Comprehensive List: 74 Phenomena Explained by Autoregressive Theory

The autoregressive brain framework proposes a single computational principle underlying cognition: each mental state is generated as the optimal continuation of the preceding trajectory. Below is the full compilation of phenomena this principle explains.

Language Phenomena

1. Garden-path sentences — Mandatory commitment to single parse, costly reanalysis when wrong (e.g., “The horse raced past the barn fell”)
Frazier & Rayner (1982). Making and correcting errors during sentence comprehension. Cognitive Psychology, 14, 178-210.

2. Incremental parsing — Each word processed as it arrives; architectural necessity, not choice
Marslen-Wilson (1973). Linguistic structure and speech shadowing at very short latencies. Journal of Verbal Learning and Verbal Behavior, 12, 585-605.

3. Structural priming — Recent syntactic structures bias subsequent generation
Bock (1986). Syntactic persistence in language production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 336-346.

4. N400 — Neural signature of prediction error; graded metabolic cost for low-probability continuations
Kutas & Hillyard (1980). Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207, 203-205.

5. Lexical ambiguity resolution (Swinney) — Brief activation of multiple meanings, then commitment to one
Swinney (1979). Lexical access during sentence comprehension: (Re)consideration of context effects. Journal of Verbal Learning and Verbal Behavior, 18, 645-659.

6. Distance-dependent decay — Longer syntactic dependencies are harder to maintain
Gibson (2000). The dependency locality theory: A distance-based theory of linguistic complexity. In Marantz et al. (Eds.), Image, Language, Brain, MIT Press.

7. Arrow of time in language — Forward generation easy, backward extremely difficult
Shannon (1951). Prediction and entropy of printed English. Bell System Technical Journal, 30, 50-64.

Memory Phenomena

8. Episodic memory as generation — “Remembering” isn’t replay of stored episodes; it’s autoregressive generation seeded by cues. Explains why memory is sequential, constructive, and shares machinery with imagination. The temporal unfolding IS the memory, not a byproduct of retrieval.
Schacter & Addis (2007). The cognitive neuroscience of constructive memory. Neuron, 56(2), 362-378.

9. False memories (DRM paradigm, War of the Ghosts) — Memory IS generation along learned trajectories; confident false memories emerge naturally
Roediger & McDermott (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(4), 803-814.

10. Priming (repetition, semantic, perceptual) — Residual activation biases subsequent generation + weight modification
Meyer & Schvaneveldt (1971). Facilitation in recognizing pairs of words. Journal of Experimental Psychology, 90, 227-234.

11. Serial position effects (primacy/recency) — Single system with activation gradient, not two stores
Murdock (1962). The serial position effect of free recall. Journal of Experimental Psychology, 64(5), 482-488.

12. Temporal contiguity in recall — Items recalled in temporal clusters; each output seeds next (this IS autoregressive generation)
Kahana (1996). Associative retrieval processes in free recall. Psychological Review, 103(3), 511-532.

13. Reconsolidation — “Reactivating” memory = regenerating it, which naturally allows updating
Nader, Schafe, & LeDoux (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406, 722-726.

14. Context-dependent memory — Generation depends on current state; context effects are inherent
Godden & Baddeley (1975). Context-dependent memory in two natural environments: On land and underwater. British Journal of Psychology, 66(3), 325-331.

15. Extended context — Discourse coherence beyond 7+/-2 items; two thresholds (explicit retrieval vs generative influence)
Cowan (2001). The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87-114.

16. Unification of STM/LTM — Not separate stores; same generative system at different timescales
Cowan (1995). Attention and Memory: An Integrated Framework. Oxford University Press.

17. Continual distractor free recall — Destroys buffer models; recency without buffer
Bjork & Whitten (1974). Recency-sensitive retrieval processes in long-term free recall. Cognitive Psychology, 6, 173-189.

18. IQ-STM correlation — Both reflect same underlying capacity for maintaining extended context
Ackerman, Beier, & Boyle (2005). Working memory and intelligence: The same or different constructs? Psychological Bulletin, 131(1), 30-60.

19. Rehearsal — Trajectory strengthening through repeated generation, not transfer between stores
Rundus (1971). Analysis of rehearsal processes in free recall. Journal of Experimental Psychology, 89(1), 63-77.

20. Amnesia dissociations (H.M., Clive Wearing) — The preserved/impaired dissociation in dense amnesia doesn’t require separate episodic and procedural memory systems. What’s lost is not the ability to generate sequences, but the ability to generate when the current step depends on information from beyond the local context window. What survives — piano playing, mirror tracing, conversational turns — are trajectories where each step depends only on immediately preceding steps (locally Markovian). What fails is any task requiring long-range dependencies. The procedural/episodic distinction is thus not about memory systems but about dependency distance.
Scoville & Milner (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery and Psychiatry, 20, 11-21.

