ICCM: Neuroscience Models

Event-related potentials (ERPs) are used to study how language is processed in the brain, including differences between native (L1) and second-language (L2) processing. A P600 ERP effect can be measured in proficient L2 learners in response to an L2 syntactic violation, indicating native-like processing. Cross-language similarity seems to be a factor that modulates P600 effect size. This manifests in a reduced P600 effect in response to a syntactic violation in the L2 when the syntactic feature involved is expressed differently in two languages. We investigate if this reduced P600 effect can be explained by assuming that ERPs reflect learning signals that arise from mismatches in predictive processing; and in particular that the P600 reflects the error that is back-propagated through the language system (Fitz & Chang, 2019). We use a recurrent neural network model of bilingual sentence processing to simulate the P600 (as back-propagated prediction error) and have it process three types of syntactic constructions differing in cross-language similarity. Simulated English-Spanish participants displayed a P600 when encountering constructions that are similar between the two languages, but a reduced P600 for constructions that differ between languages. This difference between the two P600 responses mirrors what has been observed in human ERP studies. Unlike human participants, simulated participants showed a small P600 response to constructions unique to the L2 (i.e., grammatical gender), presumably because of how this grammatical feature is encoded in the model. Our modelling results shed further light on the viability of error propagation as an account of ERPs, and on the effects of syntactic transfer from L1 to L2.

This is an in-person presentation on July 22, 2024 (12:20 ~ 12:40 CEST).

Reliably identifying relevant brain areas implicated by the simulated activity from cognitive models is still an unsolved problem for cognitive modeling, particularly when matching model output with human electroencephalography (EEG) data. We propose a new method involving post-processing of ACT-R module activity and clustered EEG component activity with generalized least squares (GLS) analysis to find matching patterns between predicted and observed data, thereby inferring neural substrates of distinct cognitive processes. This approach holds several advantages over other methods by controlling for autocorrelation and unequal variances. To exemplify its application, we used a cognitive model and EEG data from a mental spatial transformation study to show how this method finds areas involved in representational and transformational spatial processing. Parietal areas involved with spatial activity were identified, in line with prior studies on spatial cognition. In addition, previously established associations between ACT-R and brain areas were confirmed. Finally, we discuss limitations and possibilities of the approach.

This is an in-person presentation on July 22, 2024 (12:00 ~ 12:20 CEST).

Animals and humans in reinforcement learning tasks are able to learn the timing of reward delivery, even when that timing is delayed and variable, suggesting a sophisticated ability to learn the distribution of reward timings. In this work, we present two algorithms simulating the switching interval variance (SIV) task as described in Li et al. that showed mice were able to adapt their behaviour to the change of standard deviation of the reward time delays. Both algorithms implementthe wait vs stay decision by thresholding the log evidence that a forthcoming reward is likely, without assuming the specific form of the reward timing distribution. One algorithm is implemented algebraically, and the other using Spatial Semantic Pointers, a tool from Vector Symbolic Algebras for representing continuous values that have ties to hippocampal grid cells. We show that our models capture characteristic behaviour of mice on the SIV task.

This is an in-person presentation on July 22, 2024 (11:40 ~ 12:00 CEST).