Semantic Memory II
Foundations and Architecture of Semantic Memory
Taxonomy of Long-Term Memory Systems:
Long-term memory is fundamentally bifurcated into two major branches: explicit (declarative) memory and implicit (non-declarative) memory.
Explicit (declarative) memory encompasses representations that can be consciously accessed, reflected upon, and articulated through language.
Declarative memory is further subdivided into:
Episodic Memory: Stores personally experienced events situated in specific spatial and temporal contexts (autobiographical episodes).
Semantic Memory: Stores general decontextualized knowledge, factual information, concepts, and linguistic meanings.
Implicit (non-declarative) memory comprises non-conscious expressions of prior learning, including:
Procedural Memory: Motor skills, cognitive habits, and behavioral actions.
Emotional Conditioning: Conditioned emotional reactions mediated through associative learning mechanisms.
Structural Models of Semantic Storage:
Hierarchical Network Models:
Propose that semantic information is arranged in strict logical hierarchies of nodes and superordinate/subordinate categories.
Employ the principle of cognitive economy: properties are stored exclusively at the highest possible category node (e.g., "has wings" stored at the "bird" level rather than separately for "robin," "canary," and "sparrow").
Associative Network Models (Spreading Activation):
Replace rigid hierarchical trees with interconnected semantic webs where conceptual distance reflects associative strength and typicality rather than formal taxonomy.
Activation spreads outward along interconnected pathways from an activated conceptual node to related nodes, decaying over distance and time.
Better explain semantic priming effects, semantic distance variations, typicality effects (e.g., verifying a robin is a bird faster than verifying an ostrich is a bird), and category flexibility.
Empirical Techniques for Probing Semantic Architecture:
Semantic Priming Tasks: Measure reaction time facilitation when identifying a target word preceded by a semantically related prime compared to an unrelated prime.
Category Verification Tasks: Measure response latencies when participants verify sentence truth values (e.g., "An apple is a fruit" versus "A pumpkin is a fruit").
Lexical Decision Tasks: Require subjects to judge whether a given string of letters forms a legitimate word, tracking activation spread across related lexical representations.
Multilingual Organization in Associative Models:
Investigate how knowledge across multiple spoken languages is integrated within a unified or distinct conceptual system.
Cross-linguistic priming demonstrates shared conceptual representations accessed by language-specific lemma and lexeme nodes across bilingual lexicons.
Theoretical Accounts of Categorization
Overview of Conceptual Categories:
Concepts serve as mental representations of categories, grouping distinct entities, actions, or objects based on shared properties, relationships, or functional equivalence.
Categorization enables cognitive economy, predictive inference, generalization of novel stimuli, and efficient communication.
Four primary theoretical frameworks account for category formation and member assignment: Feature Comparison Theory, Prototype Theory, Exemplar Theory, and Situated Simulation Theory.
Feature Comparison Theory:
Core Mechanism: Asserts that categories are defined by structured lists of semantic features or attributes assigned to a given category.
Component Features:
Defining Features: Essential, obligatory attributes that every legitimate category member must possess (e.g., all birds must lay eggs; all triangles must possess three closed sides).
Characteristic Features: Common, typical properties that describe members of the category but are not strictly necessary (e.g., the ability to fly or construct nests, which characterizes most birds but excludes flightless birds such as penguins and ostriches).
Critical Limitations:
Natural categories rarely possess rigid, defining criteria; clear boundaries do not exist for the vast majority of biological or artifactual categories.
Fails to effectively account for boundary fuzziness, non-binary category membership, or varying degrees of typicality, making it the most easily dismissed of the four theories.
Prototype Theory:
Core Mechanism: Posits that an individual constructs and retains a single, idealized mental representation—a "prototype"—that embodies the central tendency or weighted average of all perceived members within a category.
Categorization Process: When a novel stimulus is encountered, its perceptual and conceptual features are compared against the central prototype; category assignment depends on overall graded similarity to this prototype.
Empirical Evidence:
Typicality and Naming Latencies: Highly prototypical category exemplars are named and produced significantly faster during free-generation tasks than atypical exemplars.
Priming Facilitation: Prototypical items benefit from stronger and faster semantic priming compared to peripheral category members.
Exemplar Theory:
Core Mechanism: Proposes that categories are not represented by a single synthesized abstraction, but rather by an aggregate collection of specific, previously encountered instances stored in memory (individual exemplars).
Categorization Process: Classification of a novel item proceeds by evaluating its similarity across multiple stored specific exemplars within the category.
Theoretical Advantage: Successfully accounts for atypical instances, extreme variability, and exceptions to a prototype (e.g., recognizing an emu or penguin as a bird by referencing specific stored exemplars of flightless birds rather than a generalized flying prototype).
Situated Simulation Theory:
Core Mechanism: Emphasizes grounded and embodied cognition, arguing that categorization is not a static retrieval of pre-stored feature lists or abstract representations, but a dynamic, real-time simulation constructed within a specific environmental and behavioral context.
Context-Driven Goal States: Category representations shift flexibly depending on an individual's immediate goals, physical state, and current circumstances.
Contextual Re-categorization Examples:
A conventional wooden chair is categorized as a seated dining furnishing when positioned around a table.
The identical physical object, when located in a bedroom and covered with garments, is dynamically categorized as a temporary laundry hamper or clothing storage device based on utility and situational demands.
Schemas and Scripts: Structures and Cognitive Roles
Definitions and Structural Distinctions:
Schemata (Schemas): Broad, organized, structured knowledge packages containing generalized abstract knowledge, assumptions, and expectations regarding objects, people, roles, settings, or situations.
