Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
The MBTI is a flawed measure of personality
- Paul R Matthews (21 May 2004)
-
Re: The MBTI is a flawed measure of personality - response to Paul Matthews' criticism
- Gillian B Clack, Judy Allen, The Macklin, Mansel Lacy, Herefordshire, HR4 7HQ. (19 June 2004)
-
Response to Clack & Allen [4]
- Paul R Matthews (23 August 2004)
|
|
|||
|
Paul R Matthews, Graduate Student Dept. Psychiatry, Uni. Oxford, Warneford Hospital OX3 7JX
Send response to journal:
|
Re: Clack et al (2004) This study utlises the Myers-Briggs Type Indicator (MBTI)®, a proprietary categorical classification of personality according to a Jungian typology along four independent axes, Extroversion-Introversion, Sensing-iNtuition, Thinking-Feeling and Judgement-Perception. Each dimension has a zero point such that those falling on either side are considered to be qualitatively different. This leads to 16 different and exclusive personality types with which an individual can be labelled (ENTP, ISFJ etc). An insight into the MBTI approach can be conveyed by this quotation from advocates Tieger and Barron-Tieger (1993): "A major premise of the Type model is that only one of the 16 Types best describes each person — the Type to which you are born will be the one you take to your grave. We may adjust our behaviors over time — or at a party versus a funeral — but our personal Type remains the same for life." The MBTI may well be "the most widely used personality questionnaire in the world" but its substantial limitations as a measure of personality, as well as a tool for career counselling, are too numerous to list here (see Pittenger 1993) but I would just like to highlight a few of the most relevant. The idea that there are 16 qualitatively different personality types as revealed by the MBTI is demonstrably false (McRae and Costa 1989; Stricker and Ross 1964). The only validity in the MBTI is in the continuous personality dimensions used which are related to mainstream psychometric personality measures (McRae and Costa 1989), however, in practice, the MBTI is used as a typology, not as dimensions. The MBTI personality dimensions seem to be distributed such that most people fall between the two extremes around the boundary point of the dichotomous category distinctions. Therefore people who are very small distances apart on the dimensional measure are categorised as being qualitatively different to one another because they fall either side of the cut-off point between the types and are lumped in with much more extreme scores that fall on the same side of the cut-off however far apart they are on the dimension. This flaw in the MBTI, as a measure of personality, undermines the significance of the conclusions of Clack et al as it is not possible to establish the magnitude of difference between doctors and their patients on the personality dimensions. While it seems a good idea to teach trainee doctors to try and "ameliorate the potential difficulties resulting from such personality differences, thereby improving the outcome of the interaction" it would be unhelpful to include the MBTI and personality type in the communication skills training of health care professionals, given its limited validity and potential for misuse (Pittenger 1993). Clack, G.B., Allen, J., Cooper, D. and Head, J.O. Personality differences between doctors and their patients: implications for the teaching of communication skills. Medical Education, 38(2):177-186, 2004. McRae, R.R., and Costa, P.T. Reinterpreting the Myers-Briggs Type Indicator From the Perspective of the Five-Factor Model of Personality. Journal of Personality, 57(1):17-40, 1989. Pittenger, D.J. Measuring the MBTI ... and coming up short. Journal of Career Planning and Employment, 54:48-53, 1993. Stricker, L.J., and Ross, J. An Assessment of Some Structural Properties of the Jungian Personality Typology. Journal of Abnormal and Social Psychology, 68(1):62-71, 1964. Tieger, P.D., and Barron-Tieger, B. Personality Typing: A First Step to a Satisfying Career. Journal of Career Planning and Employment, 53:50- 56, 1993. Competing interests: None declared |
|||
|
|
|||
|
Gillian B Clack, Honorary Senior Lecturer Division of Medical Education, GKT School of Medicine, Guy's Hospital, London, SE1 9RT., Judy Allen, The Macklin, Mansel Lacy, Herefordshire, HR4 7HQ.
