A major distinguishing feature of projective techniques as a measure of personality testing is to be found in their assignment of a relatively unstructured task, that is, a task that permits an almost unlimited variety of possible responses. In order to allow free play to the individuals fantasy, only brief, general instructions are provided. For the same reason, the test stimuli are usually vague or ambiguous. The underlying hypothesis is that the way in which the individual perceives and interprets the test material or 'structures' the situation will reflect fundamental aspects of his or her psychological functioning. In other words, it is expected that the test materials will serve as a type of screen onto which the individual 'projects' their characteristic thought processes, needs, anxieties and conflicts. It has therefore been suggested that the only useful personality test is a projective one, for not only the responses made richer than those given in self-report inventories or other means of personality testing, but the results cannot be faked. Thus statement, however, makes several erroneous assumptions. However, despite this, projective testing still remains popular, and is even undergoing a growth spurt. As one reviewer phrased it: "There are still enthusiastic clinicians and doubting statisticians" (Adcock, 1965).

Typically, projective instruments also represent diguised testing procedures, insofar that the test takers are rarely aware of the type of psychological interpretation that will be made of their responses. Therefore, most projective techniques represent an effective means for 'breaking the ice' between clinician and client. The task itself is usually intrinsically interesting and often entertaining. It tends to divert the individuals attention away from himself or herself and thus reduces embarassment or defensiveness. The test also offers little to no threat to the individuals prestige due to the fact that there exists no 'wrong' answer. The fact that the purpose of the projective measure is disguised would also suggest that it is less susceptible to faking than are self-report inventories. Even if the individual taking the test is somewhat psychologically sophisticated and may even be familiar with the test itself, it is still unlikely that he or she can predict the intricate manner in which the responses will be scored and interpreted. As research suggests, a comparison between objective measures and projective tests identifies an inverse relationship between the face validity of a test and its susceptibility of faking (Bornstein et al, 1994). The respondent usually becomes so absorbed in the task that they drop the disguises and restraints of interpersonal communication.

However, projective tests are not completely immune to faking, there supposed immunity to faking being one of the faulty assumptions usually attributed to them. Several experiments have shown that when respondents are asked to alter their responses to create either a favourable or unfavourable impression, they can usually give significantly different answers (Masling, 1960). Considerable experimental evidence has been collected to suggest that responses can be altered to 'fake good' or 'fake bad', particularly the latter. These results have been obtained with instruments such as the Rorschach, TAT (Thematic Apperception Test) and sentence completion tests (eg. Albert et al, 1980; Exner, 1991; Schwartz et al, 1964; Perry & Kinder, 1990). The skilled examiner, however, is usually alert to signs of faking, usually due to the response patterns demonstrated by respondents, or the inconsisitencies seen in comparison with other data about them. This highlights the need for several measures to be employed.

There is also evidence that even subtle differences in the phrasing of verbal instructions and in examiner-examinee relationships can appreciably alter performance on projective tests (Baughman, 1951; Exner, 1993). Even when employing identical instructions, some examiners may be more encouraging or reassuring, others more threatening, owing to their general manner and appearance. Such differences may affect response productivity, defensiveness, stereotypy, imaginativeness and other basic performance characteristics.

Equally serious is the lack of objectivity in scoring and interpretation. One of the most conspicuous deficiencies common to many projective instruments is the lack of adequate norms, which even when present may be based on vaguely described populations. Therapists and clinicians must then fall back on their own 'general clinical experience' to make judgements concerning test performance, and therefore the clients final evaluation.

In view of the special nature of scoring procedures and the inadequacies of normative data in projective testing, measures of reliability (particularly scorer reliability) become an important consideration. Unfortunately, in terms of internal consistency (Atkinson, 1981) and retest reliability (Lindzey & Herman, 1955) projective testing measures have not yielded particularly atractive results. This is perhaps why Holtzman's Inkblot Technique has achieved such popularity as a revision of the Rorschach as it involves parallel forms of the same test, and so fairs better than most in respect to reliability measures.

