Pith. sign in

REVIEW 5 major objections 5 minor 37 references

Can Mental Imagery Improve the Thinking Capabilities of AI Systems?

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read AI systems can reason by generating mental images, a new framework claims.

desk verdict A plausible conceptual framework, but the experiments never close the loop from generated images back into reasoning, so the paper's title question remains unanswered. read the letter →

arxiv 2507.12555 v2 pith:E2IP3GVK submitted 2025-07-16 cs.LG cs.AI

Slimane Larabi This is my paper · ORCID
classification cs.LGcs.AI
keywords machinethinkingmentalimageryreasoningcognitivearchitectureplanningtext-to-imagegenerationsketchlargelanguagemodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that an AI system can think more autonomously when mental imagery is made a first-class component of its architecture. It defines machine thinking as exploring possible ways to achieve an event by generating a time series of mental images, and builds a framework in which a Cognitive Thinking Unit draws on three auxiliary units: needs, sensory input, and mental imagery. The experiments show the pipeline can caption scenes, match a stated need to the most relevant context sentence, generate sketch sequences from action plans, and propose new actions when a plan fails. The payoff, if the claim holds, is that machines could initiate reasoning from internal goals and simulated scenarios rather than only answering explicit queries.

What carries the argument

The central object is the Mental Imagery Unit, a module that receives a high-level stimulus from the Cognitive Thinking Unit, generates sketch-style images with a pretrained text-to-image diffusion model followed by a pencil-sketch filter, and sends the convenient images back to the Cognitive Thinking Unit for further reasoning. It is what turns a sentence into a time series of mental images representing the progression from an initial state to a final state. The reasoning loop alternates textual inference with visual simulation: the cognitive unit proposes sentences, the imagery unit visualizes them, and the cognitive unit then asks feasibility questions about the depicted scenario to decide on the next action.

What would settle it

Take the paper's 'man opens the door' scenario, generate the mental image for that prompt, then edit the sketch so the door is visibly locked while keeping the prompt unchanged; if the Cognitive Thinking Unit's next proposed action is identical to the unlocked-door case, the image content is not feeding the reasoning, and if the action changes, the imagery is contributing information.

Watch

Extended reading notes

Core claim

The author's central claim is that internally generated sketch images can participate in reasoning rather than being decorative byproducts of text prompts. The paper proposes a machine-thinking framework in which a Cognitive Thinking Unit receives sentences from a Needs Unit and an Input Data Unit, sends stimuli to a Mental Imagery Unit, and then inspects the imagery the unit returns in order to refine its conclusions. Thinking is explicitly modeled as 'exploring possible ways to achieve an event by generating a time series of mental images,' with images generated from sentences and adjusted as reasoning adds new knowledge. Validation experiments cover image captioning, matching needs to context via sentence embeddings, converting a three-step action plan into a sketch sequence with a pretrained text-to-image diffusion model, and generating alternative actions for a failed plan.

Load-bearing premise

The load-bearing premise is that the Cognitive Thinking Unit can extract task-relevant information from the generated mental images and use that information to update its reasoning, even though the experiments define the action plan as text prompts and produce the reasoning steps as text.

Editorial extensions

If this is right

  • Reasoning can be initiated without an external query, because the Cognitive Thinking Unit can start from an internally stored need and generate the imagery needed to explore it.
  • Action plans become inspectable as a sequence of sketches, so intermediate states can be checked and corrected before execution.
  • Hypothetical reasoning, such as asking what happens if the keys do not open the door, can be handled by generating new mental images and deriving alternative actions from them.
  • The framework unifies perception, internal motivation, and imagination in one architecture, which is exactly the gap the paper identifies in current reactive systems.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper, the strongest version of the claim would close the loop by feeding sketch pixels back into the Cognitive Thinking Unit through a vision-language model, letting image content actually change the reasoning; the current experiments do not test this.
  • Beyond the paper, the same architecture could serve as a low-cost world model for embodied agents, using generated sketches to predict the visual outcome of an action before an agent commits to executing it.
  • Beyond the paper, a decisive test would use scenarios where the sketch necessarily contains information absent from the prompt, such as the spatial layout of objects; if the reasoning output does not change when the layout changes, mental imagery is not load-bearing.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes a machine-thinking framework in which a Cognitive Thinking Unit (CTU) is supported by three auxiliary units: a Needs Unit, an Input Data Unit, and a Mental Imagery Unit (MIU). The MIU is intended to generate sketches from natural-language stimuli, with the claim that these internally generated mental images contribute to reasoning, planning, and decision-making. The authors describe the architecture, report two experiments (sentence matching/action generation using text embeddings and a text-to-image-to-sketch pipeline), and conclude that the framework demonstrates potential to bridge perception, reasoning, and imagination.

