Jared Edward Reser Ph.D.

For more than 15 years, I have been developing a model of cognition, and I’m using this post to trace its origins through a series of published and self published work.

The model is organized around a simple observation: the contents of working memory do not disappear and reappear as a completely new configuration at each moment. Some representations remain active, some are displaced, and others are added. Each new cognitive state is therefore a partially revised iteration of the state that preceded it.

This partial inheritance allows information to accumulate across time. A goal can remain active while different possibilities are considered. A question can survive while memories and associations are retrieved. Several automatic operations can build on one another until their combined products amount to deliberate reasoning. The retained contents of working memory also act together as search parameters, helping to determine which representations, images, words, and actions are likely to enter the next state.

In 2016, I introduced state-spanning coactivity, or SSC, to describe representational content shared across successive brain states. I used incremental change in state-spanning coactivity, or icSSC, to describe the progressive turnover of this shared content. The paper connected these processes to working memory, mental continuity, progressive mental imagery, consciousness, and artificial-intelligence architecture. 

The theory has subsequently expanded in several directions. It has become an account of how working memory organizes associative search, how individually nonconscious processes compose a conscious stream, how a humanlike cognitive architecture might be constructed in software, how motives and defenses shape the transition between conscious states, and how an artificial system might eventually develop functional awareness, intentionality, and perhaps phenomenal consciousness.

The following chronology brings together the major publications, presentations, public essays, and archival materials involved in that development. The formal publications provide the principal academic sources. The website articles and essays document how individual components emerged over time. These web pages were captured by the Internet Archive during the years in which they first appeared, providing contemporaneous public records of the theory’s development.

Early foundations: belief, coactivity, and mental continuity

Reser, J. E. (2009). Psychological correlates of certainty strength: Measuring the relative influence of evidence, opinion of personal contacts, and importance to self-identity⁠ [Master’s thesis, University of Southern California].

This thesis was not yet a formulation of the Iterative Updating Model. It established the broader problem from which the model emerged: beliefs and judgments are shaped by several simultaneously operating sources of influence, including evidence, social information, and self-relevance. The later working-memory theory attempted to explain how multiple determinants can become jointly active, compete for influence, and accumulate into a consciously endorsed judgment.

Reser, J. E. (2011). Working memory, functional coactivity, and spreading activation [Poster presentation]. USC Neuroscience Graduate Symposium, University of Southern California, Los Angeles, CA, United States.

This was the earliest formal presentation centered specifically on working memory, coactive representations, and spreading activation. It proposed that the representations active together in working memory influence one another and collectively determine where activation spreads next. That premise eventually became polyassociativity, multiassociative search, and cospreading.

Reser, J. E. (2011). What determines belief? The philosophy, psychology and neuroscience of belief formation and change⁠. VDM Verlag Dr. Müller. ISBN 978-3-639-35331-0.

This book provided the broader cognitive and philosophical background for the model. It examined the relation between conscious reasoning, implicit influence, association, memory, evidence, and the organization of belief. Later work narrowed the focus from the determinants of belief to the temporal architecture through which representations are held, combined, searched, and progressively transformed.

Reser, J. E. (2011, January 1). An analogy between the neurophysiology of thought and the polypedal locomotion of an octopus⁠. Observed Impulse.

The octopus analogy supplied an early visual and intuitive description of partial state inheritance. An octopus maintains several footholds while releasing others and reaching for new ones, much as working memory may preserve several representations while replacing others. The analogy later appeared in more formal form in the 2016 paper as a way to illustrate limited capacity, staggered persistence, and incremental turnover.

Reser, J. E. (2011, January 11). Do we have an unconscious mind or simply a few unconscious processes?⁠ Observed Impulse.

This early essay questioned whether the unconscious should be treated as a cohesive hidden agent. It proposed instead that many distributed unconscious processes influence thought and behavior without collectively forming a second person inside the brain. This distinction later became central to the theory that consciousness is composed from nonconscious processes and to the framework eventually called Iterative Psychodynamics.

Reser, J. E. (2011, January 31). Poster presentation on the mechanics of thought⁠. Observed Impulse.

This post created a public record of the USC working-memory presentation. It preserved the early argument that thought depends on sustained functional coactivity and that active representations spread activation through an associative network. Its importance is partly theoretical and partly archival, since it documents the early public chronology of the model.

