Archives of Clinical Neuropsychology, Vol. 12, No. 8, pp. 711–738, 1997 Copyright C) 1997 National Academy of Neuropsychology Printed in the USA, All rights reserved 0887-6177/97 $17.00 + ,00
Methods of Estimating Premorbid Functioning Michael D. Franzen Allegheny University of the Health Sciences
E. J. Burgess University of Virginia Health Sciences Center
Laura Smith-Seemiller Allegheny University of the Health Sciences
frequently requires the comparison of obtained scores Clinical neuropsychological assessment against some estimate of premorbid level of functioning, but only necently has signj%ant attention been turned to objective methods to accomplish this objective. Clinical judgment, although useful in some circumstances, is generally insufficient. Other methods of estimating premorbid function include demographic regressionformulae, such as the Baronaformula, subtest scatter methods ,such as that suggested by Lezak, and the use of current scores on tests ofpresumabl.y spared abilities, such as the National Adult Reading Test (NART). Almost all methods predict to some general level of intellectual functioning rather than to specific neuropsychological skills. This paper reviews the suggested methods in terms of the underlying assumptions and the available empirical evidence. Suggestions for future research include the development of skill specific predictors as well as investigations regarding the relation between predictor accuracy and characteristics of the subject, such as high versus low premorbidfinctioning in the subject. Additionally, there is a great need for methods to predict premorbidfunctioning in children. O 1997 National Academy of NeuropsycholOgy.published by Elsevier Science Ltd
BACKGROUND In the area of clinical neuropsychological assessment, there is frequently a need for some baseline against which to compare current performance. Actual premorbid test scores antedating cognitive decline are rarely available. Yet diagnosis frequently requires that some decline be demonstrated or deduced. For example, the diagnosis of dementia requires that a decline in general level of cognitive functioning must be present. The forensic arena is a particularly challenging endeavor where evidence for decline must be clear and convincing. The existence of change maybe relatively easy to document in cases of severe injury and its resultant impairment, but diilcult to document in cases of mild head injury, where subtle Addresscorrespondence to MichaelD.Franzen,PhD,Department of Psychiatry andAllegheny Neuropsychiatnc Institute,Allegheny GeneralHospital,Pittsburgh, PA,15212-4772. E-mail:[email protected]
M. D. FranZen, E. J. Burgess, and L. Smith-Seemiller
impairment or no impairment are both possible. Even in cases of severe injury, the magnitude of change may be important in the determination of damages. Other diagnoses pose difficult challenges. For example, the diagnosis of specific types of early dementia requires that some qualitative and quantitative pattern of scores be demonstrated. Included in this pattern is decline in memory and cognitive functioning. Here the question is even more complicated because the clinician must ask “is the change greater than can be expected on the basis of aging alone?” As people age, there is even greater variability in the range of scores possible, especially for tests of memory. As a result, comparing the obtained score to the population average may be inappropriate, although comparing the obtained score to the average score of an age-equivalent cohort, considering possible variability, may be helpful. Comparing an individual’s performance to some population average score is appropriate only when the score is uniformly present in all individuals and when performance is not related to age, sex, race, or education (Lezak, 1995). In such an instance, the presence of a deficiency is easily determined by comparing the obtained score to the population average. Unfortunately, in many, if not most neuropsychological skill areas, performance is related to the demographic variables mentioned above as well as to the effects of experience. The use of a single population average could result in misleading inferences. The development of age, sex, and education specific normative information is a great improvement recently seen in clinical neuropsychological practice. Examples of this approach include Heaton (1992), Heaton, Chelune, Talley, Kay, and Curtiss (1993), and Heaton, Grant, and Matthews (1991). But even here, there may be some limitations. Frequently, the cell sizes are somewhat small. For example, the Heaton et al. (1991) norms have a total sample size of 486 subjects, and although the cell sizes are not given in the manual, logically, with two levels for sex, 10 levels for age, and 6 levels for education, the average cell size may be only four subjects. Another limitation is one shared with the demographic approach, namely, that the predictions are to a hypothetical average rather than to the specific individual under consideration. As with most clinical decisions, many methods for the determination of change involve subjective or clinical judgment. That is, the clinician examines the data and determines if present test scores are likely to have occurred in the individual subject in the absence of the assumed intervening injury or disorder. The clinical decision requires a complex organization of disparate sources of information and a concatenation of multiple assumptions. As a result, there is e~or that can be reduced by the application of objective or quantitative methods. Clinical judgment, and more broadly, human judgment has been the subject of multiple investigations over the past few decades. Kahneman, Slovic, and Tversky (1982) presented a compilation of workup to the last decade, and Wedding and Faust (1989) discussed some of the issues specific to clinical neuropsychological assessment. Kleinmuntz (1990) reviewed work related to clinicaljudgment in medical diagnosis,EEG interpretation,account auditing, securities investment, and managerial decisions. The particulars of this line of research are richly detailed, but we will consider only the most pertinent and simple concepts; namely, that human judgment is fallible especially when the amount of necessary information increases and that judgment can sometimes be improved by actuarial methods and decision-making aids. On the other hand, Kleinmuntz (1990) described situations in which clinical judgment can actually be preferable to the use of decision aids. One of those situations is when actuarial methods are unavailable or when they are inaccurate, a description that can apply to ~remorbid estimate methods, especially in the early stages of development. The only published study investigating clinical judgment in premorbid estimation found that althoughclinicians’estimates approximatedthose obtained from a demographic formula, the clinician’s cordidence in their predictions may have been inflated (Kareken & Williams, 1994).The role of clinical judgment is discussed in another paper in this issue (Kareken, in this issue).
