I was initially confused about why they would divide up the data this way as well (being very puzzled why declination even came out as a bin, and that it meant

I was initially confused about why they would divide up the data this way as well (being very puzzled why declination even came out as a bin, and that it meant what I thought it meant). Their measurement of the dog's orientation (or the dog's control of it) cannot even come close to this level of precision. My suspicious side says that they saw that the overall data showed no strong correlations at all -- and then used variability in declination as a bin to get some result that shows some correlation. I have never heard of variability in declination (on this timescale) as a strong determinant of anything, much less micro sensitivity of a biological physiological phenomenon. And yes, "arc" minutes is more standard.