Questions: Semantic Networks and Conceptual Organization

Question 1

A participant in a lexical decision experiment sees the word 'doctor' and then must decide whether 'nurse' is a real word. They respond faster than if they had seen an unrelated prime like 'table.' According to spreading activation theory, why?

• A. Seeing 'doctor' triggers a conscious search for related medical terms, which speeds up recognition
• B. Activating the 'doctor' node spreads activation automatically to nearby nodes like 'nurse,' pre-activating it so less activation is needed to reach recognition threshold
• C. The visual similarity between 'doctor' and 'nurse' as words reduces perceptual processing demands
• D. The participant uses prior knowledge to predict what word will come next and prepares accordingly

Answer: (B)

Spreading activation is automatic and parallel — it does not require conscious strategy or prediction. When 'doctor' is processed, its node in semantic memory is activated, and activation radiates along associative and semantic edges to neighboring nodes, including 'nurse.' When 'nurse' then appears, it is already partially activated, meaning less additional activation is required to reach the recognition threshold. The degree of priming tracks semantic proximity in the network, which is why 'doctor' facilitates 'nurse' more than it facilitates a distant word like 'bread.'

Question 2

The Collins and Quillian hierarchical network model predicts that verifying 'a canary has skin' should take longer than verifying 'a canary can sing.' What is the reasoning?

• A. Skin is a less salient feature than singing, so it is harder to recall
• B. 'Has skin' is a property stored at the animal level, requiring traversal through canary → bird → animal, while 'can sing' is stored at the canary level
• C. Negative facts are always harder to verify than positive ones in semantic memory
• D. The word 'skin' activates many competing nodes, slowing down verification

Answer: (B)

Collins and Quillian's key economy principle is that shared category properties are stored once at the category level, not redundantly at each member. 'Has skin' is true of all animals, so it is stored at the ANIMAL node. Verifying it for 'canary' requires traversing the chain: canary → bird → animal. 'Can sing,' by contrast, is distinctive to canaries and is stored directly at the CANARY node — a single step. The model predicts, and reaction-time data confirm, that the number of hierarchy levels traversed is the key predictor of verification speed.

Question 3

According to spreading activation models, a concept mainly becomes active in memory when a person consciously directs attention to it.

• T. True
• F. False

Answer: False

Spreading activation is automatic and occurs without conscious direction — it is a passive consequence of processing a related concept. This automatic nature is what gives semantic networks their explanatory power for priming effects: participants are not strategically searching for related words when they respond faster to 'nurse' after 'doctor.' The activation spreads preconsciously and pre-activates neighboring nodes, affecting processing before awareness. This distinction between automatic and controlled processes is central to cognitive psychology.

Question 4

The typicality effect — the finding that 'a robin is a bird' is verified faster than 'a penguin is a bird' — poses a challenge that cannot be explained by simple hierarchical network models.

• T. True
• F. False

Answer: True

In a strictly hierarchical IS-A model, robins and penguins are both stored as members of BIRD at the same level. The traversal cost is identical for both, so verification time should be the same. But it is not — robins, as highly typical birds, are verified faster. This typicality effect requires a different model, such as one where category membership is graded by feature similarity to a prototype, or where frequently activated connections (robin-bird) are stronger than less typical ones (penguin-bird). The typicality effect was one of the key findings that pushed researchers toward prototype and connectionist models.

Question 5

Why does the Collins and Quillian model predict that inferring a higher-level property (e.g., 'has DNA') from a concept should take longer than retrieving a property stored directly at that concept's node?

Model answer: In the hierarchical model, properties shared across a category are stored once at the category level to avoid redundancy. Inferring a higher-level property requires traversing multiple IS-A links — from the specific concept up through intermediate categories to the level where the property is stored. Each link traversal takes time. A property stored directly at the concept node (e.g., 'canary: can sing') requires only one look-up, while one stored higher (e.g., 'animal: has DNA') requires traversing canary → bird → animal — more links, more time.

This is the cognitive cost of hierarchy: efficient storage (no redundancy) comes at the price of inference time (traversal). The empirical prediction — that verification time increases with hierarchy depth — was confirmed by early reaction-time studies and established a key principle that cognitive architectures must balance storage efficiency against retrieval speed. Later models challenged this simple picture (the typicality effect, fan effect), but the core logic of traversal cost remains influential.