Wilma A. Bainbridge
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Wilma A. Bainbridge has a total of 136 co-authors. Here are the top co-authors by number of shared publications:
- Chris Ian Baker
Laboratory of Brain and Cognition, National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland 20892
13 shared publications - Cambria Revsine
Department of Psychology, University of Chicago, Chicago, USA
4 shared publications - Trent Davis
Department of Psychology, University of Chicago
4 shared publications - Xiaohan (Hannah) Guo
Department of Psychology, University of Chicago, Chicago, IL, USA
4 shared publications Annika Spottke
German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany
3 shared publicationsBrady Ryan Thomas Roberts
3 shared publications
Latest Publications
- Open Access
- 8 November 2026
- Published by Center for Open Science in PsyArXiv
Imagery Instructions Reduce False Memories Across the Visual Imagery Spectrum
Visual mental imagery is widely theorized to support memory by enhancing the richness and distinctiveness of encoded representations. However, it remains unclear how imagery ability interacts with different encoding strategies to shape memory performance. The present study examined how encoding modality (imagery vs. verbal) and processing focus (item-specific vs. relational) interact with visual imagery ability to influence veridical and false memory in the Deese–Roediger–McDermott (DRM) paradigm. A total of 173 participants (71 aphantasic participants, 12 intermediate, and 90 typical imagers) completed ten DRM lists under one of four encoding conditions in a 2 × 2 between-subjects design. Our results suggest that true recall did not differ reliably between different encoding strategies, nor was it consistently predicted by Vividness of Visual Imagery Questionnaire (VVIQ) scores, indicating that imagery ability does not broadly determine veridical memory accuracy. In contrast, imagery-based encoding reduced false memories (critical lure recall) compared with verbal encoding across the range of imagery abilities. Notably, individuals with aphantasia, despite reporting greater difficulty implementing imagery instructions, showed comparable benefits from imagery-based encoding. This finding suggests that the mnemonic benefits of imagery-based strategies may not depend exclusively on conscious visual imagery and may involve alternative representational processes. These findings challenge the assumption that imagery vividness predicts memory performance and open new avenues for understanding compensatory encoding mechanisms in aphantasia.
- 8 November 2026
- Published by Center for Open Science in PsyArXiv
Open AccessImagery Instructions Reduce False Memories Across the Visual Imagery Spectrum
Visual mental imagery is widely theorized to support memory by enhancing the richness and distinctiveness of encoded representations. However, it remains unclear how imagery ability interacts with different encoding strategies to shape memory performance. The present study examined how encoding modality (imagery vs. verbal) and processing focus (item-specific vs. relational) interact with visual imagery ability to influence veridical and false memory in the Deese–Roediger–McDermott (DRM) paradigm. A total of 173 participants (71 aphantasic participants, 12 intermediate, and 90 typical imagers) completed ten DRM lists under one of four encoding conditions in a 2 × 2 between-subjects design. Our results suggest that true recall did not differ reliably between different encoding strategies, nor was it consistently predicted by Vividness of Visual Imagery Questionnaire (VVIQ) scores, indicating that imagery ability does not broadly determine veridical memory accuracy. In contrast, imagery-based encoding reduced false memories (critical lure recall) compared with verbal encoding across the range of imagery abilities. Notably, individuals with aphantasia, despite reporting greater difficulty implementing imagery instructions, showed comparable benefits from imagery-based encoding. This finding suggests that the mnemonic benefits of imagery-based strategies may not depend exclusively on conscious visual imagery and may involve alternative representational processes. These findings challenge the assumption that imagery vividness predicts memory performance and open new avenues for understanding compensatory encoding mechanisms in aphantasia.
