Why second-language speakers sometimes, but not always, derive scalar inferences like first-language speakers: Effects of task demands

Many studies report that the computation of scalar inferences, such as the inference from 'some' to 'not all', is cognitively costly. However, a number of studies on scalar inferences in L2 suggest the opposite conclusion, reporting that, even though processing L2-input is often thought to be cognitively demanding, people are at least as likely to compute scalar inferences in L2 as L1. Here, we propose that cognitive difficulties with computing scalar inferences in L2 may be masked by the use of experimental tasks that encourage participants to take their time to process target sentences. Thus, we found that, in a self-paced pen-and-paper experiment, participants did not significantly differ in the propensity with which they computed scalar inferences in L1 and L2. However, in a computerized experiment in which sentences were flashed on screen, low-proficiency L2 speakers were significantly less likely to compute scalar inferences in L2 than L1. These results underscore the importance of analyzing task demands in research on pragmatics and multilingualism.