phenomenon · claim-type · cross-domain-transfer · mechanism · claim-relation · replication-status · limitation — grounded in the full text| phenomenon | semantic priming— “The semantic priming effect refers to the consistent observation that people respond faster to a target word (e.g., cat) when it is preceded by a semantically related prime (e.g., dog) rather than by an unrelated prime (e.g., table).” | |
| claim-type | mechanism— “The most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | |
| method — evidence | — “In the present review, I critically examine the individual studies claiming evidence of featural and associative relations in semantic memory.” | |
| mechanism | Activation spreads from the prime's concept along indirect links to concepts it is not directly associated with, producing mediated priming.— “According to spreading activation models, when a concept (e.g., LION) is activated, this activation spreads not only to directly associated concepts (e.g., TIGER), but also to indirectly related concepts (e.g., STRIPES).” | mechanisms[0] |
| mechanism | Participants notice that related pairs predict word targets and check back from target to prime, so relatedness biases the lexical decision after the target has been accessed.— “Participants who become aware of this contingency may begin to use prime–target relatedness strategically to facilitate their lexical decisions by checking back after the presentation of a target to see if it is related to the prime” | mechanisms[1] |
| mechanism | An auditory prime leaves an echoic trace that survives into target presentation, letting participants integrate or strategically match prime and target and so inflating auditory priming effects.— “Alternatively, because auditory primes leave a longer echoic trace that persists after the presentation of the target, an equally plausible explanation is that such a trace allows for some type of strategic matching or prime–target integration by participants.” | mechanisms[2] |
| mechanism | In distributed recurrent networks a featurally similar prime leaves the network nearer the target's attractor basin, so it settles into the target pattern in fewer iterations.— “If the previous word was semantically similar to the new word, the network will settle faster to the attractor basin of the new stimulus than if the previous word was dissimilar.” | mechanisms[3] |
| contradicts | Lucas (2000)— “Finally, whereas Lucas (2000) concluded that there is little to no evidence of pure associative priming, the present review leads to the opposite conclusion.” | claim_relations[0] |
| contradicts | Lucas (2000)— “First, although Lucas (2000) claims that automatic priming does occur for category coordinates, close inspection of her studies reveals that priming from these items is not automatic.” | claim_relations[1] |
| builds-on | Lucas (2000)— “It is hoped that the inclusion of pure associative studies and the present, more careful analysis of pure semantic priming, when combined with Lucas’s (2000) more global analysis of pure semantic priming, will provide a clearer overall picture of the evidence for and against both associative and featural priming.” | claim_relations[2] |
| sense | THE ANSWER IS BOTH, AND IT IS NOT A COMPROMISE - IT IS A PARTITION BY STIMULUS CLASS. `The most important conclusion is that automatic priming appears to be due to BOTH association strength AND feature overlap.` Then the split: `Mediated associates provide the strongest evidence of automatic associative priming, whereas functional associates, synonyms, and antonyms instead support priming based on feature overlap.` Different stimulus classes, different mechanisms, one label over all of them - which is precisely the condition under which a controlled vocabulary stops being informative— “Mediated associates provide the strongest evidence of automatic associative priming, whereas functional associates, synonyms, and antonyms instead support priming based on feature overlap.” | senses[0] |
| sense | AND IT CATCHES A CLAIM CHANGING MEANING ALONG A CITATION CHAIN, BY HAND. Lucas (2000) cited Hodgson (1991), who cites Neely (1991), for the claim that priming is smaller in pronunciation than in lexical decision. `However, Neely (1991) was actually referring to numerical effects in milliseconds and not to an estimate of effect size such as Cohen's d.` A claim about milliseconds became a claim about effect sizes in two citations, and the meta-analysis built on it. This is the failure the concept-record design exists to make visible, found here by one person reading a chain backwards— “However, Neely (1991) was actually referring to numerical effects in milliseconds and not to an estimate of effect size” | senses[1] |
