Development

Previous research suggests that older adults process redundant targets with greater workload capacity than younger adults. Our group previously introduced a distractor inhibition account to explain this phenomenon, proposing that age-related declines in the ability to suppress distractors lead to a violation of context invariance. Interestingly, this effect was observed in discrimination-type redundant target tasks but not in detection-type tasks. In a recent large-scale study (N = 520, female = 271, age range: 17–79), we found age-related effects in both discrimination-type redundant target tasks (color-shape task; r = 0.12, p < .001) and detection-type redundant target tasks (double dot task; r = 0.42, p < .001). To account for these findings, we propose two possible mechanisms: (a) a violation of context invariance in older adults due to slower processing of a single target when accompanied by a distractor compared to when it is part of a redundant target, and (b) increased attention to non-target locations, which prolongs response times due to serial processing. Here, we introduce a mixture model to examine workload capacity coefficients in redundant target tasks. Specifically, we investigate whether serial processing moderates age-related differences in workload capacity. Our findings contribute to a deeper understanding of cognitive aging and the underlying mechanisms driving changes in attentional processing across different task types.

Characters in dreams are generated from the dreamer’s memory, actions of spontaneous memory. If memory is veridical, properties of waking-life contacts with people would be found in occurrences of characters in dreams. A girl began journaling her dreams in high school and continued in college. Dream reports are available on DreamBank.net. Characters in the dreams were coded with the well established Hall-Van De Castle system. College dreams and an equivalent number of high school dreams were compared. Characters differed, reflecting changes in waking life between high school and college. Individual characters occurred with different frequencies in the dreams. In accord with waking-life contacts with people, properties of phone calls individuals made over time (Saramȁki, et al, 2014) were found in dream character frequencies. (i) Characters changed over time, with more change for less frequent characters, seen in Jaccard Indices. (ii) Despite character change, the probability distribution of frequencies changed little over time, seen in chi square tests and in equality of parameters of power laws fit to the distributions. Networks of aggressive, friendly and sexual interactions have a simple star-like form in high school and college. Dream social networks for high school and college were constructed by representing each character by a vertex and linking two characters if they appeared in a dream together. There was little change in measures of network form from high school to college. Over a life transition, a social network form, implicit in generation of characters by spontaneous memory, persisted despite considerable changes in people.

Evidence accumulation models allow researchers to estimate cognitive parameters from empirical data. Applying them to large datasets that include participants from diverse backgrounds is crucial for identifying patterns in parameter estimates across different subgroups. In this study, we apply a one-boundary diffusion model using to data from the EPIC-Norfolk Prospective Population Cohort Study (2021), which includes 8,623 participants aged 48–92, 5% of whom developed dementia over a 10-year follow-up period. Participants completed the Visual Sensitivity Task (VST), a perceptual detection task assessing visual processing speed. A recent study by Begde et al. (2024) reported a potential relationship between visual sensitivity and dementia risk factors. Building on this work, we aim to use trial-by-trial VST data to estimate one-boundary diffusion parameters and examine their relationship with subsequent dementia outcomes. We estimate drift rate, which maps onto mental speed (here, visual processing); boundary, separation which reflects response caution; and non-decision time, which represents the combination of encoding and motor response. We aim to examine how parameter estimates vary across demographic factors such as age, gender, education, and socioeconomic status and to explore their relationship with later dementia diagnoses.

Understanding how children acquire language requires us to disentangle the roles of input quantity and cognitive constraints in semantic and syntactic learning. To investigate this, we trained three transformer-based language models (GPT-2) with varying parameter sizes, determined through Optuna hyperparameter tuning over 100 trials (2.87M, 2.87M, and 33.71M parameters). These models were trained on progressively larger subsets of child-directed speech from CHILDES North America: speech directed to children before 12 months, before 20 months, and the full dataset. The models were evaluated using Representational Similarity Analysis (RSA)—which measures how well the models’ internal word representations align with human-derived syntactic and semantic categories—and two behavioral tests inspired by child language research: an odd-one-out task to assess semantic categorization and a wug test to evaluate syntactic generalization. Our results indicate that increasing training size and number of parameters substantially enhances semantic learning. RSA scores improved from 0.12 to 0.25 and further to 0.37, while accuracy in the odd-one-out task increased from 51.02% to 71.43% and ultimately 90.00%. In contrast, syntactic generalization showed minimal or inconsistent improvements. RSA scores for syntactic structure decreased slightly from 0.35 to 0.34 and then to 0.33, while wug test performance remained low, with scores of 5 out of 11, 4 out of 11, and 6 out of 11 across the three models, respectively. The results suggest different impacts of input and model size on semantic versus syntactic learning, offering computational insights into the potentially distinct mechanisms underlying early language acquisition.