- Open Access
Brain Decoding from a Deep Neural Network of Visual Memory Reveals the Maturation Trajectory of Memory Systems across Ages 3-7
Memories from the first four years of life are largely inaccessible by adulthood, yet these early experiences profoundly shape cognition throughout development. Understanding this phenomenon of infantile amnesia is critical for revealing the mechanisms of early memory development. Here, we developed a new content-based behavioral measure of memory that quantifies how closely children’s memory patterns resemble those of adults, which we term the memory maturation score . Using a movie-watching fMRI dataset spanning age 3 to adulthood, we trained a cross-validated model on adults’ movie-evoked brain activity to predict adult’s propensity to remember each movie snippet, estimated by ResMem, a deep neural network trained to predict visual memorability. We then applied the model to each child’s brain activity while viewing the same movie to assess their memory maturation (i.e., cross-decodability from adult patterns). Children exhibited adult-like neural patterns of memory around age 4, extending prior behavioral findings. Structural analyses showed that the volume of the left posterior hippocampus, but not age or total brain volume, significantly predicted children’s memory maturation. Memory maturation around age 4 thus likely reflects age-related neural development of hippocampal memory systems, rather than an age increase itself. Functional searchlight analyses revealed a trajectory of memory maturation: the inferior temporal gyrus reflected mature memory-related processing from age 3 onward, the left posterior hippocampus by age 5, and higher-order regions involved in social-affective processing and multimodal integration by age 7. Together, these findings suggest that hippocampal maturation between ages 3 and 5 drives the maturation of memory patterns. Significance Statement: Despite decades of research establishing potential mechanisms of infantile amnesia, little is known about the maturation of neural memory representations across development, limiting progress in developmental memory research and educational applications. This study reveals a developmental trajectory of memory: children show adult-like processing in the IT area by age 3, the hippocampus by age 5, and higher-order regions by age 7, reflecting mature perceptual, memory, and multimodal and social-affective processing, respectively. The left posterior hippocampus is crucial for the maturation of adult-like memory patterns. These findings show that (i) ResMem DNN can help educators select dynamic stimuli that are memorable to children by age 4, and (ii) our cross-validated model enables assessment of memory maturation without explicit memory measures.
Open AccessBrain Decoding from a Deep Neural Network of Visual Memory Reveals the Maturation Trajectory of Memory Systems across Ages 3-7
Memories from the first four years of life are largely inaccessible by adulthood, yet these early experiences profoundly shape cognition throughout development. Understanding this phenomenon of infantile amnesia is critical for revealing the mechanisms of early memory development. Here, we developed a new content-based behavioral measure of memory that quantifies how closely children’s memory patterns resemble those of adults, which we term the memory maturation score . Using a movie-watching fMRI dataset spanning age 3 to adulthood, we trained a cross-validated model on adults’ movie-evoked brain activity to predict adult’s propensity to remember each movie snippet, estimated by ResMem, a deep neural network trained to predict visual memorability. We then applied the model to each child’s brain activity while viewing the same movie to assess their memory maturation (i.e., cross-decodability from adult patterns). Children exhibited adult-like neural patterns of memory around age 4, extending prior behavioral findings. Structural analyses showed that the volume of the left posterior hippocampus, but not age or total brain volume, significantly predicted children’s memory maturation. Memory maturation around age 4 thus likely reflects age-related neural development of hippocampal memory systems, rather than an age increase itself. Functional searchlight analyses revealed a trajectory of memory maturation: the inferior temporal gyrus reflected mature memory-related processing from age 3 onward, the left posterior hippocampus by age 5, and higher-order regions involved in social-affective processing and multimodal integration by age 7. Together, these findings suggest that hippocampal maturation between ages 3 and 5 drives the maturation of memory patterns. Significance Statement: Despite decades of research establishing potential mechanisms of infantile amnesia, little is known about the maturation of neural memory representations across development, limiting progress in developmental memory research and educational applications. This study reveals a developmental trajectory of memory: children show adult-like processing in the IT area by age 3, the hippocampus by age 5, and higher-order regions by age 7, reflecting mature perceptual, memory, and multimodal and social-affective processing, respectively. The left posterior hippocampus is crucial for the maturation of adult-like memory patterns. These findings show that (i) ResMem DNN can help educators select dynamic stimuli that are memorable to children by age 4, and (ii) our cross-validated model enables assessment of memory maturation without explicit memory measures.