Send response to journal:
|
EDITOR: In his letter of 21 May, Matthews questions the validity of the MBTI(R) as a measure of personality compared to the Five-Factor Model, citing articles published over a decade ago. He then argues, on the basis of these, that the significance of the conclusions contained in our paper on how personality differences can affect the efficacy of doctor/patient communication, together with our recommendations for medical education, are undermined.1,2 Whilst it is, of course, true that all current measures of personality have some potential for measurement error, more recent literature provides substantial evidence for the validity of the MBTI(R) when compared to other widely-used instruments, including full discussion of the issue of bimodality raised by Matthews.3,4 It has, indeed, recently been argued that the MBTI(R) is not only a valid and reliable measure of personality differences but actually measures something extra to the 'Big Five'.5 We would, therefore, like to reassure readers that it is a both powerful and useful instrument and that our paper should be given serious consideration. Yours sincerely, Gillian Clack and Judy Allen References: 1 Matthews PR. The MBTI is a flawed measure of personality. http://bmj.bmjjournals.com/cgi/eletters/328/7450/1244#60169. 2 Clack GB, Allen J, Cooper D, Head JO. Personality differences between doctors and their patients: implications for the teaching of communication skills. Medical Education 2004; 38(2): 177-186. 3 Hammer AL. MBTI(R) applications: a decade of research on the Myers- Briggs Type Indicator(R). Palo Alto, CA: Consulting Psychologists Press, Inc., 1996. 4 Myers IB, McCaulley MH, Quenk NL, Hammer AL. MBTI(R) manual: a guide to the development and use of the Myers-Briggs Type Indicator(R) (3rd Edition). Palo Alto, CA: Consulting Psychologists Press, Inc., 1998. 5 Edwards JA, Lanning K, Hooker K. The MBTI and social information processing: an increment validity study. Journal of Personality Assessment 2002; 78: 432-50. Competing interests: We are authors of the article criticised by Matthews |
|||
|
|
|||
|
Paul R Matthews, Graduate Student Dept. Psychiatry, Uni. Oxford, Warneford Hospital OX3 7JX
Send response to journal:
|
I will divide my response into two parts; the first will concern the validity of the MBTI, and the second the specifics of the study [5]. Clack and Allen [4] dismiss my criticisms of the MBTI as a measure of personality, because I refer to “articles published over a decade ago”. They cite Hammer [6] and Myers et al [14] as “more recent literature” providing “substantial evidence for the validity of the MBTI® when compared to other widely-used instruments, including full discussion of the issue of bimodality raised by Matthews.” This is simply not the case. I have previously alluded to the failings of the MBTI as a personality measure [12], and I will go into them in more detail here, given that the BMJ has previously painted the MBTI in a positive light [10]. The fundamental weakness of the MBTI, that must be appreciated, is that it classifies people together who have very different scores on the continuous dimensional measures, and then uses this categorisation to infer characteristics of those individuals. At the same time it contrasts those falling either side of the cut-off, but scoring closely numerically, as qualitatively different. For this approach to be valid the sixteen different personality types in the MBTI should contain some predictive power over and above the continuous dimensions they dichotomise – that is they should tell us more than simply that someone scored above or below the cut-off score on the continuous dimension. This is a view endorsed by the designers of the instrument: “The eight characteristics that are defined in the MBTI are not traits that vary in quantity; they are dichotomous constructs…” [14]. “The Indicator does not attempt to measure degrees of preference; according to the theory the opposites of the dichotomies are qualitatively different categories…” [11]. They are also meant to represent more than simply the sum of the four dimensional categorisations, they are supposed to interact: “…the 16 types are greater than the sum of their parts” and “the dynamic characteristics of each type are not reflected in a simple summation of an individual’s four preferences” [14]. Bimodality of the underlying preference scores would suggest two different populations within a dimension, and that, therefore, the dimensional score might be a good way to classify someone into one of the two populations. At the very least bimodality reduces the numbers falling in the cut-off region and improves certainty in classification. However preference score methods of determining the MBTI dichotomy are not bimodal [e.g. 7, 13, 16], a point conceded by Harvey [8] in Hammer [6]. It originally seemed that a novel classification method, item response theory (IRT) scoring, might be somewhat bimodal [7] but this has proven to be an artefact of the methods used to estimate likelihood