Note also the definition of "aggressive" from their paper: "when the encounter included behaviors such as charging, agonistic vocalizations, or persistent following". People pointed out before that bears often charge as a test, without real intention of attacking. What JohnN asked was how many of these bears had a commitment to attack. The number is smaller than 25, and probably much smaller.


I guess the point is that we haven't seen anything yet which supports the assertion that bear spray *is* effective for "committed" attacks. The Smith et al. study that you cited does not show it since for all we know, none of these 25 attacks was "committed" (and the reason for that is that their definition of aggressive behavior does not necessarily indicate a commitment to attack).


I think what's interesting is not so much the fraction of all attacks they presented (as long as they were chosen at random), but rather the absolute number of attacks they analyze. The reason is that with so few attacks it is difficult to determine significance of the results. For example, they claim that for brown bears, the spray stopped the attack in 12 out of 14 cases. With so few cases, the 95% confidence interval for success is from 60% to 96%. I.e. it wouldn't be surprising if with more data their claimed effectiveness would drop to 60%.