Hooray!!!!!!! I found something. This is just a brief extract from from a site titled Investigative Interviewing Research Laboratory, url is http://iilab.utep.edu/eid.htm .

Cross-Racial Identification

Faces of one's own race are better remembered when compared with faces of another, less familiar race. This phenomenon, referred to as the "cross-race effect" or "own-race bias," has been demonstrated across a variety of memory tasks (e.g., recognition, identification, forced choice, etc.), in both adults and children, and across a variety of ethnic groups (e.g., White, Black, Hispanic, Asian, etc.). Our research has focused on explicating the cognitive or social psychological mechanisms that might underlie the phenomenon. More specifically, we have been examining the role of perceptual learning and encoding-based processes that might substantiate superior memory for own-race faces.

This is another extract from what I'm pretty sure is A German University -but as the name is in German I can't be positive. Thankfully they had an English translation underneath.

In research on the perception and recognition of human faces, the
Own-Race Bias has been the subject of numerous studies and is known as
a robust phenomenon (Meissner & Brigham, 2001). It depicts the worse
recognition performance with faces from persons of other ethnic groups
(out-group faces) compared with faces of the own ethnic group (in-group
faces). Despite some smaller gradual differences in the extent of the
Own-Race Bias (Sporer, 1992), this effect occurs with most ethnic
groups (see Meissner & Brigham, 2001), whereas Blacks and Whites are
the ethnic groups whose recognition performance has been tested in most
studies. It was the aim of this dissertation, to test Sporer's (2001)
In-Group/Out-Group Model. This model is based on earlier theoretical
explanations of the Own-Race Bias, and integrates these theories to
build a basic structure for a more complete explanation of this effect.
A central assumption of the model is the emergence of different kinds
of encoding strategies depending on the perception of a person as being
a member of the own ethnic group or another ethnic group, respectively.
The model predicts that once a face is perceived as an in-group face
it will be processed holistically, whereas out-group faces will be
processed feature-based, leading to worse recognition performance with
these faces.
...
Experiment 2 investigated the recognition performance with faces whose
outer features (eg. hair, ears) were deleted, compared with 'normal'
faces. Participants were n = 64 Turks and n = 64 Germans. By the
deletion of outer features, feature-based encoding was supposed to be
more difficult, thus leading to worse recognition performance with
out-group faces compared to in-group faces. Additionally, the interval
between encoding and test varied between immediate and three weeks.
Test stimuli were faces from African-Americans, Caucasian-Americans,
Turks and Germans. A worse recognition performance for faces without
outer features was found, independently of ethnic membership after the
delay of three weeks. When the recognition performance was tested
immediately, a Own Race effect was found for faces without outer
features, indicating feature-based encoding of out-group faces, that
disappears after longer intervals between encoding and recognition.
Experiment 3 investigated the Own-Race Bias by testing two German
groups who were supposed to have different amounts of contact with
other ethnic groups. The groups were n = 32 police officers working at
the international airport in Frankfurt/Main (high contact) and n = 32
students at the University of Giessen (low contact). Test stimuli were
faces from African-Americans, Hispanics, Turks and Germans. Both
groups showed equal recognition performance when analysing overall
performance, however, police officers performed better with black faces
than did students, whereas students showed better performance with
German faces than police officers. In a second experiment
(Delayed-Matching task) with the same participants, the view of the
test face (frontal vs. half-profile) was varied. While performance in
the frontal-view condition was almost perfect for all of the four
ethnic groups, the performance decreased with increasing social
distance from the participants (i.e. performance was worse for the
black faces, followed by the Hispanic, Turkish and German faces) in the
half-profile condition. These results are interpreted as evidence for
an influence of contact on the Own-Race Bias and as evidence for
feature-based encoding of out-group faces.