Codex Mentis: Science and technology to study cognition

Unlock the Lab: Your guide to reading science like a scientist

🪄 Created using NotebookLM, with all the benefits and blind spots of human editing. In this episode of Codex Mentis, we explore the underlying machinery of scientific truth to understand how research reaches the public and why a healthy dose of scepticism is vital for its evaluation. The conversation begins with an overview of Dr Pablo Bernabeu’s interactive web application which uses a unique peer-anchored design to help users calibrate their judgements by predicting community standards across forty-eight fictional research scenarios. We discuss how this tool trains participants to identify critical red flags such as predatory publishing models, underpowered sample sizes and overblown conclusions that often mask mundane data behind sensationalised media narratives. Transitioning to real-world research integrity, the episode reviews a systematic meta-analysis quantifying the prevalence of misconduct and explores the pervasive culture of silence revealed by the stark discrepancy between those admitting to questionable research practices and those witnessing them in colleagues. We examine granular behaviours like hypothesising after results are known (HARKing) and salami publication before explaining the randomised response technique which is a mathematical method used in large-scale surveys to elicit honest answers about sensitive misconduct. The discussion also addresses qualitative findings that characterise academia as a 'bad barrel' where systemic 'publish or perish' pressures and an over-reliance on journal impact factors actively discourage the publication of valid negative results. Finally, we analyse a massive quantitative study of over forty-one million papers revealing a structural paradox where artificial intelligence tools accelerate individual careers and impact while simultaneously contracting the collective focus of science by automating established centres of knowledge rather than exploring unknown frontiers. Further details on the app and the workshop are available at https://pablobernabeu.github.io/applications-and-dashboards/unlock-the-lab [https://pablobernabeu.github.io/applications-and-dashboards/unlock-the-lab] References (in order of appearance) Bernabeu, P. (2026). Unlock the Lab: Your guide to reading science like a scientist (Version 1.0.0) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.19153148 Xie, Y., Wang, K., & Kong, Y. (2021). Prevalence of research misconduct and questionable research practices: A systematic review and meta-analysis. Science and Engineering Ethics, 27(4), Article 41. https://doi.org/10.1007/s11948-021-00314-9 Larsson, T., Plonsky, L., Sterling, S., Kytö, M., Yaw, K., & Wood, M. (2023). On the frequency, prevalence, and perceived severity of questionable research practices. Research Methods in Applied Linguistics, 2(3), Article 100064. https://doi.org/10.1016/j.rmal.2023.100064 Gopalakrishna, G., ter Riet, G., Vink, G., Stoop, I., Wicherts, J. M., & Bouter, L. M. (2022). Prevalence of questionable research practices, research misconduct and their potential explanatory factors: A survey among academic researchers in The Netherlands. PLoS ONE, 17(2), Article e0263023. https://doi.org/10.1371/journal.pone.0263023 Bruton, S. V., Medlin, M., Brown, M., & Sacco, D. F. (2020). Personal motivations and systemic incentives: Scientists on questionable research practices. Science and Engineering Ethics, 26(3), 1531–1547. https://doi.org/10.1007/s11948-020-00182-9 Hao, Q., Xu, F., Li, Y., & Evans, J. (2026). Artificial intelligence tools expand scientists’ impact but contract science’s focus. Nature, 679, 1237–1243. https://doi.org/10.1038/s41586-025-09922-y

Modality switch effects: The brain friction of switching senses

🪄 Created using NotebookLM, with all the benefits and blind spots of human editing. This episode explores whether the human mind functions as an abstract symbol processor or a physical simulator deeply rooted in bodily experience. We delve into the 'modality switch effect', a phenomenon where shifting from one sensory modality to another, such as from sound to sight, incurs a measurable cognitive penalty. Foundational research initially suggested that people are consistently slower when verifying properties of concepts across different senses, suggesting the brain must physically reconfigure its neural resources to understand language. However, later studies proposed that our brains might be efficient rather than thorough, often relying on 'quick and fuzzy' linguistic shortcuts before booting up heavy sensory simulations. New evidence from event-related potential studies shows that this sensory activation occurs as early as 160 milliseconds after seeing a word, reinforcing the idea that grounding is a fundamental part of accessing meaning. We also discuss findings that demonstrate how even second languages, typically learned in abstract classroom settings, recruit the body's native sensory systems. Furthermore, the latest research indicates that these perceptual simulations are so automatic they activate even during 'shallow' tasks where participants are not explicitly trying to process word meaning. Finally, we consider what this means for a world increasingly dominated by flat screens and artificial intelligence, questioning if a lack of physical interaction might lead to a shallowing of human thought. References (in order of appearance) Pecher, D., Zeelenberg, R., & Barsalou, L. W. (2003). Verifying different-modality properties for concepts produces switching costs. Psychological Science, 14(2), 119–124. https://doi.org/10.1111/1467-9280.t01-1-01429 Louwerse, M., & Connell, L. (2011). A taste of words: Linguistic context and perceptual simulation predict the modality of words. Cognitive Science, 35(2), 381–398. https://doi.org/10.1111/j.1551-6709.2010.01157.x Collins, J., Pecher, D., Zeelenberg, R., & Coulson, S. (2011). Modality switching in a property verification task: An ERP study of what happens when candles flicker after high heels click. Frontiers in Psychology, 2, Article 10. https://doi.org/10.3389/fpsyg.2011.00010 Hald, L. A., Marshall, J.-A., Janssen, D. P., & Garnham, A. (2011). Switching modalities in a sentence verification task: ERP evidence for embodied language processing. Frontiers in Psychology, 2, Article 45. https://doi.org/10.3389/fpsyg.2011.00045 Bernabeu, P., Willems, R. M., & Louwerse, M. M. (2017). Modality switch effects emerge early and increase throughout conceptual processing: Evidence from ERPs. In G. Gunzelmann, A. Howes, T. Tenbrink, & E. J. Davelaar (Eds.), Proceedings of the 39th Annual Conference of the Cognitive Science Society (pp. 1629-1634). Austin, TX: Cognitive Science Society. https://doi.org/10.31234/osf.io/a5pcz [https://doi.org/10.31234/osf.io/a5pcz] Platonova, O., & Miklashevsky, A. (2025). Warm and fuzzy: Perceptual semantics can be activated even during shallow lexical processing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 51(9), 1471–1496. https://dx.doi.org/10.1037/xlm0001429 Wentura, D., Shi, E., & Degner, J. (2024). Examining modal and amodal language processing in proficient bilinguals: Evidence from the modality-switch paradigm. Frontiers in Human Neuroscience, 18, Article 1426093. https://doi.org/10.3389/fnhum.2024.1426093

