Mechanisms
Memory Systems · Non-Declarative
Priming
Example
Perceptual: You glance at a shop sign without properly reading it. An hour later a word puzzle gives you E_E_H_NT and elephant arrives immediately, whether or not you remember the sign. Present the same word spoken rather than written and the advantage shrinks, because what persists is the form. Standard paradigms are perceptual identification and word-stem or fragment completion.
Conceptual: Same sign. Later someone asks you to name some animals and elephant is among the first. It makes no difference whether you saw the word or heard it, because what persists is the meaning. Standard paradigms are category exemplar generation and conceptual classification.
Structural: A colleague says "she sent the client a revised quote." Minutes later, describing something unrelated, you say "I gave my sister the address" rather than "I gave the address to my sister." You did not notice the pattern and would not have predicted repeating it.
Newly-Formed Associations: You read the pair window-reason once, in passing. Later, given "window REA___" you complete it as reason more often than if the cue had been "officer REA___", though you cannot report having seen the pair. The link did not exist before the study episode.
Priming in Humans
Definition
Prior exposure changes how later material is processed, without awareness of the exposure or any intention to recall it. Two forms are conventionally distinguished by what carries the effect: perceptual priming, driven by the physical form of the studied item, and conceptual priming, driven by its meaning (Roediger & McDermott, 1993). These dissociate both behaviourally and neurally (Keane et al., 1995). Structural priming is a third form, in which a syntactic frame persists independently of the words carrying it (Bock, 1986). Priming is a memory phenomenon rather than a fact about pre-existing lexical structure, because it extends to associations newly formed during a single study episode and survives in amnesia (Graf & Schacter, 1985).
Neural Substrate
Lesion evidence supports a division by type. Bilateral occipital lesions impair visuoperceptual priming while leaving recognition intact, and bilateral medial temporal lesions produce the reverse (Keane et al., 1995). The same patient with occipital damage showed intact conceptual priming, and a group study using structurally parallel tasks found perceptual priming poorer after occipital lesions and conceptual priming poorer after frontal lesions (Gong et al., 2016). Together these give occipital cortex for perceptual priming and dorsolateral prefrontal cortex for conceptual priming, both dissociable from the medial temporal system supporting recognition, though only the occipital half of the group dissociation is clean, since the frontal group was also impaired on explicit memory and general cognitive measures (Gong et al., 2016).
Neuroimaging complicates this. A coordinate-based meta-analysis of 65 fMRI studies with concurrent behavioural priming found repetition suppression mainly in left inferior frontal gyrus and fusiform gyrus for both task types, with no region showing suppression selective to one and no significant difference on direct contrast (Lee, Henson, & Lin, 2020). The authors take this as questioning the simple perceptual and conceptual distinction, and propose a hybrid in which inferior frontal suppression reflects retrieval of stimulus-response bindings while fusiform suppression reflects facilitated processing. The inferior frontal result is fragile: once picture-naming studies are removed, the conjunction survives only in fusiform gyrus.
Structural priming has its own substrate. Syntactic adaptation appears in left inferior frontal gyrus (BA45), left middle temporal gyrus (BA21), and bilateral supplementary motor area (BA6), and is equally strong within and between production and comprehension, which the authors read as one neurobiological system serving syntactic processing in both speaking and listening (Segaert et al., 2012). The direction of the effect depends on whether the structure is already represented: in left inferior frontal gyrus and posterior temporal cortex, repetition of a known structure produces suppression, repetition of an infrequent novel structure produces enhancement, and a frequent novel structure falls between the two (Weber et al., 2016). Both this and the Segaert result were obtained with the verb repeated between prime and target, so neither establishes the substrate for lexically independent priming.
Left inferior frontal gyrus therefore appears in the imaging meta-analysis for both perceptual and conceptual priming and in structural priming, while the lesion evidence for conceptual priming points to dorsolateral prefrontal cortex without lateralisation (Gong et al., 2016). Whatever separates the subtypes behaviourally, the imaging converges on shared regions, which is the basis for the meta-analysis authors' doubt that the division survives at the neural level.
One subtype has no imaging account. The same meta-analysis found no hippocampal repetition suppression, which its authors attribute to the predominance of single-item paradigms and flag as requiring work on implicit relational priming (Lee, Henson, & Lin, 2020). That is the newly-formed-associations case established behaviourally by Graf and Schacter (1985), where the primed link is created within the study episode rather than drawn from existing structure. The behavioural evidence is strong, including its survival in amnesia, but the substrate is unresolved, and this row does not claim one for it.
AI at a Glance
In humans a single earlier encounter changes how later material is processed, in forms separated by whether what persists is the physical form of the material, its meaning, or its syntactic frame. Language models reproduce several of the abstract regularities governing this, including the finding that rarer primes prime more strongly, but diverge on which cues drive the effect and on which structures are most primeable.
