Richard J. Shavelson

 Stanford University, Stanford, California 94305-3096



Concept maps are network representations of important aspects of the structure of a student's science knowledge.  The nodes on a map are terms that represent concepts, the lines represent relationships between the nodes, and the labels on the lines explain how two terms (concepts) are related.  Concept maps tap an aspect of declarative ("knowing that") knowledge that goes beyond the amount of knowledge students have acquired to represent how students structure this knowledge.


We know from research on expertise in science that experts have rich knowledge structures with many connections while novices have lean knowledge structures with few connections.  Moreover, the propositions--the triple of a concept, a labeled line, and another concept--in an expert's map tend to be accurate and precise while propositions in a novice's map tend to be superficial and sometimes inaccurate scientifically.


We also have ample evidence that concept maps are effective learning tools.  In randomized experiments, as well as in less-rigorous designs, students who study and represent their learning in concept maps perform, on average, higher than students who study without using concept maps.  One possible explanation is that the effect is caused by the "mental effort" students spend in  linking concepts as they learn new material, and by the feedback the maps give to them as they progress in their learning by adding new links and improving the labels on other links.


Less research has been carried out on the use of concept maps as assessments of learning.  Nevertheless, our recent research indicates that concept maps can be scored reliably.  Moreover, these maps provide information about achievement that overlaps with, but also differs in expected ways from more traditional measures of achievement in science (e.g., short answer, essay and multiple-choice questions).  In addition, when student know that they will be responsible not only for memorizing facts and concepts, but for linking concepts together, they approach learning more broadly.