Estimating Premorbid Functioning
OVERVIEW OF VARIOUS METHODS This paper will review the various methods utilized in the clinical prediction of premorbid functioning. Excellent previous reviews include Crawford (1989, 1992), who provided a good explication of the use of the National Adult Reading Test (NART), especially in a British population, and Vanderploeg (1994), who provided an excellent discussion of the demographic methods and the use of Wechsler Adult Intelligence Scale-Revised (WAIS-R) subtest scores in the context of a discussion of the applicability of the WAIS-R to clinical neuropsychological assessment. This paper will review the more recent developments and attempt to provide direction for clinical practice and future research. In reviewing the relevant publications, it becomes obvious that most of the studies have concentrated on predicting IQ values rather than predicting memory performance or”visual-spatial skill. Therefore, the present discussion will rely heavily on concepts related to intellectual functioning. Furthermore, the populations that have seen the greatest interest are patients with closed-head injury or progressive dementia, and the present discussion will reflect this preponderance. There has been no formal survey of the methods used by clinicians in the estimation of premorbid functioning (although see Seemiller et al., this issue). There have been methods suggested in the clinical literature, and there have been limited reports of investigations in the empirical literature. The methods reported in the literature include the best performance method, use of actuarial or demographic regression formulae, and various aspects of present abilities measures where the measured skill is assumed to be relatively insensitive to the effects of the disease process or injury mechanism. The actuarial method requires that data from a large number of subjects be collected into a contingency-frequency table from which probability estimates can be calculated. As such, the actuarial method is an alternative to multiple regression prediction methods (Wiggins, 1973). Although these two terms are sometimes confused, we will consider them separately in this paper. Present abilities measures include the hold/don’t-hold strategies and tests of reading identification skill. Klesges, Wilkening, and Golden (1981) provided a review of the premorbid indices and classified these methods into three approaches. One approach utilized WAIS subtests, which yielded a derivative of the Wechsler deterioration quotient (Wechsler, 1958). A second approach involved measuring skills, such as vocabulary and reading, to predict premorbid intellectual functioning. The final approach reviewed by Klesges et al. (1981) involved the use of multiple regression techniques and demographic data. In a subsequent review, Klesges and Troster (1987) regarded the value of reliable indicators of premorbid functioning in adults as having potential and called for research to evaluate the reliability of indicators. Klesges and Troster (1987) noted that more evidence was needed to justify the use of reading and vocabulary measures. As a result, a search for ways to predict effectively premorbid abilities continues.
The Best Pe~orrnance Method Description. Muriel Lezak (1995) has provided what is probably the best articulated descrip-
tion of the clinical method in examining scores obtained on a collection of tests. Lezak’s method is based on the deficit measurement model, but it can be applied in other models as well. (The conceptual background of the deficit measurement model will not be addressed here. The reader is referred to Lezak (1995) for a more complete discussion.) The best performance method utilizes the individual’s best performance — either on current testing, nonscorable behavior (observed or reported), or evidence of premorbid achievements — as
M D. FranZen, E. J. Burgess, and L. Smith-Seemiller
the best estimate of premorbid ability (Lezak, 1995). Althclugh Lezak argues that the highest test score obtained by an individual is generally a good estimate of premorbid ability, she does acknowledge that there are some exceptions (e.g., in overachievers, or if the highest score is on a test of memory). One assumption of the best performance method is that in the absence of injury or disease, the average individual’s scores will group around some hypothetical mean level of performance. The highest score obtained on any single test in a given evaluation is used as the premorbid estimate. Scores on other tests are considered to reflect impairment if they differ by at least 1.5 standard deviations (SD) from the highest obtained score. A further assumption is that there is a single performance level that will represent the individual’scognitive ability in a range of areas. Lezak cites research on the general ability factor g as supporting the validity of this assumption. This method also assumes that marked discrepancies between the levels at which a person performs different intellectual functions or skills are evidence that some condition (e.g., neurologicalillness, emotional disturbance)is interferingwith the full expression of that person’s intellectual potential. Lezak further states that “no person functions at his endowed maximum potential” because many factors throughout development compromise intellectual effectiveness.Thus, a person’s performance on a task is seen as a floor, as opposed to a ceiling, in terms of their @uelevel of abilities. Research and Critique
Although Lezak’s method has intuitive appeal and has been widely used, there are a number of limitations to this approach as well. Most criticisms have addressed assumptions underlying the best performancemethod. Mortensen, Gade, and Reinisch (1991)note that, although a general (or g) factor can account for much of the variance in intellectualperformance,it does not account for all variance. They note that considerableintra-individualscatter is often seen in the test scores of healthy people. They reported the results of three studies in their paper, in each of which the best performance method overestimatedpremorbid level of intellectual ability in both healthy normal adults and individuals with cerebral atrophy. Examination of the WAIS-R manual (Wechsler, 1981) indicates that — for the overall standardization group as well as for various age cohorts –– the correlations among subtests can be as low as .41 (e.g., between Object Assembly and Vocabulary subtests). With correlations that low, there can be substantial differences between pairs of subtests, even without acquired impairment. Matarazzo and Prifitera (1989) presented data from the WAIS-R standardization sample demonstrating that a high degree of scatter is to be expected in normal, healthy subjects. The mean scatter, or difference between highest and lowest subtest scale score, was 6.66 in the standardization sample. Almost one third (31.990) demonstrated scatter of 8 points or more. Although demographic variables were not found to be related to the amount of scatter between tests, the magnitude of scatter increased as IQ increased. The authors also point out that correlations between each subtest score and the Full Scale IQ are far from unity, and that subtests often correlate poorly with each other. Thus, they argue that no single subtest score can be used in isolation to estimate full scale IQ. Matarazzo and Prifitera (1989) echo the argument, initially put forth by Thorndike and Thurstone, that individuals are endowed with a large number of independent specific abilities that are as important as their general level of intelligence (g). This is consistent with recent empirical and theoretical work in cognitive psychology that has posited a multifactorial view of intelligence (Steinberg, 1990) in which variability or scatter is expected as a function of individuality. McLean, Reynolds, and Kaufman (1990) make the point that examination of scatter should take into account all of the subtests scores and not just the extreme high and low
Estimating Premorbid Functioning