Reser, J. E. (2011, October 6). The brain basis for the continuity of thought⁠. Observed Impulse.

This essay stated one of the model’s foundational observations: neural activity does not change everywhere at once. Some representations persist longer than others, causing successive cognitive states to share content. Mental continuity was therefore attributed to differential persistence and gradual change within the active neural population.

Reser, J. E. (2011, December 21). How thought propagates⁠. Observed Impulse.

This article used successive configurations of letter-coded representations to show how thought could propagate through retention, displacement, and addition. The representations that remained active from the previous state joined together in determining which representation would be activated next. The scheme foreshadowed the later concepts of polyassociativity, cospreading, and successor-state construction.

Reser, J. E. (2011, December 21). An analogy for top-down to bottom-up communication: Cake baking and tasting⁠. Observed Impulse.

The cake-making analogy addressed the reciprocal relationship between higher association areas and lower sensory systems. Association areas provide abstract specifications from which sensory systems construct a more concrete image, after which the resulting image is analyzed and returned as new information. This recursive exchange later became progressive imagery modification and the broader proposal that thought advances through repeated transformations between working-memory contents and modality-specific imagery.

Reser, J. E. (2012). Assessing the psychological correlates of belief strength: Contributing factors and role in behavior⁠ (Publication No. 3513834) [Doctoral dissertation, University of Southern California]. ProQuest Dissertations & Theses Global.

The dissertation provided the first extended academic home for the emerging working-memory architecture. It included models of selection, inclusion, retention, and displacement in working memory, along with diagrams depicting reciprocal communication between association and sensory areas. The dissertation connected the dynamics of working memory to belief formation, imagery, judgment, and the continuity of cognition.

Reser, J. E. (2012, February 10). How thought propagates, part 2⁠. Observed Impulse.

This second treatment developed the reciprocal interaction between a relatively enduring set of higher-order representations and rapidly changing sensory imagery. Conscious cognition was described as a partially conserved, progressive process in which earlier contents remained available while new imagery was generated and interpreted. The account moved closer to identifying consciousness with an evolving sequence rather than any isolated state.

Reser, J. E. (2012, August 21). Firework analogy for sustained neural firing⁠. Observed Impulse.

The firework analogy offered another way to visualize temporal overlap. Traces from earlier neural events remain while later events occur, causing successive configurations to share elements rather than replacing one another instantaneously. The analogy helped make the proposed continuity mechanism concrete before the terminology of SSC and icSSC had been introduced.

Reser, J. E. (n.d.-a). On the unconscious: Is there an unconscious mind or only unconscious processes?⁠ Cognitive Mechanics. Retrieved August 6, 2013

Although the live page is undated, it was publicly archived by the Internet Archive in 2012. The essay developed a detailed distinction among conscious, preconscious, unconscious, and nonconscious processing. It also introduced implicitization, through which an explicit A-to-B-to-C reasoning sequence becomes an increasingly automatic A-to-C transition, and proposed that conscious thought may arise when selected unconscious associations remain active across successive cycles.

From mental continuity to an explicit artificial architecture

Reser, J. E. (2013, April 18). Artificial intelligence programmed to simulate mental continuity between processing states⁠. Observed Impulse.

This article made the transition from a descriptive theory of human thought to an explicit artificial cognitive architecture. It proposed a modular hierarchy modeled on association, sensory, and motor systems, with a working-memory updating function that preserved partial overlap between successive states. It also connected the architecture directly to machine consciousness, artificial general intelligence, and artificial superintelligence.

Reser, J. E. (2013, July 12-15). The neurological process responsible for mental continuity: Reciprocating transformations between a working memory updating function and an imagery generation system [Poster presentation]. 17th Annual Meeting of the Association for the Scientific Study of Consciousness, San Diego, CA, United States. Poster P2-035.

This presentation brought the model into the consciousness-science community. It framed mental continuity as the product of reciprocating transformations between a working-memory updating system and an imagery-generation system. The presentation is cited as a direct precursor in the reference list of the 2016 journal article. 

Incremental change in state spanning cortical.pdf

Reser, J. E. (2013, July 16). The neurological process responsible for internally generated thought: Short term continuity⁠. Observed Impulse.

