Search

Schema theory predicts that information consistent with previous knowledge is more easily incorporated and remembered. On the other hand, prediction error accounts in psychology and neuroscience suggest that novelty or surprise is a key driver of memory. The SLIMM model (Schema Linked Interaction between Medial prefrontal and Medial temporal regions) [1] integrates these contradictory theories by proposing a U-shaped function for memory driven by dissociable memory systems, predicting greater recall for expected and unexpected information compared to a neutral condition. Additionally, the model proposes greater recall of task-incidental information in the unexpected condition compared to the expected and neutral conditions. These predictions have been validated after training simple rules [2] and using pre-experimental knowledge in more naturalistic VR environments [3]. In the current study, we test SLIMM’s predictions for verbal memory. We will use linguistic materials that engage participants’ knowledge of idiomatic expressions, and test their memory for expected and unexpected endings to such expressions relative to those of sentences with multiple plausible endings. Memory will also be tested for task-incidental information such as the purported speakers of the sentences.

Prior knowledge or 'schemas' influence memory, with information that is consistent with predictions from schemas generally being remembered well, while memory for information that violates predictions also tending to be remembered well. This U-shaped function of memory against schema congruency - where memory is superior for highly congruent and incongruent compared to unrelated events - is consistent with a theoretical framework called Schema Linked Interaction between Medial prefrontal and Medial temporal regions (SLIMM; van Kesteren et al., 2012). This U-shape has been observed with experimentally-trained rules (Greve et al., 2019) and more naturalistic stimuli presented in immersive Virtual Reality (Quent et al., 2022). A recent study by Höltje & Mecklinger (2022) used linguistic stimuli and manipulated the contextual constraint of sentences (strong constraint: sc, weak constraint: wc) and the predictability of the final word (expected: exp, unexpected: unexp). When analyzed as a 2x2 factorial design, they found a main effect of constraint, with better memory for final words in highly constrained sentences, and a main effect of predictability, with better memory for highly expected final words. The interaction did not reach significance. However, when the four conditions are re-ordered along a continuum of schema congruency, our re-analysis of their data revealed a significant quadratic (U-shaped) component, with a pattern of sc_exp > wc_exp but wc_unexp < sc_unexp, consistent with the predictions of SLIMM. The aim of the present study is to replicate these results using an online study in English speakers, instead of in-person testing of German speakers. The study will also attempt to separate contributions of familiarity and recollection to memory performance. According to SLIMM, memory for unexpected words will be supported by recollection, whereas memory for expected words will be supported by familiarity. Additionally, the study will test memory for details that are incidental to the main sentential meaning. SLIMM predicts memory for this 'incidental' context will not follow a U-shape, but will increase monotonically from highly constrained sentences with expected endings to highly constrained sentences with unexpected endings. Furthermore, SLIMM predicts that the episodic snapshot triggered by unexpected events can include incidental context, while such incidental information is less likely to be encoded for sentences with expected words.

The goal of this study is to investigate whether prediction error drives declarative learning. We plan to replicate De Loof and colleagues’ 2018 study on the role of prediction error in learning new foreign translations of words without the use of extrinsic monetary reward. We hypothesise that prediction error will still drive memory in the absence of extrinsic reward, and consistent with the PIMMS framework (Greve et al., 2017), the magnitude of prediction error, but not the sign, will drive memory for the learned word pairs.