scores. They suggest using a profile variability index that can be interpreted as the standard deviation of subtest scores. In their examination of the WAIS-R standardization data, these investigators found that scatter increased as IQ increased. This again points to the limitations of using a general rule of thumb comparison for all subjects. In another empirical look at this question, Mitrushina and Satz (1995) investigated the scatter among subtest scores in a group of 156 healthy elderly individuals. Nearly 50’ZOof these older people exhibited age-corrected scale score scatter of 8 points when administered the Satz-Mogel short form. Remarkably, four of the subjects exhibited a scatter of 14 points. Presumably, the use of the Satz-Mogel short forms could have increased the scatter somewhat over what might have been seen with the full-length forms. However, given the concordance between Satz-Mogel and full length WAIS-R IQ values, this hypothesis cannot totally explain the degree of scatter seen. Although the sample included only elderly subjects, the results call into question use of the highest subscale as an estimate of premorbid ability, since this method would invariably result in incorrect diagnoses of acquired impairment or pathological decline in these healthy individuals. It should also be noted that the studies described here discuss only subtests of the WAIS-R. When one considers using the best performance method on data from a test battery in which a range of functional abilities are assessed using different tests, each with its own psychometric properties and normative samples, the potential dangers described above are magnified. It may be questionable to utilize an estimate of premorbid functioning derived from performance on the WAIS-R for determining whether an individual has suffered declines in memory performance. Substantial proportions of unimpaired individuals may present large intelligence-memory discrepancies similar to those discrepancies found in brain-impaired subjects when the WAIS-R FSIQ and the WMS-R are used. Alternately, for certain diagnostic categories, such as Alzheimer’s disease, both the average magnitude of the discrepancy as well as the proportion of subjects scoring above a certain minimum discrepancy may be discriminators (Bomstein, Chelune, & Prifitera, 1989). Other criticisms of the best performance method are related to psychometric aspects of tests and measurements. As noted above, one assumption underlying the use of current performance is that the test in question is reliable. Matarazzo and Prifitera (1989) note that imperfect test-retest reliabilities and the magnitude of the standard error of measurement both add to the scatter that is often observed among subtest scores. Mortensen et al. (1991) elaborate by noting that error associated with a test score is assumed to have a symmetrical distribution with a mean of O, and thus the obtained score on a test will be higher than the “true” score in half of all subjects. This situation undermines Lezak’s argument that an obtained score represents a “floor” of “true” level of ability. They note that, given the number of tests administered in a neuropsychological battery, it is quite likely that the highest score obtained by an individual may contain a large positive error component, and the lowest score may contain a large negative error component. Yet another criticism of the best performance method involves the methodology suggested for the comparison. Considering a difference of 1.5 SLI as being meaningful has two drawbacks. The first drawback is related to the standard error of difference, which is the minimum magnitude required for a difference to be considered reliable (not due simply to chance). The standard error of difference is based upon the reliability of both tests being compared. The formula is SED=SD~(
2–rll – rz2)
(Anastasi, 1988). Let us examine two numerical examples to see why a general rule of thumb of 1.5 SD can be misleading. In the first case, we wish to compare two tests whose reliability
M. D. Franz.en, E. J. Burgess, and L. Smith-Seerniller
is both .90. In this case, the standard error of difference (SED)?3? is 6.7. Multiplying by 1.96 to determine the difference required for a significance level of .05, we obtain a difference of 13.1, in which case Lezak’s suggested value of a 15-point.difference is an overestimate. In a second case, we wish to compare two tests whose reliability is .70. Here the obtained value required for a significance level of .05 is 22.7, and the suggested 15 points is insufficient. Tests with greater reliability have smaller standard errors of difference and therefore require smaller differences than less reliable tests, which have larger standard errors of difference. The second point is related to the meaningfulness of a reliable difference. A reliable difference is not necessarily meaningful. A review of the literature related to Verbal IQPerformance IQ splits suggests the problems inherent in this method. A reliable difference between VIQ and PIQ is calculated to be 9.29 on the basis of the standardization data for the WAIS-R. Yet a review of the standardization data indicates that 1% of the standardization sample had VIQ-PIQ splits of 28 points or more, and 25% of the sample had splits of 13 points or more (Ryan, 1984). Thus, a reliable VIQ-PIQ split may be seen in healthy individuals, and is not necessarily meaningful when testing for a decline. In any event, the choice of a minimum difference score is problematic. The values reported in the WAIS-R manual (Wechsler, 1981) for pairwise subtest comparisons may be misleadingly small, at least for samples of psychiatric patients, who interestingly present with larger subtest intercorrelations than the subjects in the standardization sample (Piedmont, Sokolove, & Fleming, 1989). Hypothetically, the maximum and minimum values of subscale scores can consist of any two subscales. If we consider anjr pairwise comparison of subscale scores, there are 51 possible pairwise comparisons, which unnecessarily complicates the issue. For the purposes under discussion here, the relevant comparison may not be pairwise, but instead may be comparison of an outlier subtest to the mean of the other subtests, thus avoiding problems inherent in multiple comparisons (Mc(lean et al., 1990). of best performance method in various populations. The application of best performance measures or current performancemeasures in the elderly poses special problems not present in younger populations. The issue of estimating premorbid level of functioning in the elderly is complicated by the fact that some neuropsychologicalskills show mild to moderate decline with aging, even in the absence of disease. However, IQ is based upon age-cohort comparisons, and the deviation IQ is a measure of distance from the mean in the appropriate age cohort. Therefore, longitudinal or serial IQ values should not show a decline, because a decline would mean a change in the relative position of an individual in the distributionof similarly aged individuals. One complicating factor is that most neuropsychologicalassessment instruments do not have deviation indices and insteaduse either an additivecorrectionfactor or a modified cut-off score to account for the changes associated with aging.
“Hold-Don ‘t Hold” Methods
“Hold-don’t hold” methods estimate premorbid ability based on the individual’s current performance on a measure that is considered to be relatively resistant to neurological impairment. Most such approaches can be grouped into two major types: Those based on patterns of performance on the Wechsler Intelligence scales, and those based on word reading ability, which will be discussed later in this paper. Wechsler (1958) described a Deterioration Quotient, obtained by subtracting age scale scores for “don’t hold” subtests from the sum of scaled scores for the “hold” tests. The “hold” tests are Vocabulary, Information, Object Assembly, and Picture Completion. The “don’t hold” tests are Block Design, Digit Symbol, Di,git Span, and Similarities. Other Description.