Significance. If the central claim were established, the paper would address a real gap: current AI systems do not typically use internally generated visual representations as part of a reasoning loop, despite neuroscientific evidence that mental imagery supports human cognition. The paper also usefully surveys relevant work in machine reasoning, mental imagery, and generative models, and it identifies a concrete architectural idea (a mental imagery unit communicating with a cognitive unit). However, the significance is entirely conditional because the manuscript does not provide any experimental evidence that the generated images influence the CTU's output. The only quantitative experiment (Section 6.2) is text-only, and the only mental-imagery demonstration (Section 6.3) is a text-to-image rendering with no closed loop back into the reasoning process.

major comments (5)
  1. [Section 6.3] The central claim of the paper is not tested. In the only mental-imagery experiment, the action plan is predefined as text prompts ('a man takes the keys which are on the desk', 'the man goes towards the door', 'the man opens the door'), the MIU renders Stable Diffusion sketches from those prompts, and the subsequent reasoning ('the person is nervous', 'the person breaks down the door', 'the person call the firefighters') is produced as text by the CTU. There is no mechanism, code path, or experiment in which image pixels are fed back into the CTU and shown to affect its output. The causal order is text -> image -> illustration, not image -> reasoning. Consequently, the manuscript does not demonstrate that mental imagery improves thinking; it demonstrates only that text prompts can be turned into sketches.
  2. [Section 6.2] The only experiment with quantitative similarity scores uses sentence embeddings (all-MiniLM-L6-v2) over text captions; no image data is used in the matching or action-generation steps. The sentence-matching results and the generated actions ('Pick up the keys and open the door') would be identical if the MIU were removed entirely. This experiment therefore provides no evidence for the paper's stated claim that sketches serve 'informative and decision-making purposes.'
  3. [Section 5.2] The definition of thinking in Section 5.2 ('We consider thinking as the process of exploring possible ways to achieve an event by generating a time series of mental images') makes the conclusion that mental imagery is beneficial partly definitional. If thinking is defined as generating mental images, then any reasoning process that generates them trivially 'uses' mental imagery. The paper needs an independent experimental test that compares reasoning with and without the MIU; such an ablation is absent. The statement in Section 5.1 that 'the CTU then performs further reasoning based on these visualized images' is an assumption, not a demonstrated property.
  4. [Section 6.3, Figures 7 and 8] The mental-imagery experiment has no baseline, no ablation, and no quantitative evaluation. The reported outcome is a subjective statement that the sketches 'captured the essential spatial and semantic elements of each action.' There is no accuracy metric, no downstream task, no comparison with a text-only reasoning baseline, and no measure of how much the image content contributes beyond the prompt text. The final sentence of the experiment ('This experiment demonstrates that mental imagery generation from action sequences is a viable computational approach to support machine reasoning') is not supported by the evidence presented.
  5. [Section 4.3 and Section 5.1] The framework description promises a bidirectional interaction: the MIU 'communicates the convenient ones to the CTU' and the CTU 'refine[s] the result of reasoning.' However, no algorithm is given for how the CTU inspects or interprets the generated sketches, nor for how a 'convenient' image is selected, nor for how the image content modifies the reasoning state. Section 5.1 similarly states that 'the corresponding image is modified to incorporate the additional information,' but no mechanism is described. These are architectural desiderata, not implemented or validated components.
minor comments (5)
  1. [Throughout] The manuscript contains many grammatical and typographical errors (e.g., 'the sens of touch', 'The Cognitive Unit of Thinking refers to ... responsible for Interpreting', 'generate_action_f alcon', 'the person call the firefighters'). The writing needs careful editing.
  2. [Section 2] In the related-work discussion, the text mentions 'VilBERT' but the reference list entry [8] is for LXMERT; the naming should be corrected or the citation should be aligned with the model actually discussed.
  3. [Figure 1] The caption 'Diagram of Machine Thinking algorithms' may overstate the content; Figure 1 is a conceptual diagram, and the right side appears to be a human brain image. A clearer, more precise caption would help.
  4. [Section 5.2] The phrase 'the event is scheduled as a set of elementary events corresponding to the set of generated mental images' is vague; it is not clear what 'scheduled' means computationally or how the correspondence between images and elementary events is determined.
  5. [References] Several references are cited for claims about human cognition (e.g., [29], [30], [31]) but the text does not fully unpack how these neuroscientific findings translate into the proposed computational architecture; linking each cited work to a specific design choice would improve the argument.

