Thanks so much for the reply -- although as a point of clarification, these papers were written for classes rather than as publications, which seems like it changes things significantly (I'm sorry -- this is what happens when I internet on far too little sleep.)
But for what it's worth, I'm interested in psycholinguistics, which is actually part of why the problem arises. I definitely want to choose a paper that shows I can design an experiment, and ideally a paper that shows I can do that in the context of language and the brain. However, the other term-long research papers I've done aren't focused on psycholinguistics, but instead fall in more purely linguistic areas that aren't obviously relevant to what I'm interested in studying. So I'm stuck between the relevant-but-technically-shared paper where I did the majority of the work and we turned in separate papers, and less-relevant papers that are entirely mine. Since the programs that want writing samples are primarily linguistics, it's not necessarily inappropriate to do something more linguistics-related, I just think it makes my application less cogent.
Also, the papers I think make the strongest arguments for my ability to do research weren't written for classes I took from my letter writers. I tended to pick people I worked for rather than people I had only taken classes from, which were the stronger letters but sort of leaves my writing sample to stand on its own aside from any explanation I want to add myself.