The majority of validation studies investigating projective techniques have been inconclusive because of procedural deficiencies in either experimental controls or statistical analysis, or both. The danger is that these methodological deficiencies may have the effect of producing spurious evidence of validity. One source of this spurious validity data is a failure to cross-validate (Kinslinger, 1966). However, the most common semms to be what Chapman (1967) referred to as "Illusory validation". Illusory validation is a mechanism that underlies the survival of superstitions (for example). We tend to notice and recall whatever fits our expectations and ignore or forget what's contradictory. The original research of Chapman and Chapman (1967) with the D-A-P (Machover Draw-A-Person Test) has been corroborated by similar studies with the Rorschach and with the incomplete sentences blank (Chapman and Chapman, 1969; Golding & Rorer, 1972; Starr & Katkin, 1969). It should also be noted however, that certain inadequacies of experimental design may have the opposite effect of underestimating the validity of a diagnostic instrument.

An increasing number of test users stress the importance of holisitc and integrative principles in personality assessment, such as the use of configural patterns and contextual variables. Many of them have been critical of continuing attempts to validate single indicators, isolated scores or diagnostic 'signs' derived from projective techniques. That significant correlations may result from a failure to allow for complex patterns of relationships among personality variables can be illustrated by a plethora of contradictory findings about many of the projective techniques that clinicians find most useful.

Rather than being regarded and evaluated as psychometric instruments, or tests in the strict sense of the term, most projective tests have come to be regarded as clinical tools. Thus, they may serve as supplementary qualitative interviewing aids in the hands of a skilled clinicican. Attempts to evaluate them in terms of the usual psychometric procedures may be inappropriate. By the same token, the use of elaborate scoring systems that yield quantitative scores is not only wasteful but misleading. Such scoring procedures lend the results an illusory semblance of objectivity and may create the unwarranted impression that the given technique can be treated as a test. The special value that projective techniques may have is more likely to emerge when they are interpreted by qualitative, clinical procedures than when they are quantitatively scored and interpreted as if they were objective psychometric instruments.

Borrowing a concept from information theory, Cronbach and Gleser (1965) characterized interviewing and projective techniques as 'wideband' procedures. Bandwidth, or breadth of coverage, is achieved at the cost of lowered fidelity or dependibility of information. Objective psychometric tests characteristically yield a narrow band of information with a high level of dependibility, while projective tests yield a much wider range of information with a lower dependibility. Moreover, the types of information yielded from individual or from test to test can be much different in these 'wideband' projective procedures. One person's TAT responses may reveal a great deal about their level of aggression but little about their creativity, while the Rorschach may reveal much about their creativity but not their level of aggression. Such a lack of uniformity in the information given may explain the low validities of projective tests for a single trait across a group of people.

Through an examination of projective testing it is difficult to establish an adequate measure of personality. It is often stated that personality measures should exhibit levels of stability that are consistent with their intended uses. Thus, there must be a stable, reliable relationship between the behaviour of the individual and the score he or she receives. Also, the behaviour being measured must show some consistency over a specific range of situations. Not only that but the measures must be interpretable, not only by a single clinician but by all those that use them. This would require a specificity concerning exactly what the test is intending to measure. However, as discussed, these criteria would be more suitable to a quantitative analysis of much narrower bandwidth, whereas projective testing are much more effectively described as qualititative measures. In trying to create strict criteria for trying to identify a good instrument we find that projective and objective psychometric tests are best untilized as complementary halves of the same endeavour. Just as methodological pluralism is the best way to incorporate both qualitative and quantitative analysis, so is it the manner in which projective and psychometric testing be approached. It has already been mentioned that projective tests serve as an effective method to 'break the ice' in an interview and it is perhaps for this reason that they reamin so popular. However, due to their poor psychometric properties they should be combined with more objective measures such as self-report inventories in order to validate the initial evaluation made by the administration of a projective measure. Two of the better criteria for choosing a test would therefore be related to the properties of the information being gained (whether quantitative or qualitative for example) and the level of ease with which it can be effectively employed within a clinical setting. It is in the latter that projective tests have proved themselves greatly effective as 'introductory' tools for initializing a rapport between clinician and client. While the results may be prone to the same weaknesses of other personality measures (such as faking), and in comparison fair poorly as scientifically reliable psychometric instuments they are greatly effective in suggesting leads for further exploration or hypotheses about the individual for further verification.