Functional Magnetic Resonance Imaging
Though the use of blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI) achieves a temporal resolution that is lower than EEG (seconds vs. milliseconds, respectively), the spatial resolution one can achieve from its use is highly valuable for observing subcortical activity (as shown in Figure 11.4). The temporal resolution limitation of fMRI is due to a reliance on the hemodynamic response function (hrf).
The hrf reflects a change in oxygenated blood in area of the brain due to activity in that area. Due to an initial dropout of signal in the first few seconds, temporal resolution is often limited to 3-4 seconds. It may be possible to achieve a lower temporal resolution depending on the type of experimental design one chooses to employ, that is, event-related or block design.
FIGURE 11.4 An example image output using the SPM fMRI analysis package (Wellcome Trust Centre for Neuroimaging). This image reflects activity in the Putamen.
An event-related design allows one to record activation in the brain in response to specific stimuli. Using a block design has an advantage of a decreased study time and an increase in signal-to-noise ratio. However, a huge disadvantage to using a block design is the use of cognitive subtraction, which results in a high loss of temporal resolution.
Despite the increased temporal resolution of event-related designs, concerns have been raised over the possibility of having too high of a granularity for some behavior. It has been posited that it may take up to 10s for activation in the brain to begin due to an emotional response (Liotti & Panksepp, 2004), possibly making an event-related paradigm inappropriate (and making a block design the appropriate choice). One may also raise an opposite problem, processing could be too rapid (e.g., LeDoux, 1996) to correctly capture in even an event-related design. Though these concerns and others (e.g., Kosslyn, 1999) about fMRI and its use in experiments are rightfully raised, fMRI still seems the most attractive option for observing activity in deeper structures during affective and cognitive processing given its relative availability, low invasiveness, and adequate spatial resolution (as compared to EEG).
The higher temporal resolution has been used in ACT-R to connect neural structures with functional systems specified in the ACT-R theory (Anderson, Fincham, Qin, & Stocco, 2008). This connection to actual neural structures makes ACT-R a very attractive architecture for any human-in-the loop simulations that include a measure from some physiological sensor. Though one may not be measuring the brain directly, understanding the process can lead one to a principled theory and model of behavior while using a system that provides a realistic and tractable account of both the performance and physiological data.