It's often difficult to interpret raw scores on personality instruments without reference to some normative population (or at least sample) from which your results are drawn. Just because you scored above the midpoint on the extraversion scale doesn't mean that other's don't as well. For example, what if on average, people to whom you're trying to compare yourself score 30 on this scale? You'd be right around the midpoint on Extraversion *relative* to other people.

Also, is the inventory you're describing actually the Big Five Inventory as described here? http://www.fetzer.org/sites/default/files/images/stories/pdf/selfmeasures/Personality-BigFiveInventory.pdf Goldberg's items are far more extensive (see this reference for more information: http://projects.ori.org/lrg/PDFs_papers/A%20broad-bandwidth%20inventory.pdf ) and in the public domain. However, he doesn't like to publish norms, suggesting that the most appropriate ones to use derive from a researcher's own local data.

If you have more information about what you used, I might be able to point you to published means and standard deviations that could be used as "norms" against which to interpret your scores.

I quite agree with Stephen Benning that raw scores, in and of themselves, are meaningless in this situation. If the instrument you are using lacks adequate published norms (as may well be the case here), then you have two main options.

First, you can limit yourself to correlational analyses. In most cases, this will work fine. All you lose by not using standard scores is that the scores you are using may violate one or another of the assumptions we like to make before using parametric statistics. For example, the equal interval assumption may not hold. But as long as the violations aren't too egregious, it probably won't make a substantial difference to your results - statistics like the correlation coefficient are robust in the face of modest deviations. (If you're really worried about your data set, you could convert the scores to rank-order format and use Spearman's rho instead of Pearson's r (the formulas yield the same results, so you could just plug the rank-ordered data into a regular correlation program.)

Second, you can divide your own sample into high-scoring and low-scoring subgroups. One way of doing this is a median split: top 50% vs. bottom 50%. This retains all your cases, but some members of each group will be a lot like some members of the other. Alternatively, you could leave those in the middle out of these analyses and create "upper" and "lower" groups that are separated by some range of middle scores. In the educational testing literature, the usual rule is to compare the top 27% to the bottom 27%, but anything approximating this will probably work fine.

If you are planning to keep using the instrument in the same population for a while, you might consider developing your own local norms, as Goldberg suggests.

Apologies for the delay in reply!

@ Stephen Benning: Thank you so much for your reply. I used the Big Five Inventory (44 items) by John and Srivastava, published in 1999. Here's the link where I obtained the test from: http://www.fetzer.org/sites/default/files/images/stories/pdf/selfmeasures/Personality-BigFiveInventory.pdf

Any information with respect to the normative data for this scale would be a great help :)

@ Stephen Joy: The two options you've enlisted are really helpful. Local norms may be a long way away, since my sample is relatively smaller. But thank you for the two suggestions; shall implement them if norms cannot be located..

Sharlene Fernandes

Hi,

I wish to use the same test (BFI by John & Srivastava) for a study but haven't found a relevant scoring or interpretation for Indian population. While searching about it I found your question and the subsequent answers. Any progress regarding the same? Would highly appreciate any help.

Hi,

I'd like to learn more about the Big Five 44 personality model from your perspective. Because the construct contains several reverse coded elements, I received negative outer loadings in CFA after rescaling these things and conducting the analysis. What is the best way for me to deal with this issue? Please assist.