Frontal Assessment Battery

Abbreviation: FAB

The Frontal Assessment Battery was constructed in 2000 by Bruno Dubois and colleagues and described in the journal Neurology.¹⁾ The authors' aim was to develop a brief test battery, administrable at the bedside, for assessing the “dysexecutive syndrome” in adults. The authors were aware that the clinical assessment of cognitive functions of the frontal cortex is difficult. Impairments of frontal lobe function can manifest themselves in many different ability impairments, which can be examined only by a variety of psychological test procedures. Nevertheless, the authors considered it advantageous to have available, at the very beginning of the search for frontally caused ability impairments, a brief procedure that allows a certain amount of screening at the patient's bedside. As fields of application they saw frontotemporal or vascular dementias, special forms of Parkinson's disease and conditions following brain injury.

The reception so far (in the Web of Science Core Collection there are 3291 papers in July 2026 that cite the FAB article) quite justifies the authors and shows that there is a need for such a procedure. In their original publication, Dubois and colleagues gave only rough cutoff scores, similar to what the authors of the MMSE and MoCA did. Experienced users, for whom handling the raw scores of such screening tests (and the tests themselves) is everyday practice, do not need standardized scores. Users who are not (yet) so familiar with them, on the other hand, benefit from age-dependent norming, which facilitates clinical judgment. Since about 20 of the subsequent papers provide detailed normative data for the FAB, it is now possible to include the FAB in TDB2 and TDB2Online and to report standard scores that take age into account.

The FAB consists of six subtests, namely (1) verbal concept formation, (2) lexical word fluency, (3) motor programming, (4) susceptibility to interference, (5) inhibitory control and (6) environmental autonomy. In each subtest between 0 and 3 points are awarded, which are added to a total score (0 to 18).

The FAB is completely described in the article mentioned.²⁾ On the basis of this work, a three-page record sheet including the corresponding German instructions was created for in-house clinical use, which we make available here as a pdf file. There is also a more concise German version, published by Benke et al. in 2013 in the Journal of Neurology.³⁾ The versions differ in details, which of course also applies to the other language versions that we took into account in creating the norms. Only the choice of the letter in the second subtest is critical (in the English original, as in the two German proposals, an “S”). All authors of other language versions have taken this into account, however, and chosen frequently occurring letters.

In their article in Neurology,⁴⁾ Dubois and colleagues describe very precisely how the testing is to be carried out. No explicit copyright on the procedure is mentioned. Implicitly it probably lies with the publisher of the journal. So far neither the use nor the translation of the procedure seems to have been legally restricted. (This does not mean that this cannot happen in the future; see the situation with the MMSE.)

Administration and scoring of the FAB are described with sufficient precision in the appendix of the original paper. It will rarely take more than 10 minutes.

Between 0 and 3 points can be achieved per task. After each task it is noted which answer counts for how many points. Scoring can be carried out directly after each task. The raw score consists of the sum of points over all 6 tasks.

Like the MMSE and MoCA, the FAB does not measure a normally distributed cognitive ability but grades a sample of executive cognitive abilities from a “normal state” down to a very low level of performance. The upper end of the measurement range is arbitrarily fixed at a “healthy, unimpaired” level, the lower end at the loss of executive abilities. One must therefore expect that the raw score distributions in healthy persons will be negatively skewed.

We found no published raw score distributions from random samples for the FAB. All norming studies were carried out more or less on convenience samples, some of them at least stratified according to population figures, usually for gender, age and school education, with gender never playing a role for the test results. The most extensive stratified study⁵⁾ with a total of 2508 participants reported, among other things, the raw scores at 7 percentile points (the percentile ranks 5, 10, 25, 50, 75, 90 and 95) for three education classes and 10 overlapping age groups (with means 65, 68, …, 89 and 92). From these data a raw score distribution could be back-calculated approximately. From the percentile values we calculated the corresponding standard scores for the normal distribution, smoothed them with an exponential fit, averaged them over the three education classes, interpolated the values for the age group means used in TDB2, 62 (extrapolated), 67, …, 87 and 92, and extrapolated the standard scores for raw scores below 9 (raw scores below 9 hardly occurred among the healthy persons of this study and lay far below the 5th percentile). Figure 1 shows the resulting distributions.

