The expansion, exploration, selection, and refinement (EESR) theory of plastic change
The expansion, exploration, selection, and refinement (EESR) theory of plastic change
In 2020, Lövdén et al. proposed the expansion, exploration, selection, and refinement (EESR) theory of plastic change during skill acquisition (Lövdén et al., 2020; for an update, see Hille et al., 2024; for an extended discussion of related work, see Lindenberger & Lövdén, 2019). Figure 1 presents a summary description of the current version of EESR theory. The point of departure is a situation in which the available neural resources are insufficient to meet the demands of the to-be-learned task (Lövdén et al., 2010). Driven by this mismatch, new synapses are formed in task-relevant cortical areas, and the resulting greater connectivity space is subsequently explored for neural circuits that can approximate the execution of the goal-relevant behavior. Hence, both variability of neural activity patterns and trial-to-trial behavioral variability are large. Shifts in excitatory-inhibitory (E/I) balance towards excitation favor this phase of initial expansion-exploration phase.
Eventually, the best-performing microcircuit is selected through a process of reinforcement learning that is partly mediated by the neurotransmitter dopamine, and neural and behavioral variability starts to decrease. Hence, another class of signals is needed to trigger the end of exploration and the subsequent stabilization of representations during the refinement phase of skill acquisition. In ontogeny, the formation of perineuronal nets is critical for closing critical periods. Perineuronal nets may also help to stabilize the neural substrate of skilled performance, with the ensuing retraction of structure and decreases in neural activity. After circuit selection, neural activity as well as neural and behavioral variability decreases. Synaptic remodeling within the selected neural circuit as well as myelination of selected connections continue in a final repetition-based refinement phase of task execution.
We hypothesize that the EESR sequence is reflected at the macroscopic level by several indicators. First, task-related activation as measured by the fMRI BOLD signal in task-relevant regions is assumed to be high during early phases of skill acquisition, as different and partially inefficient task representations are being probed, and to decrease with increasing task proficiency. Second, we hypothesize that three macroscopic indicators follow an inverted U-shape function: (i) E/I balance, as measured by magnetic resonance spectroscopy, tracking the opening and closing of the plastic episode; (ii) synaptic density, as measured by PET, tracking synapse formation and elimination; and (iii) regional brain volume as measured by structural magnetic resonance imaging (MRI), tracking tissue expansion and renormalization. Third, as the skill approaches asymptotic levels and competing neural ensembles have been eliminated, we expect that the neural ensemble executing the task stabilizes, indicating the selection and refinement of the underlying engram. Therefore, the self-similarity of task representations as measured using representational similarity analysis based on functional MRI or EEG data is expected to increase in the course of skill acquisition.
Clearly, the microscopic-macroscopic mapping functions hypothesized by this theoretical model need to be corroborated by empirical evidence. In some cases, such mappings might be difficult to achieve. In particular, overall changes in regional brain volume represent the net outcome of many different microscopic processes that are difficult to separate in the aggregate. At the same time, better specificity and resolution of MRI methods, including MRS at high field strengths, and improvements in PET imaging will facilitate the physiological interpretation of macroscopic measures, and inform attempts to build models and theories that connect microscopic and macroscopic levels of analysis.