That’s hard

Language learners are highly sensitive to statistical patterns in the input. When a target language provides the option to include or omit a grammatical form, learners have been shown to make decisions quite similar to native speakers. For example, learners opt to include or omit the complementizer that (as in I know (that) Steffi likes hot tea). This phenomenon has been explained in terms of a universal suite of cognitive mechanisms which support native and learner performance alike. Both learners and native speakers choose to include the complementizer when they are producing more complex or unexpected structures. The present study attempts to generalize these findings to another domain of “optional” grammatical markers, namely, relativizers (as in the hot tea (that) Steffi likes). I analyze all instances of optional relativizer use in a corpus of spontaneous learner speech produced by Spanish and German learners of English. Both of these languages have obligatory relativizers. A two-step generalized additive regression modeling technique (MuPDAR) that predicts learner choices based on native-speaker choices demonstrates that native speakers use greater shares of the relativizer in complex and disfluent environments, while learners show the exact opposite tendency: they prefer to drop the relativizer in complex and disfluent environments. These findings are discussed based on differences between complementizers and relativizers, and in terms of the limited universality of optional grammatical marking in learner speech.