If the people like Atwood who argue that some people just can't learn programming are to be taken seriously, then by the same argument and same types of evidence, there are some people who can't learn science, or math, or reading, or how to speak or write. Because what the evidence is showing is that some people have difficulties learning some topics or take longer to learn them, and in many cases have misconceptions that make it more difficult to teach the topic to them effectively. And the evidence shows that traditionally taught courses (lecture/memorization based - taught by people not trained in more effective teaching methods like active learning) often do not significantly impact conceptual change of such students.

The Bornat & Dehnadi Goats & Sheep paper is interesting in showing that we can diagnose students who have significant misconceptions or lack of knowledge needed to learn programming (see similar research on concept inventories showing the same thing in a wealth of other subject areas). Beyond that the strongest interpretation would be that students with misconceptions are less likely to succeed in traditionally taught CS courses. Not that such students can NEVER learn the topic and NEVER be a programmer. I wasn't surprised to see their research wasn't published (although it is so easy to get such research published - especially when the peer reviewers aren't trained in educational research and learning theory, either). But a line like this would never stand up to peer review by people who are even remotely familiar with research in computer science education: "change the language, change the application area, use an IDE and work on motivation. None of it works, and the double hump persists." There are decades of studies showing various strategies that DO help more students learn and better understand programming and computer science concepts. Just because Atwood and others are unfamiliar and ignorant of such research doesn't make their opinions more right.