Circularity Check

1 steps flagged · score 6.0 of 10

The paper's central claim that mental imagery improves thinking is partly self-definitional: Section 5.2 defines thinking as generating a time series of mental images, and the Section 6.3 validation only renders predefined text actions into sketches without demonstrating that image content feeds back into reasoning.

  1. self definitional [Section 5.2 ('Thinking: A Computational Perspective'), with the supporting loop in Section 4.3 and the validation in Section 6.3]
    "We consider thinking as the process of exploring possible ways to achieve an event by generating a time series of mental images. ... The Mental Imagery Unit receives stimulus from the Cognitive Thinking Unit (CTU), draws imagined images and communicates the convenient ones to the (CTU) which refine the result of reasoning."

    This stipulates that thinking is mental-image generation, so the paper's claim that mental imagery is beneficial for reasoning is true by construction. Section 6.3's validation does not break the circularity: the action plan is a predefined text sequence ('a man takes the keys which are on the desk'), Stable Diffusion renders sketches from those prompts, and the subsequent reasoning ('the person is nervous') is also generated as text. There is no measured path from image pixels back into the CTU, so the 'refine the result of reasoning' loop exists only in the framework diagram. The conclusion then credits internally generated representations with enabling reasoning, a property inherited from the definition rather than demonstrated.

full rationale

The strongest circular step is definitional: Section 5.2 defines thinking as 'the process of exploring possible ways to achieve an event by generating a time series of mental images,' and the paper's conclusion attributes reasoning and decision-making to internally generated representations. If thinking is defined as mental-image generation, then finding that mental imagery is central to thinking is not an empirical result but a consequence of the definition. The only imagery experiment, Section 6.3, uses a predefined text action plan to generate Stable Diffusion sketches and then emits new reasoning as text; the paper itself admits that 'the current system uses predefined actions,' so no dynamic loop from image content to reasoning is tested. The other validation components (image captioning in 6.1 and need-context matching via sentence embeddings in 6.2) are non-circular, but they do not involve mental imagery and therefore do not support the title claim. There are no load-bearing self-citations; the cited neuroscience and robotics work is external. The score is 6 rather than higher because the paper frames the contribution as a framework with 'potential' rather than a quantitative prediction, yet the central conceptual claim nonetheless reduces in part to the paper's own definition of thinking.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

The framework rests on definitions and assumptions rather than fitted parameters. The most important assumptions are that thinking can be equated with generating mental image sequences, that knowledge can be represented as sentences and sketches, and that the CTU can extract useful information from internally generated images. The paper provides no quantitative evidence for these assumptions. No free parameters are fitted, and the only invented conceptual entity is the Mental Imagery Unit, which lacks an isolated test.

assumptions (4)
  • ad hoc to paper Thinking can be defined as the generation of a time series of mental images.
    Section 5.2 states this definition and uses it to frame the validation.
  • domain assumption Knowledge can be stored as natural language sentences and can be represented as sketches derived from previously seen images.
    Section 5.1 assumes a neural model exists that can produce such visual representations.
  • domain assumption The Cognitive Thinking Unit can interpret internally generated mental images and reason based on their content.
    Section 5.2 claims reasoning is performed based on the content of generated images, but no experiment verifies that image content influences the CTU output.
  • domain assumption Off-the-shelf pretrained models are sufficient building blocks for the proposed reasoning framework.
    Section 6 uses Faster R-CNN, BLIP, SentenceTransformer, Falcon, and Stable Diffusion as black boxes without evaluating their failure modes.
invented entities (1)
  • Mental Imagery Unit (MIU) as a reasoning component
    purpose: Generate and visualize sketch-like images from CTU prompts, then feed them back to support reasoning.
    The paper presents the MIU as beneficial to thinking, but no experiment isolates its contribution; the examples only show text-to-image generation from predefined prompts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can Mental Imagery Improve the Thinking Capabilities of AI Systems?." pith.science (2026). https://pith.science/paper/E2IP3GVK

@misc{pith2026250712555,
  author       = {Pith},
  title        = {Pith review of: Can Mental Imagery Improve the Thinking Capabilities of AI Systems?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E2IP3GVK}},
  note         = {Machine review of arXiv:2507.12555}
}
read the original abstract