21. Hippocampus as dependency-bridging mechanism — Short-range sequential dependencies can be handled via STDP-based cortical/cerebellar chaining. But when generating the current step requires information from beyond that window, the hippocampus provides the bridging mechanism. It maintains earlier trajectory elements in active form so they can constrain current generation across temporal gaps that STDP cannot span. This reframes hippocampal function: not a gateway to a separate memory system, but an attention-like mechanism that extends effective dependency range — functionally analogous to what attention heads do in transformers.
Eichenbaum (2004). Hippocampus: Cognitive processes and neural representations that underlie declarative memory. Neuron, 44(1), 109-120. Salz et al. (2016). Time cells in hippocampal area CA3. Journal of Neuroscience, 36(28), 7476-7484.

22. Working memory as perceptual-generative loop (CSA) — The “phonological loop” and “visuospatial sketchpad” are not memory stores but self-triggered reactivation of perceptual processing. “Holding in mind” = generating perception internally = providing short-range context for continued generation. This dissolves Baddeley’s architecture: there are no slave systems feeding a central executive, just perceptual systems that can be self-triggered, feeding the same autoregressive process that handles everything else.
Baddeley & Hitch (1974). Working memory. In Psychology of Learning and Motivation, 8, 47-89. Academic Press.

Perception & Consciousness

23. Perceptual completion (blind spot filling, phonemic restoration) — Generation fills gaps based on learned patterns
Warren (1970). Perceptual restoration of missing speech sounds. Journal of the Acoustical Society of America, 48, 1490-1499.

24. Change blindness — We don’t store detailed representations; we generate on demand
Simons & Levin (1997). Change blindness. Trends in Cognitive Sciences, 1(7), 261-267.

25. McGurk effect — Top-down generation influences perceptual experience
McGurk & MacDonald (1976). Hearing lips and seeing voices. Nature, 264, 746-748.

26. Stream of consciousness — Continuity = chain of autoregressive generation; James’s “stream” is literally the generative flow
James (1890). The Principles of Psychology. Henry Holt & Co. Chapter 9.

27. The specious present (~3 sec) — Consciousness = moving context window; “now” = current span of active context
James (1890). The Principles of Psychology. Chapter 15: The Perception of Time. Poppel (2004). Lost in time: A historical frame, an elementary processing unit, and the 3-second window. Acta Neurobiologiae Experimentalis, 64, 295-301.

28. Binocular rivalry — Commitment to one interpretation, periodic switching (temperature-driven exploration)
Blake & Logothetis (2002). Visual competition. Nature Reviews Neuroscience, 3(1), 13-23.

Imagery, Memory & Thinking

Core claim: “Thinking” is self-generated autoregressive streams without (or with weak) sensory grounding. Imagery, episodic memory, dreams, deliberation, planning, daydreaming are all the same process under different conditions.

29. Mental imagery — Generation without current sensory input; same system as perception, different conditions
Kosslyn (1980). Image and Mind. Harvard University Press.

30. Mental rotation (Shepard & Metzler) — NOT rotating a 3D mental model; generating the learned visual sequence of how objects look as they rotate. Linear time cost = more sequence to generate. Dissolves Kosslyn vs. Pylyshyn debate: it feels like running a movie because you ARE running a movie. The phenomenology IS the computation.
Shepard & Metzler (1971). Mental rotation of three-dimensional objects. Science, 171(3972), 701-703.

31. Mental scanning (Kosslyn) — Generating the visual trajectory of scanning across a learned scene; time proportional to “distance” because more trajectory to generate
Kosslyn (1973). Scanning visual images: Some structural implications. Perception & Psychophysics, 14(1), 90-94.

32. Imagery-perception overlap — Same generative system; imagery = unconstrained generation, perception = input-constrained generation
Ganis, Thompson, & Kosslyn (2004). Brain areas underlying visual mental imagery and visual perception: An fMRI study. Cognitive Brain Research, 20(2), 226-241.

33. Dreams — Generation without sensory grounding and with weak executive control. Bizarreness = no input to correct implausible continuations. Narrative structure = still autoregressive. Discontinuities = context collapse and restart.
Hobson & Pace-Schott (2002). The cognitive neuroscience of sleep. Nature Reviews Neuroscience, 3(9), 679-693.

34. Daydreaming/mind-wandering — Free-running generation from own output; default mode when external input doesn’t constrain
Raichle et al. (2001). A default mode of brain function. PNAS, 98(2), 676-682. Smallwood & Schooler (2015). The science of mind wandering. Annual Review of Psychology, 66, 487-518.

35. Planning/prospection — Generating possible future trajectories; same machinery as episodic memory but running into unexperienced territory
Schacter, Addis, & Buckner (2007). Remembering the past to imagine the future: The prospective brain. Nature Reviews Neuroscience, 8(9), 657-661.

