Scaffolding is a term rooted in multiple research communities over decades of development. Customized, contingent support can be provided to learners to enable performances beyond what they can do alone. This dissertation seeks to examine how effectively scaffolds designed to promote articulation (written expressions of learner understanding) actually work, and if this effectiveness and/or the quality of the resulting models changes over time. It longitudinally examines the use of scaffolds designed into a dynamic modeling tool, as it is used by middle school science learners to create, test, and revise models of complex science phenomena like stream ecosystems. This dissertation also reviews the origins of the scaffolding construct, and summarizes conceptions of scaffolding from various lines of research. Scaffolding can be provided by both human and non-human agents, such as computers, which require specialized interface design to ensure maximum effectiveness.

In the study, learners created models in four curriculum units over the seventh and eighth grade school years. Additionally, this dissertation examines the nature of the discussion learners have while using these scaffolds and the frequency and types of interpersonal scaffolds employed during the creation of models. Model quality is also examined using a rubric developed through review of prior research on assessing models and concept maps. Learner pairs' model creation sessions on a computer are captured with screen video and learner audio, and then distilled to transcripts for subsequent coding and analysis, supported by qualitative analysis software.

Articulation scaffolds were found to succeed in promoting articulations and the quality of those articulations improved over time. Learner dialog associated with these written articulations is of reasonable quality but did not improve over time. Quality of model artifacts did improve over time. The overall use of scaffolding by each learner pair was contrasted with that pairs model quality, but no relationship was found.

Software design and classroom implementation implications of these findings are discussed. The frequency of interpersonal scaffolding provided by teachers highlights the need to consider scaffolding holistically and synergistically, with design decisions for software tools made in light of careful analysis as to what human and non-human agents can and should each provide.