Mechanisms
Metacognition · Attribution
Source Monitoring
Example
Bethany tells you what happened in Thursday's meeting, and the account stays tied to her. A week later you learn she left twenty minutes early, and because the origin is still attached you can go back and downgrade her version as incomplete rather than continuing to hold it as simply what happened. Nothing stamped her name on the memory at the time you stored it: you reconstructed the origin when you went to use it, from the fact that the account arrives in her phrasing and carries the texture of something heard in conversation rather than something you read in the minutes.
Source Monitoring in Humans
Definition
Source monitoring is the set of processes by which a person decides where a remembered piece of information came from, and keeps that origin attached to the content so the content can be re-evaluated later if the source turns out to be unreliable (Johnson et al., 1993). The framework's central claim is that people do not retrieve a stored tag naming a memory's origin; instead they evaluate qualities of the reactivated memory, such as how much perceptual detail it carries or how much mental effort it seems to have taken, and infer an origin from those qualities during remembering (Johnson et al., 1993). Because the origin is inferred rather than looked up, the process is systematically error-prone, and the errors follow from the qualities available rather than occurring at random (Johnson et al., 1993). The framework names three situations: external source monitoring, telling two outside sources apart; internal source monitoring, telling two self-generated sources apart, such as something thought from something said aloud; and reality monitoring, telling an internal source from an external one (Johnson et al., 1993).
Scope: This row covers external source monitoring and internal source monitoring. Reality monitoring is the third situation named in the same framework and is treated in its own row; it is a sibling subtype within one framework, not an independent construct, and the two rows should not be read as competing accounts (Johnson et al., 1993).
Neural Substrate
Memory for source is distributed rather than localized. Reviewing the patient and lesion evidence available at the time, Johnson et al. (1993) concluded that it is unlikely any single brain area is alone responsible for memory for source, and that frontal damage and temporal or diencephalic damage disrupt different aspects of it rather than one region owning the function. Mitchell and Johnson (2009) revisited that conclusion with functional imaging and organized their review around the contributions of subregions of the medial temporal lobes, prefrontal cortex, posterior representational areas, and parietal cortex.
A more recent activation likelihood estimation meta-analysis pools 51 fMRI studies and 864 healthy participants, and separates external source monitoring from reality monitoring rather than treating them together (Martín-Luengo et al., 2026). Across all source-monitoring contrasts it found ten clusters, the largest and highest-valued centred in the left claustrum and left middle frontal gyrus (BA 46, 9 and 10). Restricted to external source monitoring, seven clusters emerged, the largest centred on left middle frontal gyrus (BA 9) and left inferior frontal gyrus (BA 45), with further clusters in left medial and superior frontal gyri, left precuneus and bilateral cingulate. Two regions appeared only in the external source-monitoring contrasts: a left medial (BA 8) and superior frontal (BA 6, BA 8) cluster, and bilateral posterior cingulate (BA 23).
The same analysis dissociates this row's construct from row 17's. Reality monitoring alone yielded clusters unique to it in left inferior parietal lobule (BA 40) and left angular gyrus (BA 39), and in left and right caudate, which the authors link to self-agency and self-consciousness. The conjunction of the two returned only four overlapping clusters, in left precuneus and cuneus (BA 7), left middle frontal gyrus (BA 9 and 46), left precuneus (BA 39), and left inferior parietal lobule (BA 7). The authors read this as partial overlap of executive and retrieval systems within largely process-specific architecture, and they caution that ALE conjunction is conservative by construction, so restricted overlap does not by itself argue against a shared framework (Martín-Luengo et al., 2026). Two limits on that source should be carried forward. It collapses sensory domains, and it covers only retrospective source monitoring, since too few studies of online source monitoring were available to analyse separately. It also does not converge fully with the 2009 review: no cluster in hippocampus or parahippocampal cortex reached threshold in any of its three analyses, so the medial temporal contribution emphasized by Mitchell and Johnson does not appear in the pooled healthy-participant imaging data, and the authors do not address the discrepancy.
