For decades, popular culture has treated the human mind like a biological hard drive, imagining that somewhere within the dense folds of the cerebral cortex sits a file cabinet of recorded life, from the melody of a childhood lullaby to the opening measures of Beethoven’s Fifth Symphony. In this prevailing view, recalling the past is simply a matter of hitting playback. Yet across cognitive psychology and computational neuroscience, an enduring debate has destabilized that picture. Scientists and theorists are grappling with a fundamental question: does the brain actually store data representations, or has science spent half a century clinging to a seductive but misleading technological metaphor?
The fault lines broke into public view when psychologist Robert Epstein published his 2016 essay, “The Empty Brain,” in Aeon. Epstein argued that cognitive science had spent six decades trapped in an intellectual dead end by equating the brain to an information-processing computer. To Epstein, searching the nervous system for an encoded biological copy of Beethoven’s Fifth Symphony, or for stored images, words, and rules, is an exercise in futility because no such physical files exist. To illustrate his non-representational stance, Epstein pointed to a simple demonstration: when human subjects are asked to draw a one-dollar bill from memory, they inevitably produce crude, inaccurate sketches lacking serial numbers, specific lettering, and exact imagery. Only when an actual bill is placed before them can they render its details accurately. Epstein argued this disparity reveals that memory does not operate by retrieving an internal picture or symbolic blueprint. Instead, the brain simply becomes attuned to recognize and interact with its environment through experience.
Epstein’s critique sparked immediate pushback from mainstream cognitive scientists and computational neuroscientists. Complex systems theorist David Krakauer and cognitive scientist Gary Marcus countered that Epstein had attacked a cartoonish straw man of the computational model. In modern computer science and computational biology, terms like “computation” and “information” do not denote the literal architecture of a desktop computer sorting discrete MP3s and JPEGs into directory folders. Instead, computational neuroscientists conceptualize neural representations as distributed population codes, multidimensional vector spaces, and synaptic weight matrices across vast networks. Through this computational lens, physical structures corresponding to Beethoven’s Fifth Symphony do exist. They reside not as a biological audio file, but as distributed patterns of synaptic connectivity calibrated to recognize or regenerate the temporal rhythms and harmonic progressions of the piece.
This computational perspective aligns with more than a century of biological trace research. Historically pioneered by Richard Semon in 1904, the study of physical memory substrates—known as engrams—has been confirmed by modern neurobiologists including Susumu Tonegawa, Sheena Josselyn, and Tomás Ryan. Their investigations demonstrate that specific memories correspond to identifiable physical substrates. During learning, sparse, distributed ensembles of neurons undergo lasting molecular and structural changes, including dendritic spine growth, epigenetic remodeling, and long-term potentiation or depression. In rodent brains, optogenetic experiments have demonstrated that artificially reactivating specific engram ensembles triggers the retrieval of an associated learned behavior, such as fear conditioning, establishing a direct causal link between physical cell assemblies and recall. Yet these engrams do not function as internal projection screens displaying Beethoven’s score. Rather, they operate as dispositional networks that enable an organism to regenerate past perceptual or motor patterns within an active living context.
Other researchers argue that even the language of information processing obscures the biological reality. In dynamical systems neuroscience, learning does not store an inert object; it alters the geometry of the nervous system by deepening “basins of attraction.” When a person encounters a familiar cue, such as the famous opening notes of Beethoven’s Fifth Symphony, the brain’s electrical trajectory is pulled into an attractor basin. Under this dynamic, non-storage view, a memory is an emergent dynamic trajectory rather than a static record waiting to be fetched. This framework harmonizes with Radical Enactivism, articulated by thinkers such as Daniel Hutto, Erik Myin, and Anco Peeters, who contend that cognition is direct, embodied action coupled with the environment rather than internal computation over mental models. Alongside this sits Trace Minimalism, a philosophical position that accepts that the brain retains physical alterations like engrams or synaptic shifts from past events, but rejects the claim that these traces carry symbolic messages or representational content. Under these models, remembering a symphony is an active, imaginative bodily and neural reenactment rather than data retrieval.
This conceptual evolution gained further ground with neuroscientist György Buzsáki’s 2019 book, “The Brain from Inside Out.” Buzsáki challenged the classical empiricist premise that the brain passively absorbs stimuli from the outside world, encodes them into internal representations, and files them away. Instead, Buzsáki argues that the brain is an action-generating, internally organized system that creates meaning through self-generated movement and predictive testing of the environment. Within this framework, hippocampal mechanisms conventionally labeled as episodic memory never evolved to store passive historical archives. Instead, they evolved from spatial navigation networks composed of place cells and grid cells. Recalling a memory utilizes the exact same internal sequence generator used to plan a path through physical space, running the neural machinery offline to simulate possible actions rather than reading from a static repository.
Modern episodic memory research increasingly supports this prospective architecture, showing that every act of remembering is a novel reconstruction assembled from fragments, emotional context, current cues, and prospective goals. The observed vulnerability of human memory to distortion, post-retrieval reconsolidation, and confabulation reinforces the model that the brain preserves plastic recipes for reassembly rather than immutable digital snapshots. Ultimately, the theoretical debate over biological memory hinges on how researchers define “representation.” While critics charge anti-representationalists with conflating desktop hardware with formal computational theory, proponents of enactive and dynamical views counter that computational metaphors mask the continuous, living nature of neural dynamics. What is clear is that cognitive neuroscience is shifting away from warehouse and library metaphors, viewing memory instead as a generative, navigational system shaped to guide impending action rather than document history.