digit. Many used iteration; to our surprise, most iteration involved post-test loops. In 2008, 13 we described our investigation of students’ commonsense knowledge of debugging, giving beginners one of four questions designed to elicit their knowledge of debugging strategies. The questions asked them to describe the advice they would give people whose lights did not turn on when they flipped the switch; how they would locate the moment things go wrong in the children’s game “telephone”; how they would find a Starbucks if they were in a strange city where they did not speak the language; or an experience of their own involving troubleshooting.

In general, we found beginners had less commonsense knowledge of debugging than of sorting, and some of their preexisting knowledge did not serve them well. For example, real-world fixes are often easy to undo, unlike programming changes. Likewise, the real world is nondeterministic in ways CS1 programs generally are not; for example, if your car doesn’t start, should you immediately turn the key a second time?

A substantial body of work in computing and in other scientific disciplines, including physics and mathematics, involves misconceptions, or incorrect concept understandings, that must be replaced with correct ones. Clancy6 provided a survey of this work in CS, and the National Academy’s Committee on Undergraduate Science Education7 gave a more general overview. Smith et al. 15 challenged this view in the context of math and science education, arguing that misconceptions are limited mental models that can be built onto, ultimately providing a correct understanding of the concept being studied.

Most relevant to the commonsense project discussed here is Ben-David Kolikant’s 2001 work examining student preconceptions about concurrency. 2 She collected data from about 135 Israeli 12th-grade students in six classes at three different schools. The students previously studied computing and at the time of the study were taking an advanced class in concurrent and distributed computing.

At the beginning of the course she
gave them a critical-section prob-
lem in which multiple agents share a
common resource—two ticket offices
selling tickets for the same movie.
As shown in the figure, the question
was posed in a detailed pseudocode
format. It expects a relatively sophis-
ticated answer, including a hardware
system specification for supporting
multiple machines servicing sales re-
quests, pseudocode for the solution,
and an explanation of the answer and
how it avoids duplicate ticket sales.
Student answers were graded as part
of the course.

methodology

Here, we describe our process of data collection and analysis:

Data collection. In the first week of the first semester of the 2006–2007 school year, in a CS1 class, students were randomly assigned one of two tasks: the concurrency task explored here completed by 66 students from five different institutions; the other was unrelated to the study. The participating students were all beginners between the ages of 18 and 20. All but eight completed the questions online (outside of class) by typing English answers into a text box. Eight students (all at the same institution) completed the questions on paper in a laboratory setting. All were given credit for completing the assignment, though solution quality was not evaluated for credit. (The participant/subject iden-