AI at a Glance
Source monitoring keeps a remembered fact tied to where it came from, by inferring the origin at retrieval from qualities of the memory rather than reading a stored label, so the fact can be discounted later if the source turns out to be unreliable (Johnson et al., 1993). Language models reliably reproduce what was said and unreliably reproduce who said it, scoring near chance on a direct two-alternative speaker-attribution test even when handed a transcript with every speaker already labelled (Kwon et al., 2026) and asserting misattributed or unsupported memories on a majority of queries when an external memory layer supplies them (Chang & Chen, 2026); the closest named mechanisms are the context window, an external memory layer's retrieved items and their source metadata, and product-level memory features that record an item's originating conversation.
- Memory layer: an external store holding items extracted from past conversations, from which a small number are retrieved into the model's input when a new message arrives (Chang & Chen, 2026).
- Source metadata: fields attached to a stored memory item recording which conversation turn it came from, which some memory systems expose and others discard during compression (Chang & Chen, 2026).
Source Monitoring in AI
How Far It Has Been Measured
Measurement in AI: Directly Measured
AI Memory Performance: Not Yet Compared
Where It Lives in the System
What Has Been Measured
- A 2026 preprint gave six open-weight language models the reality-monitoring word-pair paradigm from human research (Johnson & Raye), in which the second word is either shown or generated, followed by an 'internal or external?' source judgment; all models identified their own generated words with 100% accuracy when that word was still in the context window, meaning what the model can see in the current conversation (Ranjan et al., 2026).
- Systems differed widely in the same preprint: on shown words, accuracy ranged from 47.9% (Gemma 3 12B) to 94.4% to 100% (Llama 3.3 70B); with a delayed source test after 20 or 40 trials, most models discriminated sources at chance (d' about 0), while the best, Llama 3.3 70B, reached d' = 2.01 and 1.62 (Ranjan et al., 2026).
- No comparison can be drawn from this study: its human reference is an anonymised prior dataset described only qualitatively (no sample size or scores), and the earlier trials stayed in the model's conversation history at test, so the best model's strong result favours the side that could still look back at the material (Ranjan et al., 2026).
- In a benchmark of 50 very long two-person conversations, long-context models were especially prone to assigning events to the wrong speaker, scoring as low as 2.1 F1 on trick questions versus 70.2 for GPT-4-turbo; a human score of 89.4 is reported without the number of people or procedure, so it cannot serve as a documented baseline (Maharana et al., 2024).
- For saved memory, a 2026 preprint built a source-monitoring probe for a robot's stored memory (for example, whether it learned about an object from a scene description or an object detector); its structured memory scored 78.6 of 100 on an AI-judged scale, and simpler memory designs fell 20 to 35 points behind on matched probes, with no human data (Rasheed & Kabtoul, 2026).
- For training, a 1.1-billion-parameter model trained on synthetic biographies with document labels answered 88.8% of questions on unseen documents correctly but named the correct source document only 47.3% of the time, a gap between content and source memory; the ability required special training, since standard models are not trained to track sources (Khalifa et al., 2024).
- For retrieved material, a human audit of four AI search engines found that only 51.5% of generated sentences were fully supported by their citations and 74.5% of citations supported the sentence they were attached to, so many claims were not backed by the source they were credited to (Liu et al., 2023).
Where a Mechanism Can Live
04 Product
The whole configuration, and the only thing a benchmark observes directly.
01
Weights
Across deployments
02
Context Window
Within a session
03
Memory Layer
Across sessions
Which of the three a result belongs to is inferred from it, not seen in it.
Read more
Angular gyrus
Cortex in the inferior parietal lobule
The inferior parietal gyrus arching around the upturned end of the superior temporal sulcus, behind the supramarginal gyrus.
The angular gyrus curves around the ascending terminal branch of the superior temporal sulcus, forming the back half of the inferior parietal lobule and corresponding roughly to Brodmann area 39. It sits at the meeting of visual, auditory and somatosensory association cortex and is connected to inferior frontal and lateral temporal regions. Left angular damage figures in Gerstmann's syndrome (agraphia, acalculia, finger agnosia, left–right confusion), though whether that cluster reflects one lesion site or the coincidental disconnection of separate pathways is disputed. Imaging also reports it in semantic processing and episodic retrieval.
Brodmann area 45
Cortex in the pars triangularis
The granular cytoarchitectonic area of pars triangularis, the anterior half of Broca's area.
