8.1 Semantic Networks and Schemas

Semantic Networks and Schemas

Semantic networks and schemas are two models cognitive psychologists use to explain how knowledge is organized in long-term memory. Understanding these models matters because they explain not just how we store information, but how we retrieve it, make inferences, and sometimes make predictable errors.

Semantic Networks and Knowledge Representation

Semantic Networks in Knowledge Representation

A semantic network is a model of knowledge representation that uses nodes (concepts) connected by links (relationships). Think of it as a web where each idea is tied to related ideas through labeled connections.

The links between nodes come in different types:

• Is-a links (hierarchical): "A robin is a bird." These place concepts into categories.
• Has-a links (compositional): "A bird has wings." These describe properties or parts.
• Can-do links (functional): "A bird can fly." These describe what something does.

This structure lets you store information efficiently. You don't need to separately memorize that a robin has feathers, a sparrow has feathers, and a canary has feathers. Instead, "has feathers" is stored at the bird node, and every node linked below it inherits that property. This is called cognitive economy, and it's a key advantage of hierarchical network models like Collins and Quillian's original proposal.

Semantic networks also support categorization and inference. If you learn that "a penguin is a bird," you can immediately infer that penguins have feathers, even if nobody told you that directly. Common real-world examples include family tree diagrams and concept maps used in education.

Structure of Schemas

A schema is a cognitive framework, a mental template that organizes your knowledge about a category, object, event, or situation. Where semantic networks focus on relationships between individual concepts, schemas bundle together clusters of expectations.

Schemas are built from slots (attributes) and fillers (the values that fill those attributes). For a "restaurant schema," the slots might include ordering method, who serves food, and how you pay. The default fillers would be from a menu, a waiter, and at the end of the meal. When you walk into a new restaurant, your schema fills in these expectations automatically.

There are several types of schemas:

• Event schemas (scripts): Sequences of expected actions, like what happens at a doctor's appointment
• Role schemas: Expectations about how people in certain roles behave (e.g., what a teacher does)
• Object schemas: Knowledge about what objects look like and how they function
• Person schemas (stereotypes): Generalized beliefs about groups of people

Schemas are also organized hierarchically. A superordinate schema for "eating establishment" sits above basic-level schemas like "restaurant" or "cafeteria," which in turn sit above subordinate schemas like "fast-food restaurant" or "fine dining."

Information Processing with Semantic Networks and Schemas

Both models explain how stored knowledge actively shapes ongoing cognition.

Spreading activation is the key processing mechanism in semantic networks. When one node is activated (say, "doctor"), activation spreads along the links to related nodes ("nurse," "hospital," "stethoscope"). This explains why thinking about one concept primes related concepts, making them faster to recognize or recall. Meyer and Schvaneveldt's classic lexical decision experiments demonstrated this: people identify "nurse" as a word faster after seeing "doctor" than after seeing an unrelated word.

Schemas shape cognition in a different but complementary way:

• Attention: Schemas direct your focus toward schema-relevant information. In a classroom, you'll notice the instructor and the whiteboard before you notice the ceiling tiles.
• Encoding: Information consistent with an active schema is easier to encode into memory.
• Retrieval: When recalling an event, schemas fill in gaps with default expectations, which can be helpful but also introduces distortion.
• Inference and prediction: Both networks and schemas let you go beyond the information given. You can understand a news article about an unfamiliar event because your existing knowledge structures supply context.

Together, these mechanisms reduce cognitive load. Instead of processing every situation from scratch, you rely on structured prior knowledge to interpret new input quickly.

Limitations of Semantic Networks and Schemas

Both models have real weaknesses, and these show up frequently on exams.

Limitations of semantic networks:

• They handle simple hierarchical relationships well but struggle with complex or context-dependent relationships. The meaning of "bat" changes depending on whether you're talking about baseball or zoology, and basic network models don't capture that easily.
• There's no standardized format for how networks should be structured, which makes them hard to apply consistently across different domains.
• They represent declarative knowledge (facts and relationships) much better than procedural knowledge (how to do things).

Limitations of schemas:

• Schemas can oversimplify complex situations. Your "restaurant script" might not prepare you for a very different dining custom in another culture, leading to confusion or misinterpretation of social cues.
• They promote confirmation bias: people tend to notice and remember schema-consistent information while ignoring or distorting information that doesn't fit.
• Stereotyping is essentially schema application gone wrong. Person schemas based on group membership can lead to overgeneralization and biased judgments.
• Schemas resist updating. Novel or contradictory information often gets distorted to fit the existing schema rather than prompting schema revision. Piaget called this assimilation versus accommodation, and the tendency to assimilate is strong.