Estimating Premorbid Functioning
variations on this approach have been described. Yates (1956) suggested that one can simply use performance on the Vocabulary subtest as an index of premorbid intellectual ability. McFie (1975) suggested using the average score of the Vocabulary and Picture Completion subtests. The choice of hold tests may depend upon the type of impairment under consideration. For example, although Picture Completion is included among the hold tests, lesions of the occipital or parietal lobes may result in relative decrements in scores on this subtest, which would render this subtest inappropriate in these instances. A related approach is to develop formulae based on hold versus no-hold tests to distinguish a specific necrologic condition from healthy controls. An example of this is the Fuld profile (also referred to as the “cholinergic profile”), which is specifically designed to differentiate subjects with Alzheimer’s disease from controls (Fuld, 1982, 1984), and which had originally been identified in younger subjects with a drug-induced cholinergic deficiency. Alzheimer’s disease is indicated if the following profile is obtained using age scale scores: A > B > C < D and A > D where A = (Information + Vocabulary) / 2; B = (Similarities + Digit Span) / 2; C = (Digit Symbol + Block Design)/ 2; and D = Object Assembly. Fuld (1984) reported 44% sensitivity and 96% specificity in a sample of 12 individuals with Senile Dementia Alzheimer’s Type (SDAT) and 28 individuals with other forms of dementia. Research and critique. Tuokko and Crockett (1987) reported that the Fuld profile was present
in only one individual in a sample of 74 healthy elderly individuals. There was no comparison of SDAT patients so the classification accuracy of this strategy could not be determined from this study. Filley, Kobayashi, and Heaton (1987) reported that only 22% of their sample of 41 patients with SDAT showed this pattern (half of the subjects had been administered the WAIS and half of the subjects had been administered the WAIS-R). Further, in a comparison sample of 42 healthy elderly and 30 patients with other neuropsychiatric disorders, 2.4q0 and 1670, respectively, showed the Fuld pattern. Logsdon, Teri, Williams, Vitiello, and Prinz (1989) reported that the profile did not occur at different rates in a sample of elderly individuals with and without SDAT. Furthermore, the profile was present in a sample of healthy elderly individuals. Goldman, Axelrod, Giordani, Foster, and Berent (1992) reported that the profile had both low sensitivity and low temporal stability in individuals with SDAT who were tested over the interval of 1 year. Satz, Van Gorp, Soper, and Mitrushina (1987) recommended cautious use of the profile to identify dementia only in samples of individuals where the prevalence of SDAT was greater than in the general population. To determine if the Fuld profile was more common in SDAT than other types of dementia, Brinkman and Braun (1987) studied WAIS results in a sample of 23 patients with SDAT and 39 patients with multi-infarct dementia (MID). They reported that the formula had a correct classification rate of 5790, and 9470 specificity. Heinrichs and Celinski (1987) reported that their sample of 50 closed-head injury patients showed the Fuld pattern 10Yoof the time. Goldman, Axelrod, Tandon, and Berent (1993) reported that 15% of a sample of young nondemented schizophrenic patients (none of whom were receiving anticholinergic medication) exhibited the profile. The profile is not frequently seen in depressed patients (Bomstein, Termeer, Longbrake, Heger, & North, 1989) or in temporal lobe seizure disorder (Bornstein & Share, 1990). The development of holdho-hold formulae that are specific to necrologic populations is theoretically and conceptually intriguing, as much of the other research addressing the issue of premorbid functioning and the identification of a decline in function ignores the fact that different neurological conditions (as well as differences in lesion location) are likely to differentially affect present ability. While the Fuld profile has been found to have adequate specificity in comparison to other progressive dementing conditions, many individuals with suspected dementia of the Alzheimer’s type do not demonstrate the profile. It has also been suggested that the Fuld profile is not specific to Alzheimer’s disease.
M. D. Franzen, E. J. Burgess, and L Smith-Seemiller
A review of the various studies conducted using the Fuld profile indicated that when the data across 18 studies were pooled, the sensitivity of the Ftdd profile was only 24.1Yo,while its specificity in clinical populations was 88.5% (Massman & Bigler, 1993). That is, neither the individuals with SDAT nor the individuals without SDAT showed the profile with any great regularity. Thus, the clinical utility of the Fuld profile may have more limitations than originally thought. Like other methods of estimating premorbid level of functioning, these hold/don’t-hold approaches are also vulnerable to criticism based on their assumption that an individual functioned at similar levels across all areas of brain-behaviorfunctions. It has also been found that some of the traditional “hold” subtests (such as Vocabulary) are more sensitive to acquired cerebral dysfunction than commonly assumed (e.g. Larrabee, Largen, & Levin, 1985).
Methods Utilizing Demographic
Description. Demographic variables, such as social class ancleducation, are known to be closely
related to scores on intelligence tests, and thus may provide the clinician with information on an individual’s premorbid level of intellectual firnctioning (Crawford, 1992). For example, Ryan, Paolo, and Dunn (1995) reported on the relationship between demographic variables and performance on the WAIS-R in a sample of 130 heahhy individuals over the age of 75 years. They found that education, age, and preretirement occupation all contributed significantly to variance in IQ values. Although these authors did not develop regression equations to predict IQ values, they did provide a table of mean IQs by levels of gender, edueation, and preretirement occupation for their overall sample. The authors suggest that values in this table can be used as rough estimates of predicted IQ in older individuals (however,caulion should be exercised because no empirical evaluation of this strategy was provided). While the use of demographic variables can be usefrd in estimating IQ scores, it may not be clear how this information should be used to predict a.specific IQ score. As a guide for combining demographic information into a specific IQ prediction, actuarial methods have been developed for predicting premorbid intellectual ability (these are referred to as actuarial methods, although actually they are multiple regression models.) Also, more recent methods have been described that utilize a combination of demographic information and performance on tests which are considered to be relatively insensitive to the effects of acquired brain injury (e.g., Grober & Sliwinski, 1991; Kareken, Gur, & Saykin, 1995). The use of objective methods, such as actuarial- or regression-based approaches utilizing demographic information for predicting premorbid intellectual ability is attractive for a number of reasons. A frequently cited reason is that objective approaches are superior to clinical judgment because of the objective methods’ higher rates of interrater reliability due to an absence of subjective judgment (Barona, Reynolds, & Chastain, 1984). Also, while it can be argued that a patient’s performance on putative “hold” tests or abilities can be affected by acquired brain injury, demographic background is unaffected. Wilson, Rosenbaum, Brown, Rourke, Whitman, and Grisell (1978)?5? provided an early attempt to utilize demographic data to develop an objective method for predicting premorbid intelligence. They developed regression equations to predict premorbid WAIS IQ scores by using data on age, sex, race, education, and occupation from the 1955 WAIS standardization sample. The resulting squared multiple correlations were .54, .53, and .42 for the WAIS Full Scale, Verbal and Performance IQ scores, respectively. W’ilsonet al. (1978) also provided a method for correcting for differences between the educational levels of the standardization sample and educational levels that were more current at the time their formulae were developed.
