Meaning and Representation

Focus on lexical meaning (sentence/compositional meaning later)

Note: Words not same as meaning

Memory


Language use depends on memory

word meanings and rules for combining them must be stored (represented) in some manner.


Long term memory (i.e., permanent)

Procedural knowledge/memory - how to perform actions (bike riding, car driving, cooking, constructing words, sentences, conversations)


Possible symbolic representation: If x....Then Y

Automatic (not controlled) process

Implicit knowledge; difficult to access and verbalize

e.g., artificial grammars; present feedback but not rules; rules acquired but not reportable.


Declarative knowledge - representation/memory for facts and experiences

Episodic memory - personal experiences/events

Semantic memory/knowledge - facts, concepts (internal lexicon)


Lexical Meaning


Denotation: Reference and Sense (what native speakers know about words)

Reference - object (real or imagined) to which lexical item refers

Specific - my dog

General - dogs (category)

Different morphemes exist for same object (e.g., John Worthen; President of BSU); so, what determines choice? (e.g., presumed familiarity, distancing)

Sense - relationship with other words; implications

Some types (open-class words only):

Synonymy- same meaning (e.g, snake - serpent)

Antonymy - opposite meaning (e.g., hot - cold)

Hyponymy - class inclusion (e.g., dog is a hyponym of animal)

Meronymy - part-whole relations (e.g., wrist is a meronym of arm)

Troponymy - manner of (e.g., to march is a manner of walking)

Connotation - implicit meaning beyond sense and reference

e.g., loaded words; pro-choice, pro-life, Republican, Democrat



Semantic (lexical) structure/representation (Psychological Models)


Hierarchical Models (e.g., Collins & Quillian, 69;70;72)

Concepts represented as nodes in network of category (hyponymy) and property relations.

Animal - eats

- breathes

Bird - wings Fish - fins

- - swims

Canary - sings Ostrich - can't fly Shark -Bites Salmon - edible

- - - -


Assumptions:

Cognitive economy - properties stored at only one level (e.g., birds breathe stored at animal level and not bird level).

Concept verification (e.g., is an ostrich a bird?) - intersection search; greater distance -> longer Rts:

e.g., bird - animal < canary - animal (category size effect)

canary - bird < canary - animal


Problems:

Category size effect could be due to semantic association

e.g., bird-wings < bird-eats; due to semantic association rather than distance

Items at one level not equal

e.g., canary-bird < ostrich-bird, even though both at same level


Overall: conceptual knowledge > classification (hyponymy) knowledge (but category relations must be included)

Semantic Features Theory (E.E. Smith)

Words represented as bundle of semantic features

defining - necessary (e.g., bird - feathers)

characteristic - typical (e.g., bird - sings)

Semantic verification based on category overlap and involves two stages:

1. Compare defining and characteristic features (if sufficient, judge same)

2. (If necessary): compare defining features only

e.g., Ostrich; sings, flys (not); must go to stage 2; not so for more typical member

Verification speed a function of feature overlap rather than distance (e.g.,,, canary shares more features with birds than animals).

Problem

defining vs. characteristic features not clear-cut

e.g., Wittgenstein - family resemblances (e.g., games)

Prototype theory


Spreading Activation Models

Lexicon = network of relationships (not only/necessarily hierarchical)

Node distance based on categorical relations and typical features

Tim Allen

Hospital Sick Medicine Tools

Hammer Screwdriver Saw

Doctor Nurse

Nails Screws

Surgeon Psychiatrist Wood

Psychologist

Bob Newhart


Automatic Spreading activation - activation of a single node spreads out in parallel throughout network;

Much evidence via semantic priming effects (lexical decisions faster for semantically related targets) attenuated over distance

brief (problem if no decay)

Automatic (no conscious control; no attentional resources)

E.g., Neely (1977); facilitation occurs quickly; inhibition takes time:

Expect type of bird to follow:

Facilitation (immediately) Inhibition (long SOA required)

xxxx - robin xxxx - arm

bird - robin bird - arm


Nonconscious priming (Marcel)

Semantic priming beyond lexical meaning (e.g., speaker meaning)