| sense | THE VERDICT ON A WHOLE CLASS OF STUDIES, STATED WITHOUT SOFTENING: `Therefore, studies of priming based only on association strength are virtually useless for distinguishing between whether such effects are due to association or to feature overlap.` The reason is the confound: associated words tend also to be similar words, so varying association alone varies both. Not `limited`, not `to be interpreted with caution`. A review willing to write that sentence about its own field's standard design is rare enough to be worth the tag on its own— “Therefore, studies of priming based only on association strength are virtually useless for distinguishing between whether such effects are due to association or to feature overlap.” | senses[2] |
| sense | `dominance` and `association strength` - REPORTED AS INTERCHANGEABLE IN THE LITERATURE, with a citation: `the terms dominance and association strength have been used interchangeably (Lorch, 1982)`. Two measures that a paper arguing about WHICH ONE DRIVES PRIMING needs kept apart.— “In fact, the terms dominance and association strength have been used interchangeably (Lorch, 1982)” | senses[3] |
| sense | `similar` - THE MEASURED QUANTITY DEPENDS ON HOW THE EXPERIMENTER DEFINED THE WORD TO THE PARTICIPANT, AND THE PAPER LISTS THAT AS ONE OF FIVE SOURCES OF VARIANCE: `there is no universal method for determining similarity in this task`, and ratings vary on (1) the other words in the norming set, (2) HOW THE RESEARCHER DEFINES SIMILAR FOR THE PARTICIPANTS, (3) presentation order, (4) instructional set, (5) association between the items. THE DEFINITION IS NOT UPSTREAM OF THE MEASUREMENT HERE - IT IS ONE OF ITS INPUTS, named alongside stimulus order.— “However, there is no universal method for determining similarity in this task, and pairs’ ratings can vary drastically depending on (1) the other words used in the norming process, (2) how the researcher defines similar for the participants, (3) the order in which the words are presented, (4) the participants’ instructional or response set… | senses[4] |
| sense | the CONSCIOUS/UNCONSCIOUS boundary - A THEORETICAL DISTINCTION WITH NO UNIVERSAL OPERATIONAL VALUE: `there is no universal SOA that distinctly separates conscious from unconscious processing, and the SOA necessary to demonstrate automatic priming may vary from task to task`. A THRESHOLD DISPUTE, not a naming one - the two categories are not in doubt and the number that divides them is, so no nomenclature body can repair it.— “Unfortunately, there is no universal SOA that distinctly separates conscious from unconscious processing, and the SOA necessary to demonstrate automatic priming may vary from task to task (see Hutchison, Neely, & Johnson, 2001).” | senses[5] |
| control | every target was counterbalanced BOTH ACROSS AND WITHIN subjects so that each appeared in the related, unrelated and neutral conditions — held — rules out that a priming effect is an item effect - that the related and unrelated sets simply contain different words— “In addition, each target was counterbalanced both across and within subjects to appear in the related, the unrelated, or the neutral condition.” | controls[0] |
| effect direction (stated by the tagger, no quotation) | association strength —causes→ automatic priming — increases | concept_edges[0] |
| note | association strength —causes→ automatic priming — HALF OF A CONCLUSION SENTENCE, AND THE OTHER HALF IS THE NEXT EDGE. The span names all three endpoints, so two triples come out of one quote and the note has to say which half each is carrying or a reader gets a partial claim as if it were the whole one. Weight 0.95: it is the review's stated conclusion, both endpoints are named, and 'due to' is unambiguous. The 'both' is the finding - either edge alone reproduces exactly the one-sided reading this paper was written to correct.— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concept_edges[0] |
| weight | association strength —causes→ automatic priming — 0.95— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concept_edges[0] |
| effect direction (stated by the tagger, no quotation) | feature overlap —causes→ automatic priming — increases | concept_edges[1] |
| note | feature overlap —causes→ automatic priming — The other half. Lucas's (2000) meta-analysis concluded feature overlap was the WHOLE story - 'implies that semantic…' - and this review keeps the mechanism while removing the 'primarily'. A finding preserved and a quantifier withdrawn, which is a shape the schema records only in prose.— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concept_edges[1] |
| weight | feature overlap —causes→ automatic priming — 0.95— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concept_edges[1] |
| audit | association strength —distinct-from→ feature overlap — Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span notes the DIFFICULTY of separating the two; distinct-from is loose.— “The second problem facing researchers is the difficulty in separating association strength from feature overlap.” | concept_edges[2] |
