Background: Discourse analysis is an important component of aphasia assessment because it can provide an insight into functional communication abilities. However, there are many unknowns regarding the levels of discourse breakdowns that occur across aphasia types. The purpose of the current study is to explore the possible differences in discourse-level communication in persons with fluent and non-fluent aphasia during picture description. Aims: To examine if persons with fluent and non-fluent aphasia differ on utterance-level discourse measures when evaluating informative content and global coherence. Additionally, to evaluate and compare the types of global coherence violations made by each group. Methods & Procedures: Data from 31 people with aphasia was collected from AphasiaBank, which included 13 people with fluent aphasia and 18 people with non-fluent aphasia. Discourse samples from three picture descriptions were analysed. Discourse outcomes included utterances with new information (UNIs--relevant utterances containing new information) and global coherence (the extent to which each utterance maintained the overall discourse theme). Additionally, seven types of errors were identified to explore the nature of breakdowns in global coherence. Outcomes & Results: People with fluent aphasia produced significantly higher proportions of UNIs and had significantly higher average global coherence ratings than those with non-fluent aphasia. Differences in global coherence violations were identified with people with fluent aphasia producing more non-specific, incorrect and off-topic utterances and people with non-fluent aphasia producing more incomplete utterances. One of the most common global coherence error types in both groups was commentary. Conclusions & Implications: Although people with fluent aphasia produced more types of global coherence errors, including incorrect, non-specific and off-topic utterances, the group was still rated significantly higher on utterance relevance and topic maintenance, indicating that the ability to produce a complete utterance plays an important role in some aspects of discourse production. Additionally, these findings provide an insight into potential targets for intervention.