There is precedent in the human cognitive neuroscience literature suggesting noisier representations in aging underlie performance decrements. Using fMRI, Samanez-Larkin et al. revealed that in older adults, greater temporal variability in nucleus accumbens BOLD response was directly related with suboptimal performance on a financial risk-taking task. Variability has been shown to be reproducible within subject across time and tasks , and has been linked to individual differences in dopamine . In aging, declines in D1 receptors have been associated with increased intraindividual variability in reaction times during an executive function task . Generally, there is consensus that intraindividual variability in neural and behavioral responses is a valuable measure for understanding the underlying neural basis of impaired performance , though it is important to note that in some cases neural variability may be beneficial . One central component of this line of research will be to distinguish between variability that is driven by imprecision in signaling versus variability that simply reflects a healthy dynamic range in neural responses that are not muted by, for example, disease processes. Important research by Garrett, Kovacevic, McIntosh, and Grady has demonstrated that, overall, procona buckets older adults show less variability in BOLD signal than younger adults, and that this reduced variability is associated with poorer cognitive performance.
Interestingly, they found subcortical structures including hippocampus and regions in the striatum that were more variable in older adults than young adults . Together, these findings suggest variability in subcortical responses may be an important, age-sensitive measure in human imaging studies that warrants further investigation to establish possible relationships with dopamine. For example, future studies could test whether individuals with highly variable cortical responses also have highly variable subcortical response, or whether these are dissociable measures potentially reflecting different underlying sources.PET imaging provides spatial information that allows for the assessment of regional differences in dopamine function. This provides unique opportunities to test hypotheses about the differential influence of region-specific measures of dopamine on discrete cognitive operations. For example, PET can be useful for specifying a role of PFC dopamine signaling in decision-making. Spatial information may be particularly relevant for studying dopaminergic mechanisms of decision-making in aging as there is growing evidence that dopamine receptor densities may decline at different rates across the brain . Here, we describe ways in which the spatial information afforded by PET imaging can be leveraged to probe the role of dopamine in agerelated changes in reliance on striatal versus extrastriatal brain regions during decision making. Evolving research has developed our understanding of complementary learning mechanisms that may trade off or interact with model-free reinforcement learning processes to affect decision-making.
These include, but are not limited to, processes for the building of deliberative internal models to guide choices , as well as processes for learning and planning that more heavily rely on prefrontal processes including working memory or medial temporal lobe episodic memory . The degree to which these processes interact, and the nature of these interactions is an area of active research . However, the multiprocess view of decision-making incorporates roles for multiple neural systems that include frontoparietal, medial temporal lobe, and limbic structures. There is general agreement that most value-based decision-making tasks accommodate multiple strategies. In some cases, these strategies can be distinguished from one another using computational modeling approaches . These lines of research have revealed profound individual differences in the extent to which people adopt one strategy over another. Age differences in the adoption of task strategy has been identified as a critical factor in considering discrepant findings in the aging literature. For example, older adults may show better or worse performance than young adults depending on whether a given task favors win-stay lose-shift strategies . Research examining how reliance on specific strategies is affected by regional differences in dopamine function is only beginning, but holds promise for informing findings in aging. Influential learning models have distinguished between model-free processes and model based processes, which can be dissociated from one another computationally using the two state Markov decision task . Briefly, this task involves two decision phases . In the first step, participants choose between two stimuli, and this selection determines a second set of choices with differing reward probabilities. Similar to single phase tasks, learning can occur slowly and incrementally via model-free mechanisms. Task performance may also rely on goal-directed, model-based strategies, for which subjects develop an internal model of the task structure. At the first decision phase, subjects may prospectively consider future reward probabilities that would occur after the second phase.
This strategy is more computationally demanding than model-free strategies, but supports rapid and flexible learning. While links between model-free reinforcement learning and striatal dopamine have been long established, model-based strategies are also powerfully modulated by dopamine. Pharmacological manipulations provide general evidence that elevating dopamine tone increases reliance on model-based processes, though do not give information about the spatial specificity of dopamine’s role. Enhancing dopamine in young adults shifts bias in task strategy toward model-based processes . Consistent with this, Parkinson’s disease patients tested off medication show selective impairment in model-based learning that is remediated by dopaminergic medication . Complementing these findings in Parkinson’s patients, model-based processing is reduced in people with disorders characterized by alteration in dopamine function such as addiction , obsessive compulsive disorder , and schizophrenia . To date, there has been only limited investigation of the spatial specificity of dopamine’s influence on model-based learning. Dopamine may modulate PFC via direct inputs from the ventral tegmental area or by indirect effects in striatum where dopamine affects PFC function via fronto-striato-thalamic loops . Pointing to a role of striatal dopamine, PET evidence in healthy young subjects demonstrated greater striatal dopamine synthesis capacity was associated with preferential reliance on model-based learning and was correlated with fMRI activation in PFC . Pointing to a role of PFC dopamine, people with genetically inferred reductions in COMT enzymatic activity show greater reliance on model-based processes . Additional research is needed to understand how direct dopamine signaling in PFC may influence model-based strategies.Though model-based processing has been linked to striatal activity , it is also associated with increased reliance on prefrontal systems. Supporting this view, higher working memory capacity is associated with greater propensity to engage model-based strategies . Further, manipulations that increase cognitive load , stress , or perturb PFC function via rTMS reduce reliance on model-based processing, particularly in participants with low working memory capacity. In aging, there is evidence of reduced reliance on model-based processing . These findings are generally consistent with age-related changes in PFC function though, somewhat perplexingly, propensity to adopt model-based rather than model-free appears to be dissociated from working memory capacity . Other studies have identified subgroups of older adults demonstrating an over reliance on model-free learning . However, the extent to which age-related shifts from model-based to model-free learning mechanisms are explained by alterations in PFC dopamine function is unknown. Accounting for alteration in dopamine function in PFC and striatum may be valuable for understanding the neural basis of shifts away from model-based processes in aging. In aging, loss of PFC D2/3 dopamine receptors may outpace losses in ventral striatum . Further, rates of D1 receptor losses may differ between nigrostriatal pathways that innervate dorsal striatum versus mesocortical/mesolimbic pathways that innervate PFC and ventral aspects of striatum, respectively . Together, such changes may alter the weighting of PFC versus striatal influences on task performance and the coherence of dorsal versus ventral striatal contributions to processes for learning, procona florida container updating and integration that support value-based decision-making. Future studies pairing task performance with neurochemical PET measures could directly test how individual differences in receptor densities and dopamine release in PFC versus ventral striatum influence variability on the adoption of model-based versus model-free strategies in aging.