| note | association strength —distinct-from→ feature overlap — THE EDGE IS A CONFOUND, WHICH IS A STRANGE THING TO BE ABLE TO SAY. 'The second problem…' They are distinct - the whole review depends on that - and in ordinary stimuli they travel together, because words people associate tend also to be words that resemble each other. Weight 0.9 rather than 0.95: the span asserts the distinction only by asserting the difficulty of drawing it. That is real evidence of distinctness and it is weaker than a sentence saying so directly.— “The second problem facing researchers is the difficulty in separating association strength from feature overlap.” | concept_edges[2] |
| weight | association strength —distinct-from→ feature overlap — 0.9— “The second problem facing researchers is the difficulty in separating association strength from feature overlap.” | concept_edges[2] |
| outcome (stated by the tagger, no quotation) | control — held | controls[0] |
| replication | contested | |
| limitation | The status of several relation types (collocates, episodic relatives, script relations) is left unresolved and needs further experiments.— “The status of other relations, such as collocates, episodic relatives, and script relations, is unclear and requires further experimentation.” | limitations[0] |
| limitation | Restricting the review to published studies leaves the direction of the resulting bias in effect-size estimates unknown.— “In the present case, it is unclear whether this should increase or decrease one’s estimates of effect size.” | limitations[1] |
| limitation | The review cannot say why auditory presentation yields so much larger priming estimates than visual presentation.— “It is unclear why presentation modality should make such a large difference in effect size estimates.” | limitations[2] |
| limitation | The available evidence cannot decide whether the associative boost reflects combined lexical and semantic connections or simply greater featural overlap.— “However, one cannot tell, on the basis of the available evidence, whether this associative boost is due to combined lexical and semantic connections, as Collins and Loftus (1975) predicted, or simply to a greater overlap in semantic features between these items (Chiarello et al., 1990; Lucas, 2000; McRae & Boisvert, 1998).” | limitations[3] |
| limitation | THE NORMS THE WHOLE LITERATURE RESTS ON ARE CONCEDED NOT TO BE WHAT THEY ARE TAKEN FOR: 'Close inspection of both associative and categorical norms therefore leads one to the uncomfortable conclusion that they are not nearly as pure as has been assumed.' This is the review's own instrument being withdrawn mid-argument, and it is what licenses a senses row's verdict on a whole class of studies.— “Close inspection of both associative and categorical norms therefore leads one to the uncomfortable conclusion that they are not nearly as pure as has been assumed.” | limitations[4] |
finding · quantity · variable · definition · key-term — grounded in the full text| finding | Automatic semantic priming is produced by both association strength and feature overlap, not by either alone.— “The most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | content_tags.findings[0] |
| finding | Mediated associates (e.g. lion-stripes) give the strongest evidence for automatic purely associative priming, while functional associates, synonyms and antonyms instead support priming based on feature overlap.— “Mediated associates provide the strongest evidence of automatic associative priming, whereas functional associates, synonyms, and antonyms instead support priming based on feature overlap.” | content_tags.findings[1] |
| finding | Category coordinates and perceptually similar items show no automatic priming when the stimuli are presented visually.— “In contrast, automatic priming does not occur for category coordinates or perceptually similar items, at least when presented in the visual modality.” | content_tags.findings[2] |
| finding | Category coordinates produced only 4 ms of priming in the pronunciation task.— “By contrast, only 4 msec of priming from these items occurs in the pronunciation task.” | content_tags.findings[3] |
| finding | In the standard lexical decision task at long SOA, backward priming (28 ms) was comparable to forward priming (34 ms), and statistically equivalent in every experiment.— “Moreover, averaged across all experiments using the standard LDT and a long SOA, the size of the backwardpriming effect (M 5 28 msec) is comparable to the size of the forward-priming effect (M 5 34 msec), and in every experiment the two priming effects were statistically equivalent.” | content_tags.findings[4] |