Brodmann area 45 corresponds broadly to pars triangularis of the inferior frontal gyrus and is granular, with a well developed layer IV, while its neighbor area 44 is dysgranular and retains only a faint layer IV. In the mapping work of Amunts and colleagues its borders cross the sulci in a substantial fraction of hemispheres, so assigning a functional activation to 45 rather than 44 on sulcal grounds alone is unreliable. Functional accounts associate it with controlled semantic retrieval more than with articulation, and it is sometimes subdivided into 45a and 45p with different connections.
Caudate nucleus
Nucleus in the dorsal striatum
A C-shaped nucleus arching alongside the lateral ventricle, the association-connected half of the dorsal striatum.
Head, body and tail follow the curve of the lateral ventricle, with the head bulging into its frontal horn. It receives from prefrontal and parietal association cortex, keeping it largely separate from the sensorimotor input that dominates the putamen. Caudate volume loss is among the earliest and most characteristic imaging findings in Huntington's disease, visible as widened ventricular horns before motor symptoms. Its activation in human studies is usually reported with feedback-based learning and with goal-directed action selection rather than movement execution.
Cingulate cortex
Cortex in the cerebral cortex
The long belt of cortex arching over the corpus callosum on the medial wall of each hemisphere.
The cingulate cortex is a collar of cortex (cingulum is Latin for belt) wrapped around the corpus callosum on the medial surface of the hemisphere, separated from the frontal and parietal cortex above it by the cingulate sulcus. It is mesocortex, with a simpler layered structure than neocortex, and it is bordered below by the callosum and the cingulum bundle that carries its long connections. Vogt's four-region model divides it into anterior, midcingulate, posterior, and retrosplenial sectors with distinct architecture and connections; the older two-part anterior/posterior split is still common in imaging papers and does not map onto it cleanly.
Claustrum
Subcortical
A thin irregular sheet of grey matter between the insula and the putamen, reciprocally connected with almost the whole cortex.
A few tenths of a millimeter thick, sandwiched between the external and extreme capsules, and among the most densely and widely connected structures in the brain relative to its size. That connectivity is the whole basis of its reputation: Crick and Koch proposed in 2005 that it might coordinate activity across cortical areas into unified conscious percepts, a suggestion made explicitly as a hypothesis. Direct evidence remains thin (a single stimulation case report, some rodent work on slow-wave activity and sleep), and its shape makes it very hard to image or lesion cleanly.
Cuneus
Cortex in the occipital lobe
The wedge of medial occipital cortex between the calcarine and parieto-occipital sulci, above the calcarine sulcus.
The cuneus is the triangular area of the medial occipital surface bounded below by the calcarine sulcus and in front by the parieto-occipital sulcus; its name is simply Latin for wedge. Its lower bank forms part of primary visual cortex, and because the retinotopic map is inverted, this cortex above the calcarine sulcus represents the lower visual field. A cuneal lesion therefore produces a lower quadrantanopia.
Diencephalon
Division in the forebrain (prosencephalon)
The deep core of the forebrain (thalamus, hypothalamus, epithalamus and subthalamus), wrapped around the third ventricle.
The diencephalon sits at the top of the brainstem and beneath the cerebral hemispheres, its four parts arranged around the third ventricle: thalamus above and lateral, hypothalamus below, epithalamus behind and above, subthalamus tucked between hypothalamus and midbrain. Almost every ascending sensory pathway and most cortical–subcortical loops pass through it, which is why small lesions here produce deficits out of all proportion to their size. It is a developmental term as much as an anatomical one: the diencephalon is one of the vesicles the embryonic forebrain divides into, and the label has stuck.
Hippocampus
Cortex in the hippocampal formation
A curved allocortical structure in the medial temporal lobe whose bilateral damage reliably produces dense anterograde amnesia in humans.
A paired, seahorse-shaped fold of three-layered cortex running front to back inside the temporal lobe, connected to the rest of the brain mainly through entorhinal cortex and the fornix. What it does is genuinely contested and has been for fifty years: the human lesion literature reads it as necessary for forming new declarative memories, the rodent recording literature as a map of space, and a third position treats both as instances of binding arbitrary elements into relations. The accounts differ on which species and which task they explain best, and the question of which is right remains open.