This post preserved a public version of the conference-stage model. It made the poster’s diagrams and principal mechanisms available outside the meeting and documented the proposal that sustained higher-order representations guide successive rounds of internally generated imagery. It therefore serves as both a theoretical source and a public timestamp for this stage of development.

Reser, J. E. (n.d.-b). The neurological process responsible for mental continuity and internally generated thought: Reciprocating transformations between a working memory updating function and multiple imagery generation systems⁠ [Web manuscript]. Cognitive Mechanics. Retrieved August 6, 2026.

This expanded manuscript was publicly archived by the Internet Archive in 2013. It presented a fuller version of the relationship among maintained association-area representations, discarded contents, newly activated representations, and multiple imagery systems. It also developed implications for intelligence, internally generated thought, the distinction between fluid and crystallized cognition, and artificial implementation.

Reser, J. E. (2016). Incremental change in the set of coactive cortical assemblies enables mental continuity. Physiology & Behavior, 167, 222-237. https://doi.org/10.1016/j.physbeh.2016.09.019⁠

This peer-reviewed article supplied the model’s principal formal terminology. State-spanning coactivity referred to content shared across consecutive brain states, while incremental change in state-spanning coactivity referred to the changing membership of that active set as some representations persisted, others deactivated, and new ones entered. The paper proposed that the contents of working memory exhibit SSC and that the progression of working memory over time exhibits icSSC. 

Incremental change in state spanning cortical.pdf

The paper also introduced polyassociativity, in which several coactive representations pool their influence to recruit a contextually appropriate addition. It proposed progressive imagery modification as a repeated exchange between enduring association-area contents and rapidly reconstructed sensory maps. Figure 4 on page 8 depicts representations being maintained, displaced, and added across successive working-memory states, while Figure 6 on page 11 depicts retained association-area contents guiding successive rounds of imagery generation. 

The article also advanced the possibility that System 2 cognition consists of System 1 operations whose products are retained and coordinated across time. On this account, individual operations remain automatic, while their sequential organization permits multistep reasoning, planning, and metacognition. The paper concluded with a proposed multimodular AI architecture capable of using icSSC for pattern recognition, prediction, decision-making, and adaptive control. 

Reser, J. E. (2018, May 29). A neural model of working memory and mental continuity: The Iterative Updating Model⁠. Observed Impulse.

This article explicitly named the framework the Iterative Updating Model. It translated the more technical language of SSC and icSSC into a direct account of working-memory updating: preserve some contents, remove others, and add contextually selected representations. It also developed variable updating rates, tight and loose iterative coupling, thought boundaries, multiassociative search, iterative inhibition, subsolution merging, and artificial implementation.

Architectural extensions, interpretability, and modular AGI

Reser, J. E. (2019, May 22). Solving the AI control problem: Transmit its mental imagery and inner speech to a television monitor⁠. Observed Impulse.

This essay applied the architecture to AI safety and interpretability. If an artificial system possessed explicit working-memory contents, imagery, and inner speech, those internal states could be translated into an externally inspectable stream. The proposal anticipated the later concept of generative interpretability, in which hidden processing is continually rendered into formats that human observers can understand.

Reser, J. E. (2019, October 30). Comparing information processing in the brain to that in the computer⁠. Observed Impulse.

This article clarified the architectural differences between biological and conventional digital information processing. The brain uses distributed representations, content-addressable retrieval, recurrent communication, and overlapping active states, whereas ordinary computers sharply separate storage, processing, and serial instruction execution. The comparison helped identify which properties would have to be added to conventional software to approximate biological working memory.

Reser, J. E. (2019, October 30). Computer short-term memory is not working memory, but it could be⁠. Observed Impulse.

This article distinguished passive temporary storage from functional working memory. A cache or context buffer becomes more analogous to working memory when its contents remain actively available, interact with one another, guide associative retrieval, and influence subsequent processing. This distinction later became important in comparisons between transformer context windows and explicit internal working-memory systems.

Reser, J. E. (2020, March). Artificial intelligence needs to utilize the process of myelination⁠. Observed Impulse.

