The study had some serious confounding issues which were not dealt with, though. For example, determining whether the bystanders were actually friends walking together with the target of the incident.

Also, the bystander effect deals with the probability of any individual helping out. It's still possible that there are two forces at play here: the declining probability of any individual helping vs the increasing probability that someone in an ever-larger group will still help despite the individual effect. It might be the case that there is a critical value N of bystanders where the second effect overpowers the first one.