In order to address the above three questions Chapters 2 and 3 of my dissertation contain experiments that employ a category learning paradigm to study the influence of mood (positive and negative), and depressive symptoms on cognitive flexibility and hypothesis- testing abilities. In the past two decades a large amount of research has been conducted with well-defined category sets that can be classified as either rule-based (RB) or non- rule-based (NRB). This research has made it possible to create category sets that require or benefit from certain processes and not others. Category learning also provides a theory of category learning that can help to motivate mood research and make predictions about the findings. The COmpetition between Verbal and Implicit Systems (COVIS) theory of category learning (Ashby, Alfonso-Reese, Turken, & Waldron, 1998) is based on
neuropsychological research and argues that there are separate category learning systems, one that is verbal, and one that is nonverbal. Table 1.1 below shows how the learning tasks used in this dissertation are theoretically related to the COVIS systems, as well as the key attributes of each type of learning task.
Table 1.1. Category learning terminology and theoretical structure
COVIS System Verbal system Nonverbal system
Learning Tasks Rule-based (RB) Non-rule-based (NRB) Key Attributes
of the learning tasks
Verbalizable;
taps cognitive flexibility, rule-selection, hypothesis- testing;
requires the separation of stimulus dimensions.
Nonverbalizable;
does not tap cognitive flexibility, hypothesis-testing;
requires the integration of stimulus dimensions.
The verbal system in COVIS is mediated by a circuit that goes from the anterior cingulate cortex (ACC) and the prefrontal cortex (PFC) to the head of the caudate nucleus, and then goes back to the PFC. This model implicates the PFC for working memory and explicit reasoning, the ACC for rule selection, and the basal ganglia for rule switching. Working memory and executive attention are crucial cognitive components of the verbal system. Category sets that are RB are learned best by the verbal system because they draw upon explicit reasoning, working memory, rule selection, and rule-switching. One key feature
of RB category sets is that the optimal rule that distinguishes category exemplars between the categories is verbalizable. RB category sets can differ in terms of how complex the verbal rule is, from simple unidimensional rules, to multi-dimensional rules that require conjunctive or disjunctive reasoning. Unidimensional verbal rules are the most salient and easy to learn, and as complexity of the rule increases beyond conjunctions the category set may still be able to be described with a verbalizable rule, but it is unclear whether participants will rely on explicit reasoning processes and verbal rules to
categorize stimuli (Ashby & Maddox, 2005). For participants to achieve high or perfect accuracy on a RB category learning task, participants must select and test rules held in working memory while they are being tested. When a rule is tested and found not to lead to correct feedback, the rule should be abandoned so that a new rule can be tested. The use of a new rule requires one to select a new possible rule, and then turn one’s attention away from the old rule to the new rule. More recently the medial temporal lobes,
specifically the hippocampus, were also added to the COVIS model’s verbal system, because while rules can be held in working memory for short periods of time, rules must also be consolidated and stored in longer-term memory (Ashby & Valentin, 2005).
According to Ashby et al. (1998) the verbal category learning system is most likely to be influenced by mood, depressive symptoms, and motivation, as will be discussed later on in this introduction. This is because mood, depressive symptoms, and motivational states have been found to influence the same cognitive processes that are used in rule-based category learning tasks.
The nonverbal system is mediated by the basal ganglia, with a focus on the striatum due to its role in dopamine reward-based feedback and because most of the visual cortex extends directly onto the tail of the caudate nucleus (Ashby et al., 1998). Within the striatum the putamen is closely connected with motor output and the caudate nucleus is associated with cognitive behavior. Ashby et al. (1998) hypothesize that for every categorization decision made a unit in the prefrontal cortex is activated by the striatum and the strength of activation is indicative of the nonverbal system’s confidence that the response was correct. With repeated trials and activations the category responses become associated with the stimuli. Feedback is immediate and the dopamine released into the
striatum causes synapses that have been activated recently to be strengthened (Ashby & Maddox, 2005). The integration of relevant stimulus information prior to making a categorization decision is crucial for successful nonverbal learning to occur. This integration of information is nearly impossible to describe accurately using a verbal rule and the learning is thought to be procedural in nature (Ashby, El, & Waldron, 2003; Maddox, Bohil, & Ing, 2004). COVIS predicts that early learning relies on the verbal system, but if the category structure is not verbalizable, responding eventually switches to the nonverbal system (Ashby et al., 1998).