Figure 1: Back-calculated distributions of the FAB raw scores in the study by Coen et al. (2016) for seven age groups

The figure shows the negative skew of the distribution: with a maximum raw score of 18, the mode in all age groups is at a raw score of 16 or 17. Values below 9 hardly occur in the younger participants (in this study, the 60-64-year-olds). The measurement range is very limited at the upper end (as with the MMSE and MoCA this is intentional: the full score is meant to be reached by “healthy” persons, not only by the especially gifted, as in ability tests that are intended to cover the whole performance spectrum). But one also sees that FAB scores decline systematically with age, and a single cutoff score for separating “healthy” from “abnormal” across all age groups would not do justice to the data. It is exactly the same with the MMSE and MoCA.

The study by Coen et al. (2016) is well suited for the linearization of the scale and we used it for that purpose.

In the original publication of the FAB from 2000⁶⁾, test results were reported only for a small, unspecified group of healthy persons and for various clinical groups. On the basis of these samples, the reliability and validity of the FAB were analyzed; no recommendations for cutoff scores are given.

In a literature search in May/June 2026 we found 19 papers that claimed to provide norms for the FAB. The studies are of varying quality, from countries at different levels of development and from different language regions. 14 of the papers give a - sometimes only rough - breakdown of the norms by age. All 14 give norms for the age range from 60 to 79, 10 take into account the age range up to 89 years, and 9 papers also give norms for younger persons. These 14 papers are characterized in more detail in the following.

The paper by Coen et al. (2016)⁷⁾ was already briefly touched upon above because we used it to derive the raw score distribution of the FAB. It arose as part of an Irish study called Trinity, Ulster, Department of Agriculture (TUDA), in which persons aged 60 years and above who lived at home and attended outpatient clinics or general practitioners were examined. For the FAB norming study only persons were included who scored more than 26 points on the MMSE, did not suffer from depression or an anxiety disorder and showed no evidence of previous strokes or TIAs. A total of 2508 persons in the age range from 60 to 95 met these criteria. Besides the percentile results already mentioned above, means and standard deviations are also given in three school education groups (primary, secondary and tertiary) for 10 overlapping age ranges at 3-year intervals, which could be converted to the 5-year groups customary in TDB2 and averaged, weighted over the education groups. With an average of 12.7 years of school education, the participants of this study are at a very high level compared with the other studies. Obviously the selection process led to a rather well-educated sample.

Ilardi et al. (2022)⁸⁾ examined a convenience sample of 1241 persons in central and southern Italy. The persons had to be older than 17, have an MMSE > 25 and have had at least 3 years of formal schooling. Mean school education was 10.55 years. The authors found that subtest 6 of the FAB contains practically no variance because almost all persons receive 3 points here. This observation is correct and has also been mentioned by many other authors. Ilardi and colleagues then reported their results, however, only for the first 5 subtests, which they called “FAB15”. They also used a regression method to create correction formulas for “Equivalent Scores”. At least the regression equations used are given, with which all this can be reversed. As usual, we included the raw scores, averaged with weighting over the education classes, in the meta-analysis of the norms. By adding 3 raw points we again arrived at the expected result of the normal FAB. More on this in the discussion.

Hurtado-Pomares et al. (2024)⁹⁾ examined 798 persons aged 19 to 91 years with the FAB. The persons were approached on the street and entered the sample stratified by age, school education and sex. Besides other inclusion criteria concerning physical and mental health, the most important was an MMSE of 28 or more. The study reports means and standard deviations for six age groups between 20 and 80 years and for three education groups. The means and standard deviations of the study, averaged with weighting over the education groups, were included in the meta-analysis. Means for school education are not available.

From Slovakia, Abrahamova et al. (2022)¹⁰⁾ have presented a norming study on the FAB. The investigation was carried out mainly in Bratislava. The MMSE had to be greater than 23 and there must be no other illness that could impair cognitive performance. The mean educational level was very high at 13.9 years. Persons aged 18 and above were included in the study. Among other things, means and standard deviations are reported for 7 age groups and two education groups. We averaged them over the education groups.