functions [1]. The venerable, four decade old, study by Stricker & Ross [16] cited, is actually highly relevant, since it was published following the first (unfavourable) review of the MBTI as a testing instrument [15]. This study of large, unselected samples found no evidence of bimodality in the continuous preference score distributions in any dimension of Form F of the MBTI. The decade old study of McRae & Costa [13] found the same using MBTI Form G, and also failed to find any evidence for interactions between dimensions (the dominant and auxiliary preferences predicted by Jungian theory). Harvey & Murray [7] found that preference scores were distributed in a centre weighted, platykurtic manner and Bess & Harvey [1] showed this to also be the case with IRT scoring. It has been proposed that non-linear relationships between preference scores and external variables might validate the dichotomous classification approach, and Myers et al [14] cite some examples of this. However there are many reasons for non-linear correlations between two variables (e.g. floor and ceiling effects). The IRT approach to scoring the MBTI is actually based upon non-linear relationships between total MBTI scores and individual test items. In summary, there is no evidence for the 16, qualitatively different, personality types: “In the absence of evidence for the typology, the instrument becomes merely a series of scales whose information is reduced, rather than increased, by dichotomous classifications” [13]. Consider the example of classifying people into ‘short’ (less than average height) and ‘tall’ (greater than average height). This is not a very meaningful categorical division of a normally distributed continuous variable. ‘Short’ people in the region of the cut-off do not have anything intrinsically about them that is qualitatively different from ‘tall’ people around the cut-off. That is, while ‘tall’ people as a group will be better able to reach things in high cupboards, this is a result of the underlying continuous variable (height), being greater on average than for ‘short’ people. The large number of ‘tall’ people in the region of the cut-off would be very little better able to reach high cupboards than the very many ‘short’ people around the cut-off – i.e. any predictive properties of the ‘tall’/’short’ dichotomy supervene upon the underlying continuous dimension. This means that, because the underlying dimension is continuous, and the majority of people score around the region of the cut-off, most people identified as being qualitatively different will actually differ very little in their scores on the continuous dimension. The continuous MBTI dimensions have reasonable split half reliability ~.84, and proponents of the MBTI have been keen to imply that this validates the MBTI [e.g. 8]. However, these test-retest measures are highly sensitive to inter-test interval (ITI). With an ITI of less than nine months test-retest reliability is ~.80, but over nine months it is ~.65 [8]. This is not what we would expect of a trait that is supposedly stable over time! In fact, you cannot judge the reliability of the MBTI typology by looking at the reliability of the continuous dimensions, because the typology dichotomises them. Similarly, attempts to validate the MBTI typology through correlating the dimensional scores with other personality scales (e.g. the five factor model) [8, 14] do not provide validity for the dichotomous classifications, only the continuous dimensions [13]. The dichotomous classifications are actually much less reliable than the measures for the continuous dimensions would imply. This is because using mid-distribution dichotomous cut-offs actually requires even more reliable continuous measures than trait instruments [8]. Only about 50% tested within nine months score the same on all four dimensional dichotomies (i.e. remain the same type) and around 36% remain the same after nine months. Within each scale ~83% retain the same categorisation when retested within nine months, and ~75% when tested after nine months [8]. This is not good for a test supposed to detect categorical type, fixed over a lifetime, and undermines the use of a typological classification, particularly given the many revisions to the scoring system. Form M of the MBTI, scored by IRT, is reported to show an overall type agreement with the previous Form G of 60% [14]. This type instability is probably due to uncertainty of type classification around the cut-off for the dimension, combined with the heavily centre weighted preference score distribution. In true-type studies (where the personal evaluation of the MBTI type is compared with the score allocated MBTI type), Carskadon & Cook [2] found that 50% of people picked their MBTI profile, while 13% picked the completely opposite profile. There are some specific reservations I have about Clack et al [5]. The first concerns the rationale of the study. Since the cohort of medical graduates used is not expected to represent the whole population of doctors, presumably we