Projective tests were originally designed to measure our subconscious thoughts  - ones that were kept hidden from others and from ourselves. Famously, tests were devised that required explanations of figures seen in inkblots (the Rorschach) or in pictures involving human situations (the Thematic Apperception Test). The idea is that the moods, fears, and beliefs expressed by the person being tested are not so much about the test images, but are actually their own inner thoughts revealed. In the day, I administered many TATs (usually about 10 images) and I can tell you that it was fascinating - I felt that I was getting a deep and privileged look unto the psyche of my patients. The trouble is that the tests did not measure up under scientific scrutiny. One issue is bias due to “social desirability”, i.e. a strong tendency to respond in a way that will be viewed favourably by others.
    A number of initiatives designed to make projective testing more objective were met with varying success. In that era (50 years ago), a fellow student, Don Ogston, and I checked out one of these - the Group Personality Projective Test. This is a 90-item test, each of which contains a stick-figure depiction of a life-situation plus 5 options that might explain what is happening. Two examples are shown below. The options tap a number of personal needs that include  aggression, achievement, affiliation, distrust, happiness, dejection, and more.
     We were interested in social desirability bias, so we tested two groups of university and nursing students. One group of 52 was administered the GPPT in the normal way to reflect how each situation was interpreted. The second group (n = 26) was  asked to choose the most socially desirable alternative (i.e. the "nicest" or most socially acceptable of the five options. For each of the 90 items the perceived choices from Group 1 were compared with the social desirability ratings provided by Group 2. These values were then summed and then averaged across the 90 items.  The resultant overall correlation proved to be a strong 0.73, but it can perhaps be better understood by squaring this number to obtain the proportion of variance accounted for by social desirability. That is, 53% of the test result is due to the wish of the subjects to  be viewed favourably.  At best, then, less than half of the scoring (47%) is available to apply to the personality factors supposedly measured by the test (plus mistakes!). This should hugely weaken our confidence in this test. To some greater or lesser extent, this problem applies to all self-report personality tests, particularly those that delve into the mysterious subconscious. Be cautious.