This article added a developmental dimension to the proposed AI architecture. Biological cortical systems do not all mature simultaneously, and higher association areas undergo prolonged development after lower sensory and motor systems have established more reliable representations. An artificial system might benefit from a similarly staged curriculum in which lower-level modules stabilize before increasingly abstract and temporally extended processing is permitted to dominate.

Reser, J. E. (2020, June 16). The stream of thought is a looping vector⁠. Observed Impulse.

This essay expanded the model beyond a single linear train of thought. It considered detours, suspended goals, recurring themes, and the ability to return to a previous cognitive trajectory after temporarily attending elsewhere. The short-term store was treated as a depot from which recently active representations could be reinstated, allowing thought to loop back toward unfinished concerns.

Reser, J. E. (2020, June 16). Why having right and left cortical hemispheres might be important for superintelligent AI⁠. Observed Impulse.

This article explored the importance of differentiated but interacting subsystems. Rather than constructing one homogeneous neural network, an advanced artificial architecture might benefit from complementary systems with partially different biases, representations, and processing strategies. Their interaction could produce richer comparison, correction, and synthesis within the common workspace.

Reser, J. E. (2021, October 7). A language model like GPT could serve as Broca’s area in a larger AGI system⁠. Observed Impulse.

This article positioned a large language model as one specialized module within a broader cognitive architecture. Language generation would interact with working memory, perception, imagery, autobiographical memory, valuation, and action rather than constituting the entire cognitive system. The proposal anticipated later efforts to embed language models within multimodal, memory-augmented, tool-using, and agentic architectures.

The two arXiv cognitive architectures

Reser, J. E. (2022a). Artificial intelligence software structured to simulate human working memory, mental imagery, and mental continuity [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2204.05138⁠

This article presented an engineering-oriented version of the architecture. It described interconnected neural networks corresponding to specialized cortical systems, a global workspace, active and latent working-memory stores, reciprocal imagery generation, and successive states with overlapping representational content. It offered a direct route from the proposed neuroscience to software capable of exhibiting mental continuity and potentially synthetic consciousness.

Reser, J. E. (2022b). A cognitive architecture for machine consciousness and artificial superintelligence: Thought is structured by the iterative updating of working memory [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2203.17255⁠

This was the most comprehensive statement of the Iterative Updating Model up to that point. It integrated sustained firing, synaptic potentiation, the focus of attention, the short-term store, multiassociative search, iterative overlap, progressive mental imagery, prediction, modularity, training, machine consciousness, AGI, and ASI. The manuscript described each working-memory state as a revised iteration of its predecessor, with part of the preceding representational configuration preserved while new contents were introduced. 

Thought iterative updating(3).pdf

The two arXiv papers are complementary. The first concentrates on the structure of the proposed software, while the second provides a broader cognitive and neuroscientific theory of how thought could be organized within such a system.

Interpretability, cognitive scenes, and temporal consciousness

Reser, J. E. (2024, October 7). Generative interpretability: Pursuing AI safety through the visualization of internal processing states⁠. Observed Impulse.

This article expanded the 2019 monitoring proposal into the named concepts of generative interpretability and generative explainability. Instead of attempting to infer cognition solely from opaque activation patterns, an artificial architecture could be designed to render its working-memory contents, imagery, intermediate conclusions, and internal language continuously. Interpretability would then become a property of the architecture rather than an analysis imposed only after training.

Reser, J. E. (2025, April 7). The specious present in machine consciousness: Why temporal continuity is the missing key to AI awareness⁠. Observed Impulse.

This article connected iterative updating to the philosophical and psychological concept of the specious present. Conscious experience appears to contain a temporally extended present in which recent events remain available while new events occur. The article proposed that a conscious machine would likewise require an internally maintained interval of overlapping states rather than a succession of computational moments joined only from an external observer’s perspective.

Reser, J. E. (2025). Hand drawn sketches⁠ [PDF archive]. Academia.edu. 

This archive preserves diagrams created across the development of the working-memory and AI architecture, including sketches that predate the formal publications. It is best treated as historical and developmental documentation rather than as an independent theoretical or empirical source. The drawings nevertheless help show that the architecture was developed visually through representations of active stores, module interactions, state overlap, imagery loops, and artificial implementations.

Reser, J. E. (2025, October 17). From context windows to cognitive scenes: Why AI thinking should be event-oriented⁠. Observed Impulse.

