Perceptual decision-making, evaluating the sensory world in order to choose an action, is often done in conditions of uncertainty. People can implicitly learn from past experiences in order to help the decision-making process but the conditions in which this happens and how metacognition impacts this is unknown. Furthermore, how this ability to implicitly learn priors changes with healthy aging has not been studied. In these three studies, participants performed a perceptual decision-making task in which different colored stimuli were associated with different prior biases (e.g., 75% of the red trials go leftward). In Study 1, we examined the differences between learning priors implicitly through experience versus explicit instruction and how that affects performance and metacognition with two computational models, Linear Ballistic Accumulator Model (LBA) and a Hierarchical Bayesian Estimation of metacognition (H-Meta- d’). Participants were able to learn priors implicitly and used them to guide decision-making. Bias primarily influenced decisions with the least sensory information, but not for stimuli with more robust information. Those who were instructed of the priors were more confident for prior- consistent (vs. inconsistent) stimuli while this was not seen in participants who experienced the priors implicitly. However, there were no differences in metacognitive efficiency between the two instruction groups. Our results suggest that implicitly learned priors can influence decision-making when sensory information is unreliable, but do not contribute when sensory information is more robust. In Study 2, older and younger adults performed the same decision-making task with prior instruction manipulation but with different prior conditions. Instead of opposite-oriented priors, one prior was biased towards a side while the other prior was equally likely to go left or right. When participants were instructed of priors, younger adults were faster and more accurate compared to older adults. Younger adults were more confident for Positive prior (biased) trials while older adults’ confidence was unaffected by prior condition. In contrast, in the implicit “experience” group, younger and older adults largely matched on speed and accuracy, but younger adults were more confident than older adults overall. LBA parameter estimates largely align with past research that suggests that older adults have a slower information processing rate, greater response caution and require more evidence before making a decision (Garton et al., 2019). In Study 3, I present preliminary results of functional magnetic resonance imaging to examine the neural correlates of implicit learning and decision-making. Activation in the putamen and thalamus was observed during prior-consistent Equal prior trials with little sensory information. Motor and visual areas, as well as frontal gyri were primarily activated for the Positive Prior trials that were prior-consistent. Findings from these studies may augment understanding of the decision-making mechanisms in healthy aging as well as in clinical patient populations and may provide insight into novel therapies or rehabilitation.

Factors contributing to the formation of habits, defined as the stimulus-response associations that form the basis of much human and animal behavior, are not well understood, and although habits are believed to underlie many negative health behaviors such as addictions, the extent to which findings from animal research on habits apply in the human is largely unknown. In Study 1 (Chapter 2), we conducted two experiments on appetitive habit formation in adults with a history of early-life stress. Using the size of the partial reinforcement extinction effect as a measure of goal-directed versus habit behavior, we found evidence of increased habit behavior in people who reported a history of early-life stress, and this effect appeared to be enhanced by the presence of distraction. In Study 2 (Chapter 3), we conducted two experiments on avoidance habit formation in this population. People with a history of early-life stress exhibited enhanced avoidance habits as measured by persistence of learned behaviors after outcome devaluation. Finally, in Study 3 (Chapter 4), we conducted a meta-analysis of fMRI studies on human habit responding to assess the contributions of striatal subregions to habit behavior. We found that activation patterns varied based on the task that was used (probabilistic classification, maze navigation, outcome devaluation, sequential decision, or motor sequence learning), with differences observed along both the anterior-posterior and medial-lateral axes. Chapter 5 summarizes the findings and makes suggestions for future research.

A crucial goal of many real-world sequence learning tasks is the transfer of knowledge to novel sequences. The current work utilized the serial reaction time task, in which participants press keys in response to visual cues that appear in sequence. The first study examined the ability to transfer to new sequences after practicing sequences in a repetitive order compared to a non-repeating interleaved order. Interleaved practice resulted in better performance on repetitive or interleaved novel sequences than repetitive practice. Interleaved training may reduce interference from sequence-specific knowledge when the learner is faced with new variations of a task. In the second study, participants practiced interleaved sequences in a functional magnetic resonance imaging (fMRI) scanner and received a transfer test of novel sequences. Transfer ability was positively correlated with cerebellar blood oxygen level dependent (BOLD) activity during practice, indicating that greater engagement of the cerebellum during training resulted in better subsequent transfer performance. The cerebellum is thought to contribute to error-based learning through the instantiation of internal models (Wolpert, Miall, & Kawato, 1998). When novel variations of a task are encountered, it is hypothesized that relevant features from other previously-formed internal models are selected and form the basis for performance on the novel task (Imamizu, Higuchi, Toda, & Kawato, 2007). Those learners in which internal models are more readily formed may be better at later tests of transfer. Based on these results, the cerebellum was targeted with transcranial direct current stimulation (tDCS) during training in order to enhance transfer learning. Results indicate that anodal tDCS can lead to better transfer to novel sequences, but this change was apparent only after a delay. On the other hand, cathodal tDCS had a detrimental online effect on learning. These results are consistent with the theory that the cerebellum plays a large role particularly in the early learning stage of fine motor skills. Overall, this work demonstrates that easily implementable and relatively inexpensive manipulations can improve transfer learning of motor sequences.

