Semantic Memory

Semantic memory consists of your general knowledge about the world. It is something you know, even though you do not remember where or when you learned it. Language and conceptual knowledge are parts of semantic memory. Semantic memory influences most of our cognitive activities such as determining locations, reading sentences, solving problems, and making decisions (Graham et al, 2000).

Categories and concepts are essential components of semantic memory. People divide the world into categories to make sense of their knowledge. A category is a class of objects that belong together. A concept refers to our mental representations of our categories. Categories have many objects included in them, for example a category of miniature dogs. Each of the objects can be called a dog. People organize the objects into categories like files using their knowledge of the world. We have a concept of what a dog is. Dogs may be different colors, shapes, and sizes yet we still know that they are dogs. People have a concept that a dog is an animal. They then make a deduction that if you have a dog you have to feed, water, love, and exercise your animal. These deductions or inferences greatly expand our knowledge. People combine similar objects into a single concept. So if you are thinking about getting a family pet you could choose between any of the categories of breed you want, German Shepard, Maltese, Doberman, or Husky. Your mental concept of a dog can be any or all of these breeds.

Theories of semantic memory include: the feature comparison model, the prototype approach, and the exemplar approach. Network models of semantic memory include: the Collins and Loftus network model, Anderson’s ACT theory, the Parallel Distributed Processing Approach (PDP). For further information please see this link: Outline of Semantic Memory Models
The feature comparison model is a concept defined as a set of features – a list. The word “bird” has a set of features such as has wings, flies, lays eggs, lives in trees, builds a nest. Sentence verification is also included in this model. People compare subject and predicate for similarity, such as “A coat is clothing.” There is a similarity between the coat and clothing because a coat is clothing.

With the prototype approach a person will make a decision if an item belongs to a category by comparing it with a prototype (the item most typical of the category). For example, when a person thinks of transportation they think of the most common forms such as a car, truck, plane, train, and bicycle. Most of the time we do not attach an item such as a wheelchair or elevator to transportation, unless it directly affects our lives. For a person who has mobility problems these examples are major forms of transportation.

The exemplar approach explains that people first learn specific examples of a concept. Then they classify each new stimulus by deciding how directly it relates to the specific examples or concepts. For example, what is the first thing you think of when someone says, “fruit?” Most people will think of a common fruit such as an apple or banana.

The Collins and Loftus network model is a netlike organization of concepts in memory with many interconnections of nodes and links. These nodes and links show that the meaning of a word is greatly affected by associations between the word and its associated concepts. The spreading activation network shows how the nodes and links are connected.

Anderson’s theory, called Adaptive Control of Thought Model (ACT), is a complex model which attempts to account for all language, learning, decision making, and human cognition. It is composed of three memory structures: declarative knowledge (knowledge about facts and things), procedural knowledge (knowledge about how to perform actions), and working memory. He proposes that a sentence can be broken down into small meaningful bits of information. For example, the sentence, “Once there was a little red hen who lived in a barnyard with her three chicks and a duck, a pig, and a cat.” can be broken into smaller sections of information. 1. Once there was a little red hen who lived in a barnyard. 2. Her three chicks lived with her. 3. A duck, a pig, and a cat lived in the barnyard.

The Parallel Distributed Processing approach (PDP) proposes that cognitive processes can be understood in terms of networks that link together neuron-like units. With this approach many operations can proceed simultaneously rather than one at a time. The PDP model can explain how the human memory can figure out missing information from a general category. People make a spontaneous generalization or a conclusion by using an assumption based on the general information. This conclusion is also called inductive thinking. Deductive thinking allows people to fill in missing information about a particular person or item in a category by making an educated guess or a default assignment. An example of PDP would be a riddle: 1. It is yellow. 2. It grows on a tree. 3. It has the shape of a quarter moon. 4. Monkeys prefer this item. Did you guess the item was a banana from my description (Matlin, 2005)?

Semantic memory storage is programmed using language. “Much of the knowledge that we store in memory is encoded using language” (Steyvers, 2006, p. 327). Three new models have been proposed to explain human memory using semantic properties: the Retrieving Effectively from Memory (REM) model, probabilistic topic models, and the Syntagmatic Paradigmatic (SP) model. The three models emphasize the task of probabilistic assumption in memory that draws on research in computer science, information retrieval, data, and computational linguistics. Probabilistic assumption is the philosophy that certainty is impossible, and therefore decisions must be based on probabilities.

Semantic memory is just one part of longterm memory. Episodic memory, the memory of those events we have experienced, requires active reprocessing of a prior event. Semantic memory does not require recall of the event to use the information. Tulving created a model of longterm memory which originally stated semantic memory and episodic memory had a hierarchical relationship. Studies done on patients with semantic dementia forced Tulving to revise his model of long term memory to include the reliance of episodic and semantic memory upon perceptual input (Graham et al, 2000).



References:

Matlin, Margaret W. (2005). Cognition Sixth Edition. Hoboken, NJ: John Wiley & Sons, Inc..

Graham, K.S., Simons, J.S., Pratt, K.H., Patterson, K., & Hodges, J.R. (2000). Insights from semantic dementia on the relationship between episodic and semantic memory. Neuropsychologia. 38, 313-324.

Steyvers, M., Griffiths, T.L., & Dennis, S. (2006). Probabilistic inference in human semantic memory. TRENDS in Cognititve Sciences. 10 No.7, 327-334.