This paper provides new empirical support for the sociolinguistic monitor hypothesis, replicating Labov et al.’s (2006, 2011) study with an experiment that examines perceptual reactions to alternating distributions of the /t,d/ variable in Brazilian Portuguese, in the context of a professional news broadcast on health and welfare. Two different simulated passages were constructed, containing tokens of /t,d/ before and after /i/, respectively, that were read with stops and palatals. The participants (N=304) rated each passage on a seven-point scale of professionalism. An analysis of variance reveals a main effect of context in mean rates and response time: the speaker evaluation on the professionalism scale for the regressive context stimuli differed significantly from the progressive context, but, interestingly, the response time to the palatal realization in the regressive context, the positively evaluated variant, is greater than for the progressive. It suggests that can there be a cognitive effort to process incoming changes, and that speakers pay more attention to the social context, which is learned at school. In the regressive context, the proportion of palatals is not significant, but in the progressive context it is, which partially corroborates the hypothesis that the sociolinguistic monitor is sensitive to frequency (in negative stereotype variables).