Describing and predicting behavior: The Big Five

Pos­tet på 28. mars 2005.


The answer to the ques­tion “What is per­so­na­li­ty?” depends very much on why the ques­tion is asked. The­re are many ways to deal with this ques­tion. Here, I will focus on the trait-based approa­ch, which is one attempt at descri­bing indi­vi­du­al dif­fe­ren­ces and pre­di­ci­ting indi­vi­du­al beha­vior.

The last twen­ty years, per­so­na­li­ty rese­arch has been cen­te­ring around a model known as the “Big Five,” and some­ti­mes referred to as the “Five-Factor Model of Per­so­na­li­ty”. This is the cul­mi­na­tion of near­ly a hundred years of rese­arch. I will sketch the his­tory of the Big 5 here, pro­vi­de a brief pre­sen­ta­tion of the factors, and men­tion the MBTI in pas­sing.

The Lexical Hypothesis

The story begins with Gor­don All­port and H.S. Odbert, who hypothe­sized that

Those indi­vi­du­al dif­fe­ren­ces that are most sali­ent and social­ly rele­vant in people’s lives will even­tual­ly become enco­ded into their lan­gua­ge; the more impor­tant such a dif­fe­ren­ce, the more like­ly is it to become expressed as a sing­le word.

This has become known as the “Lex­i­cal Hypothe­sis.”

What All­port and Odbert did, though, was to go through two of the most com­pre­hen­si­ve dic­tio­na­ries of the Eng­lish lan­gua­ge avai­lab­le at the time, and extract 18,000 per­so­na­li­ty-descri­bing words. From this gigan­tic list, they extracted 4500 per­so­na­li­ty-descri­bing adjec­ti­ves that they found descri­bing observab­le and rela­tive­ly perm­a­nent traits. And then, in 1936, they rested their case.

The case was pick­ed up again in 1946 by Ray­mond Cat­tell, who used new tech­no­lo­gy, i.e. com­pu­ters, to do data reduc­tion on the All­port-Odbert list. He orga­nized the list into 181 clus­ters and asked peop­le to rate peop­le they knew by the adjec­ti­ves on the list. Using factor ana­ly­sis, he ended up with twel­ve factors. He added four factors that he thought ought to have been the­re, and ended up with the hypothe­sis that indi­vi­duals descri­be them­sel­ves and each other accor­ding to six­te­en dif­fe­rent, inde­pen­dent factors.

With these six­te­en factors as a basis, he went on and con­structed the 16PF Per­so­na­li­ty Ques­tionn­ai­re, which is still being used in uni­ver­sities and busi­nes­ses for rese­arch, per­son­nel selection and the like. Later rese­arch has fai­led to repli­ca­te the results, how­e­ver, and it has been shown that Cat­tell retai­ned too many factors. The cur­rent 16PF, how­e­ver, take these fin­dings into account and is in fact a very good test which is still being devel­o­ped. In 1963, W.T. Nor­man redid Cat­tells work and sug­ge­sted that five factors would be suf­fi­ci­ent.

The Dark Ages

And then for sevente­en years, not­hing hap­pe­ned: It was pro­ven that per­so­na­li­ty is not stab­le, but varies wild­ly with situa­tions, so that pre­dic­tion of beha­vior by per­so­na­li­ty test is impos­sib­le: Soci­al Psycho­lo­gists demon­strated that cha­rac­ter, or per­so­na­li­ty, is somet­hing humans impose on peop­le in order to main­tain an illu­sion of con­sist­ency in the world. The final nail in the cof­fin was Wal­ter Mischel&rsqui;s 1968 book Psycho­lo­gical Assess­ment which demon­strated that per­so­na­li­ty tests can­not pre­dict beha­vior with a cor­re­la­tion of more than 0.3.

Around 1980, three things hap­pe­ned that brought the pre­sent stra­in of per­so­na­li­ty rese­arch out of the Dark Ages. These were Per­so­nal Com­pu­ters, Sta­ti­s­ti­cal Aggre­ga­tion, and The Big Five.