- 1 November 2026
- Published by American Psychological Association (APA) in Journal of Experimental Psychology: General
- Vol. 155 (11) , 2678-2695
Recognition-memory asymmetries predicted by individual item memorability.
- 1 November 2026
- Published by American Psychological Association (APA) in Journal of Experimental Psychology: General
- Vol. 155 (11) , 2678-2695
Recognition-memory asymmetries predicted by individual item memorability.
- 13 October 2026
- Published by SAGE Publications in Psychological Science
- Vol. 37 (11) , 774-790
Artists Can Design Memory, and AI Can Predict It
Can artists deliberately design artworks to be remembered—or forgotten? Across three experiments, we tested whether visual-memory outcomes can be shaped through artistic design and how they are predicted by artists, laypeople, and artificial intelligence. We launched a nationwide art contest in which artists submitted works intended to be memorable or forgettable. Using recognition tasks with adults based in the United States (119 from Amazon Mechanical Turk, 892 from Prolific, and 62 visitors from a Chicago gallery), we found that artists achieved their intended memory outcomes. However, their success was asymmetric: Memorability was more readily increased than reduced. A neural network, ResMem, consistently outperformed human observers and artists in predicting which artworks would be remembered. Data-driven analyses identified psychologically interpretable features that explained nearly half the variance in memory and aligned with ResMem’s predictions, but not with human judgments. Together, findings suggest that visual memorability can be engineered by humans, though asymmetrically, and is better captured by computational models.
- 13 October 2026
- Published by SAGE Publications in Psychological Science
- Vol. 37 (11) , 774-790
Artists Can Design Memory, and AI Can Predict It
Can artists deliberately design artworks to be remembered—or forgotten? Across three experiments, we tested whether visual-memory outcomes can be shaped through artistic design and how they are predicted by artists, laypeople, and artificial intelligence. We launched a nationwide art contest in which artists submitted works intended to be memorable or forgettable. Using recognition tasks with adults based in the United States (119 from Amazon Mechanical Turk, 892 from Prolific, and 62 visitors from a Chicago gallery), we found that artists achieved their intended memory outcomes. However, their success was asymmetric: Memorability was more readily increased than reduced. A neural network, ResMem, consistently outperformed human observers and artists in predicting which artworks would be remembered. Data-driven analyses identified psychologically interpretable features that explained nearly half the variance in memory and aligned with ResMem’s predictions, but not with human judgments. Together, findings suggest that visual memorability can be engineered by humans, though asymmetrically, and is better captured by computational models.
- Open Access
- 13 September 2026
- Published by Center for Open Science in PsyArXiv
Testing AI’s abilities at closed-loop psychology research
Large language models (LLMs) have both incredible promise and incredible risk in becoming major contributors to scientific discovery. Psychology especially is a field that could experience major transformation, as empirical experiments can be run in fully digital closed loops via crowd-sourcing platforms. As a result, LLMs can conduct every stage of the research process: generating research questions, pre-registering its research plans, coding experiments, analyzing data, and writing the manuscript of the results. This could have the potential to accelerate scientific discovery, but also alongside many risks (e.g., false information, monoculturization, high energy usage). Here, we test the degree to which ChatGPT 5.0 can conduct a psychological experiment from start to finish—specifically, testing the degree to which humans employ hexadirectional strategies in navigating a familiar conceptual space. We discuss the model’s successes and failures at each step of the research process, and provide advice on where LLMs shine and where they need extra human observation. Our work opens a discussion of what this means for the broader field of psychological research—as these models have both high potential, but also important drawbacks and ramifications to consider.