Examples of environmental schemas include the standard physical configurations and expected paraphernalia of medical examination rooms, restaurants, and offices.
Scripts: A specialized subtype of schema that represents stereotyped sequences of actions and predictable event progressions across time in common, real-world activities.
Classic examples include the script for dining at a restaurant (entering, waiting to be seated, reviewing the menu, ordering food, eating, receiving the check, paying, leaving), igniting an automobile with a physical key, or celebrating a birthday party (gathering, lighting candles, singing, blowing out candles, cutting cake).
Narrative story arcs follow structured canonical scripts characterized by an Introduction, Rising Action, Climax, Falling Action, and Resolution.
Function of Schemas and Scripts in Memory Encoding:
Facilitation of Coherence: Schemas provide interpretive scaffolds that allow incoming, ambiguous, or impoverished perceptual input to be rapidly comprehended, organized, and integrated into prior knowledge structures.
Context-Dependent Semantic Access:
Statements lacking activated schemas appear cryptic and incomprehensible.
Consider the sentence: "The haystack was important because the cloth ripped."
Without an appropriate interpretive frame, the proposition is incoherent; activating the schema of a parachute failure renders the sentence meaningful, logical, and readily memorized.
Encoding of Schema-Inconsistent Information:
While schema-consistent elements are encoded rapidly using default expectations, unexpected or schema-violating items attract focal, sustained attention.
In the Brewer and Treyens (1981) academic office study, participants spent significant processing resources attending to bizarre, schema-inconsistent objects (such as a skull or a wine bottle present within an academic workroom), leading to robust episodic memory for those specific salient violations.
Function of Schemas and Scripts in Memory Retrieval:
Reconstructive Nature of Retrieval: Memory retrieval is not an exact playback of recorded events, but a reconstructive process in which schemas are actively used to "fill in the gaps" of degraded or missing traces.
Bartlett's (1932) "War of the Ghosts" Investigations:
British participants were instructed to read a Native American folk story containing unfamiliar cultural narrative conventions and supernatural elements.
Upon repeated recall over successive retention intervals, participants systematically altered the story content through assimilation, rationalization, and omission to conform to their own cultural schemas.
Unfamiliar or culturally incongruous activities were systematically transformed into culturally conventional behaviors; specifically, descriptions of hunting seals were misremembered and reconstructed as fishing.
Generation of False Memories:
When an activated schema generates high default expectations for a context, individuals frequently endorse non-presented items as having occurred if they align strongly with the schema.
In list-learning paradigms (such as the Deese-Roediger-McDermott framework), presenting semantic associates such as nap, dream, bed, snore, tired, and rest causes participants to falsely recall and recognize the unpresented critical lure "sleep" with high subjective confidence.
Lexical Memory and Speech Production
Interface Between Conceptual Thought and Language Output:
Explicit declarative memories are consciously accessible, permitting internal mental conceptualization to be externalized as verbal communication.
Lexical memory governs the functional architecture and sequential processing stages that bridge abstract semantic concepts and physical spoken words.
The Three Stages of Lexical Access:
Conceptual Level:
Encodes the non-linguistic, abstract conceptual entity, intention, or idea.
Represents the underlying semantic attributes and perceptual characteristics associated with the target entity (e.g., retrieving the generalized visual, biological, and aquatic identity of a fish).
Lemma Level:
Encodes the intermediate, abstract syntactic and grammatical identity of the associated lexical item.
Specifies word meaning and grammatical specifications, such as part of speech (e.g., designating "fish" as a noun), grammatical gender, and syntactic subcategorization frame, without activating physical sound patterns.
Lexeme Level:
Encodes the concrete morphological, phonological, and articulatory form of the selected word.
Governs how the word is physically structured, pronounced, and sounded out, detailing its syllabic representation and phonetic transcription (e.g., mapping "fish" to its phonetic composition ).
Speech Errors as Diagnostic Windows into Lexical Architecture:
Naturally occurring slips of the tongue demonstrate the computational separation between syntactic selection (lemma) and phonological assembly (lexeme).
Word Exchange Errors:
Involve the unintended swapping of two entire words within an utterance (e.g., "I wrote a mother to my letter" instead of "I wrote a letter to my mother").
The exchanged elements almost uniformly belong to the identical syntactic grammatical class (nouns swap with nouns, verbs with verbs).
Locus of Error: Occurs at the intermediate Lemma stage, where syntactic positioning and grammatical planning take place prior to phonological realization.
Sound Substitution Errors (Phonological Slips):
Involve the accidental substitution, anticipation, perseveration, or transposition of phonemic units or word forms possessing acoustic similarity (e.g., malapropisms, swapping phonemes between adjacent words).
Locus of Error: Occurs downstream at the Lexeme stage, where phonological forms, phonemic sequencing, and motor plans are organized for vocal execution.
Analytical Frameworks and Semantic Modeling
Methodological Modeling Requirements:
Mapping semantic memory architecture requires generating concrete conceptual networks that delineate node relationships, associative connections, and property inheritance.
Strict operational definitions and directives define evaluations of semantic networks:
Name: Designate precise, explicit terminology for nodes, properties, or error types.
Describe: Characterize the functional mechanisms, properties, and developmental boundaries of schemas, scripts, or retrieval phases.
Identify: Locate precise failure points, processing stages, or categorical boundaries within empirical case studies and linguistic slips.
Draw: Visually construct accurate directed networks, hierarchical trees, or processing flowcharts modeling semantic distances, activation pathways, and categorical hierarchies.