The apparent shapes of spiral galaxies in the 2-Micron All Sky Survey Large Galaxy Atlas are used to constrain the intrinsic shapes of their disks .
When the distribution of apparent axis ratios is estimated using a nonparametric kernel method , the shape distribution is inconsistent with axisymmetry at the 90 % confidence level in the B band and at the 99 % confidence level in the K _ { s } band .
If spirals are subdivided by Hubble type , the late-type spirals ( Sc and later ) are consistent with axisymmetry , while the earlier spirals are strongly inconsistent with axisymmetry .
The distribution of disk ellipticity can be fitted adequately with either a Gaussian or a lognormal distribution .
The best fits for the late spirals imply a median ellipticity of \epsilon \approx 0.07 in the B band and \epsilon \approx 0.02 in the K _ { s } band .
For the earlier spirals , the best fits imply a median ellipticity of \epsilon \approx 0.18 in the B band and \epsilon \approx 0.30 in the K _ { s } band .
The observed scatter in the Tully-Fisher relation , for both late and early spirals , is consistent with the disk ellipticity measured in the B band .
This indicates that excluding spirals of Hubble type earlier than Sc will minimize the intrinsic scatter in the Tully-Fisher relation used as a distance indicator .