- 13 September 2026
- Published by Center for Open Science in PsyArXiv
Open AccessTesting AI’s abilities at closed-loop psychology research
Large language models (LLMs) have both incredible promise and incredible risk in becoming major contributors to scientific discovery. Psychology especially is a field that could experience major transformation, as empirical experiments can be run in fully digital closed loops via crowd-sourcing platforms. As a result, LLMs can conduct every stage of the research process: generating research questions, pre-registering its research plans, coding experiments, analyzing data, and writing the manuscript of the results. This could have the potential to accelerate scientific discovery, but also alongside many risks (e.g., false information, monoculturization, high energy usage). Here, we test the degree to which ChatGPT 5.0 can conduct a psychological experiment from start to finish—specifically, testing the degree to which humans employ hexadirectional strategies in navigating a familiar conceptual space. We discuss the model’s successes and failures at each step of the research process, and provide advice on where LLMs shine and where they need extra human observation. Our work opens a discussion of what this means for the broader field of psychological research—as these models have both high potential, but also important drawbacks and ramifications to consider.
- Open Access
- 17 August 2026
- Published by Center for Open Science in PsyArXiv
Artists Can Design Memory, and AI Can Predict It
Can artists deliberately design artworks to be remembered—or forgotten? Prior research has shown that artwork memorability is consistent across viewers and can be predicted by deep neural networks, yet it remains unclear whether memorability can be intentionally manipulated. Across three complementary experiments, we tested whether visual memory outcomes can be shaped through artistic design, and how well such outcomes are predicted by artists, laypeople, and artificial intelligence. We launched a nationwide art contest in which artists submitted works explicitly intended to be either memorable or forgettable. Using both online and in-gallery recognition tasks, we found that artist intent significantly influenced memory performance, with memorable and forgettable artworks diverging reliably across individuals and contexts. Yet when comparing predictions, a deep neural network (ResMem) consistently outperformed both human observers and artists in forecasting which artworks would be remembered. Moreover, clustering and regression analyses revealed a reduced set of psychologically interpretable features that accounted for nearly half of the variance in memory performance and aligned with ResMem’s predictions, but not with human judgments. Together, these findings suggest that visual memorability is not only a stable and generalizable construct but also one that can be deliberately crafted, though often in ways that elude human intuition and are better captured by computational models.
- 17 August 2026
- Published by Center for Open Science in PsyArXiv
Open AccessArtists Can Design Memory, and AI Can Predict It
Can artists deliberately design artworks to be remembered—or forgotten? Prior research has shown that artwork memorability is consistent across viewers and can be predicted by deep neural networks, yet it remains unclear whether memorability can be intentionally manipulated. Across three complementary experiments, we tested whether visual memory outcomes can be shaped through artistic design, and how well such outcomes are predicted by artists, laypeople, and artificial intelligence. We launched a nationwide art contest in which artists submitted works explicitly intended to be either memorable or forgettable. Using both online and in-gallery recognition tasks, we found that artist intent significantly influenced memory performance, with memorable and forgettable artworks diverging reliably across individuals and contexts. Yet when comparing predictions, a deep neural network (ResMem) consistently outperformed both human observers and artists in forecasting which artworks would be remembered. Moreover, clustering and regression analyses revealed a reduced set of psychologically interpretable features that accounted for nearly half of the variance in memory performance and aligned with ResMem’s predictions, but not with human judgments. Together, these findings suggest that visual memorability is not only a stable and generalizable construct but also one that can be deliberately crafted, though often in ways that elude human intuition and are better captured by computational models.
- 1 July 2026
- Published by American Psychological Association (APA) in American Psychologist
- Vol. 81 (5) , 662-675
What divides and unites our memories: Multifactor trial-wise predictions of memory across 6+ million trials.
