A Systematic Review of Research on Collaborative Learning with Concept Maps

Introduction

Concept maps represent knowledge as node-link networks that show concepts as nodes and the relationships among the concepts as arrows or links. Researchers and teachers have used concept maps as media for constructive learning activities, instructional presentations, and as shared materials for co-construction of knowledge (Cañas et al., 2003). Researchers have suggested that concept maps may also serve as an effective learning medium when implemented as interactive software (Cañas et al., 2003; Novak, 2002). An example of a concept map is shown in Figure 1. The map in Figure 1 has links that are directional and labeled with a relational term, but the maps used in educational research sometimes have non-directional, unlabeled links.

Figure 1.

Dansereau and his colleagues (e.g., Hall, Dansereau, & Skaggs, 1992; Lambiotte & Dansereau, 1992; McCagg & Dansereau, 1991) have used the term ‘knowledge map’ to refer to a type of concept map with directional links labeled by symbols such as P (part) or C (characteristic), drawn from a fixed set of relations. Figure 2 shows the same information in figure 1 displayed using a knowledge map.

Figure 2.

Research on the instructional applications of concept maps has proliferated since they were first used as educational tools in the early 1970s. Figure 3 shows the growth of research literature on concept maps over the past three decades. A period of rapid growth from 1985 to 1990 followed the publication of Novak & Gowin’s (1984) Learning How to Learn, which presented a comprehensive argument for the educational use of concept maps.

Figure 3.

When used for learning, concept maps can be (a) created by instructors and presented to students (Cliburn, 1986; Lambiotte & Dansereau, 1992), (b) constructed by students (McCagg & Dansereau, 1991), or (c) received and modified by students (Chang, Sung, & Chen, 2002). Recent reviews on the instructional use of concept maps show their growing prominence as a tool for supporting learning (Nesbit & Adesope, 2006; O’Donnell, Dansereau & Hall, 2002). In a recent meta-analysis, Nesbit and Adesope (2006) found that, in different instructional conditions, settings and experimental features, the use of concept and knowledge maps produced increased retention and transfer of knowledge when compared with control conditions where students studied with texts, outlines, lists or listened to lectures. The meta-analysis also found that studying with instructor- or researcher-generated concept maps rather than texts yielded an overall effect size of d = .4 standard deviations. The effect was higher (d = .82) when students constructed their own concept maps or modified maps constructed by teachers. In his influential book on statistical power analysis, Cohen (1988) suggested that an effect size of d = 0.2 could be regarded as “small”, an effect size d = .50 could be regarded as “medium”, and an effect size d = .8 could be regarded as “large”.