COVIS specifies that both systems are active and in competition with one another during learning. Eventually one system will come to dominate the response output in a particular category-learning task. The system that comes to dominate responding is the one that receives the most correct feedback. The verbal system is hypothesized to dominate all early responding while the nonverbal system takes longer to learn categories and thus longer to dominate responding.
1.4.1
Rationale for studying depressive symptoms and category
learning
As noted above, the COVIS model’s verbal system involves a circuit that extends from the PFC and ACC, the head of the caudate nucleus, and goes back to the PFC (Ashby et al., 1998). Depression represents an example of a condition that is associated with reduced activity in frontal brain regions that overlap with the verbal category learning system (Henriques & Davidson, 1991), and consequently an opportunity to evaluate the COVIS theory of category learning as well as theories of how depression influences cognitive processing.
Ashby et al. (1998) hypothesized that depression should be related to impaired RB and possibly NRB category learning. The possibility of NRB impairment is due to the
psychomotor agitation/disturbance found in some depressed patients (APA, 2011). Motor disturbances are mediated by the basal ganglia and thus the nonverbal system may be affected in some cases (Ashby et al., 1998). Only one set of studies has examined the influence of depressive symptoms on category learning. Smith et al. (1993) reported that
participants higher in self-reported depressive symptoms performed more poorly on a RB category learning task than participants lower in depressive symptoms. Depressive symptoms were not related to NRB category learning.
Chapter 2 addresses the potential relationships between self-reported depressive symptoms, hypomanic and worry symptoms, trait approach and avoidance motivation, and RB and NRB category learning performance in two studies. While Smith et al. (1993) studied the relationship between depressive symptoms and category learning, no prior research (to the best of my knowledge), has explored factors other than depressive symptoms (e.g. hypomanic and worry symptoms, approach and avoidance motivation) in this context.
Experiment 1 looks at the relationship between depressive symptoms and category learning, and is the first study to examine this relationship using a complex, RB categorization task, and the first to explore potential relationships between hypomanic symptoms, approach and avoidance motivation and category learning. This experiment’s original contribution to knowledge is that depressive symptoms are negatively related to a complex, RB learning task but unrelated to an easier, RB learning task or a NRB learning task.
Experiment 2 uses more complex category sets to further study the potential relationships between these variables and RB and NRB category learning performance. This
experiment’s original contribution to knowledge is that depressive symptoms are
negatively related to performance on a complex, RB category learning task that benefits from cognitive flexibility, but unrelated to performance on a NRB category learning task that does not benefit from cognitive flexibility. Further, this study demonstrates that history of hypomanic symptoms is negatively related to RB category learning. This experiment also makes a further unique contribution to knowledge that is described below.
1.4.2
Rationale for studying positive and negative mood and
category learning
Ashby, Isen, and Turken’s (1999) dopamine hypothesis of positive mood posits that mild to moderate positive mood is associated with enhanced cognitive flexibility due to an increase of dopamine in the frontal cortical areas of the brain, specifically the PFC and the ACC. These brain regions overlap with the COVIS model of category learning’s verbal system. Category learning involves a combination of rule learning, selective attention, inhibition of response, sensitivity to feedback, and often requires the learner to engage in hypothesis-testing, making it well suited to the study of mood and cognitive processing. Many of these functions require cognitive control and flexibility (Markman, Maddox, & Baldwin, 2007). Furthermore, there is a strong body of evidence implicating working memory and executive function in learning RB categories (see DeCaro, Thomas, & Beilock, 2008; Grimm, Markman, Maddox, & Baldwin, 2008; Maddox, Baldwin, & Markman, 2006; Minda, Desroches, & Church, 2008). No prior research has looked at the relationship between positive and negative mood states on category learning
performance, or the influence of positive and negative mood states on category learning performance.
Chapter 2’s Experiment 2 is the first study to look at the relationship between current self-reported mood state and category learning performance. This chapter’s original contribution to knowledge is that self-reported positive mood is positively related to RB category learning, but unrelated to NRB category learning. Further, this study shows that when positive mood is taken into account, the relationship between self-reported
depressive symptoms and RB category learning performance is no longer significant.
Chapter 3 looks at the influence of manipulated positive, neutral, and negative mood states on RB and NRB category learning performance, and is the first experiment to study the influence of manipulated mood states on category learning performance. This
chapter’s original contribution to knowledge is that positive mood enhances performance on a RB category set that benefits from cognitive flexibility. This has implications for category learning research and for research on positive mood.