are asking how a given cohort of doctors differs in type from their patients. This is a curious question; how does the study finding, that the frequency of personality types in one cohort of doctors differs from the population as a whole, tell us anything more than the obvious fact that any given doctor will have a personality type different to that of many of their patients? The response rate (67.5%) is quite low; it would be interesting to know how responders differ from non-responders on personality measures, and whether this confounds the results. The authors do not make it clear which version of the MBTI is used in this study, presumably the MBTI Step 1 (European English Edition) given the population norms described [11]. Differences in data collection between the sample surveyed in this study, and the norms they were compared to [11] should have been discussed (e.g. subjects who took part in the population questionnaire received a monetary incentive, this may have skewed the distribution of personality types responding). Since the number of statistical comparisons using the MBTI is potentially very large (two genders, four dichotomous dimensions plus combinations of dimensions, e.g. Sensing with Thinking), I count 24 comparisons in the paper, and presumably more were done [3], it is not clear how this study controlled the error rate. Given the limitations of the categorical MBTI I have outlined above, for any validity the means and standard deviations of the preference scores should be reported, not type and frequency. Looking at the centre weighted distributions of MBTI preference scores, it is clear that small differences in the underlying mean scores could give rise to large differences in the frequencies of the dichotomous classifications. If we look at an example of an MBTI preference score distribution [EI dimension, Form F; from ref. 9] we can see that a difference of 10 points in the mean scores on the continuous dimension could give rise to the sort of differences in category frequencies seen in Clack et al [5]. That is a difference in means of 10 points on a scale that runs from -53 to +59 points. When the frequency differences found in Clack et al [5] are recast in terms of small differences in the mean preference scores, it is unclear how significant these differences are with respect to doctor- patient interactions. In conclusion, the unsound foundations of the MBTI and type theory undermine the significance of the findings in this study and argue against its recommendations [12]. 1. Bess, T.L. and R.J. Harvey, Bimodal Score Distributions and the Myers-Briggs Type Indicator: Fact or Artifact? Journal of Personality Assessment, 2002. 78(1): p. 176-186. 2. Carskadon, T.G. and D.D. Cook, Validity of MBTI descriptions as perceived by recipients unfamiliar with type. Research in Psychological Type, 1982. 5: p. 89-94. 3. Clack, G.B., Is personality related to doctors' speciality choice and job satisfaction? 2002, University of London. 4. Clack, G.B. and J. Allen, Re: The MBTI is a flawed measure of personality - response to Paul Matthews' criticism. http://bmj.bmjjournals.com/cgi/eletters/328/7450/1244#63291, 2004. 5. Clack, G.B., et al., Personality differences between doctors and their patients: implications for the teaching of communication skills. Medical Education, 2004. 38(2): p. 177-186. 6. Hammer, A.L., ed. MBTI Applications. 1996, Consulting Psychologists Press: Palo Alto, CA. 7. Harvey, R.J. and W.D. Murray, Scoring the Myers-Briggs type indicator: Empirical comparison of preference score versus latent-trait methods. Journal of Personality Assessment, 1994. 62: p. 116-129. 8. Harvey, R.J., Reliability and Validity, in MBTI Applications, A.L. Hammer, Editor. 1996, Consulting Psychologists Press: Palo Alto, CA. p. 5- 29. 9. Harvey, R.J. and L.A. Thomas, Using Item Response Theory to Score the Myers-Briggs Type Indicator: Rationale and Research Findings. Journal of Psychological Type, 1996. 37: p. 16-60. 10. Houghton, A., Using the Myers-Briggs type indicator for career development. BMJ Career Focus, 2000: p. 320. 11. Kendall, E., The Myers-Briggs Type Indicator (MBTI): [European English Edn.] Manual (Suppl.). 1998, Oxford: Oxford Psychologists Press. 12. Matthews, P.R., The MBTI is a flawed measure of personality. http://bmj.bmjjournals.com/cgi/eletters/328/7450/1244#60169, 2004. 13. McRae, R.R. and P.T. Costa, Reinterpreting the Myers-Briggs Type Indicator From the Perspective of the Five-Factor Model of Personality. Journal of Personality, 1989. 57(1): p. 17-40. 14. Myers, I.B., et al., MBTI Manual: A guide to the development and use of the Myers-Briggs type indicator. 3rd ed. 1998, Palo Alto, CA: Consulting Psychologists Press. 15. Pittenger, D.J., Measuring the MBTI. and coming up short. Journal of Career Planning and Employment, 1993. 54: p. 48-53. 16. Stricker, L.J. and J. Ross, An Assessment of Some Structural Properties of the Jungian Personality Typology. Journal of Abnormal and Social Psychology, 1964. 68(1): p. 62-71. Competing interests: None declared |
|||