The superior colliculus (SC) is a laminated oculomotor structure in the midbrain. In non-human primates SC has long been known to contain a retinotopically-organized map of visual stimulation in its superficial layers, which is aligned to a map of saccadic eye movements in the deeper layers. Microstimulation and electrophysiology experiments have shown that SC also plays a key role in covert visuospatial attention and suggest that attentional modulation also occurs in a retinotopic manner. Retinotopic organization of the visual field can be non-invasively mapped in humans using functional MRI with a technique called phase-encoded retinotopy. In this technique, rotating wedges and expanding rings of visual stimuli are used to map the polar angle and eccentricity dimensions of a polar coordinates system, respectively. A similar technique can also be used to map spatial attention by keeping the visual stimulus constant and cueing subjects to attend to apertures of rotating wedges and expanding rings within the stimulus. A previous study using fMRI has shown the polar angle representation of visual stimulation in human SC but was unable to find a representation of eccentricity. This work uses high-resolution fMRI along with special surface analysis techniques developed in our lab to demonstrate maps of both polar angle and eccentricity for visual stimulation. Moreover, visual attention is also shown to be topographically organized within SC and in registration with visual stimulation. Finally, in human visual cortex, fMRI is known to show activity for sustained spatial attention even in the absence of a significant visual stimulus, an attentional "base response". In this work, SC is shown to exhibit a similar sustained attention base response using a threshold-contrast detection paradigm. This base response was compared with a response for attention with visual stimulation. The peak amplitude of the base response occurred more deeply within SC tissue than the peak for attention with stimulation. It is proposed that this reflects the specific attentional enhancement of the deeper visuomotor neurons, which are hypothesized to be a direct neuronal correlate of the oculomotor theory of attention. / text

<p>The brain’s dopaminergic system is critical to adaptive behaviors, and is centrally implicated in various pathologies. For decades, research has aimed at better characterizing what drives the mesolimbic dopamine system and the resulting influence on brain physiology and behavior in both humans and animals. To date, the dominant modes of research have relied on extrinsic approaches: pharmacological manipulations, direct brain stimulation, or delivering behavioral incentives in laboratory tasks. A critical open question concerns whether individuals can modulate activation within this system volitionally. That is, can individuals use self-generated thoughts and imagery to invigorate this system on their own? This process can be referred to as “cognitive neurostimulation” -- a precise and non-invasive stimulation of neural systems via cognitive and behavioral strategies. And if not, can they be taught to do so? Recent technological advances make it feasible to present human participants with information about ongoing neural activations in a fast and spatially precise manner. Such feedback signals might enable individuals to eventually learn to control neural systems via fine-tuning of behavioral strategies. The studies described herein investigate whether individuals can learn to volitionally invigorate activation within the mesolimbic reward network. We demonstrate that under the right training context, individuals can successfully learn to generate cognitive states that elicit and sustain activation in the ventral tegmental area (VTA), the source of dopamine production within the mesolimbic network. Although participants were explicitly trained to increase VTA activation, multiple mesolimbic regions exhibited increased connectivity during and after training. Together, these findings suggest new frameworks for aligning psychological and biological perspectives, and for understanding and harnessing the power of neuromodulatory systems.</p> / Dissertation

<p>The ability to re-experience the past is a defining feature of episodic memory. Yet we know that even the most detailed memories are distinct from the initial experiences to which they refer. This relationship between the initial encoding and subsequent retrieval of information is central to our understanding of memory and its capacity to connect us to the past. Past research has shown that neural signatures present during perception are reactivated during later memory, but the correspondence between this reactivation and various aspects of memory function remains unclear. This dissertation attempts to connect behavioral measures of memory to the reinstatement and modification of neural information that takes place when memories are retrieved. In the first two studies reported, functional magnetic resonance imaging (fMRI) is used to assess event-specific cortical patterns from encoding that are reinstated during retrieval (encoding-retrieval similarity, ERS). Increases in this fine-grained of reinstatement are found in occipitotemporal cortex (OTC) during detailed memory for scenes (Study 1), and in the medial temporal lobes (MTL) for the recovery of relational information (Study 2). In addition to reflecting encoding-related content, retrieval is also found to strengthen previously encoded information via hippocampally-mediated mechanisms in Study 3. Together, these studies demonstrate the detailed nature of information that is recovered across varying degrees of memory and show how retrieval can also alter stored representations, emphasizing the interactive nature of memory processes.</p> / Dissertation

