Attentional control, memory and intelligence are cornerstones of human higher-order cognition. Research on the neurobiological underpinnings of these fundamental cognitive functions can be approached from different directions, including experimental manipulation and inter-individual differences research. This symposium aims to integrate these perspectives, bringing together four talks that highlight different methodological approaches to assess neural correlates of higher-order cognition. The first talk will focus on the interplay of selective attention and multisensory integration in audiovisual working memory, demonstrating across two EEG studies that task-irrelevant cross-modal features are automatically encoded into working memory. The second talk investigates retroactive interference between working memory and episodic memory and shows that episodic memory representations are particularly vulnerable to interference from subsequent working memory processing when the way relevant information is probed is similar across tasks. The third presentation will examine the inter-individual differences in intelligence and its relationship to working memory capacity and neurocognitive processing speed, showing that the mental speed hypothesis cannot be generalized to WM encoding processes. The fourth talk introduces a MATLAB application that simplifies the recovery of ERP component latencies to enhance accuracy and efficiency in measuring the speed of neural mechanisms — an approach widely used in both inter-individual differences research and experimental cognitive neuroscience. Finally, we will close with an outlook on how cognitive neuroscience can inform inter-individual differences research and vice versa. Together, this symposium provides a comprehensive perspective on how diverse methodological approaches deepen our understanding of the neural mechanisms underlying higher cognitive functions.

Efficient use of limited working memory capacity requires prioritizing task-relevant information. However, when attending to objects with multiple features, task-irrelevant features are encoded into working memory in a largely automatic fashion. The extent to which these irrelevant features persist, especially in multisensory contexts, remains unclear. This study provides both behavioral and electrophysiological evidence for the persistence of task-irrelevant cross-modal features in an audiovisual delayed-match-to-sample paradigm. Critically, participants were cued to selectively attend to auditory, visual or audiovisual features. Behaviorally, we observed partial repetition costs for multisensory objects: that is, consistent with prior unisensory findings, changes in task-irrelevant features impaired performance. This interference was particularly strong when auditory features were presented in lateralized positions, i.e., spatially disparate from visual features rather than spatially compatible. Electrophysiological results from two sets of representational similarity analyses (RSA) further clarified these effects. First, comparing activity patterns of attend-auditory and attend-visual conditions to unisensory controls (auditory-only, visual-only) and conjunction conditions revealed partial filtering of task-irrelevant visual features during memory maintenance. In contrast, task-irrelevant auditory features persisted more robustly. In line with behavioral results, the persistence of task-irrelevant auditory features in neural activity patterns was more pronounced with spatially disparate sound presentation. Second, neural activity patterns indicated orientation-specific and frequency-specific representations regardless of attended modality, with task-irrelevant features represented as robustly as task-relevant ones. Together, these findings suggest that multisensory integration strongly promotes feature binding, resulting in memory representations that include even those features deemed irrelevant by task demands.

This study explores how long-term memory (LTM) is vulnerable to retroactive interference from working memory (WM). Participants learned associations between visual objects and circular locations, which they later had to report from LTM. Between encoding and retrieval, the same objects were used in a WM task, with locations shifted by ~120°. We further manipulated WM task conditions to investigate retroactive interference mechanisms: In two conditions, participants were tested on the associated location and either moved the object itself to the desired location or used spatial placeholders for report. Additional conditions assessed the effect of shorter versus longer WM storage durations without testing. A baseline condition involved objects only presented in the LTM task. The target confusability competition model (TCC) estimated memory strength (d-prime) for target reports (LTM retrieval of the associated location) and swap errors (erroneous LTM report of the interfering WM location). Results revealed that swap errors were more prevalent for objects tested during the WM task, while testing method or storage duration did not have an effect. Critically, testing object-location associations in WM increased the memory strength of the interfering locations without affecting memory strength for the target location. This shows that retroactive interference did not result from a weakening of the target representation, but from the fact that WM retrieval of the interfering location led to a more competitive LTM trace. These results provide new insights into the mutual influence of WM and LTM, offering a new perspective for a better understanding of forgetting through retroactive interference.

The relationship between intelligence, working memory capacity (WMC), and information processing speed (IPS) is central to cognitive psychology. While extensive research has explored their interrelations, the specific role of IPS in explaining intelligence differences remains debated. In line with the mental speed hypothesis, prior studies demonstrate moderate but robust associations between reaction time measures and intelligence. To obtain more precise measures of mental speed, researchers have analyzed the latencies of event-related potential (ERP) components associated with higher-order cognitive processes, providing strong evidence for the mental speed account with latent correlations ranging from -.49 to -.89 (Schubert et al., 2017; 2022). However, these findings predominantly stem from decision-related processes, leaving open questions about their generalizability to non-decisional phases, such as WM encoding. This talk provides insights into the relationship between intelligence, WMC and IPS based on the analysis of ERP latencies during WM coding. Using data from 141 participants and a latent state-trait modeling approach, we found no significant association between IPS during WM encoding and intelligence or WMC. These findings challenge the universality of the mental speed hypothesis and suggest that cognitive speed advantages in intelligence primarily emerge during decision-making rather than encoding. Additionally, we observed a relationship between P3 amplitude and WMC, reinforcing its role as a neural marker of WM processing. Our results contribute to the broader discussion of how interindividual differences in intelligence are shaped by distinct neurocognitive mechanisms and highlight the need for an integrated approach combining cognitive neuroscience and differential psychology.

Neurocognitive processing speed is a fundamental correlate of intelligence and higher-order cognition, making accurate ERP latency extraction essential for studying individual differences and cognitive mechanisms. However, existing methods often require trade-offs between efficiency, objectivity, and reliability. Automatic latency extraction using peak latency algorithms or fractional area latency algorithms often yields unreliable latency values (Sadus et al., 2024). Manual extraction, while accurate, is time-consuming and subjective, whereas traditional automated approaches can be inconsistent.

In this talk, we introduce a MATLAB application that provides a user-friendly implementation of a novel template-matching algorithm for ERP latency extraction (Lesche et al., under review). The app streamlines the process by offering an intuitive interface that allows researchers to apply the algorithm with minimal effort. It includes automated rejection or confirmation of latencies based on a user-defined fit threshold, ensuring high data quality while reducing manual workload. Additionally, the app presents results from other selected algorithms for latency extraction, enabling users to compare multiple approaches and select the most suitable method for their data.

By integrating automated quality control with interactive visualization, this tool significantly improves efficiency and replicability in ERP analysis. We will demonstrate its functionality and discuss its potential applications for both experimental and individual-differences research, highlighting how it facilitates more precise measurement of neural processing speed.