| finding | Mediated priming is strongest when strategic processing is minimised and disappears when it is encouraged.— “In fact, mediated priming is strongest when strategic processing is minimized and disappears when strategic processing is encouraged.” | content_tags.findings[5] |
| finding | LSA similarity of Balota and Lorch's mediated pairs (.20) was significantly higher than when the items were randomly re-paired (.10).— “The mean LSA similarity value of .20 for Balota and Lorch’s mediated items was significantly greater than the .10 value given to these items when they were randomly re-paired to form unrelated controls [t(47) 5 4.27, p , .001].” | content_tags.findings[6] |
| finding | Auditory-presentation studies gave more than twice the effect size of visual ones (d = .51 vs .25).— “In fact, if one simply averages her unweighted effect sizes from the five studies in which auditory presentation was used, the resulting Cohen’s d of .51 is more than twice as large as the d of .25 obtained from the remaining visual presentation studies.” | content_tags.findings[7] |
| finding | Published studies averaged an effect size of .35, almost twice the .18 of unpublished studies.— “An informal examination of the effect sizes appears to favor the first explanation, since the unweighted mean effect size from published studies was .35, an estimate almost twice as large as the .18 effect size from unpublished studies.” | content_tags.findings[8] |
| finding | Contrary to Lucas's claim, close inspection of her studies shows that priming from category coordinates is not automatic.— “First, although Lucas (2000) claims that automatic priming does occur for category coordinates, close inspection of her studies reveals that priming from these items is not automatic.” | content_tags.findings[9] |
| quantity | 31 msec — average priming from category coordinates across studies using the standard lexical decision task— “Across studies, approximately 31 msec of priming from these items is found using the standard LDT.” | content_tags.quantities[0] |
| quantity | 4 msec — average priming from the same category-coordinate items in the pronunciation task— “By contrast, only 4 msec of priming from these items occurs in the pronunciation task.” | content_tags.quantities[1] |
| quantity | t(47) 5 4.27, p , .001 — test comparing LSA similarity of mediated pairs (.20) with randomly re-paired controls (.10)— “The mean LSA similarity value of .20 for Balota and Lorch’s mediated items was significantly greater than the .10 value given to these items when they were randomly re-paired to form unrelated controls [t(47) 5 4.27, p , .001].” | content_tags.quantities[2] |
| quantity | .51 — mean Cohen's d for the five auditory-presentation studies, against .25 for the visual ones— “In fact, if one simply averages her unweighted effect sizes from the five studies in which auditory presentation was used, the resulting Cohen’s d of .51 is more than twice as large as the d of .25 obtained from the remaining visual presentation studies.” | content_tags.quantities[3] |
| quantity | .35 — unweighted mean effect size of published studies, against .18 for unpublished ones— “An informal examination of the effect sizes appears to favor the first explanation, since the unweighted mean effect size from published studies was .35, an estimate almost twice as large as the .18 effect size from unpublished studies.” | content_tags.quantities[4] |
| quantity | 26 studies (accounting for 116 experiments) — size of the Lucas (2000) meta-analysis that this review re-examines— “Lucas (2000) examined the overall effect size from 26 studies (accounting for 116 experiments) that claimed to somehow control for association strength” | content_tags.quantities[5] |
| quantity | over 6,000 participants — norming sample behind the Nelson et al. (1999) association norms used to recheck every study's stimuli— “the Nelson et al. values were normed on over 6,000 participants, making for more reliable estimates.” | content_tags.quantities[6] |
| quantity | 160 million words — size of the Usenet corpus from which HAL builds its co-occurrence matrix— “The HAL model of semantic memory assesses a word’s meaning through the construction of a co-occurrence matrix based on a sample of 160 million words taken from Usenet newsgroups.” | content_tags.quantities[7] |
| quantity | 14 — number of association types into which the author classified each Palermo-Jenkins norm pair, showing association norms are heterogeneous— “On the basis of previous studies and an informal analysis, I classified each stimulus and its primary associate in these norms into 1 or more of 14 possible relations.” | content_tags.quantities[8] |
| variable | priming effect (dependent)— “For instance, in the standard LDT, the mean perceptual priming of 20 msec is reduced to 4 msec when the RP is lowered from .75 to .30.” | content_tags.variables[0] |
