From altered synaptic plasticity to atypical learning: A computational model of Down syndrome

•We present a neurocomputational model of Down syndrome biologically motivated.•Our model shows that certain training schedules mitigate atypical learning in Down syndrome.•The model predictions were borne out in a study with Down syndrome participants.•Our model bridges a gap between research with mouse models and observed human behaviour.•Implicit learning in Down syndrome is highly susceptible to interference effects. Learning and memory rely on the adaptation of synaptic connections. Research on the neurophysiology of Down syndrome has characterized an atypical pattern of synaptic plasticity with limited long-term potentiation (LTP) and increased long-term depression (LTD). Here we present a neurocomputational model that instantiates this LTP/LTD imbalance to explore its impact on tasks of associative learning. In Study 1, we ran a series of computational simulations to analyze the learning of simple and overlapping stimulus associations in a model of Down syndrome compared with a model of typical development. Learning in the Down syndrome model was slower and more susceptible to interference effects. We found that interference effects could be overcome with dedicated stimulation schedules. In Study 2, we ran a second set of simulations and an empirical study with participants with Down syndrome and typically developing children to test the predictions of our model. The model adequately predicted the performance of the human participants in a serial reaction time task, an implicit learning task that relies on associative learning mechanisms. Critically, typical and atypical behavior was explained by the interactions between neural plasticity constraints and the stimulation schedule. Our model provides a mechanistic account of learning impairments based on these interactions, and a causal link between atypical synaptic plasticity and associative learning.