While the validity of a single test is difficult to determine without more information on how the test was constructed, the approach itself is valid and has been used in other peer reviewed studies.
Is There a "Language of the Eyes"? Evidence from Normal Adults, and Adults with Autism or Asperger Syndrome
The results of Experiment 1 strongly support both predictions. First, normal adult subjects are able to detect a range
of mental states (both basic and complex) from whole facial
expressions, showing strong agreement. This replicates our earlier
study which used paintings and drawings of whole faces (Baron-Cohen et
al., 1996), but shows this ability under tightly controlled,
standardized conditions (photographs of the same actress). Second,
whilst for basic mental states the whole face provides significantly
more information than either the mouth or the eyes, for complex mental
states the eyes (but not the mouth) provide as much information as the
full face. This may be because complex mental states are not easily
expressed just by the mouth, unlike basic ones (happy, sad, etc).
These results are consistent with Nummenmaa’s (1964) result, but
demonstrate it for a wider set of mental states. They are also
consistent with the idea that there is a language of the eyes. In
Experiment 2, we tested if the eye-region effect would replicate if
photos of a male face were used, in order to test the robustness of
Impaired Recognition of Social Emotions following Amygdala Damage
Recognition of Basic Emotions and Complex Mental States from the Eyes
In general, it is more difficult to recognize emotions from
only a small region of the face, such as the eyes, than from the whole
face. In our analysis, we controlled for this effect since correctness
scores were calculated on the basis of the distribution of
performances given by normal subjects to the stimuli. Figure 4
summarizes these data, in the same format previously shown for Figure
Most of the peer review studies I was able to find were related to fields of neural science; however, they did use the technique of showing the area around the eyes to test for emotional recognition. However, the studies also take steps to control the images that are used by ensuring the same actor or actress and also using standard references such as "Handbook of Emotions" when preparing the images to be used for the tests.
Since the website provided doesn't actually provide information on how the images were selected and there are different actors and actresses in the pictures, the best we can say is that while the approach is valid, the test is unlikely to pass peer review.