| variable | RP (moderator)— “As can be seen in Table 5, priming from perceptually similar items in both LDTs and pronunciation tasks depends critically on the RP.” | content_tags.variables[1] |
| variable | presentation modality (moderator)— “In fact, if one simply averages her unweighted effect sizes from the five studies in which auditory presentation was used, the resulting Cohen’s d of .51 is more than twice as large as the d of .25 obtained from the remaining visual presentation studies.” | content_tags.variables[3] |
| variable | association strengths (covariate)— “In order to confirm the association strength assigned by the authors to their stimulus pairs, I have calculated the forward and backward association strengths of the items in each study using the Nelson et al. (1999) norms.” | content_tags.variables[4] |
| variable | type of task (moderator)— “The experiments in this table are organized according to the type of task used (standard LDT, masked LDT, continuous LDT, standard pronunciation, and pronunciation with a visually degraded target) and the SOA between the presentation of the prime and the presentation of the target (greater than or less than 300 msec), on the assumption that… | content_tags.variables[5] |
| variable | SOA (moderator)— “The experiments in this table are organized according to the type of task used (standard LDT, masked LDT, continuous LDT, standard pronunciation, and pronunciation with a visually degraded target) and the SOA between the presentation of the prime and the presentation of the target (greater than or less than 300 msec), on the assumption that… | content_tags.variables[6] |
| definition | semantic priming effect— “The semantic priming effect refers to the consistent observation that people respond faster to a target word (e.g., cat) when it is preceded by a semantically related prime (e.g., dog) rather than by an unrelated prime (e.g., table).” | content_tags.definitions[0] |
| definition | automatic— “In line with Posner and Snyder, the term automatic here suggests that the process occurs quickly and unintentionally, even if one is simultaneously engaged in other activities.” | content_tags.definitions[1] |
| definition | association strength— “The association strength of a given pair (e.g., dog–cat) refers to the percentage of people responding with the target word (cat) when given the cue (dog).” | content_tags.definitions[2] |
| definition | expectancy generation— “The expectancy generation component involves the use of the context provided by the prime to anticipate consciously the approximate area in the network (or collection of features, if a distributed representation is assumed) containing the target word.” | content_tags.definitions[3] |
| definition | instrument relations— “Instruments (e.g., hammer–nail) are related in that the intended function of one of the items is to perform some action on the other.” | content_tags.definitions[5] |
| definition | script relations— “Script relations (e.g., restaurant–wine) involve items that are part of the same script or schema” | content_tags.definitions[6] |
| definition | attractor basins— “These learned patterns are referred to as attractor basins, because the network is biased to settle into one of these known patterns when given partial information.” | content_tags.definitions[7] |
| definition | semantic distance— “According to Collins and Loftus (1975), semantic distance is the “shortest path [direct or indirect] between two nodes” (p. 412, note 3).” | content_tags.definitions[8] |
| definition | Hyperspace Analogue to Language (HAL)— “The HAL model of semantic memory assesses a word’s meaning through the construction of a co-occurrence matrix based on a sample of 160 million words taken from Usenet newsgroups.” | content_tags.definitions[9] |
| concept | association strength — not defined in this paper's own words; how often people produce one word given the other - a fact about co-occurrence in usage, not about meaning. 'bread' primes 'butter' without the two being alike in any respect. This is the variable meyer1971 built the original effect on in 1971, taken straight from free-association norms— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concepts[0] |
| concept | feature overlap — not defined in this paper's own words; shared semantic features - 'robin' and 'sparrow' are both small, both fly, both nest. The reading that would justify calling the effect SEMANTIC. Lucas's meta-analysis concluded this was the whole story; this review finds it is half of it— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concepts[1] |