Estimating Premorbid Functioning
The original WAIS was revised in 1981. In addition to replacing approximately 20% of the specific test items and altering the order of subtest adrninistration,the revised WAIS or WAIS-R (Wechsler, 1981) was standardized on a sample of 1,880 individuals, selected to more precisely reflect the demographic composition of the population of the United States. The IQs obtained from the WAIS-R were not directly comparable to those obtained with the original WAIS, differing by 7 to 8 IQ points (Wechsler, 1981). Thus, IQ estimates obtained using Wilson’s method would be expected to have attenuated validity in predicting WAIS-R IQs. In response to this concern, Barona et al. (1984) utilized a similar approach to develop regression equations for predicting IQ scores on the Wechsler Adult Intelligence ScaleRevised. In addition to age, sex, race, education, and occupation (as Wilson et al., 1978 used), they included urbanhural residence and geographic region. They initially included handedness as a predictor variable, but dropped it from the equations when their study failed to find a relationship between handedness and WAIS-R IQ scores. Education, race, and occupation were found to be the most powerful predictors in each equation. The squared multiple correlations between the equations obtained and actual WAIS-R IQ scores were .38, .24, and .36 for the Verbal, Performance, and Full Scale IQ scores, respectively. Research and critique. Attempts to empirically evaluate the Wilson et al. (1978) formula have been reported with equivocal results. For example, Karzmark, Heaton, Grant, and Matthews (1985) used a sample of 491 normal control subjects to evaluate the match between predictions made using the 1978 Wilson formula and actual WAIS FSIQ values. These investigators reported that when a standard of matching the actual IQs within one standard error of estimate (SEE’)was used, 7070 accuracy was achieved. (There was underprediction in 20% of the subjects and overprediction in 10% of the subjects.) Klesges, Sanchez, and Stanton (1981) reported that although the Wilson predictions and obtained WAIS IQs correlated significantly in their samples of 60 psychiatric inpatients and 106 outpatients referred for Workman’s Compensation evaluations, there were significant differences in the exact IQ values predicted. Using the educational correction factor suggested by Wilson et al. (1978) resulted in accurate predictions for the outpatient sample but continued overprediction for the inpatient sample. A later study by Klesges, Fisher, Vasey, and Pheley (1984) was similarly pessimistic regarding the use of the Wilson demographic prediction formulae in individual clinical cases. Karzmark et al. (1985) found that there was 66% accuracy when using only education as a predictor, and they recommended using the simpler formula. Wilson, Rosenbaum, and Brown (1979) compared the 1978 Wilson formula with the Wechsler deterioration index in predicting which subjects could be classified as impaired or normal using discriminant function analysis. These investigators reported that the Wechsler method resulted in 62% accuracy of classification, and the 1978 Wilson formula resulted in 73~0 accuracy. Taken together, these seemingly disparate studies supported the use of demographic methods over the use of current subtest performance methods. As noted, the Wilson equations were developed in 1978to predict WAIS IQ scores, while the Barona equations were developedin 1984to predictWAIS-RIQ scores.Attempts to crossvalidate the 1984 formula have generally found that IQ tends to be overestimated in normal subjects (Eppinger et al., 1987), especially when FSIQ is less than 89. There may also be a tendency to underestimate IQ when FSIQ is above 110 (Ryan & Prifitera, 1990). However, it should be pointed out that use of the Barona estimates in a discriminantfunction allowed generally accurate classification of healthy and demented elderly (Eppinger et al., 1987; Mittenberg, Thompson, Schwartz, Ryan, & Levitt, 1991). Interestingly,the Mittenberg et al. (1991) study also indicated that use of the Barona formulae was not enhanced by the inclusion of an index of qualitative aspects of performance (e.g., degree of scatter in subtest scores).
M. D. FranZen, E. J. Burgess, and L. Smith-Seemiller
There are now actually two Barona formulae. The first was based on a regression prediction of IQ using the entire WAIS-R standardization sample (Barona et al., 1984) and the second was based on a regression prediction using Black and White subjects over the age of 19 years from the WAIS-R standardization sample (Barona & Chastain, 1986). In a study comparing these formulae, Paolo and Ryan (1992) administered the WAIS-R to a sample of 75 healthy elderly males and 20 elderly males with neurological disease. The 1984 Barona formula significantly underestimated VIQ and FSIQ in the healthy subjects while the 1986 Barona formula significantly underestimated only VIQ. In subjects with neurological disease, both the 1984 and 1986 formulae resulted in significantly greater predicted IQs than obtained IQs, which is the desired direction of difference when some deterioration has taken place. The authors concluded that the 1984 formula is probably superior to the supposedly improved 1986 formula when predicting IQ in healthy elderly individuals. A number of limitations are present in the use of either the Wilson and Barona regression equations. As noted by Barona et al. (1984), the equations will tend to artificially lower or raise the estimated scores for individuals who fall outside of 1 SD of the population mean IQ due to regression toward the population mean. They note that this is particularly likely if premorbid IQ scores are above 120 or below 69. In acknowledgingthe limitations of their equations, they state that the estimates of premorbidIQ should not be seen as “exact predictions” and reported that the utility of the procedure may be limited for individual cases. As further evidence that the Wilson method may underestimate IQ in the higher range, Hale, Dingemansj Wekking,and Cornelissen (1993) found that the Wilson formula underpredicted IQ in a sample of 27 depressed and 34 nondepressed Dutch subjects with average and above IQ values. Due to the restriction of range inherent in the use of the Barona and Wilson estimates, most IQ predictions will cluster in the average to high average range (Sweet, Moberg, & Tovian, 1990). Sweet et al. (1990) found that neither the Barona nor the Wilson formulae performed any better than chance in classifying subjects into IQ ranges (using the standard IQ ranges presented in the WAIS-R Manual) and that both formulae predicted IQs significantly greater than obtained for a sample of neurological and psychiatric patients. Ryan and Prifitera (1990), in a study of the Barona estimates’ accuracy in predicting short form IQ scores, found it to be fairly reliable in predicting IQ scores between 90 and 109. However, this also raises questions regarding the ability of the equations to predict IQ scores in the upper and lower ranges of the distribution. Eppinger et al. (1987) found a strong correlation (r = .78) between formula-estimated IQ and obtained WAIS-R IQ in a sample of neurologically normal patients. The percentage of subjects who were correctly classified (as indicated by obtained IQ scores within 1 SEE of the predicted score) ranged from 69 to 75%. While judging the Barona estimate to have a “respectable” level of accuracy, they also noted the problems with estimating intelligence at the extremes of the normal curve. Eppinger et al. (1987) noted some additional practical problems and criticisms of the Barona equations. They observed that the formula’s occupational classification system is unsatisfactory, as many occupations could not be easily classified based on Barona et al. ’s (1984) system. Although some attempts have been made to address this (i.e., Karzmark et al., 1985), problems remain. Eppinger and colleagues also noted that the method of classifying education has limitations. For example, there are no differences in rating between an individual who obtains a bachelor’s degree and an individual who completes a master’s or doctoral degree. Also, it is unclear how one might rate an individual with 12 years of special education, compared with an individual who is mainstreamed for 12 years of education.
