Info

Note: The total number of clusters was 1,436 for concrete concepts, 1,539 for intermediate concepts, and 1,346 for abstract concepts. " Averages are weighted by the number of observations in each concept type x cluster type cell of the design.

Note: The total number of clusters was 1,436 for concrete concepts, 1,539 for intermediate concepts, and 1,346 for abstract concepts. " Averages are weighted by the number of observations in each concept type x cluster type cell of the design.

Because one participant's data were lost prior to Analysis 2, only 20 participants were included, 10 with situations and 10 without. For these 20 participants, the average proportion of properties produced in each of the 14 coding categories was established across the 3 concepts at each level of concept abstractness. These proportions were subjected to an arcsin transformation that normalized variance, and were then submitted to a mixed ANOVA having context, concept abstractness, and coding category as factors. Unless noted otherwise, all reported tests were significant at the .05 level. Because all tests assessed a priori predictions, post hoc corrections were not employed. The large size of most F and t values further reduces the probability of a Type I error. MSEs are reported in arcsin units.

As Table 7.2 illustrates, the results from Analysis 2 converge on the same basic conclusions as Analysis 1. On the one hand, the distributions of cluster types are similar across concrete, intermediate, and abstract concepts, indicating that similar types of situation information underlie all three concept types (Hypothesis 1). On the other hand, differential emphasis exists on the different types of situational information for concrete vs. abstract concepts (Hypothesis 2). There is also evidence for greater complexity in abstract concepts (Hypothesis 3).

In the ANOVA, cluster type exhibited a main effect, indicating that some cluster types were more likely than others (F( 13,234) = 53.96, MSE = .007 arcsin). Some of the more common cluster types included agentive events (.19), objects (.14), persons (.11), and evaluations/affects (.11). The importance of a given cluster type was often similar for the three concept types. For example, agentive events and evaluations/affects were important for all three, whereas times, non-agentive events, and goals were relatively unimportant for all three. Clearly, though, important differences occurred, as indicated by a significant interaction between concept type and cluster type (F(26,234) = 16.84, MSE = .007 arcsin). As will be seen, the specific differences underlying this interaction are consistent with Hypotheses 2 and 3.

Cluster Content: Dominant Cluster Types for Concrete Concepts. First we consider cluster types that tended to be more important for concrete concepts than for abstract ones (intermediate concepts will be included as relevant). As Table 7.2 illustrates, object clusters were more important for concrete and intermediate concepts than for abstract concepts (concrete vs. abstract, F(1,234) = 11.57, MSE = .007 arcsin; intermediate vs. abstract, F(1,234) = 48.29, MSE = .007 arcsin). This pattern supports the prediction that concrete and intermediate concepts both focus on physical objects in situations. Interestingly, more object clusters occurred for intermediate than for concrete objects (F(1,234) = 11.57, MSE = .007 arc-sin). This contrasts with the finding from Analysis 1 that entity properties were produced equally often for both concept types. The analysis of cluster length, however, will show that object clusters were longer for concrete concepts than for intermediate ones (Table 7.3). This pattern suggests the following conclusion. For concrete concepts, a single object tends to be salient (e.g., SOFAS), such that participants describe it at length. For intermediate concepts, a configuration of objects is often salient, such that participants describe each of them briefly (e.g., TO COOK refers to a cook, food, a stove, utensils, etc.). As a result, more clusters occur for the intermediate concepts, albeit shorter.

Locations were also more important for concrete and intermediate concepts than for abstract ones (concrete vs. abstract, F(i,234) = 14.29, MSE = .007 arcsin; intermediate vs. abstract, F(i,234) = 9.14, MSE = .007 arcsin). Concrete and intermediate concepts did not differ. This pattern suggests that people often think of physical objects in particular physical locations. In contrast, abstract concepts appear less tied to particular settings, and more tied to particular types of events, as we will see.

Finally, characteristic behaviors were important only for the concrete concepts (concrete vs. abstract, F( 1,234) = 5.14, MSE = .007 arcsin; concrete vs. intermediate, F(i,234) = 5.14, MSE = .007 arcsin). When describing these objects, participants often discussed the behaviors that these entities typically exhibit (e.g., for BIRDS, they fly, nest, etc.).

In summary, this pattern is consistent with Hypothesis 2. Although concrete concepts are associated with much of the same situational information as abstract concepts, concrete concepts focus more on the physical aspects of situations, including objects, locations, and typical behaviors. Intermediate concepts also tend to focus attention on objects and locations.

Cluster Content: Dominant Cluster Types for Abstract Concepts. We next consider cluster types that tended to be more important for abstract concepts than for concrete ones. As Table 7.2 illustrates, person clusters were more important for abstract concepts than for concrete and intermediate concepts (abstract vs. concrete, F(i,234) = 51.57, MSE = .007 arcsin; abstract vs. intermediate, F(i,234) = 41.29, MSE = .007 arcsin). This suggests that abstract concepts may often have a more social character than concrete concepts, drawing attention to the properties of people and the relations between them. Further evidence for this conclusion is the trend towards significance for communicative events. As Table 7.2 illustrates, descriptions of communication tended to be more likely for abstract concepts than for concrete and intermediate ones (abstract vs. concrete, F( 1,234) = 2.29, MSE = .007 arcsin, p < .25; abstract vs. intermediate, F( 1,234) = 3.57, MSE = .007 arcsin, p < . 10). As Table 7.2 further illustrates, social institutions were mentioned most often for abstract concepts (abstract vs. concrete, F( 1,234) = 5.14, MSE = .007 arcsin; abstract vs. intermediate, F( 1,234) = 3.57, MSE = .007 arcsin, p < .10). Together, these three findings suggest that social table 7.3. Average Cluster Length and Hierarchical Level for Cluster Types Across Concept Types from Analysis 2

Cluster Type

Space-Time Entities Event Introspection

Contingency/

Concept Taxonomic Characteristic Non- Communi- Evaluation/ Complex Social

Type Category Location Time Object Person Behavior Agent Agentive cation Goal Affect Belief Relation Institution

Average length

Concrete 2.59

0 0

Post a comment