I've discovered that there is an argument between Stubbs and Widdowson about how much detail corpus linguistics can really pick up. Even Stubbs - the corpus linguist - says that you can really only look for the language features you already suspect are there.
So .... suppose I start with the 5 lyrics that mention Cohen or Leonard by name. Analyse them in huge detail. Then look for the same patterns in the whole Cohen corpus. And find that corpus linguistics can't do it with the current state of the software. So my dissertation is formed. And there is so much work for the IT people to do in corpus linguistics. A word count and a concordance is just the tip of the iceberg.
I'm seeing S on Wednesday, so I'll see what he thinks.
I'm making bean stew for tonight's meal and, if I have time, banana bread for when more of our visitors arrive on Thursday. I've never made it before but it'll be fine.