Although existing models can interact with humans and provide satisfactory responses, they lack the ability to act autonomously or engage in independent reasoning. Furthermore, input data in these models is typically provided as explicit queries, even when some sensory data is already acquired. In addition, AI agents, which are computational entities designed to perform tasks and make decisions autonomously based on their programming, data inputs, and learned knowledge, have shown significant progress. However, they struggle with integrating knowledge across multiple domains, unlike humans. Mental imagery plays a fundamental role in the brain's thinking process, which involves performing tasks based on internal multisensory data, planned actions, needs, and reasoning capabilities. In this paper, we investigate how to integrate mental imagery into a machine thinking framework and how this could be beneficial in initiating the thinking process. Our proposed machine thinking framework integrates a Cognitive thinking unit supported by three auxiliary units: the Input Data Unit, the Needs Unit, and the Mental Imagery Unit. Within this framework, data is represented as natural language sentences or drawn sketches, serving both informative and decision-making purposes. We conducted validation tests for this framework, and the results are presented and discussed.

Figures

Figures reproduced from arXiv: 2507.12555 by the authors.

Figure 1
Figure 1. (Left)Diagram of Machine Thinking algorithms. (Right) Output of Human Thinking. 4.1 The "Needs" Unit This unit contains data that may include:(1) A set of actions to be performed, typically scheduled as a result of prior reasoning. These actions are generated by the Cognitive Thinking Unit (CTU) and transferred to the Needs Unit for exe￾cution planning. (2) Knowledge acquired by the machine, representing internal go… view at source ↗
Figure 2
Figure 2. Reconstruction of imagined images from brain activity [33]. The ability to decode and reconstruct these imagined visuals offers a wide range of promising possibilities from aiding individuals with disabilities, to en￾hancing creative processes [34]. In [34], authors train a modified version of an fMRI-to-image model and demonstrate the feasibility of reconstructing images from two modes of imagination: from memory a… view at source ↗
Figure 3
Figure 3. Results from strong imagination. The subject is given an instruction on the fMRI screen, such as ”Imagine a portrait representing optimism”. Then, the subject purely imagines such image, giving an oral description. Then, the associated brain data is fed to the fMRI-to-Image model to produce a reconstruction [34] [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Set of mental images generated from a scenario and reasoning queries associ￾ated. An emerging challenge is determining how a computational system can in￾terpret and engage with internally generated mental images. The understanding aspect is relatively straightforward, …
Figure 5
Figure 5. Figure 5: (Left) Some of captions are: "a car is seen in this image", "a bus is driving down the street", "a red car is parked in a parking", "a man in a white shirt and black pants is seen through a red square". (Right) A sample of used image for captioning. Detected objects ar…
Figure 6
Figure 6. Figure 6: A sample of used image sequence for captioning [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Mental images generated from the three prompts [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: The new mental images generated receiving from CTU unit new prompts. 7 Conclusion In this paper, we addressed the ambitious challenge of enabling machine think￾ing by proposing a comprehensive framework inspired by recent advancements in neuroscience, deep learning, an…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 28 canonical work pages

  1. [1]

    Monk, Gareth R

    Anna M. Monk, Gareth R. Barnes and Eleanor A. Maguire, The Effect of Object Type on Building Scene Imagery—an MEG Study, Front. Hum. Neurosci., 2020

    work page 2020

  2. [2]

    Redmon, S

    J. Redmon, S. Divvala, R. Girshick and A. Farhadi, You Only Look Once: Unified, Real-Time Object Detection, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.779-788, 2016. 14 S. Larabi

    work page 2016

  3. [3]

    Jean-François Bonnefon, and Iyad Rahwan, Machine Thinking, Fast and Slow, Trends in Cognitive Sciences, December , Vol. 24, No. 12, 2020

    work page 2020

  4. [4]

    Bartley,J.E.,Riedel,M.C.,Salo,T.etal.,Brainactivitylinksperformanceinscience reasoning with conceptual approach, npj Sci. Learn. 4, 20, 2019

    work page 2019

  5. [5]

    Li Wang, Meng Zhang, Feng Zou, Xin Wu, Yufeng Wang, Deductive-reasoning brain networks: A coordinate-based meta-analysis of the neural signatures in deductive reasoning, Brain and Behavior, Volume10, Issue12, 2020

    work page 2020

  6. [6]

    Noveck and J

    Vinod Goel, Gorka Navarrete, Ira A. Noveck and J. Prado, Editorial: The Reasoning Brain: The Interplay between Cognitive Neuroscience and Theories of Reasoning, Front. Hum. Neurosci., 2017

    work page 2017

  7. [7]