Estimating Premorbid Functioning
In reviewing studies that have assessed the validity of the Wilson and Barona equations, it is sometimes difficult to make interpretations regarding the clinical utility of the equations. For example, while some studies find a significant correlation between the demographic equations and IQ scores, correlation coefficients only provide information on the linear relationship between two variables, they do not specifically provide information about the accuracy of the individual IQ estimates (Bolter, Gouvier, Veneklasen, and Long, 1982). In a study addressing the clinical utility of the Wilson equations, Bolter et al. (1982) found a significant correlation between IQ estimates and measured IQ scores. However, they found the method’s accuracy to be limited in terms of predicting groups of head-injured patients and controls, using the percentage the criterion of controls and recovered patients’ obtained FSIQ scores that were within 1 SEE of estimate of the predicted FSIQ score. For the control subjects, 67Y0of the subjects had accurate IQ estimates, but 5070 of the patient sample had IQ estimates at least one SEE greater than obtained FSIQ values. Bolter et al. (1982) argued against the use of the Wilson equations with head trauma patients. In an examination of premorbid estimates of PIQ and VIQ in the same subjects, Gouvier, Bolter, Veneklasen, and Long (1983) reported similar results. More empirical work is needed regarding the accuracy of demographic prediction models, especially in relation to the differential validity of the methods as a function of subject population. of demographic formulae in various population. The issue of estimating premorbid level of functioning in children has received much less interest and activity than has the counterpart issue in adult populations. There maybe many reasons for this relative dearth of material. One of the reasons may be that premorbid estimates in children are fraught with even more uncertainty because of the shorter lifespan preceding the injury. There are fewer pieces of data on which to base empirical estimates, resulting in greater reliance on clinical judgment. Reynolds and Gutkin (1979) modeled an approach on the same methods used in the adult literature, namely the use of demographic data to predict IQ in the standardization sample of the relevant IQ test, in this case, the Wechsler Intelligence Scale for Children-Revised (WISC-R). They suggested that a discrepancy of at least 21 points between predicted and obtained IQ values be used as an indicator of organic brain impairment. Unfortunately, Klesges and Sanchez (1981) failed to replicate these results in a sample of 76 brain-impaired and 23 unimpaired children. There were low and nonsignificant correlations between predicted and obtained IQs. Only 35% of the impaired subjects had IQ discrepancies consistent with Reynolds and Gutkin’s (1979) suggestions, and 12V0of the unimpaired children had similar discrepancies. Klesges (1982) reported similarly disappointing findings in a separate sample of 35 nonimpaired and 26 brain-impaired children. Sellers, Burns, and Guyrke (1996) attempted to derive demographic predicting formulae for the Wechsler Preschool and Primary Scale of Intelligence-Revised based on the standardization sample data. Unfortunately, they too found disappointingly low multiple correlations of .54 for Full Scale IQ, .52 for Verbal IQ, and .44 for Performance IQ. There is the possibility that the greater variability in longitudinal childhood IQ may play a role here. One promising note is that in the Sellers et al. (1996) study, the strongest predictor was level of parental education, indicating that parental IQ may predict child IQ fairly well for these purposes. Application
Word Reading Tasks Description. In addition to using the WAIS scores and demographic variables to estimate premorbid abilities, word reading tasks have been shown to be effective for this purpose. As
M. D. FranZen, E. J. Burgess, and L. Smith-Seemiller
discussed by Willshire, Kinsella, and Prior (1991), these tests are based upon four premises: reading is highly correlated with intelligence level in the general population; reading ability is more resistant to dementia than is the WAIS Vocabulary subtest; the reading of irregular words is more resistant to cognitive decline than is reading of regular words; and word reading taps previous knowledge while minimizing the demands on current cognitive capacity. The fiist popular reading test used to estimate premorbid abilities was the Schonell Graded Word Reading Test (GWRT) (Schonell, 1942).The GWRT proved to be a reasonable predictor of premorbid intellect in adults with average intelligence. However, it was originally designed to assess children’s reading ability, and a ceiling effect occurred when used with adults of above average intelligence (predicted IQs had a ceiling of 115) (Schonell, 1942). As the GWRT proved to be too narrow to evaluate adults, the National Adult Reading Test (NART) (Nelson& McKenna, 1975) was developed. One of the original applications was to aid in the assessment of elderly who might be showing a disease-related decline. The NART is a reading test for irregularly spelled words. It was proposed that a reading test of this type would be a better predictor of premorbid ability than a vocabulary test alone, since it would assess the level of experience achieved before the onset of brain impairment. It was hypothesized that while the ability to decode phonological relations may decline with disease, recognition of words with which one was familiar would not decline. Using a group of 98 subjects with extracerebral disease, Nelson and McKenna (1975) regressed scores on the GWRT against IQs estimated from the Arithmetic, Similarities, Digit Span, Vocabulary,Picture Completion, Block Design and Picture Arrangement subtests of the Wechsler Adult Intelligence Scale. This regression formula was then used to estimate IQs in a group of 45 subjects with a dementing disease. The authors interpreted the lower correlation between the GWRT and the prorated scores in the dementia group as being evidence for the relative sparing of word reading ability. However, the difference between these two correlation coefficients was not tested for significance. Furthermore, it is important to realize that the correlation between those variables had a value of .75 in the normal control group and a value of .61 in the dementia group, and both correlations were significant at the p < .001 level. Clearly, this is an issue of being relatively insensitive to the effects of dementia rather than an issue of sparing. The authors themselves point out that the use of the GWRT would result in a conservative estimate of premorbid IQ. Ruddle and Bradshaw (1982) attempted a replication of Nelson and McKenna’s (1975) study with a sample of 78 healthy subjects and 75 patients with heterogeneous cortical diseases, and reported generally consistent findings. They also regressed the GWRT scores against the Raven’s Progressive Matrices test in order to obtain premorbid estimates of that test as well. Fifty percent of the patient sample had a discrepancy between predicted and obtained IQ scores at the p 91. Karzmark,P., Heaton, R. K., Grant, I., & Matthews, C. G. (1984).Use of demogmphicvariablesto predict overall level of performance on the Hafstead-ReitanBattery.Journal of Consultingand Clinical Psychology,52, 663-665. Karzmark,P., Heaton, R. K., Grant, I., & Matthews,C. G. (1985).Use of demographicvariablesto predictFull ScrdeIQ: A replication and extension.Journal of Clinicaland ExperimentalNeurvpsychology,7, 412=420. Kleinnmntz,B. (1990). Why we still use our heads instead of formulas:Towardan integrativeapproach.Psychological Bulletin, 107, 296-310. Kfesges, R. C. (1982). Establishingpremorbid levels of intellectualfunctioningin children:Arr empirical investigation. ClinicalNeumpsychology, 4, 15-17. Klesges, R. C., Fisher, L., Vasey, M., & Pheley, A. (1985). Predicting premorbid functioning levels: Another look. InternationalJournal of ClinicalNeumpsychology, 7, 1-3. Klesges, R. C., & Sanchez,V. C. (1981). Cross-validationof an index of premorbid tlrnctioningin children.Journal of Clinical and ConsultingPsychology,49, 141. Kfesges, R. C., Sanchez,V.C., & Strmton,A. L. (1981).Cross validationof an adultpremorbidfunctioningindex. Clinical Neuropsychology,3, 1%15. Klesges, R. C., & Troster,A. I. (1987).A review of premorbidindicesof intekctual and neuropsychologicalfunctioning: What have we learned in the past five years? The InternationalJoumAl of ClinicalNeumpsychology, 9, 1–11. Kfesges, R., Wilkening, G. N., & Golden, C. J. (1981). Premorbid estimates of intelligence: A review. Clinical Neuropsychology,3, 32–39. Klitzke, M. J. (1994). Prediction of behavioral dysfiorctionfollowing traumatic brain injury by observer report of prerrrorbidcharacterologicalstructure. Paper presented at the 14tb Annual Meeting of the NationrdAcademy of Nenropsychology.November, 1994, Orlando, Florida. Krull, K. R., Scott, J. G., & Sherer, M. (1995). Estimation of premorbid intelligencefrom combined performance and demographicvariables. The Clinical Neumpsychologist,9, 83-88. Larrabec, G. J., Largen, J. W., & Levin, H. S. (1985). Sensitivity of age-declineresistant (“hold”) WAIS subtests to Alzheimer’sdisease. Journal of Clinical and ExperimentalNeumpsychology, 7, 497–504. Lees-Haley,P. (1990).Vocationalneuropsychologicalrequirementsof U.S. occupations.Pemeptual and Motor Skills, 70, 1383-1386. Lezak, M. D. (1995). Neumpsychologicalassessment (3rd cd.). New YorklOxford:Oxford UniversityPress. Logsdon, R. G., Teri, L., Williams, D. E., Vitiello,M. V., & Prinz, F. N. (1989).The WAJS-Rprofile:A diagnostictool for Alzheimer’s disease?Journal of Clinical and ExperimentalNeurvpsychology,11, 892-898. Massman,1?J., & Bigler,E. D. (1993).A quantitativereview of the diagnosticutility of the WAIS-RFuldprofile. Archives of Clinical Neumpsychology, 8, 417-428. Mattamzzo, J. D., & Prifitera, A. (1989). Subtest scatter and premorbid intelligence: J_essonsfkom the WAIS-R standardizationsample. PsychologicalAssessment:A Journal of Consultingand Clinical Psychology, 1, 186-191. McCleart,J. E., Reynolds, C. R., & Kaufman,A. S. (1990). WAIS-R subtest scatter using the profile variability index. PsychologicalAssessment:A Journal of Consultingand Clinical Psychology,2,289-292. McFie, J. (1975).Assessment of organic intellectualimpaimrerrt.New York:Academic Press. Mitrushina, M., & Satz, P. (1995). Base rates of the WAJS-R irrtersubtestscatter and VIQ-PIQ discrepancy in normal elderly.Journal of Clinical Psychology,51, 7&78. Mittenberg, W., Thompson, G. B., Schwartz, J. A., Ryan, J. J., & Levitt, R. (1991). Intellechud loss in Alzheimer’s dementia and WAIS-R intraaubjectscatter.Journal of Clinical Psychology,47,544-547. Mortensen, E. L., Gade, A., & Reinisch,J. M. (1991).A critical note on Ixzak’s “best performancemethod” in clinical neuropsychology.Jourrralof Clinical and E.sperirnerrtal Nerovpsychology,13, 361–371.
Estimating Premorbid Functioning
Moss, A. R., & Dowd, T. (1991). Does the NART hold afkr head injury?: A case report. British Journal of Clinical Psychology,30, 179–180. Nebes, R. D., Martin, D. C., & Horn, L. C. (1984). Sparing of semantic memory in Alzheimer’s disease. Jourmzl of Abnormal Psychology,93, 321-330. Nelson, H. E. (1982). NationalAdult Reading test (NMT) manual. Windsor,U.K.: NFER-Nelson Publishing Co. Nelson, H. E., & McKerma,P. (1975).The use of current reading ability in the assessmentof dementia.British JourmIl
of Socialand ClinicalPsychology,14,259-267. Nelson, H. E., & O’Connell, (1978). Dementia: the estimation of premorbid intelligence levels using the New Adult Reading Test. Cortex, 14,234-244. Nurmally,J. (1978). Psychometrictheory. New York: McGraw-Hill. O’Carroll, R. E. (1987). The inter-rater reliability of the National Adult Reading Test (NART): A pilot study. British Journal of Clinical Psychology,26, 229-230. O’Carroll,R. E., Baikie, E. M., & Wbhtick, J. E. (1987).Does the NationalAdult ReadingTest hold in dementia? British Journal of Clinical Psychology,26, 315–316. O’Carroll, R. E., & Gillead, C. J. (1986). Estimation of premorbid intelligencein dementia. British Journal of Clinical Psychdon, 25, 157–158. O’Carroll, R. E., Moffoot, A., Ebmeier, K. P., & Goodwin, G. M. (1992). Estimating pre-morbid intellectual ability in alcoholic Korsakoff syndrome.PsychologicalMedicine, 22,903-909. Paolo, A. M., & Ryan, J. J. (1992). Generrdizabilityof two methods of estimatingpremorbid intelligencein the elderly. Archives of Clinical Neuropsychofogy,7, 135–143. Patterson, K., Graham, N., & Hodges, J. R. (1994). Reading in dementia of the Alzheirner’stype: A preserved ability? Neurvpsychology,8, 395407. Piedmont, R. L., Sokolove, R. L., & Fleming, M. Z. (1989). On WAIS-R difference scores in a psycbiatic