fMRI: problems / W5L2 Flashcards
How can statistical tests obscure accurate findings?
Due to a significant concern with ‘false positives’ we only put forward the most robust finding which are often very specific.
This may result in an overlooking of broader networks at play.
How can using t-test be challenging for understanding the brain ?
- > 50,000 voxels in the brain
- Because we are essentially running a t-test on every voxel and there is a % chance of a false-postive on every test.
- A bonferroni on this number of test leaves us with an extremely low significant value (p-value).
How is reverse inference a problem when interpreting neural processes?
Because our inference can only be based on other studies.
Therefore we run the risk of inaccurately attributing specific brain activity (x) during a task to a particular cogntive process because that was shown in other studies study.
In reality (x) may be due to a range of different tasks or cog processes
What does Duncan’s (2010) Multiple Demand Network imply for understanding neuroimaging?
Showed that the prefrontal cortex was active during a range of high-level cog processes.
This shows us that it simple cause and effect analysis can be very inaccurate.
What are some dissagreements to Duncan (2010) research on the pre-frontal cortex (PFC)?
While Duncan suggests that the PFC activation is to broad to be understood other have found evidence for task specification in specfic areas of PFC
duncan conceeds there may be relative specialisation, not absolute
What is the problem with task specificity ?
Researches may assume because their is a relationship between task and activation that it is the ideal measure of a cognitive process. This may not be true.
What are 3 factors to look out for when reading neuroimaging research
- What kind of post-hoc tests have been run. Does the finding therefore represent something to explain brain activity
- Is there some reverse inference going on? i.e attributing a certain activation to a cog process demonstrated in another study
- How confident are you that the chosen task accurately captures the cog process?
How did Poldrack (2006) consider learning from fMIR results?
In probabalisitic terms.
The probability that we can learn from fMIR results depends on specificity of task and specificity of brain region
What is a problem in interpreting null findings ?
That we cannot interpret them. i.e we cannot say that the null is true because the H1is not shown.
Limiation of BOLD: Spatial resolution
What is the size of a voxel
A voxel is the smallest unit we can measure. Therefore we cannot know what occurs inside one:
* cannot know what is occuring within the voxel
* cannot compare all the differences between voxels
standard size: 3 x 3 x 3 mm / 3D pixel
Limitation of BOLD: Temporal Resolution
how long does it take to scan the entire brain?
“poor temporal resolution” means that we cannot see any changes that take place before ~1-2 seconds
What method provides good spatial resolution?
fMRI
weak temporal resolution
What method provides good temporal resolution?
EEG or MEG
