The Neural Oscillation Profiles And Machine Learning Of A Premise Monotonicity Effect During Semantic Category-based Induction

Category-based induction involves applying knowledge of categorical relations to generalize a novel property from one or a set of categories to a conclusion category.Category-based induction is involved in categorization,probability judgment,analogical reasoning,and decision-making and other cognitive activities,which playing an essential role in learning and adaptation.The premise monotonicity effect is one of the typical psychological effects during category-based induction.This robust effect reflects the observation that individuals are more likely to generalize properties shared by many instances than those shared by few instances.Similarity and background knowledge were the main explanations for the cognitive mechanisms that drive the premise monotonicity effect during category-based induction in the classical theories.However,category-based induction involves flexible information representation processing,these models do not reveal when these processes are involved during category-based induction.The speed of the premise monotonicity effect during category-based induction was hypothesized via exemplar-based linear ballistic accumulator model.However,this model was a computation model,the direct evidence for their hypothesis is lacking.Event-related potentials(ERPs)have superior temporal resolution and can provide real-time information about cognitive processes beyond reaction times.Previous studies employed ERPs to explore the temporal course of brain activity underpinning the premise monotonicity effect during semantic category-based induction.In traditional ERP analysis,time-locked and phase-locked activity are averaged as event-related EEG signals,while the time-locked and non-phase-locked activity,which provides important information related to cognitive processing,are lost.Thus,the current study aims to reveal the phase-locked and non-phase-locked power of neural oscillations of the premise monotonicity effect during semantic category-based induction via the time-frequency technology.EEG measurements may still be somewhat unstable due to individual differences between participants,machine learning algorithms are adopted as their ‘learning' feature may increase both the stability of ERP data and classification accuracy.In current study,the effectiveness of the average amplitudes of ERP components,time-locked and phase-locked power and time-locked and non-phase-locked power to classify the single premise arguments and the two premise arguments were evaluated by k-Nearest Neighbor,Naive Bayesian Mode,Support Vector Machine,and Random Forest algorithms.In the present study,the premise monotonicity effect were measured by manipulating the premise sample size(single instance vs.two instances)in a semantic category-based induction task,with the conclusion categories either including the premise categories(congruent induction)or not(incongruent induction).A pilot experiment was implemented to prove the task could validity measure the premise monotonicity effect during category-based induction.The behavior and EEG signal was recorded during the formal experiment.The behavioral and ERPs results replicated the premise monotonicity effect,the arguments with two instances in premise(T arguments)had more “definite” responses,high “correct” response rates,and shorter reaction times than those had one instance.And the arguments with single instances(S arguments)in premise elicited larger FN400 amplitudes than did T arguments,which was linked to reduced global similarity,decreased cognitive relevance,and attenuated conceptual fluency,and S arguments elicited larger SN amplitudes than did T arguments,which is related to more inference-driven integration and interpretive processes.For phase-locked power,the results illustrated that the premise monotonicity effect was revealed by anterior delta power,suggesting differences in working memory updating.The results also illustrated that T arguments evoked larger posterior theta-alpha power than S arguments,suggesting that T arguments led to enhanced subjectively perceived inductive confidence than S arguments.For non-phase-locked power,the results illustrated that the premise monotonicity effect was indicated by anterior theta power,suggesting that the differences in sample size were related to a change in the need for cognitive control and the implement of adaptive cognitive control.Moreover,the results illustrated that the premise monotonicity effect was revealed by alpha-beta power,which suggested the unification of sentence and inference-driven information.Therefore,the neural oscillation profiles of the premise monotonicity effect during semantic category-based induction were elucidated,and supported the connectionist models of category-based induction.One of the limitations of the EEG technique is the poor spatial resolution in brain activity.Further studies can use magnetoencephalography(MEG)and functional magnetic resonance imaging(fMRI),which have good spatial resolution,to explore the spatial activity of the brain in the premise monotonicity effect during category-based induction.