Inferior frontal gyrus
Cortex in the frontal lobe
The lowest lateral frontal gyrus, divided into pars opercularis, triangularis and orbitalis by the ascending rami of the lateral fissure.
The inferior frontal gyrus runs above the lateral fissure and is cut into three parts by that fissure's ascending and horizontal rami: pars opercularis behind, pars triangularis in the middle, pars orbitalis in front. In the left hemisphere the first two correspond roughly to Broca's area, and damage there is associated with non-fluent aphasia. The right IFG is repeatedly reported in response-inhibition tasks such as stop-signal, though whether its contribution is specifically inhibitory or reflects detecting behaviorally relevant signals is a live argument. Its cytoarchitectonic boundaries do not align neatly with the gross parts.
Inferior parietal lobule
Cortex in the parietal lobe
The parietal cortex below the intraparietal sulcus, comprising the supramarginal and angular gyri.
The inferior parietal lobule lies beneath the intraparietal sulcus and above the lateral fissure, made up of the supramarginal gyrus in front and the angular gyrus behind, with the temporoparietal junction at its lower border. It is disproportionately expanded in humans relative to other primates and is among the last cortical regions to myelinate. Left-sided lesions contribute to apraxia and to phonological and semantic deficits; right-sided lesions are the commonest cause of hemispatial neglect. Its parcellation is unsettled: cytoarchitectonic studies report five to seven areas where the gross anatomy shows two gyri.
Medial temporal lobe
Cortex in the temporal lobe
The inner fold of the temporal lobe (hippocampus, the cortices wrapped around it, and the amygdala), removed together in the classic amnesia cases.
A regional grouping rather than a single structure: the hippocampal formation, the entorhinal, perirhinal and parahippocampal cortices that feed it, and, on some definitions, the amygdala. It became the canonical "memory region" because the surgeries and infarcts that produce dense anterograde amnesia (H.M. most famously) take out most of it at once. Whether the parts constitute one memory system or several with different jobs is an open argument, and lesions in patients often cross the boundaries that would settle it.
Middle frontal gyrus
Cortex in the frontal lobe
The gyrus between the superior and inferior frontal sulci, occupying most of the lateral prefrontal surface.
The middle frontal gyrus lies between the superior and inferior frontal sulci and extends forward to the frontal pole. Much of what is called dorsolateral prefrontal cortex falls on it, along with the frontal eye field near its junction with the precentral sulcus. It is reported across a wide range of imaging tasks, activated during working memory, attentional control and rule-guided behavior. That ubiquity is a caution rather than a finding: the gyrus is large, its sulcal pattern varies considerably between individuals, and group-average coordinates on it can correspond to different cytoarchitectonic areas in different brains.
Parahippocampal cortex
Cortex in the parahippocampal gyrus
The posterior part of the parahippocampal gyrus, responding strongly to scenes and spatial layouts and feeding medial entorhinal cortex.
Cortex behind the entorhinal region on the medial temporal surface, connected to retrosplenial and posterior parietal areas and projecting into medial entorhinal cortex. Human imaging finds a patch here that responds far more to images of places than of objects or faces: the parahippocampal place area. What that response encodes is unsettled: candidate accounts include scene geometry, spatial context for navigation, and the presence of large landmark-like objects, and each explains some of the contrast that named the region.
Parietal lobe
Lobe in the cerebral hemisphere
The lobe between the central sulcus and the parieto-occipital sulcus, above the lateral fissure.
The parietal lobe begins at the central sulcus and ends at the parieto-occipital sulcus, with the lateral fissure below it. Its front edge is the postcentral gyrus, carrying primary somatosensory cortex; behind that the intraparietal sulcus splits the remainder into the superior parietal lobule and the inferior parietal lobule, the latter containing the angular and supramarginal gyri. Medially it includes the precuneus. Parietal regions are consistently engaged by spatial attention, reaching and grasping, numerical tasks and the coordinate transformations between eye-, head- and hand-centred frames; right parietal damage is the usual cause of hemispatial neglect.
Posterior cingulate cortex
Cortex in the cingulate cortex
The rear section of the cingulate, above the splenium of the corpus callosum and above the retrosplenial cortex in the callosal sulcus.
