Semantic Development in Children With Language Impairments

The Semantic Network Model

In the semantic network model, Allan Collins posits that long-term semantic knowledge is organized into networks of interconnected conceptual nodes. For instance, the concept of school bus is linked to related concepts such as fire truck and yellow, which are linked to associated concepts such as hose and banana. A key processing property of semantic networks is spreading activation, which describes how semantic information is retrieved during online processing. The retrieval process is initiated by thinking about a concept, which sends spreading activation to related concepts. The activation starts off strong but abates in strength as it travels away from the core of the network. This change in activation strength is referred to as the rippling effect.

Semantic neighborhood: Theories diverge on how they define semantic relatedness. Lori Buchanan divided semantic models into two broad classes: feature-based models and association- or co-occurrence-based models. Feature-based models suggest that concepts that share many features are semantic neighbors (dog and cat are both furry pets) and that spreading activation between neighbors is a function of the number of shared features. The second type of models defines semantic neighbors based on human association responses or through calculations of global co-occurrence using a large corpus of natural language. Semantic neighborhood can be measured through the number of semantic associates to a target word (i.e., semantic size) and the mean distance from the target word to its 10 nearest neighbors (i.e., semantic distance). A large neighborhood is associated with a high semantic size value and a low semantic distance value. Both types of models depict important semantic relationships, but the latter models are more amenable to precise definition.

Experimental methods: To directly and explicitly evaluate children's semantic representations, researchers have used drawing (e.g., Draw a picture of a strawberry), definition (e.g., What is a strawberry? Tell me all you know about it.), picture naming, and repeated word-association tasks (e.g., Tell me three words that go with the word strawberry). To assess semantic category knowledge, researchers have used category inclusion (e.g., Are fish animals?), semantic fluency (e.g., Tell me all the animals you can think of), and contrast question tasks (e.g., While pointing to the picture of a rose, ask, Is this a dandelion? A tree? An animal?). To measure the more automatic and implicit aspects of semantic processing, studies have employed the [Page 535]priming technique and false memory paradigms. Semantic priming utilizes prime-target pairs (e.g., carrot-rabbit) to examine if and how the presentation of the prime influences the accuracy and speed of recognizing or producing the target. False memory paradigms typically take the form of a list recall task and investigate the incidences of false recall of a nonpresented critical word (e.g., sleep) that relates semantically to the presented words (e.g., bed, doze, pillow, and dream). Finally, in word learning studies, researchers have manipulated the number and type of semantic cues and semantic-neighborhood properties of the targets to study how these factors affect word learning.

...