Striatal activity during episodic memory retrieval is a robust phenomenon observed across various stimulus types and experimental paradigms. Previous research suggests that the ventral striatum and the caudate contribute differentially to retrieval processes: ventral striatal activity is posited to vary with the intrinsic reward of successful retrieval, whereas caudate activity is attributed to effortful cognitive control. The present fMRI study aimed to test the overarching hypothesis that these striatal subregions make distinct contributions to episodic retrieval and to further specify their functional roles. To this end, we employed a word–picture paired-associates learning paradigm in which words served as cues in two consecutive cued-recall tests that were separated by 24 h. Participants indicated the specificity of their associative memories, enabling characterization of striatal response profiles as a function of associative memory specificity and treatment of memory specificity as a proxy for the intrinsic reward associated with retrieval success. Rather than being not readily compatible with a role of the caudate in effortful post-retrieval control, the results suggest that elevated caudate activity during successful recall reflects pre-retrieval control or involuntary orienting to salient mnemonic information. In contrast, rather than reflect intrinsic reward itself, ventral striatal activity varied with the discrepancy between expected and actual intrinsic reward elicited by retrieval outcomes—that is, a metacognitive reward prediction error. Moreover, ventral striatum signaled the relevance of retrieved information and modulated its re-encoding to facilitate future retrievals.
Numeracy, the ability to understand and manipulate numbers, is a crucial foundational ability for many daily life skills. Numeracy tasks have been categorized using different taxonomic approaches, and understanding of the relationship between tasks categories and neural models of numeracy remains poor. In this study, we use the lesion method to address this key gap. The study employs a data-driven approach with support vector regression lesion-symptom mapping (SVR-LSM) of extracted factors from two left hemisphere chronic stroke survivor datasets: one with data from a battery of math and language tests (N = 100) and the other with data from the EC301 math battery (N = 64). We find that our data best support a separation of number tasks requiring verbal vs. magnitude number knowledge, aligning with a broad theme that emerges across theories in different variations. This is supported by distinct neuroanatomical correlates that link verbal coding of number to the language-centered regions including left perisylvian area and insula, and that link magnitude coding to the parietal lobe, as described in prior models. Our findings indicate that these two different sources of representation variously contribute to a range of math tasks, including different types of arithmetic calculation tasks, in ways which cannot be cleanly defined by one singular theoretical model of math cognition.
Comprehension for spoken sentences and discourse is differentially impacted by left- (LH) and right-hemisphere (RH) brain damage, such as that caused by stroke. Comprehension deficits can have disabling effects on one's ability to process information, participate in everyday life, and create and maintain relationships. This cross-sectional, observational study evaluated differences in sentence- and discourse-level comprehension between LH and RH stroke survivors across different prosodic conditions and examined relationships between comprehension and damage to key LH and RH regions. Participants included 43 LH and 46 RH acute stroke survivors and 39 controls. All participants completed an in-house clinical assessment that included short passages presented in three conditions: neutral prosody, affective prosody, and affective prosody plus facial expression cues. Stroke survivors also completed magnetic resonance imaging. Consistent with expectations, stroke participants achieved lower comprehension accuracy than controls across conditions. LH participants' accuracy in the ‘affective prosody plus facial expressions’ condition was greater than in the neutral prosody condition, and total LH lesion volume was associated with performance. RH participants' accuracy in the ‘affective prosody’ and ‘affective prosody plus facial expressions’ conditions was greater than in the neutral prosody condition. Overall, both groups' scores increased linearly from Condition 1 to 3. None of the prespecified LH or RH brain regions were significantly associated with performance. Findings suggest that some acute LH and RH stroke survivors with comprehension deficits can benefit from affective prosody and facial expression cues, but different or additional cues may be needed for greatest benefit.
Referential semantic processing tasks, such as word-picture matching (WPM), require the matching of stimuli that represent the same concept, while relational semantic processing tasks, such as Pyramids and Palm Trees (PPT), require the processing of relationships between concepts. Compared to referential tasks, relational tasks may be likely to place a greater demand on working memory and/or executive functioning. In the current study, we examined referential and relational semantic processing in all variants of primary progressive aphasia (PPA). We predicted that backward digit and spatial span tasks and/or Trails B would be significant predictors of accuracy on relational tasks, while similar tasks that place less significant demands on working memory or executive functioning, such as forward digit and spatial span tasks and Trails A, would not be significant predictors of accuracy on relational tasks. In behavioral regression analyses, neither backward span task was a significant predictor of accuracy on referential tasks, but accuracy on 5 out of 6 relational tasks was predicted by at least one backward span task. In a second set of behavioral analyses, neither Trails task was a significant predictor of accuracy on referential tasks. Conversely, accuracy on 5 out of 6 relational tasks was predicted by Trails B. In regression analyses involving grey matter volume, three ROIs (left middle temporal gyrus, left insula, right middle temporal pole) were significant positive predictors of accuracy on a referential task (Auditory WPM). The same three ROIs were significant positive predictors of accuracy on a relational task (PPT-Pictures), along with one additional ROI – the left pars triangularis. The findings of this study suggest that performance on relational semantic processing tasks is affected by working memory capacity and executive functioning, which are supported by the left pars triangularis.
Attention-Deficit/Hyperactivity Disorder (ADHD) has been associated with altered integration across large-scale brain networks. Temporal fluctuations in functional organization and differences between acquisition centers complicate classification based on resting-state functional magnetic resonance imaging (rs-fMRI). This study used BrainNetCNN as the reference architecture to examine how region-of-interest (ROI) dimensionality, connectivity representation, network design, window configuration and acquisition site influence ADHD discrimination. The sample included 465 participants from four ADHD-200 centers: New York University (NYU; n = 177), Peking University (Peking; n = 183), NeuroIMAGE (n = 39) and Oregon Health & Science University (OHSU; n = 66). Functional connectivity was derived from regional blood-oxygen-level-dependent (BOLD) time series using prespecified 12, 18, 39 and 116-ROI panels from the Automated Anatomical Labeling atlas (AAL116). The reference window was 120 s with an approximately 12-s step. Within each center, classification was assessed through participant-level stratified 10-fold cross-validation repeated five times, with area under the receiver operating characteristic curve (AUC) as the primary metric. BrainNetCNN was configured using NYU data and then kept fixed for evaluation at the remaining centers. Across site-by-ROI combinations, AUC ranged from 50.1% to 64.5%. The 12-ROI configuration reached 63.6% in both NYU and Peking, while the strongest panel differed across centers. Comparisons involving static connectivity, long short-term memory (LSTM), gated recurrent unit (GRU), DeepSets, logistic regression and alternative window settings revealed no consistent advantage. Leave-one-site-out analyses yielded AUCs from 44.4% to 63.9%, with 14 of 16 pointwise 95% confidence intervals including 0.50. Overall, the models showed modest site-dependent discrimination and limited cross-site transportability.
