| phenomenon | semantic memory retrieval and semantic priming— “The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept.” |
| claim-type | mechanism— “1. When a concept is processed (or stimulated), activation spreads out along the paths of the network in a decreasing gradient.” |
| method — evidence | — “In the first section, we briefly review the original theory while trying to correct a number of the common misunderstandings concerning it.” |
| extends | Quillian's (1966, 1969) spreading-activation theory of semantic memory— “We have extended Quillian's spreadingactivation theory of semantic processing in order to deal with a number of experiments that have been performed on semantic memory in recent years.” | claim_relations[0] |
| contradicts | the Smith, Shoben and Rips (1974) feature-comparison model— “But there is a fundamental disagreement, because they argue that all categorizations judgments are made by comparing features of the instance and category, whereas we argue that people use whatever evidence they find, including superordinate links.” | claim_relations[1] |
| contradicts | Conrad's (1972) argument that the Collins and Quillian (1969) results are confounded by property frequency— “In conclusion, the differences between the two experiments and the fact that the only relevant data do not particularly support the conjecture about a confounding of property level and property frequency make Conrad's argument rather tenuous.” | claim_relations[2] |
| sense | THE SENTENCE THE PRIMING LITERATURE IS BUILT ON: `The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept.` Every semantic-priming record in this corpus argues about, extends or tests this - and the parenthetical `(or priming)` is the term arriving inside the theory that would carry it. The corpus now holds the model itself and not only the papers that assume it— “The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept” | senses[0] |
| sense | AND THE ORIGIN NOTE, WHICH IS THE MOST USEFUL SENTENCE HERE FOR A PROJECT LOOKING AT LLMs: `Rather than a theory to explain data, it was a theory designed to show how to build human semantic structure and processing into a computer.` Quillian wanted to implement it. A computational proposal became the field's psychological explanation, and the paper that completed the conversion says so in its second paragraph. The traffic between computation and psychology has run both ways here since 1962, and this record marks the turn— “Rather than a theory to explain data, it was a theory designed to show how to build human semantic structure and processing into a computer” | senses[1] |
| sense | AND SEARCH IS DEFINED AS TWO WAVEFRONTS MEETING: `The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found.` Not a lookup and not a strength comparison - **a geometry.** That is what makes semantic DISTANCE meaningful in this model and why the whole downstream literature measures relatedness in path terms— “The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found” | senses[2] |
| sense | AND THE MODEL PAYS FOR ITS OWN GENEROSITY, WHICH IS WHAT MAKES IT TESTABLE: `Assumptions 2 and 3 impose a limitation on the amount of activation that can be allocated in priming more than one concept, because the more concepts that are primed, the less each will be primed.` A conservation constraint, derived from two numbered assumptions rather than added to fit a result. It predicts that priming cannot scale freely - a cost an unconstrained associative account does not carry, and a prediction later work can collect— “Assumptions 2 and 3 impose a limitation on the amount of activation that can be allocated in priming more than one concept, because the more concepts that are primed, the less each will be primed” | senses[3] |
| sense | `semantic relatedness` vs `semantic distance` - DISTINGUISHED HERE AND CONFLATED IN THE FIELD, in one sentence: `a slightly different notion ... though the two terms are sometimes used interchangeably`. The paper draws the distinction and reports that its field does not keep it.— “Semantic relatedness is a slightly different notion from semantic distance, though the two terms are sometimes used interchangeably” | senses[4] |
| differs in | the Smith, Shoben and Rips (1974) feature-comparison model — THE AXIS IS WHETHER ALL JUDGEMENTS USE ONE MECHANISM. The span names it as a 'fundamental disagreement' and states both sides in one clause: 'they argue that all categorizations judgments are made by comparing features of the instance and category, whereas we argue that people use whatever evidence they find, including superordinate links.' Feature-comparison as the sole route, against evidence of any available kind.— “But there is a fundamental disagreement, because they argue that all categorizations judgments are made by comparing features of the instance and category, whereas we argue that people use whatever evidence they find, including superordinate links.”— “they argue that all categorizations judgments are made by comparing features of the instance and category, whereas we argue that people use whatever evidence they find” | claim_relations[1] |
| effect direction | spreading activation —causes→ priming — increases— “The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept”— “explained in terms of spreading activation from the node of the primed concept” | concept_edges[0] |
| note | spreading activation —causes→ priming — THE FOUNDATIONAL EDGE OF THIS CORPUS'S SUBJECT. Activation spreads from the primed concept's node and the effects of priming follow. Weight 0.95: it is the theory's own statement of its own mechanism, in the paper that established it, cited 5,530 times. Note the verb is 'explained' - this is a proposed mechanism, not a measurement, and everything measurable about it is downstream. The corpus holds the measurements elsewhere; this is where they get their meaning.— “The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept” | concept_edges[0] |
| weight | spreading activation —causes→ priming — 0.95— “The effects of preparation (or priming) in semantic memory were also explained in terms of spreading activation from the node of the primed concept” | concept_edges[0] |
| audit | concept nodes —causes→ intersection — Audit 2026-09-28: DOUBTFUL - read against its span and kept, but the span describes search spreading from concept nodes until an intersection is found; causes is loose.— “The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found” | concept_edges[1] |
| effect direction | concept nodes —causes→ intersection — increases— “The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found”— “activation spreading from two or more concept nodes” | concept_edges[1] |
| note | concept nodes —causes→ intersection — SEARCH AS A GEOMETRY: two wavefronts, and the search ends where they meet. The edge records that intersections are PRODUCED by activation spreading from multiple nodes rather than being looked up. Weight 0.9 rather than 0.95 - the quote is Collins and Loftus reporting Quillian's original formulation, and their own paper then adds the threshold that makes an intersection cost something (fourth concept).— “The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found” | concept_edges[1] |
| weight | concept nodes —causes→ intersection — 0.9— “The theory viewed memory search as activation spreading from two or more concept nodes in a semantic network until an intersection was found” | concept_edges[1] |
| replication | untested |
| limitation | The authors flag that their elaboration of Quillian's theory may itself be wrong.— “The elaboration may itself be wrong, so our mistakes should not be held against Quillian's theory.” | limitations[0] |
| limitation | They do not weigh alternative explanations of the experiments they reinterpret.— “Our objective is to show how a spreading-activation theory can handle these results, not to consider all the possible alternative explanations of the experiments.” | limitations[1] |
| limitation | The set of evidence types postulated is not exhaustive.— “It is beyond the scope of this paper, however, to consider all the different ways people use evidence to make semantic decisions.” | limitations[2] |
| limitation | As a computational prescription the theory omits many decision-strategy details.— “though at that level many details are omitted about decision strategies for different judgments that arise in language processing” | limitations[3] |