There is a need for empirical evaluation of programming languages for unski
lled users, but it is more effective to compare specific features common to
many languages than to compare complete languages. This ran be done by dev
ising micro-languages stressing the feature of interest, together with a su
itable subject matter for the programs. To illustrate the power of this app
roach two conditional constructions are compared: a nestable construction,
like that of Algol 60, and a branch-to-label construction, as used in many
simpler languages. The former is easier for unskilled subjects. Possible re
asons for this finding are discussed.