Psychometric Assessment of Deaf People
There are very very few psychometric instruments available for use with prelingually profoundly deaf people. When testing such people therefore, there is a great deal to be borne in mind.
The most important point to consider is that of the validity and reliability of the tests used, both in and of themselves, and as a result of any changes that would need to be made to the formal administration in order to communicate the task to the client. Four main questions can help with this:
1. Does the test consist of verbal test items or performance items? Verbal items are inappropriate for deaf people, particularly prelingually deafened. Such people usually have difficulty with English syntax and vocabulary independent of cognitive function, as a result of English being a purely visual (and unspoken, and therefore considerably less frequently encountered) language for deaf people. Deaf people’s mean reading age (compared with hearing peers) has been estimated to be at the 3rd or 4th grade level irrespective of intellectual function.
2. Do instructions for the test require verbal communication? This is important even if the tasks themselves are considered to be “performance” tasks. For those measures that offer little or no flexibility in the way in which instructions are presented (such as the WAIS-III), departure from the formal administration departs also from the standard administration with the normative sample and comparison with that sample then becomes misleading in ways that cannot be predicted or allowed for. The same problem is more frequently met when comparing hearing people tested in the formal manner with those tested with more flexible instruction.
3. Do any test items discriminate against (or for) an individual with an auditory impairment? Test items may discriminate directly or indirectly against, or for, deaf people. Those that discriminate against are more common, but both for and against present problems of comparison with a normative sample. An immediately obvious example would be from the Vineland Social Maturity Scale which contains an item “Makes Telephone Calls”. Clearly this would underestimate the individual’s functioning. Perhaps less obviously, but equally critical, an attempt to address the cultural validity of this item by converting it to “Makes Minicom (Textphone) Calls” still underestimates the ‘true’ level of functioning since a deaf person’s experience of using a textphone differs entirely from a hearing person’s with a telephone (in terms of when one is first seen used, how often they are seen on TV, and so on.) Similarly there is some evidence that BSL users are advantaged when using the block tapping tasks (such as Corsi’s) – but advantaged in such a way that disappears when the person sits alongside the examiner instead of opposite.
4. Are deaf people included in the normative sample provided by the test developer? Apart from a very select handful of tests the answer to this question is always no. Norms for deaf people are rarely provided. There are good arguments for and against comparing this deaf person with deaf or ‘hearing’ norms, and the answer depends on the reason for testing. However there is only ever an argument for ‘hearing’ norms if that normative sample contained deaf people in equal proportion to the population of concern and that the results from that sub-sample are demonstrably as reliable as those of the hearing subjects. Such norms are rarely if ever established, and prelingually profoundly deaf people are more likely to be formally excluded from the normative study. The interpretation of results of clients who do not mirror individuals for whom the test was designed must therefore be explicitly cautious.
These main areas make the assumption that communication with the client, in sign, is unproblematic – that is, that the assessor is able to sign. When that is not the case, further issues of the reliability of the interpretation are raised:
5. How reliable is your communication support? The ability of the interpreter(s) must be stated in the report and considered when interpreting the results. Even if the interpreter is Level IV accredited and a registered Sign Language Interpreter, which ought really to be the only acceptable standard for psychometry, the reliability of the interpretation will decline significantly after 35 minutes. Two interpreters should be used and regular breaks should be introduced (which can itself conflict with a formal test administration). Interpretation will always add to the error in the estimated level of function, and it is impossible to say whether it contributes to an over- or an under-estimation. The extent of this error is, in part, a function of the competence of the interpreter(s).
6. How reliable is perfect communication support for this measure? Even if the communication support is 100% perfect (and this is in truth only available in our imagination) then it is important that each item of each test be discussed in advance with the interpreters. (There is no such thing as a perfect translation. There can be a best translation but that is hard to find and is only a “working-best” until a better one is noted in the future.) However, even supposing perfect interpretation, the assessor must be fully aware of the nature of the signed administration in order that the psychometric equivalence of the standard and the signed administration of that item can be judged. For example, the Similarities subtest of the WAIS-III contains the question, in English, “In what way are a coat and a suit alike?” It is fairly easy to translate this into a British Sign Language equivalent with little debate. However the nature of the most common signs for coat and suit make it self-evident that both are articles of clothing – a response which receives maximum points. Conversely the signs ‘rhyme’ in terms of both being bimanual and symmetrical, sharing the same location in space, employing very similar movements, and handshapes differing only in terms of thumb position. It could be argued that these phonological similarities are unreasonably misleading and/or that they imply alternative false answers (the location of both signs is commonly used for emotions) in a way that the spoken English items do not.
These concerns all contribute unknown amounts of error to the estimated level of whatever psychological construct is under scrutiny (such as intellectual function). In addition the ways in which these concerns may be addressed (that is by departing from standard administration in a variety of ways) also contribute error to the estimate. The magnitude of this error and the overall direction of it are both unknowns and lower the reliability of the obtained results.
The validity of obtained results remains in question when the normative sample is exclusively hearing.
Generally, examiners should be advised to:
• Assess the deaf person with support from qualified British Sign Language Interpreters.
• Discuss each item of each test in advance with the interpreters, and afterwards such that any instances of note may be raised and accommodated in the interpretation of the results.
• Consider which, if any, tests to use in the light of points one to four above.
• Interpret results with extreme caution in the light of points one to six above.
• Make each of these shortcomings explicit in the report such that future readers will not jump to erroneous conclusions.
• And generally to assume that obtained results reflect a hypothetical minimum value, and never imply a ceiling to the individual’s ability or potential.
Download the original article.