Lexical Access


Processing of retrieving lexical information

1. Semantic priming - related concepts activated (see above)

2. Morphological structure

Debate: each word has separate entry (e.g., bar, bars, barroom,etc.)

vs.

decompositional; morphemes have separate entries and are composed or decomposed when used (e.g., bars = bar + s)


Note: decompositional more economical for memory/representation (fewer entries)

separate entry more economical for use (not necessary to compose and decompose)

Representation appears to vary over morpheme type:

decompositional for inflectional suffixes (e.g., 's')

separate entry for derivational morphemes (e.g., teacher)

semantically opaque (e.g., butterfly)

3. Variables

Frequency - high frequency recognized more quickly; not clear if due to lexical access or post-access decision processes (e.g., lexical decision = access then decision).

Phonology - tip of the tongue (TOT); implicates phonological retrieval

Grammatical class; closed-class (function words); no frequency effects; open-class; frequency effects occur. May be organized separately.

Summary: appears to be flexible organization of lexicon; semantic, phonological, grammatical, all seem possible.


4. Models of lexical access

 Must explain recognition and production of words based on orthographic and phonetic properties.

Automatic process; not open to introspection

Access models have implications for organization


Serial Search

Search lexical list sequentially, one item at a time

e.g., Foster (1976) autonomous search model:

3 access files/codes: orthographic, phonological, syntactic/semantic (no overlap in use) connected to one master list

2 stages: relevant access file activated

search master code for location

if no match, post-access check

plausible nonwords reconsidered (via serial check; clear nonwords rejected)


Frequency effects and semantic priming handled with bin storage system (most frequent on top of bin & so easier to get to) and cross-referenced associates in master list.

Parallel Access

Candidate entries activated simultaneously rather than sequentially.

Logogen model (Morton, 1969; 1979) (first parallel access model)

Access = location; Access = activation

A word's logogen consists of a set of features; if logogen threshold is crossed, word is activated If multiple items activated (thresholds crossed), item with most features that overlap input is activated.

Explains frequency effects (lowered threshold for recently activated)

Explains semantic priming (lowered threshold for related items (existing in a separate cognitive system).

Connectionist (PDP) (Rumelhart & McClelland)

Associationism - Neural nets - nodes and connections (like spreading activation models described above, but not symbolic, nodes and connections have no 'meaning').

Neural nets consist of input nodes, hidden nodes, output nodes (neuron analogy) and their connections

Connections either excitatory (activation forwarded) or inhibitory (activation stopped)

Meaning resides in pattern of connections (nodes and patterns alone are meaningless; unlike spreading activation models)

Organization is layered

E.g., visual features -> letters -> words -> semantics

Major advantage: system can learn; reconfigure based on experience. But somewhat limited beyond lexical level.

Lexical Ambiguity


words can have multiple meanings (words = meanings)

e.g., polysemous words > 1 logogen

Issue: are all word meanings accessed at comprehension?

Two views: all meanings activated vs. context restricts to appropriate meaning only.

Early study: Foss (1970) - phoneme monitoring; longer following ambiguous words suggesting all meanings activated. Effect is very brief; disappears after two syllables.

E.g., The men started to drill before they were ordered to do so.

The sergeant order the men to drill before they were ready to do so.

Much evidence for activation of multiple meanings.

Even a biasing context can fail to restrict multiple activation.

E.g., Swinney (1979) cross-modal priming study:

Hear: Rumor had it that, for years, the government building has been plagued with problems. The man was not surprised when he found several spiders, roaches, and other bugs in the corner of his room.

Lexical decision: bug, spy, sew. Facilitation for bug and spy immediately, but not after 4 syllables.


Meaning frequency and context may interact:

Duffy et al (1988) eye gaze technique

For balanced (e.g., right) words, both meanings activated when context occurs after word; for polarized (e.g., yarn) only dominant meaning is activated.

When context occurs before word and favors subordinate meaning, both meanings of a polarized term are activated.


Overall:

Multiple meanings appear to be activated in many situations

Appears to be automatic process

Ambiguities appear to be resolved during post-accesss checking.