Design Optimization

Note

This section is still under construction; please come back soon.

Overview

fMRI experiments are designed slightly differently than typical behavioral experiments. If we take our Flanker experiment as an example, we can see the following differences between the behavioral experiment and the fMRI experiment:

Behavioral:

  • Dependent measure of interest is reaction time
  • Can have the same amount of time between trials
  • Subject is in front of a computer, uses a keyboard to make responses

fMRI:

  • Two dependent measures: reaction time and the BOLD response
  • Will need differing amounts of time between trials
  • Subject is in the bore of the MRI, uses a button pad to make responses

Let’s take a look at each of these differences in turn.

Dependent Measures

In both versions of the experiment, we have two conditions: Incongruent and Congruent trials. (See here <fMRI_02_ExperimentalDesign> for more details). We collect several trials, or instances of each condition. For example, if we have 5 trials in each condition, there would be 5 trials in which the subject saw an Incongruent stimulus, and 5 trials in which they saw a Congruent stimulus.

In the behavioral version, we would collect reaction times for each trial, and average them over each condition. Let’s say that the reaction times (in milliseconds) for each trial were as follows:

Incongruent: 650, 670, 681, 591, 623 -> Average RT of 643ms Congruent: 560, 571, 603, 580, 575 -> Average RT of 577.8ms

This is what we would expect from a typical subject: The average reaction time for Incongruent trials is greater than for Congruent trials. Or, to put it another way: Subjects are slower on average at making responses to the Incongruent trials as compared to the Congruent trials.

With fMRI data, on the other hand, we measure the BOLD response in addition to reaction time. The goal is to detect a difference in brain activity between the two conditions, if such a difference exists.

Practicals/Incongruent-Congruent_fMRI.png

Amount of Time between trials

The behavioral version of the experiment would include some time between each trial, usually a few seconds. The experimenter may choose to introduce random variation, or jitter, between trials in order to add some variety to the experiment and make it more engaging. This is not, however, strictly necessary.

For an fMRI experiment, on the other hand, jitter is necessary in order to distinguish betwee the BOLD responses of multiple conditions. Recall that the BOLD response unfolds over a period of several seconds, typically peaking around 4-6 seconds and falling back to baseline around 12 seconds. (There is also an undershoot that occurs later, but we will ignore that for now.) Thus, for trials that occur within 10 seconds of each other, their BOLD responses will overlap and will convolve.