Testing dog heroism

From ANIMAL PEOPLE, September 2010:
(published October 5, 2010)
Do dogs have an innate capacity for heroism on behalf of
their people? Do dogs instinctively know how to fetch help for a
person in crisis?
Hal Herzog in Some We Love, Some We Hate, Some We Eat
acknowledges the abundance of heroic dog stories–“Just Google ‘dog
saves owner,'” he challenges–but cites a 2006 study by University
of Western Ontario psychologist Bill Roberts and dog breeder/trainer
Krista Macpherson which found that none of a dozen dogs they tested
responded at all to either a man who was faking a heart attack or a
man who was pinned to the floor by a fake falling bookshelf.

Both tests, however, omitted recognition of the subject
dogs’ senses of smell–their dominant sense. The stimulus in either
case was auditory and visual, but a dog could easily smell the
hormone level of the alleged victim, and could probably tell thereby
whether the victim was genuinely in trouble or just playing a game,
even if the subtle differences in tone between a person acting and a
person who is actually in distress did not tip the dog off to the
ruse.
Having logged details about 454 heroic dog episodes since
1993, ANIMAL PEOPLE compared the Roberts/Macpherson findings to real
life, sorting cases according to the primary sensory cues that told
the dog that there was danger and suggested a response.
Multiple primary cues were involved in about half the cases,
but in 270 cases, or 60%, the dog smelled a hazard which was not
easily seen or heard, such as smoke from a night housefire or
leaking gas, and took immediate action.
In 261 cases, 58%, a sound–most often a cry for help–was
a primary cue, and vision was a primary cue in 244 cases (54%).
However, in almost all of these cases, multiple senses were used to
confirm the first information that the dog received. Few cases
descriptions eliminated the possibility that odor was involved in the
dog’s response.