The Minnesota Multiphasic Personality Inventory

1
The Minnesota Multiphasic Personality Inventory
2
What is the MMPI?

• The Minnesota Multiphasic Personality Inventory
  is "The most thoroughly researched objective
  personality assessment instrument yet produced"
• Minnesota - Because it was developed there
• Multiphasic - Applied to tests or investigations
  designed to reveal various phases or aspects of
  personality, health, etc.
• Coined by the MMPI designers
• Personality- Because it tests personality traits
• Inventory- List of things

3
Early Personality Tests

• The first personality tests were devised by our
  old friends, Sir Francis Galton (1884) James
  Cattell (1890)
• In WWI, Woodworth (1920) compiled a list of 116
  symptoms of 'psychological nervousness - the
  Woodworth Psychoneurotic Inventory
• This and related tests were composed of symptom
  sets taken from books had high face validity

4
Problems with Earlier Personality Tests

• There were three main problems with early
  inventories
• i.) 'Load' on any trait was considered to be a
  function of the number of items endorsed- a bad
  idea (Why?)
• ii.) Criterion groups were established by
  comparison with extreme responses from other
  similar symptom sets, unvalidated by any
  empirical checks
• iii.) It was assumed that people told 'the whole
  truth and nothing but the truth' and (therefore)
  that their endorsements of items were predictive
  of their real behavior

5
Self-Perception Versus Reality

• - It became apparent that it would be better to
  treat endorsement as subjective opinions, rather
  than as objective facts
• - The Humm-Wadsworth Temperment Scale (1935)
  tried to control for subject temperment and to
  assess reliability by taking a 'no' count
• - Too few or too many symptoms denied was a
  clue about how to read the subject's responses
• - Alas, it was also pretty bad due to poor
  criterion groups

6
"Out of the psychometric wilderness..."

• Despite this failure, Stark Hathaway J.C.
  McKinley (1940) decided to give the empirical
  approach another shot
• Their ambition aimed high but was focused they
  wanted to make one single comprehensive test for
  measuring all behaviors (inferred from items)
  that might be of interest psychiatrically
• It took many years to make all the scales
• Papers were published on the individual scales as
  they were completed, with final versions not out
  until 1956

7
Emphasis on Empirical Validation

• They took as an assumption that the link between
  item endorsement and its significance was a
  purely empirical question (hurray!), rather than
  a matter of theory or opinion
• This meant empirical validation, using
• The population that the test was to be used with-
  psychiatric patients
• A recognition that people told untruths, for many
  reasons
• Item analysis (of sortsbut flawed)

8
Initial Item Guidelines

• Notwithstanding the problems that had beset
  earlier tests, the initial 550 (566) item MMPI
  set was drawn from many of the same old sources
  clinical intuition, the medical (especially
  psychiatric) literature, and even previous tests
• i.) Stated as brief, clear declarative sentences
  (true/false)
• ii.) Written in simple everyday idiomatic
  language
• iii.) Written in the first-person singular
• iv.) Avoiding negatives
• v.) Assuming only common knowledge
• - Takes about 90 minutes to complete

9
Normal Sample

• The test was normed on 1040 relatives and
  friends of hospital patients, students, and
  white-collar workers (any problem with that?), as
  well as 254 non-psychiatric hospital patients
• The group was not very carefully controlled on
  sex, age, or SES

10
Patient Samples

• The beauty of the MMPI design came in the way it
  was validated on pre-chosen patient samples
  (concurrent validity)
• Patients were selected who had relatively pure
  uncomplicated diagnoses of different traits that
  the authors of the MMPI wished to diagnose
• Although it has problems, it is perhaps the best
  clinical assessment tool out there today (others
  disagree!)
• It is also highly researched, found to be used in
  5043 published studies by one 1978 review

11

• "It is essential note that the details of scale
  development have involved many tentative trials
  with subsequent validating studies and the
  finally the adoption of the best scale for
  inclusion."
• McKinley Hathaway, 1944

12
Example 1 Scale 1 Hs (Hypochondriasis)

• Hypochondriasis is characterized by an abnormal
  concern about physical symptoms in the absence of
  any evidence of physical infirmity or illness
• 50 'pure' hypochondriacs were identified
• Items were included in the Hs scale if there was
  significant difference in the frequency of
  agreement between the hypochondriacs and non-
  hypochondriacs
• Difference gt 2 SEproportions of those endorsing
  the item