36. Deliberation — Controlled generation with evaluation; running multiple trajectory branches before committing
Kahneman & Tversky (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263-292.

37. Inner speech — Linguistic trajectory generation without vocalization; thinking in words is generating words
Vygotsky (1934/1962). Thought and Language. MIT Press.

Everyday Phenomenology

38. Tip-of-tongue states — Stalled trajectory; partial activation insufficient to complete generation
Brown & McNeill (1966). The “tip of the tongue” phenomenon. Journal of Verbal Learning and Verbal Behavior, 5(4), 325-337.

39. Deja vu — Near-miss reactivation; current input partially matches learned trajectory, generating false sense of continuation
Brown (2003). A review of the deja vu experience. Psychological Bulletin, 129(3), 394-413.

40. Freudian slips — Wrong trajectory completion; highly activated alternative wins over intended continuation
Fromkin (1973). Speech Errors as Linguistic Evidence. Mouton Publishers.

41. Musical earworms — Attractor basin loops; melodies that reenter their own continuation space, self-perpetuating
Jakubowski et al. (2017). Dissecting an earworm: Melodic features and song popularity predict involuntary musical imagery. Psychology of Aesthetics, Creativity, and the Arts, 11(2), 122-135.

42. Sudden insight — Trajectory suddenly finds high-coherence path after exploration; “aha” = landing on strong attractor
Jung-Beeman et al. (2004). Neural activity when people solve verbal problems with insight. PLoS Biology, 2(4), e97.

System 1/2 & Deliberation

43. System 1 (fast/automatic) — Default autoregressive generation; follows strongest learned trajectories
Kahneman (2011). Thinking, Fast and Slow. Macmillan.

44. System 2 (slow/deliberate) — Controlled autoregression; holding elements in workspace, generating multiple branches before committing
Evans (2008). Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255-278.

45. Cognitive load effects — Limited capacity for maintaining multiple trajectory branches simultaneously
Sweller (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285.

Learning & Pedagogy

46. The testing effect / generation effect — Passive review is not generation. Retrieval practice forces the system to bootstrap from degraded context, actually running the generative process and thereby strengthening the underlying structure. “Desirable difficulty” is just the phenomenology of effortful generation under sparse context.
Roediger & Karpicke (2006). The power of testing memory: Basic research and implications for educational practice. Psychological Bulletin, 132(3), 331-346.

47. Confirmation bias — Committed interpretation biases what gets generated; trajectory attracts consistent evidence
Nickerson (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2), 175-220.

48. Belief perseverance — Costly to regenerate from different starting point; sunk cost in current trajectory
Ross, Lepper, & Hubbard (1975). Perseverance in self-perception and social perception. Journal of Personality and Social Psychology, 32(7), 880-892.

49. Functional fixedness — Committed to one object interpretation; hard to regenerate alternative uses
Duncker (1945). On problem-solving. Psychological Monographs, 58(5), 1-113.

50. Cognitive dissonance — Conflicting committed trajectories create tension; resolution = aligning them
Festinger (1957). A Theory of Cognitive Dissonance. Stanford University Press.

51. Post-choice spreading (Sharot) — After commitment, trajectory divergence increases; chosen option’s trajectory strengthened
Sharot, De Martino, & Dolan (2009). How choice reveals and shapes expected hedonic outcome. Journal of Neuroscience, 29(12), 3760-3765.

52. Anchoring — Initial value seeds generation; subsequent estimates are continuations from anchor
Tversky & Kahneman (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124-1131.

53. Hindsight bias — Current knowledge contaminates regeneration of past state; can’t uncommit from what you now know
Fischhoff (1975). Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288-299.

Aesthetics

54. Musical tension and release — Controlled trajectory curvature; tension = departure from expected path, release = return
Salimpoor et al. (2011). Anatomically distinct dopamine release during anticipation and experience of peak emotion to music. Nature Neuroscience, 14(2), 257-262.

55. Rhythm — Temporal scaffolding for generation; provides predictable structure to deviate from
Large & Palmer (2002). Perceiving temporal regularity in music. Cognitive Science, 26(1), 1-37.

56. Humor/jokes — Low-probability continuation under coherence constraint; punchline as trajectory jump that retrospectively fits
Suls (1972). A two-stage model for the appreciation of jokes and cartoons. In Goldstein & McGhee (Eds.), The Psychology of Humor. Academic Press.

57. Catchiness — Melodies with high autogenerative potential; they complete themselves
Jakubowski et al. (2017). Dissecting an earworm: Melodic features and song popularity predict involuntary musical imagery. Psychology of Aesthetics, Creativity, and the Arts, 11(2), 122-135.

Clinical Phenomena

58. Schizophrenia (positive symptoms) — Dysregulated generation; temperature too high, weak constraint from input
Powers, Mathys, & Corlett (2017). Pavlovian conditioning-induced hallucinations result from overweighting of perceptual priors. Nature Neuroscience, 20(4), 565-567.

