Using Graphology to Predict Performance?

I recently read an article in the Atlanta Journal Constitution (Friday, October 15, 2004) about using graphology as a hiring tool. A bespectacled lady wearing a business suit was photographed carefully examining a written document with a magnifying glass. It looked very professional. The article stated that graphology:

Reveals volumes about job applicants’ character and integrity

Gives insights into passions, strengths, weaknesses, and personality
Measures a person’s thought patterns and behaviors
Picks the right person at the right time

Folks, these stories rank right up there with the tabloids’ claims that Elvis lives in a small town in Iowa (we all know it is actually Mississippi) or that “My Dear Friend, Mr. Bio from Ghana” will transfer $1.5 million to me once I send him my bank account number Good Sense and Nonsense Graphologists believe their claims. But it takes replication and confirmation by disinterested third parties before claims become “facts” (remember cold fusion?). As usual, we’ll compare graphology claims with facts drawn from professional research conducted by leading scientists at major universities. In this case, we’ll use the published work of Roy King and Derek Koehler of the University of Waterloo. King and Koehler wanted to determine why otherwise intelligent people persisted in using graphology, even though there was scant, if any, evidence that it predicted job performance (Journal of Experiential Psychology: Applied, 2000, Vol.6, No.4, 336-348). Here are a few of their experimental findings, including some of their background research. Graphology Predicts Personality? Graphologists claim that handwriting predicts personality. If that was true, then handwriting analysis should correlate with scores on widely known personality tests, right? They don’t.

Graphologists were unable to predict scores on the Eysenck personality questionnaire using writing samples from the same people (Furnham and Gunter, 1987)
Graphologists were unable to predict scores on the Myers-Briggs test using writing samples from the same people (Bayne and O’Neill, 1988)
Using meta-analysis drawn from over 200 studies, graphologists were generally unable to predict any kind of personality trait on any personality test, let alone predict job performance (Jennings, Amabile & Ross, 1992).

Graphology Predicts Job Performance? Graphologists say graphology predicts performance. But does the slant, shape or form of letters really have anything to do with job performance? Again, there is no substantial research evidence to support that claim. For example:

Professional graphologists using handwriting analysis were just as ineffective as lay people at predicting performance (Neter & Ben-Shakar, 1989).
A broad literature screen done by King and Koehler confirmed dozens of studies showing the mechanical aspects of graphology (slant, slope, etc.) are essentially worthless predictors of job performance.

A Controlled Experiment Good research requires controlling for all possible conditions that could affect the experiment, so King and Koehler set up two experimental conditions to test 1) whether subjects “discovered” relationships that did not exist and 2) whether “positive” relationships were “discovered” as often as “negative” ones. The materials for Experiment 1 included:

Casebooks that included a real handwriting sample on the left page and seven fake personality scores on the right side
Casebooks with the same handwriting sample on the left page and seven reversed fake personality scores on the right side
An excerpt taken from a graphology handbook citing a list of handwriting features and examples “of interest to graphologists”
A list containing brief descriptions of the seven personality traits that subjects were told may or may not apply to the casebook handwriting examples

Experiment 2 included exactly the same conditions, except four of the seven personality “scores” in the casebooks were nearly identical to the graphology handout examples. Casebooks were randomly assigned to 58 to 78 subjects, who were then asked to examine the materials and evaluate whether particular handwriting features were indicative of certain personality traits. Many subjects were certain they could see strong relationships between scores on the profile and handwriting samples (i.e., even though all the personality scores were faked). Furthermore, they tended to “discover” more relationships when words in the personality profile were positively associated than when the words were negatively associated with words in the graphology handbook. Bottom line? Subjects seemed to feel compelled to report a relationship between handwriting and personality ó even when none existed ó driven solely by word-association. This led subjects to make subjective, completely inaccurate assumptions about personality traits. Basically, subjects “discover” non-existent relationships totally independent of handwriting data! But let’s not just trust King and Koehler’s experiment. People interested in using graphology might also want to check out the subject on this site: http://www.ntskeptics.org/factsheets/graphol.htm. What about Graphology as a Hiring Tool?

Graphology was, is, and probably always will be, a generally worthless job performance predictor.
Graphology will persist because interpreters tend to make illusory associations based on semantics, not hard data. Their illusions tend to become self-confirming beliefs.
If a graphologist has some insight into a person’s personality (say from an autobiographical writing sample or other source) he or she will tend to “discover” relationships within the writing sample even though none exist. That is, the graphologist will guess personality traits from the autobiographical information and then “discover” similar data in the handwriting samples.
Even though a writing sample may contain wildly incongruent information, a graphologist will tend to “discover” relationships that lead to the illusion of congruent personality traits.
Some of the personality traits “discovered” are so general they could apply to a poodle (e.g., “at times he is extroverted, affable, and sociable; while at other times he is wary and reserved”).

What can we learn from this article? First, newspaper journalists often don’t know enough about hiring tests and employment laws to get the facts straight (I emailed the journalist but never received a response). Second, many people are still selling snake oil to naive organizations. Third, what appears to be legitimate information might be pure nonsense. Finally, we always have to check out the facts by seeing if there is any controlled research than confirms the data. The Last Laugh The article noted that Pilot Pens just hired someone as “Chief Graphology Officer.” This person’s role will be to use graphology to help the company hire better people. Presumably, the CGO will be given an office located between the Chief Astrology Officer and Chief Necromancy Officer. Watch the papers closely folks. Once the company replicates King and Koehler’s experiment (at about $50,000 per bad hire), the whiz who made this decision will soon be looking for a new job!