Evidence demonstrates that individuals with schizophrenia have impaired corticostriatal functioning. Non-psychotic relatives of patients with schizophrenia exhibit a subset of the deficits present in schizophrenia patients, suggesting that there are cognitive endophenotypes of this disorder and that these deficits appear to be associated with genetic liability for schizophrenia. In order to determine if a deficit in corticostriatal functioning is an endophenotype for schizophrenia, it is important to investigate the idea that genetic risk for schizophrenia is associated with impairment in striatal dependent tasks. Therefore, we will use both behavioral measures and fMRI to assess differences in performance level and pattern of brain activation associated with different striatal-dependent tasks to investigate the corticostriatal function of relatives of patients with schizophrenia. Both behavioral and fMRI studies will be conducted in healthy adolescent and adult relatives schizophrenic individuals and controls using cognitive and motor skill learning tasks. The behavioral studies will measure performance and function of two different corticostriatal loops and will investigate the developmental course for the behavior pattern of relatives of schizophrenia patients. The fMRI studies will expand the results of the behavioral studies and determine whether there is a difference in the pattern of brain activation between the adolescent relatives, adolescent controls, adult relatives, and adult controls. This research makes a novel contribution in that it demonstrates that individuals with genetic liability for schizophrenia exhibit deficits in corticostriatal function. Because these individuals do not have the disease, their performance gives insights into whether corticostriatal function is an endophonotype of schizophrenia, independent of the effects of medication and the illness itself and also leads to an increased understanding of the neural mechanisms of schizophrenia and advances the search for the treatment of this severe disorder.

Aging is associated with declines in memory performance and profound functional and structural changes in the brain, though older adult’s ability to selectively learn and retrieve valuable information appears to be well preserved. Prior research suggests that older adults differentially employ elaborative encoding strategies when encoding valuable items and selectively allocate attention to these items, as reviewed in Chapter 1. The underlying aims of the current study were to examine how this selective learning affects different aspects of memory and what cognitive and neural mechanisms may support older adult’s effective learning of valuable information.

Chapter 2 describes findings from a project aimed at determining how value-directed remembering affects different aspects of memory. This research suggests that selectively encoding valuable items is associated with greater recognition performance, higher confidence, and increased recollection in both younger and older adults. Although age-related declines in recollection are typically observed, value’s enhancement of item recollection appears to be intact in healthy aging. Selectively encoding valuable items also resulted in stronger binding between items and incidentally learned task-relevant details, perhaps at the expense of encoding extraneous details. Older adults did not demonstrate greater item-location memory for valuable items, suggesting that value’s influence on associative binding may be diminished in aging.

Chapters 3 and 4 describe two projects designed to investigate how strategic and automatic processes contribute to value-directed remembering, and whether differences in the structural integrity of white matter tracts predicts differences in older adult’s ability to selectively learn based on value. The findings of these two studies suggest that younger adults are relying on a more automatic reward circuit involving the nucleus accumbens and ventral tegmental area, whereas older adults rely on semantic processing in the inferior fronto-occipital fasciculus (IFOF). Older adults may be relying on the IFOF in a compensatory manner. The implications of these findings for theories of aging and learning are discussed in Chapter 5.

Recall performance is greatly affected when older adults are presented with stereotype threat regarding memory (See Lamont et al., 2015 for a review). Stereotype threat is a concern that one’s performance will confirm negative stereotypes about one’s group. In Study 1 (Chapter 2), we examined the effect of threat on metamemory processes and memory selectivity for high value information. Our findings showed that threat affected calibration (i.e., bets-recall, p=.045) and total score (p=.03), both which require metacognitive control. Our threat manipulation did not affect value. That is, both groups placed bets and recalled more high-value words than low and medium value words (p>.05). Metacognition goes hand in hand with executive control, thus it is possible that threat burdened both cognitive processes. Moreover, the effect of value was stronger than that of threat.

In Study 2 (Chapter 3), we were interested in testing the effect of threat when the manipulation was done after encoding (i.e., prior to retrieval). We found no differences across groups in free recall or cued recall with this manipulation. However, we cannot refute the plausible effect of threat on retrieval given differences in experimental conditions (i.e., encoding time) between the current study and previous research. Taken together, our results provide partial support for executive function (See Chapter 2) as a possible mechanism of stereotype threat’s effect on memory performance.

We were interested in contributing to the growing body of literature on the cognitive mechanisms that underlie this disruption in performance. Study 3 (Chapter 4) aimed to assess the effect of different types of threat on executive function. Older adults were assigned to one of three conditions: a neutral, memory threat, or processing speed threat condition. Participants completed a task-switching paradigm, in which global, local, and alternating switch costs were examined. Reaction time and accuracy were used as dependent measures. Overall, participants displayed a pattern of performance that is consistent with the task switching literature. That is, older adults showed global and local switch costs (e.g., Mayr, 2001). Participants in the memory threat condition did not differ greatly from those in the neutral condition, while participants in the processing speed threat condition were significantly faster than the other two groups, (p=.03). We did not observe an interaction between trial type and group for local switch costs (p>.05). Group differences only emerged for global switch trials. It is possible that our processing speed threat manipulation may have prompted a reminder about the objective of the task. Taken together, incorporating value-based tasks into neuropsychological assessments would provide an improved objective measure of memory performance. Also, as suggested by Study 3 deemphasizing memory prior to a task of executive function may improve performance.