qualitylab

the station

On the Effectiveness of the Test-First Approach to Programming

tier I/2005/IEEE TSE 31(1)

https://www.cs.unm.edu/~joel/cs351/paper/IEEE-Effectiveness_of_Test-First_Approach_to_Programming.pdf

Method

"The experiment was conducted with undergraduate students. While the experiment group applied a test-first strategy, the control group applied a more conventional development technique, writing tests after the implementation."

Population

"Thirty-five subjects participated in the study... we retained data from 24 subjects. Eleven of those subjects were in the Test-First (TF) group and 13 in the Test-Last (TL) group."

What it does not show

Students, one small task, 30% attrition. Critically for this site: it tests test-order in greenfield feature work, NOT the narrower practice of reproducing a reported bug in a failing test before fixing it. It is adjacent evidence about TDD generally, not direct evidence about bug-reproduction-first.

Hakan Erdogmus, Maurizio Morisio, Marco Torchiano

Test-first students wrote more tests, and writing more tests predicted higher productivity. On quality — the benefit most often claimed — the result was null: “there are no significant quality differences between the two groups”, with the test-last group’s median quality numerically higher.

Tier I: Genuine controlled experiment with assignment to test-first and test-last groups; undergraduate subjects on one small task, which limits external validity but not the design.

Cited by