| concept | automatic priming — not defined in this paper's own words; priming that survives the controls designed to strip out strategy - short SOAs, low relatedness proportion, tasks without a binary decision. The thing both mechanisms turn out to feed, which is why isolating it does not settle what causes it— “In terms of models of semantic priming, the most important conclusion is that automatic priming appears to be due to both association strength and feature overlap.” | concepts[2] |
| concept | mediated associates — not defined in this paper's own words; 'lion' and 'stripes', related only through 'tiger'. The cleanest evidence for a purely associative route, because a mediated pair shares almost no features - whatever primes here cannot be feature overlap. The stimulus class is doing the work a control condition usually does— “Mediated associates provide the strongest evidence of automatic associative priming, whereas functional associates, synonyms, and antonyms instead support priming based on feature overlap.” | concepts[3] |
| concept | semantic matching — not defined in this paper's own words; a POSTLEXICAL strategy: having accessed both words, the subject checks whether they are related and lets the answer bias the word/nonword decision. Not a memory effect at all, a decision effect - and it is confounded with the nonword ratio in the standard task, which is why the review keeps returning to which task was used— “Two particular conscious strategies that could potentially obscure the detection of any automatic activation in priming tasks have been suggested: expectancy generation and postlexical semantic matching (see Neely, 1991, for a review).” | concepts[4] |
| concept | stimulus onset asynchrony — not defined in this paper's own words; the interval between prime onset and probe presentation, used as the operational divider between conscious and unconscious processing— “Unfortunately, there is no universal SOA that distinctly separates conscious from unconscious processing, and the SOA necessary to demonstrate automatic priming may vary from task to task (see Hutchison, Neely, & Johnson, 2001).” | concepts[5] |
| concept | semantic similarity — not defined in this paper's own words; participants rate prime-target pairs for similarity; the procedure has no standard form— “In other studies, researchers give participants prime– target pairs and ask them to rate each pair in terms of the semantic similarity of its components. However, there is no universal method for determining similarity in this task, and pairs’ ratings can vary drastically depending on (1) the other words used in the norming… | concepts[6] |
| t-test t | as written: [t(47) 5 4.27, p , .001] | content_tags.statistics[0] |
| effect size d | as written: unweighted effect sizes ... Cohen’s d | content_tags.statistics[1] |
| mean M | as written: averaged across all experiments ... (M 5 28 msec) | content_tags.statistics[2] |
| p-value p | as written: (all ps , .05) | content_tags.statistics[3] |
When we have changed our mind. Some entries in the table begin with a word in capitals and the reason for it. EVIDENCE CORRECTED — the finding still stands, but we had quoted the wrong sentence from the paper, and the quotation has been replaced with the right one. RETRACTED — the entry itself was wrong; it stays on the page, marked, with the reason, rather than being deleted. WITHDRAWN AND REPLACED — the entry was wrong, and a corrected version of it appears in the entries below. Each mark carries its own reason, in brackets, because the reasons differ from one entry to the next. An entry we have WITHDRAWN is removed from this page and kept in the record, with its reason. A line reading Audit and a date is a later re-reading of the entry against its whole sentence: DOUBTFUL means it was kept, and the reason says what is loose about it. Quotation marks inside the tagger's sentence. Where a tag is the tagger's own sentence, a phrase in quotation marks inside it is borrowed from the paper — the tagger's wording around it is not. The paper's own sentence is the evidence quote for that row; on this page it is withheld with the article, and the article is at its publisher. Nothing inside quotation marks anywhere on this page is ours: if it is quoted, it is the paper's. Whose words a concept carries. A concept entry records a term the paper itself uses, so the term, the sense given for it, and its evidence are the paper's own words, quoted where the paper fixes them. This is the opposite of a sense entry in the tag table, which is the tagger's sentence about how the paper uses a term. Where a concept entry could not be grounded in the paper's words, the entry says so on its own row.
Tag layer © the Meta-Tagging Project, CC BY 4.0. The paper text is © its authors/publisher
and is linked, not reproduced. Excerpts shown are short quotations for identification and verification.
· back to index
· concepts · authors
· עברית
© 2026 Shir Sivroni · ORCID 0009-0000-2597-2824