brain activity” say about future possibilities in decoding brain activity?

Answer 6

Results suggest it may soon be possible to reconstruct a picture of a person’s visual experience from measurements of brain activity alone

Question 7

What are the stages of the experiment in the paper “identifying natural images from human brain activity”?

Answer 7

Stage 1: estimate a receptive-field model for each voxel. Stage 2: measure brain activity for an image and predict brain activity for a set of images using receptive-field models. Select the image whose predicted brain activity is most similar to the measured brain activity

Question 8

What was the best performance in the experiment in “identifying natural images from human brain activity”?

Answer 8

For one subject, 92% of images were identified correctly, based on results from activity patterns averaged across 13 repeated trials. Single trial performance was 51% for the same subject

Question 9

What were the results about the retinotopy only model in “identifying natural images from human brain activity”?

Answer 9

The retinotopy-only model performed worse than the Gabor wavelet pyramid model in identification performance

Question 10

What did the results from “identifying natural images from human brain activity” say about stimulus’s over time?

Answer 10

The stimulus-related information that can be decoded from voxel activity remains largely stable over time

Question 11

What was the hypothesis of the paper “Attention during natural vision warps semantic representation across the human brain”?`

Answer 11

Hypothesis: attention causes tuning changes to expand the representation of attended stimuli at the cost of unattended stimuli

Question 12

How was the hypothesis tested in “Attention during natural vision warps semantic representation across the human brain”?

Answer 12

Used fMRI to measure how semantic representation changed during visual search for different object categories in natural movies

Question 13

What were 3 results of the paper “Attention during natural vision warps semantic representation across the human brain”?

Answer 13

• Many voxels across the cortex shifted their tuning toward the attended category. These tuning shifts expanded the representation of the attended category and of semantically related but unattended categories, and compressed the representation of categories that were semantically dissimilar to the target
• Attentional warping of semantic representation occurred even when the attended category was not present in the move, this means the effect was not a target-detection artefact
• Results suggest that attention dynamically alters visual representation to optimise processing of behaviourally relevant objects during natural vision

Question 14

What is the semantic system according to “Natural speech reveals the semantic maps that tile human cerebral cortex”?

Answer 14

The meaning of language is represented in regions of the cerebral cortex collectively known as the semantic system

Question 15

What did the experiment of “Natural speech reveals the semantic maps that tile human cerebral cortex” involve?

Answer 15

• Systematically map semantic selectivity across the cortex using voxel-wise modelling of fMRI data collected while subjects listened to narrative stories
• Then the paper used a novel generative model to create a detailed semantic atlas

Question 16

What were the 2 main results of the paper “Natural speech reveals the semantic maps that tile human cerebral cortex”?

Answer 16

• Shows that the semantic system is organised into intricate patterns that seem to be consistent across individuals
• Suggests that most areas within the semantic system represent information about specific semantic domains, or groups of related concepts, and the atlas shows which domains are represented in each area

Question 17

What did the paper “Natural speech reveals the semantic maps that tile human cerebral cortex” use model estimation for and what were the results?

Answer 17

After measuring BOLD responses using fMRI for 2 hours of narrative stories, it used model-estimation to predict new BOLD responses for new 10-minute story. The result showed good predictive performance in many brain areas

Question 18

In “identifying natural images from human brain activity”, what do the graphs of marginal orientation and spatial frequency curves show?

Answer 18

As spatial frequency increases, the positive response starts low and increases almost linearly before flattening.
As orientation degree increases, the positive response remains around a similar level that is about halfway on the spatial frequency plot

Question 19

In “identifying natural images from human brain activity”, which factors impacted identification performance?

Answer 19

• difference in performance across subjects is due to intrinsic differences in the level of noise
• repeated trials gave better performance than individual trials

Question 20

In “identifying natural images from human brain activity”,, how does performance extrapolate to 10% correct?

Answer 20

In all cases performance scaled well with set size. At least 10^3.5 set size required for performance to decline to 10%

Question 21

In “Attention during natural vision warps semantic representation across the human brain”, how does semantic and voxel representation change across passive viewing and target viewing according to the tuning-shift hypothesis?

Answer 21

During passive viewing semantically similar categories project to nearby points in semantic space. Attention to a specific category expands representation of attended and nearby categories, voxels shift their tuning towards attended categories, fewer voxels are tuned for distant categories

Question 22

In the results from “Attention during natural vision warps semantic representation across the human brain”, how does a single voxel change during searches for humans and vehicles?

Answer 22

The voxel is strongly tuned, produces above mean responses, for the attended category in both conditions. weaker tuning is observed for unattended categories

Question 23

In the results from “Attention during natural vision warps semantic representation across the human brain”, how is the overall tuning shift explained?

Answer 23

The degree of the tuning shift is positively correlated with the fraction of variance explained by tuning changes for attended categories