In a simple transcription task in which sections of Java programcode are copied by freehand writing, it is demonstrated thatchunk related temporal signals are sufficiently robust to permitthe measurement of programming competence. An experimentwith 24 participants revealed that the number of views of thestimulus per trial and the duration of writing per stimulus vieware both strongly correlated with independent measures of Javacompetence.

Studies have shown that temporal chunk measures in transcrip-tion tasks can be used to assess competence in various domains.However, in other tasks, chunking strategies and thus perfor-mance differences can be highly variable across participants.If such individual differences are also large in transcriptiontasks this would undermine the use of chunk-based competencemeasures. Using four stimuli with fixed spatial structures thisexperiment demonstrates that there is good consistency inchunking strategy across 52 participants in two types of tran-scription tasks. The experiment spans 16,000 data points.