The dead salmon problem: Multiple tests, minimality and data-driven alternatives

🪄 Created using NotebookLM, with all the benefits and blind spots of human editing. In 2009, a deceased Atlantic salmon was placed inside a functional magnetic resonance imaging scanner to test its calibration parameters. Although the subject was undeniably dead, the standard statistical software produced results suggesting the fish was actively contemplating human emotions. This bizarre outcome highlights a systemic fragility in modern science known as the multiple tests trap, where conducting thousands of tests without adjustment guarantees that random noise will eventually look like a discovery. Just as flipping a coin enough times will inevitably produce a streak of ten heads, asking too many questions of a large dataset ensures that a researcher will find significant results purely by luck. Escaping this trap requires rigorous pre-planning and methodological self-restraint to avoid the statistical cheating known as hypothesising after the results are known. While the classical Bonferroni correction acts as a 'sledgehammer' by dividing the significance threshold by the total number of tests, more sensitive sequential procedures like the Holm-Bonferroni method offer a more refined approach. Modern researchers often prefer sophisticated data-driven strategies such as permutation testing, which shuffles experimental labels thousands of times to build a custom noise map specific to the dataset rather than relying on broad theoretical assumptions. Choosing between the precise spatial localisation of maximum t-statistic testing and the sensitive yet fuzzy cluster-based methods reveals that statistical truth is often a philosophical judgement call. Ultimately, the decision of how to define a family of tests depends on the logical structure of a scientific claim and the intent of the investigator. By embracing the principle of test minimality, researchers can move beyond mere p-value adjustments and toward a more robust, transparent and honest scientific practice. References Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57(1), 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x Bennett, C. M., Miller, M. B., & Wolford, G. L. (2009). Neural correlates of interspecies perspective taking in the post-mortem Atlantic Salmon: An argument for multiple comparisons correction. Neuroimage, 47(Suppl 1), S125. https://doi.org/10.1016/S1053-8119(09)71202-9 Cumming, G. (2014). The new statistics: Why and how. Psychological Science, 25(1), 7–29. https://doi.org/10.1177/0956797613504966 Frane, A. V. (2021). Experiment-wise type I error control: a focus on 2× 2 designs. Advances in Methods and Practices in Psychological Science, 4(1), 2515245920985137. https://doi.org/10.1177/2515245920985137 García-Pérez, M. A. (2023). Use and misuse of corrections for multiple testing. Methods in Psychology, 8, 100120. https://doi.org/10.1016/j.metip.2023.100120 Groppe, D. M., Urbach, T. P., & Kutas, M. (2011). Mass univariate analysis of event‐related brain potentials/fields I: A critical tutorial review. Psychophysiology, 48(12), 1711-1725. http://doi.org/10.1111/j.1469-8986.2011.01273.x Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6(2), 65–70. https://www.jstor.org/stable/4615733 Rubin, M. (2021). When to adjust alpha during multiple testing: A consideration of disjunction, conjunction, and individual testing. Synthese, 199(3-4), 10969–11000. https://doi.org/10.1007/s11229-021-03276-4