Priming in AI
How Far It Has Been Measured
Measurement in AI: Directly Measured
AI Memory Performance: Comparable to Humans approximate
Where It Lives in the System
What Has Been Measured
- Run on a classic 1998 sentence-completion experiment (32 prime and target pairs, 1,000 runs each), ChatGPT and Vicuna reused the primed structure (ChatGPT 0.71 vs 0.58; Vicuna 0.81 vs 0.51), more so when prime and target shared a verb, as people do. Against the human means plotted in the paper (an effect of about 5 points with different verbs and 18 with the same verb), ChatGPT's effect (about 9 and 14) sits within the human values and Vicuna's (about 21 and 38) above them (Cai et al., 2024).
- The comparison is approximate: materials and procedure differed, the human means come from the original study with no reported spread, and no test compared people with the models. The within-range reading still holds because the models could still see the prime in the conversation, so Vicuna's larger effect cannot count in its favour, and ChatGPT, the stronger system (it matched the human pattern in 10 of 12 experiments, Vicuna in 7), sits among the human values (Cai et al., 2024).
- Seven multilingual models (XGLM, 564 million to 7.5 billion parameters, and PolyLM, 1.7 and 13 billion) were run on the materials of 12 human structural priming experiments across six languages, including priming from one language to another. All models showed priming on the single-language materials and most showed it on 8 of the 12 cross-language sets, with larger models usually showing larger effects; the authors compared only the direction of effects with people, not their size (Michaelov et al., 2023).
- Across GPT-2 Large and 7-billion-parameter Llama-2, Falcon, Mistral and Zephyr models (base and tuned versions), priming was far more lopsided than in people: sentences that strongly primed one dative structure tended to prime the other weakly or negatively (correlation -0.72 to -0.77), with only 26 to 38% priming in both directions, whereas people generally show positive priming for both. Like people, the models were primed more by rarer structures (the inverse frequency effect) (Jumelet et al., 2024).
- Using the same English dative sentences that people read during EEG recording, 13 transformer language models showed priming, but at both noun phrases after the verb, whereas people showed it only at the second. Across the full set of models, model priming scores were not good predictors of human priming effects (Sinclair et al., 2026).
- Using the materials of a human study in which solving bracketed arithmetic shifted relative-clause attachment in 36 French speakers (in eye movements, among the math-skilled), five current models (Llama 3.3 70B, Qwen3 8B, Kimi K2, gpt-oss-120b and gpt-oss-20b) showed no robust cross-domain priming, even Qwen3, which solved every equation. This one condition falls below people, though the human and model measures differ (eye tracking versus an explicit choice) (Lee et al., 2026).
- Beyond the conversation, a 2025 preprint found a training-level analogue it calls priming: after fine-tuning (further training that changes the model's stored weights) on one new fact, PaLM-2, Gemma and Llama models applied it in unrelated contexts, predictably from the fact's pre-training word probabilities, but with no human comparison. No study was found that tests priming from AI assistants' saved memory (Sun et al., 2025).
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
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.
Dorsolateral prefrontal cortex
Cortex in the prefrontal cortex
The upper outer surface of the prefrontal cortex, on and around the middle frontal gyrus, roughly Brodmann areas 9 and 46.
The dorsolateral prefrontal cortex sits on the lateral convexity of the frontal lobe, centered on the middle frontal gyrus and covering approximately Brodmann areas 9 and 46; its borders are drawn differently by different labs and no sulcus marks them. It is densely connected with posterior parietal cortex, the dorsal striatum, and the mediodorsal thalamus. Activity here tracks working-memory delays, task switching, and rule maintenance, and it is the usual target for prefrontal transcranial stimulation. Whether it stores the contents of working memory, holds the rules for acting on them, or simply reflects task difficulty is genuinely contested, and "executive function" names the dispute rather than settling it.
Fusiform gyrus
Cortex in the temporal lobe
A long spindle-shaped gyrus on the ventral surface, between the collateral sulcus medially and the occipitotemporal sulcus laterally.
The fusiform gyrus stretches along the underside of the temporal and occipital lobes, bounded by the collateral sulcus medially and the occipitotemporal sulcus laterally, and split lengthwise in most brains by a mid-fusiform sulcus. Its lateral mid-portion contains the fusiform face area, which responds more strongly to faces than to other objects, and nearby patches respond preferentially to words and to scenes. Whether these are domain-specific modules or the peaks of a graded map shaped by visual expertise has been argued since the late 1990s and is not settled.
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.
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 temporal gyrus
Cortex in the temporal lobe
The temporal gyrus between the superior and inferior temporal sulci, running most of the lobe's length.