Personal Computers

Tra­ditio­nal­ly, Psycho­lo­gists who nee­ded com­pu­ters had to rent access to a main­frame. Sud­den­ly, they could do their sta­ti­s­ti­cal ana­ly­sis on their desk­tops. This ment that any­body could, for instan­ce, re-exa­mi­ne the All­port-Odbert list. But why would they, if per­so­na­li­ty is an illu­sion?

Statistical Aggregation

It was argued that per­so­na­li­ty psycho­lo­gists had look­ed at beha­vior from the wrong level. Inste­ad of try­ing to pre­dict sing­le instan­ces of beha­vior, which just didn’t work, rese­ar­chers should try to pre­dict pat­terns of beha­vior. Cor­re­la­tions soared from .3 to .8. Hey presto, we’­ve got per­so­na­lities any­way! Soci­al Psycho­lo­gists still argue that we impose con­sist­ency on the world, but that the­re is in fact more con­sist­ency to begin with than they once thought.

The Big Five

At a sym­po­si­um held in Hono­lu­lu in 1981, the four pro­mi­nent rese­ar­chers Lewis Gold­berg, Nao­mi Takamo­to-Chock, And­rew Com­rey, and John M. Dig­man reviewed the per­so­na­li­ty tests avai­lab­le at the day, and deci­ded that most of the tests that held any pro­mi­se seemed to measure a sub­set of five com­mon factors – in fact the same as Nor­man had dis­covered in 1963.

We’ve got triangulation!

Both the lex­i­cal hypothe­sis, then, and more theory-laden rese­arch con­ver­ge on a model that sta­tes that per­so­na­li­ty can be descri­bed in terms of five aggre­gate-level trait descrip­tors. Now, any per­so­na­li­ty rese­ar­cher with an oun­ce of self-respect has his own model with its own nice featu­res, but when rese­ar­chers talk to each other, they usu­al­ly trans­late their model into the one pro­po­sed by Nor­man in 1963. These are

The Factors

Sur­gency or Extra­ver­sion
How ener­ge­tic one is. Peop­le who sco­re high on this factor like to work in coope­ra­tion with others, are tal­ka­ti­ve, ent­hus­i­a­s­tic and excite­ment seeking. Peop­le who sco­re low on this factor pre­fer to work alo­ne, and can be per­ce­i­ved as cold, dif­fi­cult to read, even a bit eccent­ric.
One’s level of ori­en­ta­tion towards other peop­le. Those who sco­re high on this factor are usu­al­ly co-ope­ra­ti­ve, can be sub­mis­si­ve, and con­cerned with the well-being of others. Peop­le who sco­re low on this factor can be chal­len­ging, com­pe­ti­ti­ve, some­ti­mes even argu­men­ta­ti­ve.

These are the two “soci­al” factors. Peop­le who sco­re more than midd­le on both plus more than midd­le on IQ tend to have high “emo­tio­nal intel­li­gen­ce.”

How “struc­tu­red” one is. Peop­le who sco­re high on this factor are usu­al­ly pro­duc­ti­ve and disci­plined and sing­le tas­king. Peop­le who sco­re low on this factor are often less struc­tu­red, less pro­duc­ti­ve, but can be more flex­ib­le and inven­ti­ve, and can be multi­ta­s­king.
Neu­ro­ti­cism or (inverse­ly) Emo­tio­nal Sta­bi­li­ty
Tendency to wor­ry. Peop­le who sco­re low on this factor are usu­al­ly calm, relaxed and ratio­nal and may some­ti­mes be per­ce­i­ved as lazy and inca­pab­le of taking things serious­ly. Peop­le who sco­re high on this factor are alert, anxious, some­ti­mes wor­ried.
Cul­tu­re or Open­ness to Expe­ri­en­ce or Open­ness to Ideas
Tendency to be spe­c­u­la­ti­ve and ima­gi­na­ti­ve. Peop­le who sco­re high on this factor are neo­p­hi­le and curious and some­ti­mes unrea­li­s­tic. Peop­le who sco­re low on this factor are down-to-earth and prac­ti­cal and some­ti­mes obstruc­ti­ve of chan­ge.