The digital parrot or the universal machine? Debating the mind in the model

🪄 Created using NotebookLM, with all the benefits and blind spots of human editing. Can a machine that writes Shakespearean sonnets about traffic jams actually help us understand the human soul? In this episode of Codex Mentis, we dive into a 'potential bomb' thrown into the heart of cognitive science: the rise of Large Language Models (LLMs) and their challenge to how we think humans learn to speak. For fifty years, the 'nativist' view, championed by Noam Chomsky, argued that children are born with a 'built-in grammar' because the speech they hear is too messy and 'impoverished' to learn from scratch—a concept known as the Poverty of Stimulus. However, new research suggests LLMs provide an 'existence proof' that complex grammar can indeed be mastered through pure statistical patterns alone, potentially refuting half a century of linguistic theory. But are these models truly 'brain-like,' or are we looking at a 'Cessna vs. Bird' problem? While both an aircraft and a bird achieve flight, their internal mechanisms are worlds apart. We explore the rigorous 'Four Questions' framework from ethologist Niklas Tinbergen to see where the comparison between silicon and synapse breaks down—from the lack of biological evolution to the 'unimodal' nature of text-only training. We also tackle the 'Grounding Problem' and the 'Spanish Dictionary' thought experiment: can a model truly 'understand' a sunset if it has only ever read descriptions of one? We discuss the fascinating dissociation between formal linguistic competence (grammar) and functional competence (thought), and why the model’s greatest failures—like its inability to handle unwritten sign languages or pass the BabyLM Challenge—might be its most important scientific gifts. Join us as we determine if LLMs are a new theory of the mind or simply the sharpest tool cognitive science has ever been handed. References (in order of appearance) Chomsky, N. (1980). Rules and representations. MIT Press. https://doi.org/10.1017/S0140525X00001515 Contreras Kallens, P., Kristensen-McLachlan, R. D., & Christiansen, M. H. (2023). Large language models demonstrate the potential of statistical learning in language. Cognitive Science, 47(3), e13256. https://doi.org/10.1111/cogs.13256 Piantadosi, S. T. (2024). Modern language models refute Chomsky’s approach to language. In E. Gibson & M. Poliak (Eds.), From fieldwork to linguistic theory: A tribute to Dan Everett (pp. 353–414). Language Science Press. https://doi.org/10.5281/zenodo.12665933 Cuskley, C., Woods, R., & Flaherty, M. (2024). The limitations of large language models for understanding human language and cognition. Open Mind: Discoveries in Cognitive Science, 8, 1058–1083. https://doi.org/10.1162/opmi_a_00160 Tinbergen, N. (1963). On aims and methods of ethology. Zeitschrift für Tierpsychologie, 20(4), 410–433. https://doi.org/10.1111/j.1439-0310.1963.tb01161.x Schrimpf, M., Blank, I. A., Tuckute, G., Kauf, C., Hosseini, E. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2021). The neural architecture of language: Integrative modeling converges on predictive processing. Proceedings of the National Academy of Sciences, 118(45), e2105646118. https://doi.org/10.1073/pnas.2105646118 Goldstein, A., Zada, Z., Buchnik, E., Schain, M., Price, A., Aubrey, B., Nastase, S. A., Feder, A., Emanuel, D., Cohen, A., Jansen, A., Gazula, H., Choe, G., Rao, A., Kim, C., Casto, C., Fanda, L., Doyle, W., Friedman, D. … Hasson, U. (2022). Shared computational principles for language processing in humans and deep language models. Nature Neuroscience, 25, 369–380. https://psycnet.apa.org/doi/10.1038/s41593-022-01026-4 Bender, E. M., & Koller, A. (2020). Climbing towards NLU: On meaning, form, and understanding in the age of data. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (pp. 5185–5198). https://doi.org/10.18653/v1/2020.acl-main.463 Further references available at https://youtu.be/7lOVAkCk-sc

Beyond the cloud: Reclaiming data sovereignty in speech transcription

🪄 Created using NotebookLM, with all the benefits and blind spots of human editing. In this episode of Codex Mentis, we explore the critical intersection of generative AI and research methodology, focusing on a production-ready, open-source workflow for secure speech transcription developed by Dr Pablo Bernabeu. While OpenAI’s Whisper models have set a new gold standard for speech-to-text accuracy, relying on consumer-grade cloud interfaces like ChatGPT or Google Gemini often proves incompatible with the rigorous demands of academic and clinical research. We dissect the three primary limitations of these cloud-based tools—restrictive file size caps, a lack of methodological reproducibility, and the significant privacy and GDPR risks inherent in transmitting sensitive human data to third-party servers. The discussion highlights a sophisticated alternative that leverages high-performance computing environments to achieve complete data sovereignty by running transcription entirely offline within a secure institutional perimeter. We break down the engineering behind this transition, including the use of SLURM job scheduling for unlimited scalability across GPU nodes and the implementation of advanced quality controls to fix common AI hallucinations such as spurious repetitions and accidental language switching. Furthermore, we examine the system's intelligent, multi-tiered approach to personal name masking and speaker diarisation, which ensures participant anonymity and structured dialogue without compromising the semantic integrity of the research data. This episode provides a comprehensive look at how researchers can balance the power of modern AI with the non-negotiable requirements of ethical compliance and long-term scientific sustainability. Sources and related content can be consulted at https://pablobernabeu.github.io/2025/speech-transcription-python