Successful recognition memory decisions depend on mnemonic and decision making processes that are computed by multiple, distributed brain areas. However, little is known about what computations these areas perform or how these areas are connected. Here, I collected behavioral and functional magnetic resonance imaging data from humans during the performance of an old-new recognition memory task with retrospective confidence judgments. Across runs, choice bias was successfully manipulated by providing rewards for correct responses that were either symmetric (equal reward for hits and correct rejections) or asymmetric (one response worth more than the other). Successful recognition memory was associated with activation in anterior prefrontal, parahippocampal, posterior cingulate, and parietal cortex. Resting state functional connectivity demonstrated that these brain areas are organized into two distinct networks. The first network includes parahippocampal cortex and angular gyrus. The second network includes lateral prefrontal cortex and intraparietal sulcus. The hippocampal-cortical network was most active during old vs. new decisions, did not differentiate hits from false alarms, and was differentially active during low confidence old and new judgments. In contrast, while the frontoparietal network was robustly activated by hits, it was not activated during either false alarms or low confidence old judgments. Thus, these two distinct networks can be distinguished by their relative connectivity to the medial temporal lobe vs. lateral prefrontal cortex and their responses during uncertain old judgments and errors. The choice bias manipulation had opposing effects on the parietal components of these networks, which further suggests these networks make distinct contributions to mnemonic decision making. / Psychology

Alcohol-induced blackouts, or periods of anterograde amnesia without loss of consciousness, were a diagnostic indicator in Jellinek’s (1952) theory of alcoholism and have been correlated with alcohol use problems (Campbell & Hodgins, 1993; Goodwin, Crane, & Guze, 1969; Ryback, 1970; Tarter & Schneider, 1976). Other findings suggest that blackouts are a warning sign of problem drinking, but not a predictor of alcohol use disorders (Anthenelli, Klein, Tsuang, Smith, & Schuckit, 1994). Most published research on blackouts focuses on cognitive deficits among older alcohol-dependent adults, yet recent research indicates prevalence rates for blackouts as high as 50% among college students (White, Jamieson-Drake, & Swartzwelder, 2002). In addition, young adults who reported experiencing a blackout were later told that they had vandalized property, driven a car, or engaged in other risky behaviors without remembering (Buelow & Koeppel, 1995). Despite their high prevalence and associated negative consequences, relatively little is known about alcohol-induced blackouts or their neural, social, and behavioral correlates among non-dependent populations. The current research explored individual variation in memory functioning under sober and intoxicated conditions and alcohol’s effects on neural activation during memory processes. / text

A substantial body of previous research suggests that the medial prefrontal cortex (MPFC) plays an important role in multiple aspects of emotion. These aspects include, but are not limited to, (1) generating, (2) experiencing, and (3) regulating one's own emotional state, as well as (4) facilitating the use of emotion-related information within goal-directed cognition and action selection. However, there is considerable controversy with regard to the distinct functional roles of various MPFC subregions. In this dissertation, I first provide a review of the theoretical and experimental literature to date in order to defend a plausible model of the hierarchical neural processes associated with each of the aspects of emotion highlighted above. This model proposes that different MPFC subregions each play distinct, but interactive, roles at or near the top of the respective hierarchical systems associated with those aspects of emotion. After reviewing this model, I then provide a description of four experiments that test the predictions of this model's claims regarding the roles of three distinct MPFC subregions: the rostral anterior cingulate cortex (rACC), the dorsomedial prefrontal cortex (DMPFC), and the ventromedial prefrontal cortex (VMPFC). These experiments provide independent support for the claims that (1) rACC plays an important role in representing the conceptual meaning of one's felt emotional reactions, (2) DMPFC plays an important role in maintaining representations of one's own emotions within a consciously accessible state, and (3) VMPFC plays an important role in both appraising the emotional significance of one's current situation and triggering the somatic/visceral reactions associated with the generation of an emotional response. In the concluding section of the dissertation, I then integrate these findings together with the larger model and discuss important directions for future research as well as ways in which the model might be extended to include insights from recent advances in theoretical neuroscience associated with predictive coding.

This thesis presents an evaluation of algorithms for classification of functional MRI data. We evaluated the performance of probabilistic classifiers that use a Gaussian model against a popular non-probabilistic classifier (support vector machine, SVM). A pool of classifiers consisting of linear and quadratic discriminants, linear and non-linear Gaussian Naive Bayes (GNB) classifiers, and linear SVM, was evaluated on several sets of real and simulated fMRI data. Performance was measured using two complimentary metrics: accuracy of classification of fMRI volumes within a subject, and reproducibility of within-subject spatial maps; both metrics were computed using split-half resampling. Regularization parameters of multivariate methods were tuned to optimize the out-of-sample classification and/or within-subject map reproducibility. SVM showed no advantage in classification accuracy over Gaussian classifiers. Performance of SVM was matched by linear discriminant, and at times outperformed by quadratic discriminant or nonlinear GNB. Among all tested methods, linear and quadratic discriminants regularized with principal components analysis (PCA) produced spatial maps with highest within-subject reproducibility. We also demonstrated that the number of principal components that optimizes the performance of linear / quadratic discriminants is sensitive to the mean magnitude, variability and connectivity of simulated active signal. In real fMRI data, this number is correlated with behavioural measures of post-stroke recovery , and, in a separate study, with behavioural measures of self-control. Using the data from a study of cognitive aspects of aging, we accurately predicted the age group of the subject from within-subject spatial maps created by our pool of classifiers. We examined the cortical areas that showed difference in recruitment in young versus older subjects; this difference was demonstrated to be primarily driven by more prominent recruitment of task-positive network in older subjects. We conclude that linear and quadratic discriminants with PCA regularization are well-suited for fMRI data classification, particularly for within-subject analysis.