    Towards combining commonsense reasoning and knowledge acquisition to guide deep learning

    Sridharan, M., Mota, T. Towards combining commonsense reasoning and knowledge acquisition to guide deep learning. Auton Agent Multi-Agent Syst 37, 4 (2023). https://doi.org/10.1007/s10458-022-09584-4

    work page doi:10.1007/s10458-022-09584-4 2023

  8. [8]

    arXiv:1908.07490

    Hao Tan, Mohit Bansal LXMERT: Learning Cross-Modality Encoder Representa- tions from Transformers. arXiv:1908.07490. 2019

    arXiv 1908

Show all 37 references
  1. [9]

    Machine Reading Comprehension

    Chenguang Zhu. Machine Reading Comprehension. Algorithms and Practice. 2021, Pages 3-26. Elsevier 2021

    2021

  2. [10]

    ATOMIC: An Atlas of Machine Commonsense

    Sap, M., et al. ATOMIC: An Atlas of Machine Commonsense. AAAI. Pages 3027-

  3. [11]

    Antoine Bosselut, Hannah Rashkin, Maarten Sap, Chaitanya Malaviya, Asli Ce- likyilmaz, Yejin Choi COMET: Commonsense Transformers for Automatic Knowl- edge Graph Construction. 2019. https://doi.org/10.48550/arXiv.1906.05317

    2019 doi arXiv

  4. [12]

    Robyn Speer and Catherine Havasi. 2012. Representing General Relational Knowl- edge in ConceptNet 5. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), pages 3679–3686, Istanbul, Turkey

    2012

  5. [13]

    Adina Williams, Nikita Nangia, Samuel R. Bowman. (2018). A Broad-Coverage Challenge Corpus for Sentence Understanding through Inference. NAACL. https://doi.org/10.48550/arXiv.1704.05426

    2018 doi arXiv

  6. [14]

    Alec Radford et al. (2021). Learning Transferable Visual Models from Natural Language Supervision. https://doi.org/10.48550/arXiv.2103.00020

    2021 doi arXiv

  7. [15]

    What Does BERT Look at? An Analysis of BERT’s Attention

    Clark, C., et al. What Does BERT Look at? An Analysis of BERT’s Attention. (2019). https://doi.org/10.48550/arXiv.1906.04341

    2019 doi arXiv

  8. [16]

    Alayrac, J.-B., et al. (2022). Flamingo: A Visual Language Model for Few-Shot Learning. DeepMind. 36th Conference on Neural Information Processing Systems (NeurIPS 2022)

    2022

  9. [17]

    DenseCap: Fully Convolutional Localization Networks for Dense Captioning

    Justin Johnson, Andrej Karpathy, Li Fei-Fei. DenseCap: Fully Convolutional Localization Networks for Dense Captioning. 2016 IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2016, Pages: 4565-4574 DOI: 10.1109/CVPR.2016.494

    2016 doi

  10. [18]

    Colin Raffel et al., Exploring the Limits of Transfer Learning with a Unified Text- to-Text Transformer, Journal of Machine Learning Research, 2020, 21(140), pages 1–67

    2020

  11. [19]

    Human-like machine thinking: Language guided imagina- tion

    Feng Qi, Wenchuan Wu. Human-like machine thinking: Language guided imagina- tion. 2019. https://doi.org/10.48550/arXiv.1905.07562

    2019 doi arXiv

  12. [20]

    Deep Multimodal Data Fusion

    FEI ZHAO, CHENGCUI ZHANG, BAOCHENG GENG. Deep Multimodal Data Fusion. ACM Comput. Surv., Vol. 56, No. 9, Article 216. Publication date: April 2024

    2024

  13. [21]

    Heinrich Jiang, Ofir Nachum Identifying and Correcting Label Bias in Machine Learning. 2019. https://doi.org/10.48550/arXiv.1901.04966 Can Mental Imagery Improve the Thinking Capabilities of AI Systems? 15

    2019 doi arXiv

  14. [22]

    Bias in Machine Learning: A Literature Review

    Mavrogiorgos, K.; Kiourtis, A.; Mavrogiorgou, A.; Menychtas, A.; Kyriazis, D. Bias in Machine Learning: A Literature Review. Appl. Sci. 2024, 14, 8860. https://doi.org/10.3390/app14198860

    2024 doi

  15. [23]