13
What is standard error of the proportions?

• Recall Standard error of the mean (NOT the SE of
  measurement) is the SD divided by the square root
  of N
• Standard error of the proportions is a way of
  computing the standard error for proportions,
  which is what we have whenever we have true/false
  questions like on the MMPI
• The variance for data with p 1s and q 0s
  simply pq
• So the SD (pq)0.5
• And SEproportions (pq)0.5 / N 0.5

14
What is standard error of the proportions?

• SEproportions (pq)0.5 / N 0.5
• What properties does this have?
• Just like every SE, it diminishes as N increases
• It is smaller at the extremes than in the middle
  (why?)
• p 0.99 q 0.01 (pq)0.5 0.099
• p 0.1 q 0.9 (pq)0.5 0.3
• p 0.5 q 0.5 (pq)0.5 0.5
• It is indifferent to directionality (qp pq)
• P.S. pq is indeed the real variance that is, you
  could also compute it by taking the average
  squared difference from the mean

15
Standard error of the proportions Example

• According to a CNN/USA Today/Gallup survey, on
  October 27, 2004, the Dark Lord ( Dubya Jr.)
  held a slight lead over Kerry in the November,
  2004 presidential election. Among committed
  voters, GWB received 51 support, while JK
  received 46. These results are based on
  interviews with 1,538 adults. What is the SE?
  What is the 95 confidence interval? Should Kerry
  have been worried?
• SEproportions (pq)0.5 / N 0.5
• (0.46 0.51) 0.5 / (1538) 0.5
• 0.012
• The SE is 1.2. The 95 confidence interval is
  2.4 in either direction.
• Should Kerry have been worried?
  We should all have been worried.

16
Hs Item Exclusion

• Some were excluded for lacking face validity
  e.g. obviously based on marital status or
  attitude towards ones children
• Others were excluded by more defensible empirical
  means contrasting a non-hypochondriac patient
  group with high scores against true
  hypochondriacs (also with high scores)

17
Hs Cross-Validation

• The whole scale was then cross-validated by
  trying it on a second set of hypochondriacs,
  patients with physical problems, and normals, and
  showing that the scale differentiated between the
  three

18
Hs Examples

• Positives
• 23. I am troubled by attacks of nausea and
  vomiting
• 161. The top of my head sometimes feels tender.
• 189. I feel weak all over much of the time.
• Negatives
• 2. I have a good appetite
• 7. My hands and feet are usually warm enough
• 190. I have very few headaches
• 243. I have few or no pains.

19
Example 1 Scale 2 D (Depression)

• Depression is (a state rather than trait)
  characterized by depressed feelings, low morale,
  dissatisfaction
• 50 diagnosed depressive patients served as the
  reference group
• There were also a group of depressed normals, and
  normal groups

20
D Item Selection

• Items were included in the D scale if
• They showed a progressive increase in frequency
  of endorsement from the normal through the
  depressed normals to the patients
• The age difference between the normal and
  reference group was 2.5 times the standard error
• 60 items were selected
• 11 of served as correction items to identify
  special conditions (suppressor variables),
  including physical illnesses and religious fervor

21
D Validity

• Normals averaged 18
• Physically ill averaged 22
• Random psychiatric cases averaged 24
• Depressed group averaged 37

22
D Examples

• Positives
• 43. My sleep is fitful and disturbed
• 182. I am afraid of losing my mind
• 259 I have difficulty in starting to do things
• Negatives
• 2. I have a good appetite
• 8. My daily life is full of things that keep me
  interested
• Correctors (Negative)
• 98. I believe in the second coming of Christ
• 154. I have never had a fit or convulsion

23
The Clinical MMPI Scales

• The neurotic triad
• 1 Hs Hypochondriasis (33) Abnormal concern
  with bodily functions
• 2 D Depression (60) Depressive affect
• 3 Hy Hysteria (60) Neurotic symptoms for
  avoiding or hiding from conflict and
  responsibility
• - High religious Bible students also score high
• 4 Pd Psychopathic Deviate (50) Measure
  disregard for social customs
• 5 Mf Masculinity/Femininity (60) Gender roles/
  Homoeroticism

24
The Clinical MMPI Scales

• The psychotic tetrad
• 6 Pa Paranoia (40) Suspiciousness, delusions
  of grandeur and persecution
• 7 Pt Psychasthenia (48) General neurotic
  symptoms, including obsession, compulsion, fear,
  and guilt
• 8 Sc Schizophrenia (78) Bizarre thoughts and
  behavior, delusions, hallucinations
• 9 Ma Hypomania (46) Emotional excitement and
  over-activity
• 0 Si Social Introversion (70) Shyness,
  insecurity, low social interest