This article distinguished a large textual context window from a bounded cognitive scene. A cognitive scene is an internally coherent, capacity-limited configuration containing the objects, agents, goals, relations, and action possibilities relevant to the present event. This proposal added event segmentation and scene construction to the architecture, helping explain how an AI might maintain a usable focus without treating its entire available context as equally active.

Reser, J. E. (2025, November 20). A temporal theory of consciousness via iterative updating and state-spanning coactivity⁠. AI Thought.

This article explicitly presented iterative updating as a response to the snapshot problem in theories of consciousness. A theory that explains only why a representation is conscious at one instant does not yet explain how conscious moments become one continuous experience. SSC and icSSC were proposed as the mechanisms through which a present state inherits context, meaning, and causal organization from its predecessors.

Reser, J. E. (2025, December). Language models are trapped in token-bound time with token-locked receptive fields⁠. Observed Impulse.

This essay introduced Token-Bound Time and Token-Locked Receptive Fields. Language models process information through a sequence of token-indexed operations, while biological cognition can preserve multimodal representations across continuously unfolding sensory, bodily, emotional, and motor activity. The contrast clarified why a large context window may approximate continuity without possessing an explicit, self-maintaining, embodied workspace.

Explicit overlap substrates and artificial intentionality

Reser, J. E. (2026, January 16). Why transformers approximate continuity, why we keep building prompt workarounds, and what an explicit overlap substrate would change⁠. Iterated Insights.

This article developed the contrast between external record-based continuity and intrinsic state continuity. Prompt summaries, scratchpads, retrieval systems, and agent-state files allow transformers to recover prior information, but they remain compensatory mechanisms unless the system itself maintains and selectively updates an internal representational coalition. An explicit overlap substrate would implement keep, drop, add, and reinstate operations as native cognitive dynamics.

Reser, J. E. (2026, June 16). Valenced autobiographical control memory: Innate valence priors, combinatorial incentive templates, and working-memory sustain as a basis for artificial intentionality⁠. Iterated Insights.

This article added motivation, value, and autobiographical organization to the architecture. Valenced memories and incentive templates would influence which goals and concerns remain active, while a dopamine-like sustain mechanism would regulate their persistence in working memory. Artificial intentionality was therefore linked to the capacity to preserve valued goals, compare anticipated outcomes, and integrate action with an autobiographical record.

State-Spanning Workspace Theory

Reser, J. E. (2026, July 10). State-spanning workspace theory of consciousness: Iterative updating, diachronic unity, and the progressive evolution of experience⁠. Iterated Insights.

State-Spanning Workspace Theory, or SSWT, integrated SSC and icSSC with global workspace theory. It described consciousness as a functionally global, recurrently grounded, self-conditioning trajectory of partially overlapping representational coalitions. Each state inherits selected contents, broadcasts them to specialized systems, receives transformed candidate contributions, and helps construct its successor.

The theory added a more explicit transition law to the earlier model. Multiassociative search selects additions in relation to the retained configuration as a whole, while progressive imagery modification connects abstract workspace contents to sensory, interoceptive, affective, linguistic, and motor representations. SSWT also distinguished basal phenomenality, conscious access, stream consciousness, reflective consciousness, and self-consciousness rather than treating them as one indivisible capacity. 

State-Spanning_Workspace_Theory_of_Consciousness.docx

Reser, J. E. (2026, July 21). Iterative updating and compositional generalization in artificial intelligence⁠. Iterated Insights.

This article applied iterative updating to compositional generalization. Retained relational structures can serve as scaffolds while individual objects, roles, values, and operations are replaced, allowing familiar components to be recombined in unfamiliar ways. The model also provides a way to preserve intermediate subsolutions and later merge them into a larger solution without losing the constraints established earlier in the reasoning sequence.

Reser, J. E. (2026, July 21). The Iterative Artificial Consciousness Scale: A working-memory framework for functional awareness, sentience, and artificial superconsciousness⁠. Iterated Insights.

This article translated the architecture into a graded scale of artificial consciousness. The proposed levels extend from nonconscious computation through global access, iterative continuity, self-modeling, affective valuation, sentience, and possible artificial superconsciousness. The scale also suggested measurable dimensions such as representational overlap, update rate, working-memory persistence, multimodal reach, metacognitive access, autobiographical continuity, and valenced control.