The posterior cingulate cortex sits above and behind the splenium of the corpus callosum, bordering the precuneus above and the retrosplenial cortex below; its dorsal and ventral halves have different connections and different response profiles. It has one of the highest resting metabolic rates in the brain and is among the first regions to show reduced glucose metabolism in Alzheimer's disease. It is a core node of the default mode network and is engaged during autobiographical memory retrieval and imagining the future. Whether its activity reflects internally directed thought, the breadth of attention, or a more general scene-construction role is not settled.
Precuneus
Cortex in the parietal lobe
The medial parietal cortex in front of the parieto-occipital sulcus, above the posterior cingulate and behind the paracentral lobule.
The precuneus occupies the medial surface of the parietal lobe, bounded by the parieto-occipital sulcus behind, the marginal branch of the cingulate sulcus in front, and the subparietal sulcus below. It is densely connected to posterior cingulate and retrosplenial cortex, to the superior parietal lobule, and to the thalamus. Metabolically it is one of the most active regions of the resting brain and among the earliest to show hypometabolism in Alzheimer's disease. It is reported in episodic retrieval, mental imagery, perspective-taking and self-referential tasks; no single account of what unites those is established.
Prefrontal cortex
Cortex in the frontal lobe
The frontal cortex in front of the motor areas, defined classically by its connections with the mediodorsal thalamus rather than by any one function.
The prefrontal cortex occupies the front of the frontal lobe, anterior to premotor and motor cortex, and accounts for roughly a sixth to a quarter of human cortex depending on where the posterior border is drawn. Three competing definitions are in use (projection from the mediodorsal thalamus, the presence of a granular layer IV, and the absence of movement on electrical stimulation), and they do not enclose the same tissue, which is why rodent-to-primate homology is still argued over. It is reliably engaged by tasks with delays, rules, and conflicting options; what it contributes to those tasks is the open question, not whether it participates.
Superior frontal gyrus
Cortex in the frontal lobe
The uppermost of the three long gyri on the lateral frontal surface, continuing over the midline onto the medial wall.
The superior frontal gyrus runs from near the frontal pole back to the precentral sulcus, bounded below by the superior frontal sulcus, and wraps over the hemisphere's upper edge onto the medial surface. Its posterior medial portion contains the supplementary motor area and pre-SMA; more anteriorly it takes in dorsomedial prefrontal cortex. Dorsolateral portions are routinely reported in working-memory and task-switching studies. The gyrus is an anatomical landmark spanning several functionally distinct fields, so a coordinate "in the superior frontal gyrus" localizes weakly unless the subdivision is named.
Chang & Chen (2026)
When Users Don't Ask: Benchmarking Context-Driven Memory Retrieval in Conversational Agents
In
arXiv
Abstract
Large language models (LLMs) are increasingly deployed as long-horizon conversational agents, motivating growing interest in mem- ory systems. However, existing benchmarks primarily evaluate memory through QA-style probing rather than in-situ conversational usage. We introduce LOCOMO-CONV, a conversational memory benchmark derived from Lo- CoMo with four query styles: dialog, implicit, counterfactual, and composed. Across five representative memory systems, we evaluate both retrieval recall and end-to-end response quality. Our experiments show that conversational framing exposes substantial retrieval gaps over- looked by QA benchmarks, especially on implicit and composed queries, which...
Find it
Johnson et al. (1993)
Source monitoring.
In
Psychological Bulletin, 114(1), 3-28
Abstract
A framework for understanding source monitoring and relevant empirical evidence is described, and several related phenomena are discussed: old–new recognition, indirect tests, eyewitness testimony, misattributed familiarity, cryptomnesia, and incorporation of fiction into fact. Disruptions in source monitoring (e.g., from confabulation, amnesia, and aging) and the brain regions that are involved are also considered, and source monitoring within a general memory architecture is discussed. It is argued that source monitoring is based on qualities of experience resulting from combinations of perceptual and reflective processes, usually requires relatively differentiated phenomenal...
Find it
Khalifa et al. (2024)
Source-Aware Training Enables Knowledge Attribution in Language Models
In
arXiv
Abstract
Large language models (LLMs) learn a vast amount of knowledge during pretraining, but they are often oblivious to the source(s) of such knowledge. We investigate the problem of intrinsic source citation, where LLMs are required to cite the pretraining source supporting a generated response. Intrinsic source citation can enhance LLM transparency, interpretability, and verifiability. To give LLMs such ability, we explore source-aware training -- a recipe that involves (i) training the LLM to associate unique source document identifiers with the knowledge in each document, followed by (ii) an instruction-tuning stage to teach the LLM to cite a supporting pretraining source when prompted....
