“The Inheritance of Shyness in Dogs”, Frederick C. Thorne1944 (dog genetics)⁠:

…The original experimental group [Thorne1940] consisted of 181 dogs, but 3 animals died and are not included in the final report. 82 or 46% of the 178 experimental animals showed consistent withdrawal behavior which was unmodified by incentives to taming. Of the 82 shy animals, genetic analysis revealed that 43 or 52% were descendants of a single Bassett bitch named Paula who was extremely shy and known as a bad “fear-biter.” In spite of her extremely fearful and generally undesirable social behavior, this animal was extensively crossbred with other morphological types because of her great fecundity.

n analysis of the genetic records indicates that this Bassett bitch was successively mated with a Saluki, a Dachshund, an English bulldog, and a German Shepherd. In each case the male animals were considered to be normal friendly dogs who had never shown excessive shyness or withdrawal behavior. The offspring of these various ratings are graphically tabulated in Figure 1 together with notations from the experimental records indicating whether each animal was rated as being friendly or shy. In several instances siblings in the second generation were mated, and in two instances the 12 animals were mated with unrelated animals of the same breed.

…Analysis of the genetic data concerning an experimental group of 178 dogs, used in a study of approach and withdrawal behavior, reveals that a sampling error was responsible for the abnormally high percentage of shy animals reported in the experiment.

Of 82 shy animals, 43 or 52% were descendants of an exceedingly shy Bassett hound who was known as a fear-biter. 59 descendants of this shy dog were traced and 43 or 73% were also shy unfriendly animals.

It is suggested that this excessive shyness is caused by the inheritance of a dominant characteristic and is therefore insusceptible to modification through learning and training.

[Almost certainly wrong, and actually polygenic (possibly liability-threshold). As usual in a pre-1960s American genetics paper, everything (eg. catnip response) is shoehorned into a Mendelian framework and the biometric alternative ignored completely. If it doesn’t make sense, just assume dominance/recessiveness until the tiny n fits!]