Reser, J. E. (2026, July 30). Iterative phenomenalization: A working-memory account of qualia and conscious experience⁠. Iterated Insights.

This article extended the model from conscious access and mental continuity to phenomenal character. Sensory, bodily, emotional, and conceptual systems possess different representational geometries, and their contributions acquire temporally extended presence when they are repeatedly incorporated into the evolving workspace. Qualitative experience was therefore associated with the iterative global presence of modality-specific and affective contents rather than with an abstract amodal signal alone.

Reser, J. E. (2026, July 30). Time as iterative causal updating: State inheritance across physical, biological, and cognitive systems⁠. Iterated Insights.

This article generalized state inheritance beyond working memory. Physical systems, organisms, and cognitive processes all persist through transformations in which later states inherit causal structure from earlier states. The broader framework helped place mental continuity within a general account of process, while preserving the additional organizational requirements needed for consciousness.

Reser, J. E. (2026, August 2). Paniterativism: Iterative state inheritance as a general principle of physical, biological, and cognitive continuity⁠. Iterated Insights.

Paniterativism developed the philosophical implications of iterative causal updating. It proposed that inheritance with transformation is a widespread feature of temporally extended processes, from physical systems to organisms and conscious streams. The position remains materialist and does not imply that every iterating system is conscious; consciousness requires a special form of integrated, differentiated, globally consequential iterative organization.

Reser, J. E. (2026, August 5). Consciousness from nonconscious processes: State-Spanning Workspace Theory and the iterative composition of the conscious stream⁠. Iterated Insights.

This article addressed how consciousness could be composed entirely from operations that are individually nonconscious. Distributed perceptual, mnemonic, affective, predictive, linguistic, and motor processors generate candidate contributions without themselves possessing a unified point of view. Their products become constituents of consciousness when they participate in a recurrent global coalition and help construct a temporally connected sequence of successor states.

The article distinguished synchronic composition, through which multiple processors contribute to one integrated present; diachronic composition, through which selected contents are inherited across states; and recursive composition, through which the current coalition conditions the processing that constructs its successor. Consciousness was identified with the organization of the resulting trajectory rather than with a separate conscious ingredient added to otherwise unconscious computation.

Reser, J. E. (2026, August 6). Iterative psychodynamics: How distributed unconscious processes compose and shape the conscious stream⁠. Iterated Insights.

Iterative Psychodynamics returned to the questions raised in the early work on the unconscious. It retained the reality of unconscious motives, conflicts, defensive patterns, denial, rationalization, ambivalence, and incomplete self-knowledge while describing the unconscious as a distributed ecology rather than a second cohesive agent.

The theory proposed that unconscious dispositions are partly encoded in the transition structure of cognition. Motives alter which successors are likely to be constructed, defenses alter which contents are admitted or maintained, denial involves defensive non-integration with the self-model, and repetitive patterns reflect compiled transition policies. Trajectory compilation describes the conversion of repeatedly traversed conscious sequences into automatic structure, while trajectory reconstruction describes the process through which previously omitted causal steps become explicit and available for revision.

A cumulative research program

Taken together, these works describe a cumulative research program. The earliest writings identified differential neural persistence, functional coactivity, reciprocal cortical processing, and the partial conservation of mental content. The 2016 article supplied the formal concepts of SSC, icSSC, polyassociativity, and progressive imagery modification. The 2018 and 2022 works translated these mechanisms into an explicit theory of working memory and a proposed artificial cognitive architecture.

The later articles extended the architecture into interpretability, modular language-model integration, event-oriented cognitive scenes, temporal consciousness, artificial intentionality, compositional generalization, graded artificial consciousness, phenomenal experience, process philosophy, and psychodynamics. Each extension addressed a different consequence of the same underlying organization.

The central proposition has remained stable throughout:

Thought is a self-conditioning sequence of partially overlapping representational states. Some contents remain active, others are displaced, and new contents are recruited in relation to the retained configuration. This iterative process gives cognition continuity, permits automatic operations to accumulate into deliberation, allows nonconscious processes to compose a conscious stream, and provides a possible architectural basis for humanlike and superhuman artificial intelligence.