Methodological Concerns

First, most neuroimaging studies use a contrast or subtraction method (see Sartori and Umilta, 2000), which critically relies on assumptions of linearity and additivity as well as on the critical choice of comparison control states. Experimental models in vogue in cognitive neuroscience have been applied to the study of emotion, but it is debatable whether this is legitimate, considering certain fundamental differences in the nature of cognitive and emotional processes (Panksepp, 1998a, 2000, 2003). Emotional affects impose pervasive influences on cognitions, in part as a consequence of different arousal states generated subcortically and broadcast widely via a variety of "state-control" pathways (e.g., acetylcholine, dopamine, norepinephrine, serotonin). To understand emotions, large-scale network properties and organic controls (neurohormonal effects) are as important as the flow of information through discrete neuronal pathways. There is no assurance that such brain changes are well reflected in blood flow dynamics. In contrast, cognitive processes may rely on more modular cortical representations that arise from more discrete informational "channel-control" functions of the brain. These brain changes may be more apparent under imaging procedures.

It is important to note that the statistical analysis of fMRI images using the most common method [statistical parametric mapping (SPM); Hammersmith, London] applies a general linear algorithm that assumes a typical time course of the hemo-dynamic function (peak at 5 sec, decay by 12 sec), but it has been shown that the time course of signal changes associated to emotional states generally starts later and has more protracted effects (over 30 sec) (Marcus Raichle, personal communications). In summary, methods perfected on the study of cognitive processes may not transfer automatically to the study of affective states and emotion/cognition interactions.

The choice of a control state is also of critical importance in evaluating most published functional imaging studies. Several studies have used as a baseline state an "eyes-closed rest" (ECR) condition. Others have adhered to the additive method tradition by proposing the use of a higher-level control state (e.g., nonemotional imagery) that matches more closely the active psychological condition (e.g., emotional imagery), thereby controlling, for example, basic sensory, motor, and attentional components of a complex psychological process. Use of the ECR condition has been criticized on the basis that the individual's mental state is typically not monitored, with random thoughts and feelings producing background "noise" (Andreasen et al., 1995). Greater interpretive leeway would emerge if all studies employed at least two control conditions (one being ECR and another a carefully selected "active" control), but that remains rare in the field.

Another limitation of early neuroimaging studies of emotion was the tendency to focus on selected anatomical regions of interest while ignoring many other areas of the brain. More recent studies tend to report more effects, due to the overall acceptance of a view of cognitive functions as represented in distributed networks of regions rather than discrete centers. Also, it used to be common to only report blood flow increases, while ignoring decreases or deactivations, which more recent studies highlight as important in emotional processing and emotion/cognition interactions (Drevets and Raichle, 1998; Mayberg et al., 1999).

Across existing studies, considerable variability in the regional findings derives from the variable choice of statistical thresholds employed in reporting effects (Liotti et al., 2000a). Another major source of variability arises from differences in data transformation steps and statistical processing software (e.g., AFNI [Analysis of Functional Neural Image], CDA [Change Distribution Analysis], etc.) used across laboratories. Most neuroimaging studies employ a considerable amount of filtering and smoothing of the data, resulting in large blobs with a final resolution of 12 to 20 mm (SPM, Hammersmith, London). While this may be an acceptable spatial resolution to localize cortical effects, it does not enable adequate identification of effects in smaller and more compact subcortical structures involved in triggering emotional reaction, especially critical hypothalamic, midbrain, and brainstem structures. Hence, many studies, especially those using fMRI, have found no participation of subcortical structures, long implicated by animal research, in the constitution of human emotions. Some PET and fMRI studies, particularly those with an a priori regional hypothesis, have employed less filtering of the image volumes, and thereby approach the actual resolution of the current state-of-the-art PET (6 to 7 mm) and MRI (2 to 3 mm) machines, which allows better localization of effects in subcortical structures (e.g., Liu et al., 2000; Tracey et al., 2002). The increasing use of high-field MRI scanners (3 to 7 T) and higher resolution PET cameras, combined with less filtering will improve spatial resolution of subcortical emotional effects.

Another important difficulty in the neuroimaging of human emotions arises from the wide discrepancy in experimental situations, mood induction paradigms, emotional tasks, and instructions employed. While some studies have been carefully designed to address a specific domain of emotional processing (emotion recognition, subjective feeling or affect, emotion expression), others have used a combination of these. Importantly, studies of affect (e.g., anxiety and sadness) have generally been biased toward cognitive processing of the emotion-inducing materials (perceptual, mnemonic, visual imagery tasks), resulting in widespread activations of cortical structures probably involved in the nonaffective components of the emotional task. Only a few studies have imaged the induced affect after the initial affect-induction phase, when the subjective experience had reached a desired intensity (see Damasio et al., 2000; Liotti et al., 2000a). A closely related problem reflects the timing of activation of critical brain structures participating in emotional processing. As an example, amygdala activations have been common in studies of facial expressions (particularly fear). However, it has been found that such activity: (1) habituates over multiple cycles of presentation (Breiter et al., 1996) and (2) shifts to a deactivation in the case of emotion categorization tasks (Hariri et al., 2000). In addition, (3) significant changes (increases or decreases) are typically absent in this brain territory in those studies that have taken pains to focus on the subjective affective feelings (Damasio et al., 2000; Liotti et al., 2000a). This suggests that the time courses of activity (both activation and deactivation) are critical variables, but all too commonly ignored, in studies of emotional processing.

Still, there is now an abundance of evidence that emotional stimuli can have regionally specific effects on the brain. Regional effects have been shown for both emotions generated externally by viewing films or emotional scenes, as well as internally generated, memory-driven emotional reminiscences. However, the two approaches often yield different results (Lane et al., 1997; Reiman et al., 1997; reviewed in Phan et al., 2002).

0 0

Post a comment