The­re is a lot of rese­arch avai­lab­le on the Big Five. The pro­blem is that very litt­le of it is pub­lis­hed in books with pic­tu­res – most of it is rela­tive­ly uncom­pi­led in rese­arch jour­nals. In order to make use of the Big Five, then, one needs to be up to date on the lite­ra­tu­re.

When con­flicts are due to dif­fe­ren­ces in per­so­na­li­ty, it is usu­al­ly due to dif­fe­ren­ces in Open­ness to Expe­ri­en­ce. The factors that pre­dict job per­for­mance are Con­scien­tious­ness (which should be high) and Neu­ro­ti­cism (which should be low). Lea­ders should be high on all five factors except Neu­ro­ti­cism and Agre­ea­b­le­ness which should be low – and sale­spe­op­le should be simi­lar but with even hig­her sco­res on Sur­gency. Inci­den­tally, this is cul­tu­re-depen­dent – in most countries out­side the USA (inclu­ding Cana­da), lea­ders and sale­spe­op­le should be hig­her on Agre­ea­b­le­ness than in the USA


What about the MBTI, then, which I assu­me most of you have heard of? Well, it seems that it measu­res four of the factors – Neu­ro­ti­cism is left out, but has been pro­po­sed as added in newer revi­sions, against much pro­tes­ting. The MBTI is not based on the lex­i­cal hypothe­sis but on a theory which is, to Jung’s credit, fai­r­ly head-on with cur­rent rese­arch. The­re is one issue with the MBTI, though, and that is the notion that all the 16 types it covers are equal­ly good but dif­fe­rent – this, the Big Five rese­arch has shown, is not true. Most ENTJ’s are bet­ter lea­ders than most INTP’s, not just dif­fe­rent types of lea­ders. This one issue does of cour­se not inva­li­da­te all of the MBTI, which is a very good test.

Further research

Cur­rent rese­arch con­cen­tra­tes on three areas. The first is: Are the five factors the right ones? Attempts to repli­ca­te the Big Five in other countries with local dic­tio­na­ries have succe­e­ded in some countries but not in others, appa­rent­ly, for instan­ce, Hun­ga­ri­ans don’t have Open­ness to Expe­ri­en­ce. Of cour­se they do, others say, the pro­blem is that the lan­gua­ge does not pro­vi­de enough vari­an­ce of the related terms for pro­per sta­ti­s­ti­cal ana­ly­sis. Some have found seven factors, some only three.

The second area is: Which factors pre­dict what? Job out­come as lea­ders and sale­spe­op­le has alre­ady been measu­res, and lots of rese­arch is cur­rent­ly being done in expan­ding the list. You need to search the lite­ra­tu­re to find just what you’re look­ing for, unfor­tu­nate­ly. Bar­rick and Mount’s rese­arch are good places to start.

The third areas is to make a model of per­so­na­li­ty. Cos­ta and McCreae have built what they call the Five Factor Model of Per­so­na­li­ty which is an attempt to use the Big Five to pro­vi­de a model of per­so­na­li­ty that can explain issues in per­so­na­li­ty from the crad­le to the gra­ve. They don’t follow the lex­i­cal hypothe­sis, though, but favor a theory-dri­ven approa­ch.


Eve­rybody is inte­re­sted in per­so­na­li­ty, inas­much as eve­rybody inte­racts with them­sel­ves and other peop­le. The Big Five is a con­ve­ni­ent way to redu­ce per­so­na­li­ty into mana­ge­ab­le bits. I hope this litt­le pie­ce has been enter­tai­ning and infor­ma­ti­ve, and would like to rece­i­ve lots of ques­tions for a follow-up article.