sample. PsychologicalAssessment:A Journal of Consultingand ClinicalPsychology, 1, 155–159. Reynolds,C. R., & Gutkin,T. B. (1979).Prdcting the p~morbid intelleetuaJstatus of childrenusing demographicdata. ClinicalNeuropsychology,L 3G38. Ruddle, H. V., & Bradshaw,C. M. (1982). On the estimationof premorbid intellectualfunctioning:Validationof Nelson & McKenna’s formula and some new normative data. British Journal of Clinical Psychology,21, 159-165. Ryan, J. (1984). Abnormality of subtest score and verbal-performanceIQ differences on the WAIS-R. Zrrternational Journal of Clinical Neuropsychology,6, 97-98. Ryan,J. J., & Paolo,A. M. (1992).Ascreerringprcm&rreforestimating premorbidintelligencein the el&rly. The Clinical Neurupsychologist,6,53-62. Ryan, J. J., Paolo, A. M., & Bmngardt, T. M. (1990). Standardizationof the WAIS-R for persons 75 years and older. PsychologicalAssessment:A Journal of Consultingand Clinical Psychology,2,404-411. Ryan,J. J., Paolo,A. M., &Dnnn, G. E. (1995).Analysisof aWAIS-Rold-age normativesampleon terms of gender,years of education, and preretirementoccupation.Assessment, 2, 225-231. Ryan, J. J., & Prifitera,A. (1990).The WAIS-Rindex for estimatingpremorbidintelligence:Accuracy of predicdng short form IQ. The IntematiortalJournal of Clinical Neumpsychology, 12,20-23. Satz, P.,Van Gorp, W. G., SoPer,H. V., Mitmsbina,M. (1987).WAIS-Rmarker for dementiaof the Alzheimertype?:An empirical and statisticalinduction test. Journal of Clinical and ExperimentalNeumpsychology, 9, 767–774. Scblosser,D., & Ivison,D. (1989).Assessingmemory deteriorationwith the WeehslerMemory Scafe,the NationalAdrdt Reading Test, and the SchonellGmded Word Reading Test. Journal of Clinicaland EaperinwntalNeuropsychology, 11,785-792. Schlottmarm,R. S., & Jobnsen,D. E. (1991).The intellectualcorrelatesscale and the predictionof premorbidintelligence in brain-damagedadults. Archives of ClinicalNeurvpsychology,6, 363–374. Scblosser,D., & Ivison, D. (1989).Assessingmemory deteriorationwith the WechslerMemory Scale, the NationalAdult Reading Test, and the SchonellGraded Word Reading Test.Journal of Clinicaland Expen”mentalNeuropsychology, 11,785-792. Schlottman,R. S., & Johnsen,D. E. (1991).The IntellectualCorrelatesScale and the predictionof premorbidintelligence in brain-damagedadults.Arehives of Clinical Neumpsychofogy, 6, 363–374. Schonell, F. J. (1942). Backwardnessin the basic subjects. Edinburgh:Oliver & Boyd. Sellers, A. H., Bums, W. J., & Guyrke, J. S. (1996). Prediction of premorbid intellectualfunctioningof young children using demographicinformation.Applied Neuropsychology,3, 21–27. Sharp, K., & O’Crrrroll,R. (1991).Estimatingpremorhidintclleetnallevel in dementiausing the NationalAdult Reading Test: A Canadian study.British Journal of Clinical Psychology,30, 381–384. Spreen, O., & Strauss, E. (1991).A compendiumof neurupsychologicaltests:Administration, norms, and commentary. New York: Oxford University Press.
M D. FranZen, E. J. Burgess, and L. Smith-Seemiller
Starr, J. M, Whalley, L. J., Inch, S., & Shering, P. A. (1992). The quantification of the relative effects of age and NART-PredictedIQ on cognitive function in healthy old people. InternationalJournal of Geriatric Psychiatry, 7, 153-157. Stebbins, G. T., Gilley, D. W., Wilson, R. S., Bernard, B. A., & Fox, J. H. (1990). Effecfi of language disturbanceson premorbid estimates of IQ in mild dementia. The ClinicalNeuropsychologist,4, 64+8. Stebbins,G. T., Wilson, R. S., Gilley,D. W., Bernard,B. A., & Fox, J. H. (19!)0).Use of the NationalAdult ReadingTest to estimate premorbid IQ in dementia. The ClinicalNeuropsychologist,4, 18-24. Steinberg,R. J. (1990).Metaphorsofmind: Conceptionsof the nalureof intelligence.Cambridge:Cambridge University Press. Storandt, M., Stone, K., & LaBarge, E, (1995).Deficits in reading performancein very mild dementiaof the Alzheimer type. Neumpsychology, 9, 176176. Sweet, J. J., Moberg, P. J., & Tovian, S. M. (1990). Evaluationof WechslerAdult IntelligenceScale-Revisedpremorbid IQ formulas in clinical populations.PsychologicalAssessment: A Journalof ConsultingandclinicalPsychology,2,
4144. Tuokko, H., & Crockett,D. (1987).CentralcholinergicdeficiencyWAJSprofilesin a nondementedaged sample.Journal of ClinicalNeuropsychology,9, 225–227. Vanderploeg,R. D. (1994).Estimatingpremorbid level of functioning.In R. D. Vanderploeg(Ed.), Clinician’sguide to neuropsychologicalassessment (pp. 43-68). Hillsdale, NJ: Lawrence Erl.baumAssociates. Vanderploeg,R. D., & Schinka,J. A. (1995).PredictingWAIS-RIQ premorbid ability: Combining subtest performance and demographicvariables predictors.Archives of ClinicalNeuropsychology,10, 225–239. Vanderploeg,R. D., Schinka, J. A., & Axelrod, B. N. (1995). Validationof [email protected]
Schinka premorbid IQ prediction srrategy. Paper presented at the annual meeting of tJreAmerican PsychologicalAssociation,New York. Wechsler,D. (1958). The measurementand appraisal of adult intelligence(Mr cd.). Baltimore:Williams and Wilkins. Wechsler,D. (1981). WechslerAdult IntelligenceScale- Revised: Manual. San Antonio, TX: The PsychologicalCorporation. Wedding, D., & Faust, D. (1989). Clinical judgment and decision making;in neuropsychology.Archives of Clinical Neuropsychology,4, 233–265. Wiens, A. N., Bryan, J. E., & Crossen, J. R. (1993). Estimating WALS-RFSIQ from the National Adult Reading Test-Revisedin normal subjects. The Clinical Neuropsychologist,7, 70-84. Wiggins, J. S. (1973).Personalityand prediction:Principlesofpersonality assessment.Reading, MA: Addison-Wesley. Wilkinson, G. S. (1993). Wide Range Achievement Test-3manual. Wihnington, DE: Wide Range, Inc. Willshire, D., Kinsella, G., & Prior, M. (1991). Estimating WAIS-R IQ from the National Adult Reading Test: A cross-validation.Journal of Clinical and Expen”mentalNeuropsychology,13, 2&h216. Wilson, R. S., Rosenbaum, G., & Brown, G. (1979). The problem of premorbid intelligence in neuropsychological assessment.Journal of ClinicalNeurupsychology,1, 49-53. Wilson, R. S., Rosenbaum, G., Brown, G., Rourke, D., Whitman, D., & Grisell, J. (1978), An index of premorbid intelligence.Journal of Consultingand Clinical Psychology,46, 1554-1555. Yates,A. (1956). The use of vocabukwyin the measurementof intellectual deterioration:A review. Journal of Mental Science, 102,409-440. Zachary,R. A., Cnrmpton,E., & Spiegel,D. E. (1985).EstimatingWAIS-RIQ from the ShipleyInstituteof Living Scale. Journal of Clinical Psychology,41,532-540.