The middle temporal gyrus lies between the superior and inferior temporal sulci, running much of the lobe's length. Its anterior and middle portions on the left are consistently implicated in lexical-semantic retrieval, while lesion evidence ties comprehension deficits more to the posterior part of the gyrus; posterior MTG connects to inferior parietal and inferior frontal regions through the arcuate and superior longitudinal fasciculi. It should not be confused with area MT (the motion-sensitive visual area also called V5), which was named for the owl monkey's middle temporal gyrus and in humans sits further back at the occipitotemporal junction.
Occipital lobe
Lobe in the cerebral hemisphere
The hindmost lobe, behind the parieto-occipital sulcus, containing primary visual cortex around the calcarine sulcus.
The occipital lobe occupies the back of each hemisphere behind the parieto-occipital sulcus, tapering to the occipital pole. Most of it is visual: primary visual cortex lines the calcarine sulcus, with the cuneus above and the lingual gyrus below, and successive extrastriate areas wrap outward from there. It receives its main input from the lateral geniculate nucleus of the thalamus, and the mapping is retinotopic and inverted, so the upper visual field projects below the calcarine sulcus. Damage produces field defects whose shape follows that map closely enough to localize the lesion.
Supplementary motor area
Cortex in the motor cortex
The medial part of Brodmann area 6, on the inner surface of the hemisphere just in front of the leg area of motor cortex.
The supplementary motor area lies on the medial wall of the superior frontal gyrus, anterior to the primary motor leg representation and above the cingulate sulcus. It projects directly to the spinal cord as well as to primary motor cortex, and stimulation evokes complex postures rather than single twitches. It is engaged by internally generated and sequential movements, by bimanual coordination, and by timing. The readiness potential preceding voluntary action arises largely here, which is why the region features in arguments about volition that go well beyond what the recordings themselves establish.
Temporal lobe
Lobe in the cerebral hemisphere
The lobe below the lateral fissure, running forward to the temporal pole and back into parietal and occipital cortex without a sharp boundary.
The temporal lobe sits beneath the lateral fissure, folded into superior, middle and inferior temporal gyri laterally, with the fusiform and parahippocampal gyri on its ventral surface and the hippocampal formation curled inside its medial wall. Heschl's gyrus, buried in the fissure's floor, carries primary auditory cortex. Its posterior boundary against parietal and occipital cortex is conventional rather than anatomical. The lobe is associated with auditory processing, object and face recognition along its ventral stream, semantic knowledge at its pole, and, through the medial structures, declarative memory.
Bock (1986)
Syntactic persistence in language production
In
Cognitive Psychology, 18(3), 355-387
Abstract
Activation processes appear to have an important impact on the mechanisms of language use, including those responsible for syntactic structure in speech. Some implications of this claim for theories of language performance were examined with a syntactic priming procedure. On each priming trial, subjects produced a priming sentence in one of several syntactic forms. They then viewed an unrelated event in a picture and described it in one sentence. The probability of a particular syntactic form being used in the description increased when that form had occurred in the prime, under presentation conditions that minimized subjects' attention to their speech, to the syntactic features of the...
Find it
Cai et al. (2024)
Do large language models resemble humans in language use?
In
Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics, 37-56
Find it
Gong et al. (2016)
Dissociation between Conceptual and Perceptual Implicit Memory: Evidence from Patients with Frontal and Occipital Lobe Lesions
In
Frontiers in Human Neuroscience, 9
Abstract
The latest neuroimaging studies about implicit memory (IM) have revealed that different IM types may be processed by different parts of the brain. However, studies have rarely examined what subtypes of IM processes are affected in patients with various brain injuries. Twenty patients with frontal lobe injury, 25 patients with occipital lobe injury, and 29 healthy controls (HC) were recruited for the study. Two subtypes of IM were investigated by using structurally parallel perceptual (picture identification task) and conceptual (category exemplar generation task) IM tests in the three groups, as well as explicit memory (EM) tests. The results indicated that the priming of conceptual IM...
Find it
Graf & Schacter (1985)
Implicit and explicit memory for new associations in normal and amnesic subjects.
In
Journal of Experimental Psychology: Learning, Memory, and Cognition, 11(3), 501-518
Abstract
Two experiments, with 64 undergraduates (Exp I) and 12 20–67 yr old amnesic patients and 2 control groups (12 22–81 yr olds and 24 undergraduates with no memory deficits \[Exp II\]), investigated whether repetition priming effects on a word comparison completion task are influenced by new associations between unrelated word pairs that were established during a single study trial. On the word completion task, Ss were presented with the initial 3 letters of the response words from the study list pairs and completed these fragments with the first words that came to mind. The fragments were shown either with the paired words from the study list (same context) or with other words (different...