Schema theory predicts that information consistent with previous knowledge is more easily incorporated and remembered. On the other hand, prediction error accounts in psychology and neuroscience suggest that novelty or surprise is a key driver of memory. The SLIMM model (Schema Linked Interaction between Medial prefrontal and Medial temporal regions) [1] integrates these contradictory theories by proposing a U-shaped function for memory, with greater recall for highly expected and highly unexpected information compared to less (un)expected information. Additionally, the model proposes greater recall of task-incidental information in unexpected conditions compared to expected conditions. These predictions have been validated after experimental training of simple rules [2] and using pre-experimental knowledge in more naturalistic virtual reality (VR) environments [3]. In the current study, we test SLIMM’s predictions for verbal memory. We will use linguistic materials that engage participants’ knowledge of idiomatic expressions, and test their memory for expected and unexpected endings to such expressions relative to those of sentences with multiple plausible endings. Memory will also be tested for task-incidental information such as the purported speakers of the sentences. The background, predictions, and materials of the current experiment are identical to those of our previous experiment pre-registered here: https://osf.io/v8r24. In this previous version of the experiment, the foils in the forced-choice recognition tests allowed participants to use foil familiarity instead of associative memory to make accurate responses, confounding results. The current experiment addresses this shortcoming by ensuring all foils are studied, but with different associations (see “Study Design” section). References: [1] van Kesteren, Ruiter, Fernández & Henson (2012). Trends in Neurosciences [2] Greve, Cooper, Tibon & Henson (2019). Journal of Experimental Psychology: General [3] Quent, Greve & Henson (2022). Psychological Science

Fast Mapping (FM) has been proposed as a cortical-based learning process that allows for rapid memory formation, independent of the typical, hippocampally-dependent episodic memory system. This incidental learning procedure was originally proposed in adults and involves presenting a familiar reference picture alongside an unfamiliar object and its label. The known referent is thought to be mechanistically important in supporting the integration of new information into pre-existing semantic knowledge, especially in patients suffering from Hippocampal amnesia (Sharon et al., 2011). Coutanche and Thompson-Schill (2014) found that FM learning in healthy young adults showed an advantage for implicit memory tests but not for explicit memory tests. More specifically, they demonstrated that participants show better same-day lexical learning under the FM condition compared to the explicit encoding (EE) condition. However, our previous methodologic replication attempt of their study with over 80% statistical power and a sample of 56 young adults did not reproduce their same-day lexical competition finding. Instead, in a post-hoc analysis we found a facilitatory semantic priming effect instead of inhibitory lexical competition, which we attempted to replicate in a subsequent study but found no effects, i.e., neither semantic priming nor lexical competition. Nonetheless, Coutanche and Koch (2017) suggested that the typicality of the known referent might play a crucial role in FM learning. Their findings suggest that when a known referent is more typical, there is less same-day lexical learning. If our referents were too typical for our sample (despite using very similar stimuli to Coutanche and Koch, 2017), this could explain our failure to replicate their lexical competition effect. Therefore, in the experiment proposed here, we will explicitly manipulate the typicality of the referent item, and investigate whether this affects the degree of lexical competition (or semantic priming), e.g, only being found for less typical referents.

Data and R scripts or FM with implicit memory measure project. # Cooper,E., Tibon, R., Greve, A.,& Henson, R.N.(2021)"No evidence of a fast mapping advantage in healthy adults using an implicit memory measure: Failures to replicate the results of Coutanche and Thompson-Schill (2014)"

216 p. Functional specialisation and plasticity are fundamental organising principles of the brain. Language is a uniquely human phenomenon that requires a delicate balance between neural specialisation and plasticity, and language learning offers the perfect window to study these principles in the human brain. Though the human brain exhibits a remarkable ability to support a variety of languages that may be acquired at different points in the life span, the capacity for neural reorganisation decreases with age. Further, language is a complex construct involving linguistic as well as visual, auditory, and motor processes. The current doctoral thesis asked two main questions: (1) Do large-scale functional changes accompany language learning in adulthood? and (2) Are these neural changes similar across different language systems such as reading, speech comprehension, and verbal production? These questions were investigated in three fMRI experiments with adult language learners. In Experiments I and II, comprehension and production were examined in 30-to-60-year-old intermediate and advanced language learners and functional learning-related changes in each modality were comprehensively characterised. In Experiment III, hemispheric lateralisation of reading, speech comprehension, and verbal production were compared and contrasted, and the analyses were extended to a second longitudinal study with a contrasting participant sample. Robust evidence was found for significant functional plasticity well into adulthood, and results showed that different language systems exhibited different patterns of hemispheric specialisation and plasticity. The results have theoretical and practical implications for our understanding of fundamental principles of neural organisation of language, language learning in healthy populations, and language testing and recovery in patients.