Another study from Italy, by Aiello et al. (2022)¹¹⁾, likewise provided norms for the FAB, this time from northern Italy. 475 healthy adult examinees without signs of a mental or physical disorder were included in the study. The mean educational level was 11.7 years. As in the study by Ilardi, the results here too were processed by regression analysis and converted into “Equivalent Scores”. Since the regression equations were given, everything could be calculated back to raw scores. The age-appropriate weighted means could thus be included in the meta-analysis.

A study by Benke et al. (2013)¹²⁾ comes from Austria. Here 401 cognitively healthy persons between 50 and 95 years were examined with a German version of the FAB. At 13 years, the mean educational level was again relatively high. The MMSE had to be above 23. The results are available as means and standard deviations for four 10-year groups and two education groups and were averaged over the two education groups for TDB2.

A third Italian study is by Appollonio et al. (2005)¹³⁾. Here 364 persons aged 20 to 94 years were examined with the FAB; the mean educational level was 10.4 years. The results are available as means and standard deviations for eight 10-year age groups and five education groups (classified by years of education) and were averaged with weighting over the education groups.

A relatively recent study from Chile is available by Grandi et al. (2022)¹⁴⁾. The authors examined 344 persons with a mean age of 55, a mean educational level of 12.6 years and an MMSE > 24 points, as well as further exclusion criteria, with a Spanish version of the FAB that is printed in the appendix of the paper. All examinees were examined in a dementia clinic in Santiago; they probably came from the city or its immediate surroundings. The norms are available, among other things, as means and standard deviations for 3 relatively coarse age groups and 3 education classes. They were averaged with weighting over the education groups for TDB2.

Lima et al. (2008)¹⁵⁾ examined 122 healthy adult persons with a Portuguese version of the FAB, with a mean educational level of 8.7 years and an age between 20 and 81 years. Means and standard deviations of four coarse age groups and four education classes were averaged with weighting over the education classes for inclusion in TDB2.

The nine studies mentioned so far can all be used for a meta-analysis of the normative data of the FAB. For the five studies described in the following, there are in each case one or more obstacles to inclusion.

Salinas-Lopez et al. (2026)¹⁶⁾ have recently presented a study from Chile that has so far been published only online. A relatively large convenience sample of 1022 healthy persons aged between 56 and 101 years, all able to read and write, was examined with the FAB with the help of a senior citizens' organization. Besides many psychometric results on the FAB-15 (that is, without the 6th subtest, which was omitted in all analyses although it was administered), the study also contains the means and standard deviations for two education groups and five age groups. A weighted averaging over the education groups was not possible because the number of persons per combination of education and age is not given. As in Ilardi et al. (2022), 3 points were added to the means of the FAB-15. The test scores of this study are clearly below those of the studies discussed so far, especially in the older persons of the sample. They are also below the values of the study by Lima et al. (2008), which likewise comes from Chile. More precise information on the educational background of the examinees is lacking. The standard deviations of the age groups are clearly higher than in the studies mentioned so far. This gives us reason not to include the study in the meta-analysis.

Kim et al. (2010)¹⁷⁾ translated the FAB into Korean and used it to test, among others, 635 healthy persons aged over 54 from Korea. The participants belonged to a Korean longitudinal study. In this study the Clinical Dementia Rating¹⁸⁾ was also carried out. Persons of the normative sample were not allowed to have a CDR score > 0. The participants had on average 9 years of education. Means and standard deviations are available for four education groups and three age groups. If they were averaged over the education groups, values were obtained that lay outside the values of studies 1 to 9, which is not unexpected given the low mean educational level. We did not include this study in the meta-analysis.

From Brazil (de Paula et al., 2013)¹⁹⁾ there is a study of 391 healthy persons aged 60 to 99 years from Belo Horizonte. In this study too, the average educational level was only 6.5 years, and the mean FAB scores were clearly below the values of studies 1 to 9.