The feasibility of using real-time functional magnetic resonance imaging (fMRI) feedback regarding the level of activation in rostromedial prefrontal cortex (rMPFC) to learn improved regulation of this brain area was examined in a group of 5 young adults. Subjects received real-time feedback from the target brain region while engaging in a blocked-design task involving alternating blocks of attempted up-regulation and down-regulation of the target brain region. A transient negative emotional state was induced prior to each scanning session. Subjects completed 6 scanning sessions (a pre-training session, 4 feedback sessions and a post-training session - no feedback was provided for pre and post-training sessions). The guideline strategy provided to subjects of engaging in emotional awareness during up-regulation and bodily awareness during down-regulation was found to consistently regulate the region in the pre-training session prior to the fMRI feedback sessions. This finding is in line with the previously proposed role of the rMPFC in emotional awareness. In contrast to previous real-time fMRI findings, greater recruitment of the region was observed in the pre-training session compared to the post-training session, with a non-significant negative trend observed across feedback sessions. These results suggest that there may be limitations to which the feedback techniques successfully employed for other brain regions extend to yet unexplored brain regions.

Investigating the Neural Correlates of Crossmodal Facilitation as a Result of Attentional Cueing: An Event-Related fMRI Study. Degree of Masters of Science, 2008 Zainab Fatima Institute of Medical Science, University of Toronto ABSTRACT Attentional cueing modulated neural processes differently depending on input modality. I used event-related fMRI to investigate how auditory and visual cues affected reaction times to auditory and visual targets. Behavioural results showed that responses were faster when: cues appeared first compared to targets and cues were auditory versus visual. The first result was supported by an increase in BOLD percent signal change in sensory cortices upon cue but not target presentation. Task-related activation patterns showed that the auditory cue activated auditory and visual cortices while the visual cue activated the visual cortices and the fronto-polar cortex. Next, I computed brain-behaviour correlations for both cue types which revealed that the auditory cue recruited medial visual areas and a fronto-parietal attentional network to mediate behaviour while the visual cue engaged a posterior network composed of lateral visual areas and subcortical structures. The results suggest that crossmodal facilitation occurs via independent neural pathways depending on cue modality.

This thesis presents an evaluation of algorithms for classification of functional MRI data. We evaluated the performance of probabilistic classifiers that use a Gaussian model against a popular non-probabilistic classifier (support vector machine, SVM). A pool of classifiers consisting of linear and quadratic discriminants, linear and non-linear Gaussian Naive Bayes (GNB) classifiers, and linear SVM, was evaluated on several sets of real and simulated fMRI data. Performance was measured using two complimentary metrics: accuracy of classification of fMRI volumes within a subject, and reproducibility of within-subject spatial maps; both metrics were computed using split-half resampling. Regularization parameters of multivariate methods were tuned to optimize the out-of-sample classification and/or within-subject map reproducibility. SVM showed no advantage in classification accuracy over Gaussian classifiers. Performance of SVM was matched by linear discriminant, and at times outperformed by quadratic discriminant or nonlinear GNB. Among all tested methods, linear and quadratic discriminants regularized with principal components analysis (PCA) produced spatial maps with highest within-subject reproducibility. We also demonstrated that the number of principal components that optimizes the performance of linear / quadratic discriminants is sensitive to the mean magnitude, variability and connectivity of simulated active signal. In real fMRI data, this number is correlated with behavioural measures of post-stroke recovery , and, in a separate study, with behavioural measures of self-control. Using the data from a study of cognitive aspects of aging, we accurately predicted the age group of the subject from within-subject spatial maps created by our pool of classifiers. We examined the cortical areas that showed difference in recruitment in young versus older subjects; this difference was demonstrated to be primarily driven by more prominent recruitment of task-positive network in older subjects. We conclude that linear and quadratic discriminants with PCA regularization are well-suited for fMRI data classification, particularly for within-subject analysis.