Can you believe your ears?
What do you think of when you think of a frightening growl? Chances are, it's probably a deep growl from a large dog. Most mammals using low-pitched signals in order to appear larger and more threatening, and recognize the sounds the same way. It appears that this tacit understanding also leads to mistakes.
People are generally pretty good at estimating a dog's size based on a growl, but it turns out, they are not so good at gauging aggression from the same signal.
Researchers took a set of recorded growls and digitally manipulated both the frequency (pitch) and formant frequency dispersion (resonance) of the growls, using a technique to that has been used in a few different studies that tested both how well people and how well dogs were able to estimate based on the sounds.
The growls were from 20 different dogs that had been coaxed into growling by an approaching stranger. They were then played for 30 different subjects, who were asked to rate them on a seven point scale for aggression. After the test they were asked to rate their own experience with dogs.
I really wish I could just reprint the graphs included in the research. It really boils down to this: regardless of the subject's experience with dogs, they always rated growls that were manipulated with lower pitches as more aggressive than those with higher. If the unchanged growl started with a low resonant frequency, the difference is even more striking. Since resonance is usually how people accurately estimate size, this implies that people are generally more willing to correlate size with aggression.
Manipulating the resonant frequency had similar results, and at the extreme boundaries of the resynthesizing, the resonant frequency still seemed to override the pitch. Again, perceived size seems to be the deciding factor in interpreting a growl:
Across both experiments [manipulating pitch and then resonance - EG], the stimuli perceived as being most aggressive were thus those at the level of Df resynthesis that simulated the largest dogs (mean aggressiveness rating: 6.32) [ that's out of 7! - EG] and the stimuli perceived as being least aggressive were those at the level of Df resynthesis simulating the smallest dogs (2.59). In contrast, the aggressiveness ratings for the stimuli with the lowest and highest F0 were more conservative (mean ratings of 5.86 and 2.89, respectively), providing strong support for the interpretation that participants in our experiment relied on the size-related acoustic variation in Df to infer caller attributes other than, or in addition to, body size.
It's possible that when we interpret aggressive intent what we are really doing is evaluating the potential threat or danger. An 80 pound-plus dog is a lot more intimidating than a 10 pound-minus dog. Of course, this can lead to dismissing a potentially dangerous threat: multiple bites from any sized dog can be pretty devastating.
Similar to last week's post, the takeaway here is that our interactions with dogs are heavily influenced by things that we are not always completely aware of. However dogs, as you might expect, are pretty good at interpreting growls. Over on my blog I discuss how dogs can infer a lot from just hearing a growl.