    Venkatesh Babu

    Sunandini Sanyal, Ashish Ramayee Asokan, Suvaansh Bhambri, Ak- shay Kulkarni, Jogendra Nath Kundu, R. Venkatesh Babu. Domain- Specificity Inducing Transformers for Source-Free Domain Adaptation. 2023. https://doi.org/10.48550/arXiv.2308.14023

    2023 doi arXiv

  16. [24]

    Putting GPT-3’s Creativity to the (Alternative Uses) Test

    Claire Stevenson, Iris Smal, Matthijs Baas, Raoul Grasman, Han van der Maas. Putting GPT-3’s Creativity to the (Alternative Uses) Test. 2022. https://doi.org/10.48550/arXiv.2206.08932

    2022 doi arXiv

  17. [25]

    Forrest Huang, Eldon Schoop, David Ha, John Canny Scones: Towards Conversa- tional Authoring of Sketches. 2020. https://doi.org/10.1145/3377325.3377485

    2020

  18. [26]

    Sim- ulated Mental Imagery for Robotic Task Planning

    Shijia Li, Tomas Kulvicius, Minija Tamosiunaite, and Florentin Worgotter. Sim- ulated Mental Imagery for Robotic Task Planning. arXiv:2211.08321v3 [cs.RO] 27 Jul 2023

    2023 arXiv

  19. [27]

    Visual mental imagery: A view from artificial intelligence

    Maithilee Kunda. Visual mental imagery: A view from artificial intelligence. Cor- tex, Volume 105, August 2018, Pages 155-172

    2018

  20. [28]

    Representations of imaginary scenes and their properties in cortical alpha activity

    Stecher, R., Kaiser, D. Representations of imaginary scenes and their properties in cortical alpha activity. Nature, Sci Rep 14, 12796 (2024). https://doi.org/10.1038/s41598-024-63320-4

    2024 doi

  21. [29]

    Holmes, Stephen M

    Joel Pearson, Thomas Naselaris, Emily A. Holmes, Stephen M. Kosslyn. Mental Imagery: Functional Mechanisms and Clinical Applications. Volume 19, Issue 10, pp. 590-602, October 2015

    2015

  22. [30]

    Imagining predictions: mental imagery as mental emulation Philos

    Moulton, S.T., Kosslyn, S.M. Imagining predictions: mental imagery as mental emulation Philos. Trans. R. Soc. Lond. B: Biol. Sci. 2009; 364:1273-1280

    2009

  23. [31]

    Prospection: experiencing the future Science

    Gilbert, D.T., Wilson, T.D. Prospection: experiencing the future Science. 2007; 317:1351-1354

    2007

  24. [32]

    Elena Commodari, Jasmine Sole, Maria Guarnera, Valentina Lucia La Rosa. Men- tal imagery in education: What impact on the relationships with visuospatial pro- cessing and school performance in junior high school students? Thinking Skills and Creativity, Volume 54, December 202...

    2024

  25. [33]

    Mental image reconstruction from human brain activity: Neural decoding of mental imagery via deep neural network-based Bayesian estimation

    Naoko Koide-Majima, Shinji Nishimoto, Kei Majima. Mental image reconstruction from human brain activity: Neural decoding of mental imagery via deep neural network-based Bayesian estimation. Neural Networks 170 (2024) 349–363

    2024

  26. [34]

    Mind-to-Image: Projecting Visual Mental Imagination of the Brain from fMRI

    Hugo Caselles-Dupre, Charles Mellerio, Paul Herent, Alizee Lopez-Persem, Benoit Beranger, Mathieu Soularue, Pierre Fautrel, Gauthier Vernier, Matthieu Cord. Mind-to-Image: Projecting Visual Mental Imagination of the Brain from fMRI. arXiv:2404.05468v4 [q-bio.NC] 4 May 2024

    2024 arXiv

  27. [35]

    M., Ball, T

    Kosslyn, S. M., Ball, T. M., Reiser, B. J.. Visual images preserve metric spatial information: Evidence from studies of image scanning. Journal of Experimental Psychology: Human Perception and Performance, 4(1), 47–60.(1978)

    1978

  28. [36]

    Modelling mental imagery in the ACT-R cognitive architecture In A

    David Peebles. Modelling mental imagery in the ACT-R cognitive architecture In A. Goel, C. Seifert, C. Freksa (Eds.), Proceedings of the 41st Annual Meeting of the Cognitive Science Society. 2019. Montreal, Canada

    2019

  29. [3035]

    https://doi.org/10.1609/aaai.v33i01.33013027

    2019. https://doi.org/10.1609/aaai.v33i01.33013027

    2019 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.