Lecture 10 - fMRI Research: What do we learn about cognition? Flashcards
How does the strict statistical analysis of fMRI results affect our ability to understand what is actually happening in the brain?
In order to avoid false positives only the most robust results are trusted. What this means when it comes to reading fMRI scans is that we run the risk of excluding valuable data that can give a better indication of what is happening in the brain. We only see the “tip of the iceberg” when it comes to observing brain activation using fMRI scanning.
What is the Bonferroni-Correction ?
Divide the overall desired significance value by number of tests.
How do we reduce the risk for false positives when presenting and interpreting fMRI results?
There are around 50,000 voxels/signal points measured during fMRI scanning. Each voxel’s level of activation is determined by whether it is significantly activated compared to a control. This means around 50,000 t-tests. To reduce percentage of risk for false positives for due to all these tests some people use the Bonferonni-correction so that the overall significance value used is 1%. This is a very conservative approach and means that only highly significantly activated brain regions will show up.
What is REVERSE INFERENCE?
Reverse Inference refers to the phenomenon of taking prior research or inferences and applying those to your results and then concluding that the prior inference can be applied to this data given that it similar data. E.g. if a brain region has been said to be activated during a specific cognitive task, such as the FFA for facial recognition, and then our data shows activation in FFA and then we say that this person must be recognising a face, then we are using a prior inference globally without foundation. It may be that that area is activated for multiple different cognitive tasks or that the cognitive task attributed to that task is more complex than being assigned to one brain region.
What are two competing beliefs about prefrontal cortex activation seen in fMRI scans for different cognitive tasks? Hint Duncan (2010) - Demand, and Badre et al (2009) - abstractness.
Duncan (2010) proposes a theory of Multiple Demand Network that describes the prefrontal cortex being activated to a larger extent and across more areas based on the complexity or demand of the task, as opposed to specialised areas or networks for specific tasks.
Badre et al (2009) agree that there may not be specialisation in brain regions for specific tasks, but do propose that there may be specialisation based on how abstract the cognitive ask is, with the more abstract the cognitive process the further to the anterior (front) of the prefrontal cortex the activation is.
What is task specificity and why is it important to consider this when interpreting fMRI (and all cognitive experiments) results?
Task specificity when it comes to cognition research refers to the extent to which a task engages the cognitive process being studied.
When studying cognitive processes, such as using fMRI, it is important that the task being used to engage a specific cognitive task actually engages that task. What can happen if this is not the case is that the way the data is interpreted is based on the assumption that the brain regions being activated are representative of the cognitive task being studied. This is part of the problem of reverse inference.
Furthermore, NO task involves only one cognitive process, so it is very difficult to study a cognitive process in isolation.
What are some problems faced when trying to interpret fMRI results?
Does the task actually engage the cognitive task being studied? Is reverse inferencing happening? Has the statistical analysis been conservative enough or has it been too conservative?
Can Null Results be interpreted? Why or why not?
Classical Statistical Tests are based on the Null Hypothesis (Null Hypothesis Significance Testing). When we get a significant result this means that it is highly unlikely that the result we got could be true if the Null hypothesis is true. If we do not get a significant result however, this doe not mean that we have evidence to show that the Null Hypothesis is true. NHST is based on making it difficult for the experimental hypothesis to be true, not for the null hypothesis to be true.
Do fMRI and MRI scans have high spatial resolution?
Compared to other brain imaging techniques fMRI and MRI scans have high spatial resolution. However, the smallest unit of measurement is a voxel. Voxel’s are 3mmx3mm and are representative of information from >100,000 neurons. What is happening within a voxel cannot be determined with the currently resolution capacity of MRI scanning, and therefore a lot information about brain activation patterns within these voxels is lost. Another way of saying this is that the resolution of MRI scanning may not be sensitive enough to pick up differences between different conditions.
What is the quality of the temporal resolution of BOLD fMRI scanning?
The temporal resolution of BOLD fMRI is poor. This is because it takes a relatively long time to measure all the points in the brain that are required to build a 3D image of the brain. Cognitive processes occur very quickly and so the fact that it takes so long to obtain the signals from the brain using BOLD signal fMRI we cannot infer or capture information about the timing of cognitive processes.
What are two techniques that could be used to determine timing of cognitive processes?
Electroencephalography (EEG) and Transcranial Magnetic Stimulation (TMS) have high temporal resolution.