25
Other MMPI Scales

• - Over 500 other experimental scales have been
  constructed
• - Most have been little researched, although a
  few are fairly widely used

26
The Question Scale

• 4 scales were designed specifically to enable
  real-time validity checks of individual MMPI
  profiles
• i.) The question (Q or ?) scale Any items
  answered both true and false ( "cannot say")
• - High scores in general neurotic
  ('psychoasthenic') patients
• - This scale has never been well-validated, and
  is now little used

27
The Lie Scale

• 4 scales were designed specifically to enable
  real-time validity checks of individual MMPI
  profiles
• ii.) The lie (L) scale 15 false-scoring items
  that try to catch people attempting to put
  themselves in a good light
• - It was never validated or cross-validated,
  but based on a priori notions
• - A high L scale does not totally invalidate the
  profile, but suggests that the interpretation
  must be cautiously made
• - Sophisticated subjects will not be fooled by it

28
Scale L Examples

• 15. Once in a while I think of things too bad to
  talk about
• 45. I do not always tell the truth
• 75. I get angry sometimes
• 150. I would rather win than lose in a game

29
The Validity/Frequency Scale

• 4 scales were designed specifically to enable
  real-time validity checks of individual MMPI
  profiles
• iii.) The Validity (F Frequency) Scale
• - 64 disparate items answered almost always in
  one direction by normals
• - A high F usually indicates carelessness,
  inattention, or lack of comprehension or
  cooperation, although subjects high on the
  schizophrenic spectrum may also score high on the
  F Scale (and lying is usually considered
  pathological in itself.)

30
The Validity/Frequency Scale Examples

• True
• 27. Evil spirits possess me at times
• 151. Someone has been trying to poison me
• 210. Everything tastes the same
• False
• 65. I loved my father
• 220. I loved my mother
• 272. At times I am full of energy

31
The Correction (K) Scale

• iii.) The Correction (K) Scale
• - Items endorsed by deviants with relatively
  normal profiles false negatives
• - Using a weighting procedure and correction
  items, it was possible to define a means to use
  this to increase validity of many clinical scales
  (Hs, Pd, Pt, Sc, Ma)

32
The Correction (K) Scale

• True
• 6. I have very few quarrels with members of my
  family
• False
• 160. I have never felt better in my life than I
  do now
• 142. I certainly feel useless at times
• 170. I worry over money and business
• 383. People often disappoint me
• 398. I often think "I wish I was a child again".

33
Problems with the Correction (K) Scale

• The K scale is also loaded on SES
  college-educated subjects are 1.5 SDs higher on
  this scale than non-college-educated subjects
• According to John Reddon, the K scale contains
  two negatively correlated subscales
• The weights for the K scale are fairly arbitrary

34
MMPI-2

• The MMPI was re-normed on 2600 people, and
  updated in the 1980s
• All items were retained, but 14 were re-worded
• 154 new experimental items were included, for
  'new' disorders such as eating disorders, Type A
  personality, drug abuse
• The t-scores were fixed up, to be better
  comparable between scales

35
MMPI-2

• Test/re-test r of individual scales .58 to .92
• SE is large
• Nevertheless, studies (mostly on the old MMPI)
  indicate it works to some degree a high score on
  a scale predicts the proper clinical diagnosis
  60 of the time
• Depending on the disorders base rate and the
  uncertainty in diagnosing it by other means, this
  may or may not suffice.

36
Problems with the MMPI

• There was minimal item analysis
• Many scales have overlapping items
• This lowers discriminability (diagnoses are
  correlated), thereby reducing the specificity of
  the diagnoses, and also makes measurement error
  correlated
• The K scales excerbates this, by adding a further
  chunk of overlap to several scales
• The original keying was never cross-validated on
  a large scale- some of the variance is
  group-specific
• The original diagnostic groups were fairly small
  and varied widely in size (N 20- 78)

37
Problems with the MMPI

• Categorical diagnoses are either/or there is
  weak assessment of severity
• Diagnostic practices have changed substantially
  since the scale was formed
• There are no strong tests for suppressor
  variables (e.g. drug use and off-beat belief
  systems are confounded with psychopathology)