59. Depression (rumination) — Stuck in negative attractor basins; same trajectories regenerate repeatedly
Nolen-Hoeksema (1991). Responses to depression and their effects on the duration of depressive episodes. Journal of Abnormal Psychology, 100(4), 569-582.

60. OCD — Compulsive trajectory completion; incomplete actions demand finishing
Foa & Kozak (1986). Emotional processing of fear: Exposure to corrective information. Psychological Bulletin, 99(1), 20-35.

61. Anxiety — Threat-biased generation; negative continuations more accessible
Grupe & Nitschke (2013). Uncertainty and anticipation in anxiety. Nature Reviews Neuroscience, 14(7), 488-501.

62. PTSD intrusions — Trauma trajectories have extremely strong weights; easily triggered, hard to escape
Brewin, Dalgleish, & Joseph (1996). A dual representation theory of posttraumatic stress disorder. Psychological Review, 103(4), 670-686.

63. Confabulation — Generation proceeds without adequate constraint; produces confident false content
Johnson, Hashtroudi, & Lindsay (1993). Source monitoring. Psychological Bulletin, 114(1), 3-28.

Consciousness

64. Unity of consciousness / binding problem dissolved — Unity is not constructed from parts; it’s intrinsic to global autoregressive computation. Each generation step is holistically unified. There’s nothing to bind because there are no separate fragments. The binding problem was an artifact of the wrong computational model.
Crick & Koch (2003). A framework for consciousness. Nature Neuroscience, 6(2), 119-126.

65. Subjectivity / first-person perspective — Arises from recursive self-input. The system’s own outputs become part of its input context. “The ‘I’ is the felt shape of recursion.” Consciousness is what recursive global computation feels like from within.
Chalmers (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200-219.

66. The hard problem reframed — Subjective experience isn’t added to computation. The recursive, self-reading, holistic generation IS the experience. The phenomenology is the computation (same insight as mental imagery).
Chalmers (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200-219.

67. Attention as weighting, not binding — Attention doesn’t unify separate streams; it tunes the balance of contributions within already-unified global computation.
Desimone & Duncan (1995). Neural mechanisms of selective visual attention. Annual Review of Neuroscience, 18, 193-222.

68. Phenomenology as generative input — Conscious experience is the processed, interpreted output of generative inference in its role as input to the next generative step. The format of phenomenology is the format of generative context. Subjectivity doesn’t “arise” from this computation — it is this computation.
Hohwy (2013). The Predictive Mind. Oxford University Press. Clark (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181-204.

Neural Validation

69. Schrimpf et al. (2021) — GPT-2 predicts brain activity during language processing
Schrimpf et al. (2021). The neural architecture of language: Integrative modeling converges on predictive processing. PNAS, 118(45), e2105646118.

70. Goldstein et al. (2022) — Shared computational principles between LLMs and neural responses
Goldstein et al. (2022). Shared computational principles for language processing in humans and deep language models. Nature Neuroscience, 25(3), 369-380.

71. Caucheteux & King (2022) — Autoregressive models predict comprehension
Caucheteux & King (2022). Brains and algorithms partially converge in natural language processing. Communications Biology, 5(1), 134.

72. Hosseini et al. (2024) — Works even with child-scale training data
Hosseini et al. (2024). Artificial neural network language models predict human brain responses to language even after a developmentally realistic amount of training. Nature Neuroscience, 27(3), 412-422.

73. Ruchkin et al. — Sustained activation predicts LTM encoding
Ruchkin et al. (2003). Working memory retention systems: A state of activated long-term memory. Behavioral and Brain Sciences, 26(6), 709-728.

74. Daume et al. — Delay period activity predicts subsequent memory
Daume et al. (2024). Persistent activity during working memory maintenance predicts long-term memory formation in the human hippocampus. Neuron, 112(17), 2779-2790.

Summary: 74 Phenomena, One Principle

All of these are autoregressive generation under different conditions:

With/without sensory input (perception vs. imagery vs. thinking)
With/without executive control (System 2 vs. System 1)
With different temperature settings (exploration vs. exploitation)
With different context lengths (simple vs. complex tasks)
With intact vs. disrupted mechanisms (normal vs. clinical)
With recursive self-inclusion (consciousness, subjectivity)

Key Unifications

Imagery, episodic memory, dreams, daydreaming, planning, deliberation, and inner speech are all “thinking” = self-generated autoregressive streams without sensory grounding.

Unity and subjectivity of consciousness are intrinsic to recursive global autoregressive computation, not added features.

Phenomenology is the format of generative context — egocentric, action-formatted, serving behavioral generation.

The theory’s power is unification: these aren’t 74 separate explanations. They’re 74 manifestations of one process.