Gygax, Oakhill and Garnham (2003) showed that, contrary to the assumption of earlier research, readers do not infer specific emotions such as "guilt" or "boredom". This paper presents evidence for the non-specificity of emotional inferences regardless of the nature of the stories. In Experiment 1 and 2, Gygax et al.'s stories were made longer. In Experiment 1 (off-line), people rated specific emotions as more likely, but in Experiment 2 (on-line), there was no difference between target sentences containing different matching emotions, although participants took longer to read sentences containing emotions mismatching the stories. In Experiment 3 and 4, the stories included a coherence break resolvable by inferring the main character's emotional state. In Experiment 3 (off-line), people rated specific emotions as more likely, but in Experiment 4 (on-line), there was, again, no difference between target sentences containing different matching emotions.