Find it
Kwon et al. (2026)
M3-SLU: Evaluating Speaker-Attributed Reasoning in Multimodal Large Language Models
In
Proceedings of the Language Resources and Evaluation Conference, 5722-5736
Abstract
We present M3-SLU, a new multimodal large language model (MLLM) benchmark for evaluating multi-speaker, multi-turn spoken language understanding. While recent models show strong performance in speech and text comprehension, they still struggle with speaker-attributed reasoning, the ability to understand who said what and when in natural conversations. M3-SLU is built from four open corpora (CHiME-6, MELD, MultiDialog, and AMI) and comprises over 12,000 validated instances with paired audio, transcripts, and metadata. It includes two tasks: (1) Speaker-Attributed Question Answering and (2) Speaker Attribution via Utterance Matching. We provide baseline results for both cascaded pipelines...
Find it
Liu et al. (2023)
Evaluating Verifiability in Generative Search Engines
In
Findings of the Association for Computational Linguistics: EMNLP 2023, 7001-7025
Find it
Maharana et al. (2024)
Evaluating Very Long-Term Conversational Memory of LLM Agents
In
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 13851-13870
Abstract
Paraphrase, not the publisher's wording. Prior work on long-term open-domain dialogue evaluates model responses over no more than five chat sessions, leaving the effectiveness of long-context models and retrieval augmentation in genuinely long-term conversation untested. The authors introduce a machine-human pipeline that generates high-quality very long-term dialogues by grounding LLM agent conversations in personas and temporal event graphs, with agents able to share and react to images. Human annotators verify and edit the output for long-range consistency and grounding. The resulting dataset, LoCoMo, supports an evaluation benchmark covering question answering, event summarisation,...
Find it
Martín‐Luengo et al. (2026)
External Source Monitoring and Reality Monitoring: Meta‐Analysis of <scp>fMRI</scp> Studies
In
Human Brain Mapping, 47(11)
Abstract
Being able to remember the origin of a particular memory is for times even more important than the memory itself. Yet, although memory for the information and memory for the source are intrinsically connected, the particular neural underpinnings for the source-memory monitoring remains unclear. We run an fMRI meta-analysis to help to clarify the brain structures necessary for source memory monitoring. Source monitoring (SM) relates to distinguishing the origin of the information of two external sources, whereas reality monitoring (RM) refers to distinguishing between one internal (e.g., thoughts) and one external source, or two internal sources. We used PRISMA to select the manuscripts...
Find it
Ranjan et al. (2026)
Reality Monitoring in Large Language Models: Self-Knowledge That Transforms with Conversation Memory
In
arXiv
Abstract
A conversational AI that cannot tell its own output from what a user said will treat its own mistakes as user-provided facts. In humans, this capacity is called reality monitoring, and its failures are linked to hallucinations, delusions, and confabulation, yet whether LLMs possess it remains untested. Here we show, across two experiments and six LLMs, that source attribution depends on how conversational memory is structured: ceiling accuracy for self-generated content under minimal memory demands reverses to a fragile external-item advantage once episodic delay removes that shortcut. Feedback exposes two failures: in some models, internal and external judgments swap; in others, accuracy...
Find it
Rasheed & Kabtoul (2026)
eMEM: A Hybrid Spatio-Temporal Memory System For Embodied Agents
In
arXiv
Abstract
We present eMEM (Embodied Memory), a hybrid graph-based memory system for embodied agents operating in physical environments. Current agent memory architectures, such as Generative Agents, MemGPT, and A-MEM, treat memory as text streams or knowledge graphs, but embodied agents require memory that is simultaneously searchable by meaning, space, and time. eMEM fills this gap with a multi-index architecture (SQLITE for structured storage, hnswlib for approximate nearest neighbour semantic search, and an R-tree for spatial queries) unified behind a single graph model. A tiered consolidation pipeline transforms raw perceptual observations into compressed summaries, mirroring...
Find it