While prior work emphasizes that organizations can use language to engage attention, the effects of language are often confounded by contextual factors and the topic and timing of communication. To isolate the effects of language on attention, we exploit a novel quasi-experiment where pairs of tweets are posted by the same organization on the same topic within a short amount of time. We empirically analyze how using vocabularies with different amounts of positive and negative sentiments between a pair of tweets attracts different types of attention garnered.