A majority of memory research examines a single or a few factors that influence memory, but a broader perspective is needed to form a comprehensive understanding of the mechanisms of memory. Prior work has largely focused on observer effects (e.g., individual differences, internal states) and environmental effects (e.g., task, context), with only more recent work focusing on item effects (e.g., the typicality or memorability of stimuli). Here, I analyze 6,223,953 memory decisions made by 13,946 participants on a continuous recognition task for 26,107 naturalistic object images to test how these different factors influence long-term episodic memory. The individual participant and the memorability of the stimulus emerge as the most important factors for predicting successful hits, with surprisingly little influence of the experimental context. In contrast, while less variance is captured by these factors for false alarms, a wider variety of factors are important, including image factors, observer factors, and experimental context. These results suggest that much of what drives our memory is shared and predictable and demonstrate the need for more research that focuses on multiple factors simultaneously in more representative and naturalistic tasks. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
- 1 July 2026
- Published by American Psychological Association (APA) in American Psychologist
- Vol. 81 (5) , 662-675
What divides and unites our memories: Multifactor trial-wise predictions of memory across 6+ million trials.
A majority of memory research examines a single or a few factors that influence memory, but a broader perspective is needed to form a comprehensive understanding of the mechanisms of memory. Prior work has largely focused on observer effects (e.g., individual differences, internal states) and environmental effects (e.g., task, context), with only more recent work focusing on item effects (e.g., the typicality or memorability of stimuli). Here, I analyze 6,223,953 memory decisions made by 13,946 participants on a continuous recognition task for 26,107 naturalistic object images to test how these different factors influence long-term episodic memory. The individual participant and the memorability of the stimulus emerge as the most important factors for predicting successful hits, with surprisingly little influence of the experimental context. In contrast, while less variance is captured by these factors for false alarms, a wider variety of factors are important, including image factors, observer factors, and experimental context. These results suggest that much of what drives our memory is shared and predictable and demonstrate the need for more research that focuses on multiple factors simultaneously in more representative and naturalistic tasks. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
- Open Access
- 13 June 2026
- Published by Center for Open Science in PsyArXiv
FalseResMem: A Neural Network to Predict False Memories in Visual Recognition Tasks
Predicting false alarms, i.e., when novel images are mistakenly identified as familiar, is crucial for understanding visual memory bias. Prior studies show that some images consistently elicit more false alarms than others, so to model these patterns, we introduce FalseResMem, the first neural network designed to predict image-level false alarm rate (FAR). The model combines ImageNet-pretrained ResNet50 features with a retrained AlexNet-like architecture to capture image properties that drive false recognitions. Trained on a large-scale object memory dataset, FalseResMem achieves consistent performance (Spearman’s rank correlation = 0.48 ± 0.13; max = 0.58, 10-fold cross-validation) and FalseResMem’s predictions successfully generalize to other image categories, including scenes, art, faces, and symbols, demonstrating its robustness. This contribution highlights the potential of integrating error-based metrics into visual memory modeling through a simple FalseResMem package, offering researchers a practical way to predict false alarms and investigate the image features that systematically shape false memories.
- 13 June 2026
- Published by Center for Open Science in PsyArXiv
Open AccessFalseResMem: A Neural Network to Predict False Memories in Visual Recognition Tasks
Predicting false alarms, i.e., when novel images are mistakenly identified as familiar, is crucial for understanding visual memory bias. Prior studies show that some images consistently elicit more false alarms than others, so to model these patterns, we introduce FalseResMem, the first neural network designed to predict image-level false alarm rate (FAR). The model combines ImageNet-pretrained ResNet50 features with a retrained AlexNet-like architecture to capture image properties that drive false recognitions. Trained on a large-scale object memory dataset, FalseResMem achieves consistent performance (Spearman’s rank correlation = 0.48 ± 0.13; max = 0.58, 10-fold cross-validation) and FalseResMem’s predictions successfully generalize to other image categories, including scenes, art, faces, and symbols, demonstrating its robustness. This contribution highlights the potential of integrating error-based metrics into visual memory modeling through a simple FalseResMem package, offering researchers a practical way to predict false alarms and investigate the image features that systematically shape false memories.
