Studies investigating the possible effects of age at immigration (a proxy for age at onset of second language learning) on second language acquisition among immigrants often explicitly take the effect of length of residence in the destination country (a measure of exposure to opportunities to learn the second language) into account. A third variable, age at testing, recently has caught the attention of some scholars. However, the three variables "age at immigration," "length of residence," and "age at testing" are linearly dependent. The easiest and most common response to this problem is to ignore one or two of the variables. In this article I show that this strategy can result in potentially misleading conclusions and then suggest some strategies for dealing with the problem.