Objectives and Theoretical Frame: Previous research has established the role of proactive interference (PI)--when old knowledge interferes with acquiring new knowledge--in forgetting simple experimental materials like lists of words (Wickens, 1972), nonsense syllables (Keppel & Underwood, 1962), and word pairs (Postman & Keppel, 1977). Additionally, retrieval practice has been shown to reduce PI using similar materials (Szpunar, McDermott, & Roediger, 2008). However, research has yet to investigate the role of PI and how retrieval practice may reduce it when learning from complex argumentative texts, such as reading multiple documents. Thus, the current research examined whether PI occurs when reading multiple argumentative texts on a topic and whether retrieval practice successfully reduces PI in this setting. A secondary goal was to rule out fatigue as an alternative explanation using a PI release condition (i.e., testing whether a lack of conceptual overlap between sets of information eliminates PI-related forgetting; Wickens, 1972). Methods: A total of 108 participants (69.4% female) completed this study. Argumentative texts had three components: source, claim, and evidence. Participants in the PI condition read 24 texts on the same topic. The retrieval practice condition did as well, although they practiced retrieval of the first 16 texts. The PI release condition read 16 texts on a different topic followed by the eight target texts. The no-PI condition only read eight texts on the target topic. After a delay, memory for the last eight texts was tested using free recall of the texts and recognition tests of the evidence and sources. Results: During free recall, the no-PI condition exhibited the highest claim recall accuracy (better memory for target texts and less intrusions from previous non-target texts), which did not significantly differ from that of the retrieval practice condition. The release condition did not significantly differ from the retrieval practice condition, but it was significantly lower than the control condition. The PI condition was the lowest condition on both claim recall accuracy and recognizing which evidence statements were from the final set of eight texts. For source memory, however, the retrieval practice condition was significantly worse than all other conditions at recognizing which sources were from the target set, suggesting a trade-off: Claim and evidence memory was strengthened at the expense of source memory. Significance: The current experiment contributes uniquely to research on readers' memory for arguments presented across multiple documents. Results suggest PI decreases memory for information from multiple texts and that this forgetting is not simply due to reading fatigue. The detrimental effect of PI in this context can be reduced through retrieval practice (i.e., taking breaks to remember what was previously read), though the results stress the importance of practicing retrieval of "all" and not just a subset of information (e.g., only claims and evidence). These findings have educational implications for how readers interact with a series of texts as they try to learn from them.