Find it
Jumelet et al. (2024)
Do Language Models Exhibit Human-like Structural Priming Effects?
In
Findings of the Association for Computational Linguistics ACL 2024, 14727-14742
Abstract
We explore which linguistic factors—at the sentence and token level—play an important role in influencing language model predictions, and investigate whether these are reflective of results found in humans and human corpora (Gries and Kootstra, 2017). We make use of the structural priming paradigm—where recent exposure to a structure facilitates processing of the same structure—to investigate where priming effects manifest, and what factors predict them. We find these effects can be explained via the inverse frequency effect found in human priming, where rarer elements within a prime increase priming effects, as well as lexical dependence between prime and target. Our results provide an...
Find it
Keane et al. (1995)
Double dissociation of memory capacities after bilateral occipital-lobe or medial temporal-lobe lesions
In
Brain, 118(5), 1129-1148
Abstract
Memory for recently encountered information can be reflected in conscious recall and recognition of that material, or in facilitated reprocessing of that material, an effect known as repetition priming. Repetition priming may be perceptual (form-based) or conceptual (meaning-based). A patient with bilateral occipital-lobe lesions (L.H.) and a patient with bilateral medial-temporal lobe lesions (H.M.). L.H. showed a double dissociation between visuoperceptual priming (impaired in L.H. and intact in H.M.) and visual recognition memory (intact in L.H. and impaired in H.M.). L.H. showed intact conceptual priming for visually presented words; his pattern of impaired visuoperceptual priming and...
Find it
Lee et al. (2020)
Neural Correlates of Repetition Priming: A Coordinate-Based Meta-Analysis of fMRI Studies
In
Frontiers in Human Neuroscience, 14
Abstract
Repetition priming is a form of implicit memory, whereby classification or identification of a stimulus is improved by prior presentation of the same stimulus. Repetition priming is accompanied with a deceased fMRI signal for primed vs. unprimed stimuli in various brain regions, often called “repetition suppression,” or RS. Previous studies proposed that RS in posterior regions is associated with priming of perceptual processes, whereas RS in more anterior (prefrontal) regions is associated with priming of conceptual processes. To clarify which regions exhibit reliable RS associated with perceptual and conceptual priming, we conducted a quantitative meta-analysis using coordinate-based...
Find it
Lee et al. (2026)
Do Language Models Show Structural Priming Across Different Domains?
In
Proceedings of the 1st Workshop on Computational Developmental Linguistics (CDL), 37-51
Find it
Michaelov et al. (2023)
Structural Priming Demonstrates Abstract Grammatical Representations in Multilingual Language Models
In
Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 3703-3720
Find it
Roediger (1993)
Implicit memory in normal human subjects
Segaert et al. (2012)
Shared Syntax in Language Production and Language Comprehension--An fMRI Study
In
Cerebral Cortex, 22(7), 1662-1670
Abstract
During speaking and listening syntactic processing is a crucial step. It involves specifying syntactic relations between words in a sentence. If the production and comprehension modality share the neuronal substrate for syntactic processing then processing syntax in one modality should lead to adaptation effects in the other modality. In the present functional magnetic resonance imaging experiment, participants either overtly produced or heard descriptions of pictures. We looked for brain regions showing adaptation effects to the repetition of syntactic structures. In order to ensure that not just the same brain regions but also the same neuronal populations within these regions are...
Find it
Sinclair et al. (2026)
Structural priming in humans and large language models
In
Journal of Memory and Language, 149, 104713
Find it
Sun et al. (2025)
How new data permeates LLM knowledge and how to dilute it
In
arXiv
Abstract
Large language models learn and continually learn through the accumulation of gradient-based updates, but how individual pieces of new information affect existing knowledge, leading to both beneficial generalization and problematic hallucination, remains poorly understood. We demonstrate that when learning new information, LLMs exhibit a "priming" effect: learning a new fact can cause the model to inappropriately apply that knowledge in unrelated contexts. To systematically study this phenomenon, we introduce "Outlandish," a carefully curated dataset of 1320 diverse text samples designed to probe how new knowledge permeates through an LLM's existing knowledge base. Using this dataset, we...
Find it
Weber et al. (2016)
fMRI Syntactic and Lexical Repetition Effects Reveal the Initial Stages of Learning a New Language
In
Journal of Neuroscience, 36(26), 6872-6880
Abstract
When learning a new language, we build brain networks to process and represent the acquired words and syntax and integrate these with existing language representations. It is an open question whether the same or different neural mechanisms are involved in learning and processing a novel language compared with the native language(s). Here we investigated the neural repetition effects of repeating known and novel word orders while human subjects were in the early stages of learning a new language. Combining a miniature language with a syntactic priming paradigm, we examined the neural correlates of language learning on-line using functional magnetic resonance imaging. In left inferior...
Find it