Such studies may identify those subgroups of older adults most reliant on model-free processes are those with most marked reductions in PFC dopamine measures. Such lines of research are ripe for cross-species comparisons which could assess relationships between age and performance while providing critical information as to the temporal dynamics of dopamine signaling in PFC versus striatum using microdialysis techniques and voltammetry. Together, such studies hold promise for promoting our basic understanding of PFC dopamine’s involvement in goal-directed decision-making while characterizing age differences in the recruitment of distinct dopamine pathways underlying specific task strategies.Models of complex decision-making support the view that multiple systems for learning, memory, and attention interact to shape our choices. We have argued that in vivo dopamine imaging is a powerful tool for understanding the neural basis of individual differences in performance in aging. This line of research can illuminate the extent to which neurochemical traits affect decision-making. However, it is important to also consider how neurochemical influences on behavior interact with other attentional or state measures to affect decision making. A recent study by Kircanski et al. illustrates the mutual influence of these factors on value-based choices. First, they modified the monetary incentive delay task by including trials in which participants unexpectedly gained or lost relatively large amounts of money . They found monetary loss and gain modifications induced negative and positive affective arousal respectively, which they assessed with self-report rating scales. Following the affective arousal manipulation, participants performed a separate decision-making task in which they chose whether or not to purchase items with misleading advertisements. They found deficits in choice performance following both arousal conditions relative to a neutral condition. Together, these findings demonstrate the capacity for financial reward and punishment manipulations to impact self-reported affective state, which in turn modulates the quality of subsequent value-based decisions. This study found no evidence for differences in the induction of arousal between young and older adults or in the detrimental effect of arousal on later choice performance. Other studies have shown differential age effects such that following positive affect induction , older adults, but not young adults, are more likely to make risky decisions than when in neutral states . In the following section, we describe possible systematic changes in affective attention over the life span that may influence decision-making performance. We consider how these attentional changes may interact with influences of dopamine on performance, and suggest strategies for empirically examining these interactions. Older adults report reduced levels of negative affective experience as they age, but preserved levels of positive experience . These observations have been linked specifically to memory and attention where older adults are more likely to remember positive events than young adults . It is possible that there are age-related changes in neurophysiology and neurochemistry that drive these effects, though there has been little direct investigation of this possibility model first described by Mather, Clewett, Sakaki, & Harley, 2016). Functionally, there appears to be relative preservation of the networks supporting emotional processing in aging , though in some cases the engagement of these networks by older adults may be more effortful . The dominant interpretation is that at the end of life there is a shift in motivational goals derived from changes in the perception of time horizons . This view is supported by evidence that positivity effects have been reported in young people diagnosed with terminal illness, and people on death row, where systematic neurophysiologic and neurochemical changes mirroring those that occur in aging are unlikely . Further, positivity can be enhanced in young subjects experimentally in task scenarios in which they are encouraged to think about a limited future . The positivity effect has been implicated in fMRI findings that responses associated with reward anticipation are intact in aging, but that responses associated with the anticipation of monetary losses in the insula are muted in older adults . Other studies have examined relationships between subjective ratings of emotional stimuli and fMRI responses in PFC and amygdala. Behaviorally, older adults report lower arousal for negative stimuli than young adults , report less unpleasantness for “lowarousal” negative stimuli, and greater pleasantness for “low-arousal” positive stimuli in a study that manipulated the levels of stimulus valence . These behavioral differences are accompanied by age-group differences in amygdala responsivity. Older adults consistently show greater enhancement of amygdala activation for positive relative to negative stimuli . Further, there is evidence that the relative suppression of amygdala activation for negatively valenced stimuli in aging is associated with greater recruitment of rostral anterior cingulate cortex when viewing unpleasant stimuli .