From Wang et al. (2016)²⁰⁾ there is a study from Taiwan. Here 301 healthy persons aged between 15 and 86 years were examined with a Chinese version of the FAB. For inclusion in the meta-analysis it is critical that - although the mean duration of education was 12.8 years - the older persons over 60 in particular often had only a few years of education. Accordingly, the FAB means in this study were roughly at the level of studies 1-9 for the young up to about age 40 (as far as those studies included such young persons at all), but for the older persons over 60 they were far below. This study was therefore also excluded from the meta-analysis.

Another study is available from Iran. Asaadi et al. (2016)²¹⁾ examined a convenience sample of 149 healthy persons aged between 48 and 81 years with the FAB. Persons with fewer than 5 years of education or an MMSE below 24 were excluded. On average the sample had an educational level of 11.3 years. The norming results are available in the form of mean and standard deviation for three age groups and four education groups. If one compares the results with studies 1 to 9, the means lie at the lower margin of these studies, but the standard deviations lie far outside all the other studies. Since in addition the age groups are only very coarse and not differentiated above 70 years (the most interesting range for the FAB), we also omitted this study from the meta-analysis.

Figure 2 shows the means of all the studies mentioned over the course of age in five-year periods from 20 to 94 years (green: the nine studies used for the TDB2 norms with their mean in black; orange: the five samples not used). Since the FAB was also normed for younger adults in many studies, we show the entire age course of all studies. As with all procedures used predominantly in older age, however, we will relate the performance scores to the age group 60 to 64 for the sake of simpler interpretation. Since the case numbers in the studies are again very unequal, we did not calculate the means across the studies with weighting, because otherwise individual studies would have received too great a weight.

Figure 2: Means of the FAB raw scores across age in nine studies used for the meta-analysis (green) with their mean (black) and five unused studies (orange)

Compared with other tests, the norming studies on the FAB include many from less developed countries. In these, the younger persons in the sample have school education trajectories similar to those in developed countries, but the older persons do not. In cross-section (with test norming we are always dealing with cross-sectional studies) this manifests itself in different age trajectories with greater losses in older age. For use in developed countries, including such data is not sensible.

Figure 3 shows, in the same way as Figure 2, the standard deviations of the 14 studies, as far as they were reported. Here, as described above, the study by Asaadi stands out through implausible values.

Figure 3: Standard deviations of the FAB raw scores across age in nine studies used for the meta-analysis (green) with their mean (black) and five unused studies (orange)

We used the study by Coen et al. (2016) mentioned above for the linearization of the measurement scale. Since the means of this study are somewhat higher across the age groups than the means of studies 1 to 9, on which we base the norms, a correction factor was applied. Figure 4 shows the performance scores for the FAB raw scores from 5 to 18, based on an percentile-to-normal transformation of the percentile values of the study by Coen et al. (2016).

Figure 4: Percentile-to-normal transformation of the FAB raw scores into IQ-scaled standard scores for the 60-64-year-olds of the study by Coen et al. (2016)

see the detailed test documentation

The FAB tests important and fundamental executive functions that are normally available to healthy persons of almost any age. In the original paper, the FAB correlated relatively highly, in a larger (N=120) heterogeneous sample, with the Wisconsin Card Sorting Test and the Mattis Dementia Rating Scale, both of which are sensitive to disorders of executive functions. The correlations with patients' age or the MMSE, by contrast, were low, which shows that the executive functions captured by the FAB hardly enter into the MMSE score.

In healthy persons practically only the range of values from about 9 to 18 occurs; lower values occur only in people with severe executive impairment. Below a FAB raw score of 9 one is therefore moving in an ability range in which population-standardized measured values can no longer find any meaningful use. The conversion of such low raw scores into performance scores given in TDB2Online is merely extrapolated and is no longer based on empirical data.

In two of the available studies (and those used for the norming), a “FAB-15” was used instead of the usual FAB (which in this notation would be a “FAB-18”). In the last subtest of the FAB, that is, in the assessment of environmental autonomy, practically all healthy persons reach three points. This subtest is so “easy” that on the one hand it has hardly any variance and interferes with measurement (e.g., in factor analyses), and on the other hand very rarely contributes anything to measurement. We simply added 3 points in each case to the FAB-15 results in the two studies. Only very severely impaired persons reach fewer than 3 points in this subtest.