- Open Access
- 8 March 2026
- Published by Center for Open Science in PsyArXiv
Reverse-engineering what makes a symbol memorable
Symbols may represent the first form of human visual communication, yet little is known about the cognitive and neural mechanisms supporting memory for these pervasive graphics. By investigating memory for everyday symbols, we can understand how abstract concepts are concretized with simple referents and later processed in visual memory systems. Recently, symbols have been found to be highly memorable, especially relative to words, but it remains unclear what drives their heightened memorability. We identified the key visual and conceptual attributes driving high memorability for symbols. Participants were tested on their memory for conventional symbols (e.g., !@#$%) before sorting them based on visual or conceptual features. Principle component analyses performed on the sorting data revealed which of these features predict memory for symbols. Generative artificial intelligence was then used to accentuate or downplay these predictive features to create a set of memorable and forgettable novel symbols. A memory test revealed that symbols designed to be memorable were not only better recognized than those designed to be forgettable, but they also afforded superior recall of associated abstract words. This work demonstrates that certain features drive memory for images and offers clear evidence that memory can be intentionally engineered.
- 8 March 2026
- Published by Center for Open Science in PsyArXiv
Open AccessReverse-engineering what makes a symbol memorable
Symbols may represent the first form of human visual communication, yet little is known about the cognitive and neural mechanisms supporting memory for these pervasive graphics. By investigating memory for everyday symbols, we can understand how abstract concepts are concretized with simple referents and later processed in visual memory systems. Recently, symbols have been found to be highly memorable, especially relative to words, but it remains unclear what drives their heightened memorability. We identified the key visual and conceptual attributes driving high memorability for symbols. Participants were tested on their memory for conventional symbols (e.g., !@#$%) before sorting them based on visual or conceptual features. Principle component analyses performed on the sorting data revealed which of these features predict memory for symbols. Generative artificial intelligence was then used to accentuate or downplay these predictive features to create a set of memorable and forgettable novel symbols. A memory test revealed that symbols designed to be memorable were not only better recognized than those designed to be forgettable, but they also afforded superior recall of associated abstract words. This work demonstrates that certain features drive memory for images and offers clear evidence that memory can be intentionally engineered.
- Open Access
- 6 March 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (10)
Reverse engineering what makes a symbol memorable
Symbols may represent the first form of human visual communication, yet little is known about the cognitive and neural mechanisms supporting memory for these pervasive graphics. By investigating memory for everyday symbols, we can understand how abstract concepts are concretized with simple referents and later processed in visual memory systems. Recently, symbols have been found to be highly memorable, especially relative to words, but it remains unclear what drives their heightened memorability. We identified the key visual and conceptual attributes driving high memorability for symbols. Participants were tested on their memory for conventional symbols (e.g., !@#$%) before sorting them based on visual or conceptual features. Principal component analyses performed on the sorting data revealed which of these features predict memory for symbols. Generative AI was then used to accentuate or downplay these predictive features to create a set of memorable and forgettable novel symbols. A memory test revealed that symbols designed to be memorable were not only better recognized than those designed to be forgettable, but they also afforded superior recall of associated abstract words. This work demonstrates that certain stimulus features drive memory for images beyond distinctiveness or context and offers clear evidence that memory can be intentionally engineered.
- 6 March 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (10)
Open AccessReverse engineering what makes a symbol memorable
Symbols may represent the first form of human visual communication, yet little is known about the cognitive and neural mechanisms supporting memory for these pervasive graphics. By investigating memory for everyday symbols, we can understand how abstract concepts are concretized with simple referents and later processed in visual memory systems. Recently, symbols have been found to be highly memorable, especially relative to words, but it remains unclear what drives their heightened memorability. We identified the key visual and conceptual attributes driving high memorability for symbols. Participants were tested on their memory for conventional symbols (e.g., !@#$%) before sorting them based on visual or conceptual features. Principal component analyses performed on the sorting data revealed which of these features predict memory for symbols. Generative AI was then used to accentuate or downplay these predictive features to create a set of memorable and forgettable novel symbols. A memory test revealed that symbols designed to be memorable were not only better recognized than those designed to be forgettable, but they also afforded superior recall of associated abstract words. This work demonstrates that certain stimulus features drive memory for images beyond distinctiveness or context and offers clear evidence that memory can be intentionally engineered.