- Copyright and creating other forms
- Approved short forms
- Gender and using the BSQ with men
- Psychometric properties
- Cautionary note: DON’T ask me for the reliability let alone the validity of the BSQ!
- Original PhD thesis about the development of the BSQ
The BSQ is a self-report measure of the body shape preoccupations typical of bulimia nervosa and anorexia nervosa. It was first reported in: Cooper, P.J., M.J. Taylor, Z. Cooper & C.G. Fairburn (1986). The development and validation of the Body Shape Questionnaire. International Journal of Eating Disorders 6: 485-494.
The full detail is in Melanie’s PhD thesis: “The Nature and Significance of Body Image Disturbance”. Melanie J Taylor. Wolfson College Cambridge 1987. A paper copy is kept in the Library for Experimental Psychology at the university but, with Melanie sacrificing one of her paper copies and me cutting it up and putting it through a scanner, an electronic copy is now online. Go to the BSQ thesis directory to get it in total or parts.
Copyright and creating other forms
As of October 2023 the management of the BSQ has been transferred to MAPI trust, see:
(I think MAPI/ePROVIDE will be mounting information about the short forms shortly.)
This means that all requests for permission to use or translate the BSQ must go to them at https://eprovide.mapi-trust.org/my-eprovide/my-requests/new.
You must sign up with them, which is free, this will ensure an optimized service. It does not commit you to purchasing a questionnaire and if your work has no specific funding for the use of the questionnaire you will be able to download the form you want directly from ePROVIDE, using the “online distribution” process. For any further help, please consult the FAQ, or contact email@example.com.”
Approved shortened forms of the BSQ
The approved short forms came into existence when I did some work on the psychometric properties of the full BSQ based on data largely from women with bulimia that Bridget Dolan had collected. I suggested that four near parallel eight item short forms and two near parallel 16 item forms could be extracted from it with very little loss of internal reliability. That work was published in: Evans, C. & Dolan, B. (1993). Body Shape Questionnaire: derivation of shortened “alternate forms”. International Journal of Eating Disorders 13(3): 315-321.
Mapping of items for approved short forms
My paper with Bridget Dolan, showed that, for the data we had from white British women attending a family planning clinic, two 16 item shortened forms of the BSQ had Cronbach’s coefficient alpha values in the range .93 to .93 with non-significant differences in mean scores (paired t-tests). Four eight item scales had alpha values ranging from .87 to .92 and very nearly equivalent scores. All scales showed similar correlations with other correlated and discriminant variables (see paper for details).
The six shortened scales and the original 34 item version are enclosed. The mapping of the items from the shortened forms to the original 32 items is as follows. Items 26 and 32 from the BSQ-34, which refer to vomiting and use of laxatives, were dropped from all shortened forms.
Gender and using the BSQ with men
The BSQ was designed for use with women but Melanie has recently confirmed approval for changes to three items allowing the BSQ to be used with men given the increasing prevalence of, and recognition of, eating disorders and body shape concerns, in men. The changes are:
Item 9 now reads – “Has being with thin men made you feel self-conscious about your shape?”
Item 12 now reads – “Have you noticed the shape of other men and felt that your shape compared unfavourably?”
Item 25 now reads – “Have you felt that it is not fair that other men are thinner than you?”
This form of the BSQ, and the derivative shortened forms for men, will be made available here when I have time to make the changes. Results from men should be explored psychometrically and scores not just compared directly to referential scores from women as the psychology of body shape concerns may differ between men and women even in men with clear anorexia nervosa or bulimia. There is is clearly now an empirical issue about using the measure with people who don’t identify as male or female and about another change to those three items to use “people” instead of “women” or “men”. Please contact me if you want to do that so I can discuss the copyright issue with Melanie.
Why there cannot be a meaningful ungendered BSQ
The authors created a gendered instrument. Later they allowed that an adaptation for men, but still gendered, could be created and that’s explained immediately above.
However, the question of “using the BSQ with women and men” or of having an ungendered version is coming up. I am not convinced, and I believe Melanie agrees with me, that you can.
The issue is whether
“How do you compare your body with that of other [men|women]?”
is the same as
“How do you compare your body with that of other people?”
They are not for me, I identify as male and I really am not sure how I would answer the question:
“How do you compare your body with that of other women?”
For me it’s grammatically flawed and if I were to turn it into:
“How do you compare your body with that of women?”
I am into completely different territory from trying to answer:
“How do you compare your body with that of other men?”
Perhaps that shows I have a rather binary, gender stereotyped position (I am in my 60s!) However, I think there are issues here that are important to the kinds of self-appraisal the BSQ is designed to measure and which will get very muddled if people try to create an ungendered form. As we move more and more to online rather than paper forms this can be handled by having a branching form that asks for the gender of the respondent early on and then asks the items that are gendered with the same gender form … however, that begs the question of what should happen if you also want a non-binary gender identification. These issues matter and can’t just be solved by using an ungendered form with both/all genders. I suspect ultimately they will need new methods for us to handle them well.
Scoring the BSQ and short forms
People often ask about the scoring. Each item is scored 1 to 6 with “Never” = 1 and “Always” = 6 and the overall score is the total across the 34 items, i.e. a theoretical score range from 34 to 204.
In my own use of the BSQ, years ago now, I think we had essentially no omitted items in our data. However, much of that work was with motivated people seeking treatment so such a low omission rate may not always be the case. I tend to use a “≤10% prorating” approach to all measures where someone has missed out an item. That’s to say that for the BSQ16 I would prorate (multiply by 16/15 the total across the 15 they did answer) but I wouldn’t do that if someone missed out more than one item as that would take me over the 10% criterion.
That “≤10%” prorating recommendation was never “official”. I don’t think anything was said in our paper or the original papers on the BSQ. I am clear from my own reading that this rule is used by a number of other researchers and measure developers but I’m not aware of any canonical work on prorating and maximum proportions of items to prorate. (Do contact me if you know of any such work or have done empirical or simulation work on this issue: I’d love to hear from you.)
We have now agreed ongoing scoring rules with MAPI:
Missed items should be pro-rated. However, if a great number of items are omitted then it would invalidate the questionnaire.
We recommend prorating, i.e., taking the mean of the scores per completed item and multiplying by the total number of items, 34 for the full BSQ, 16 or 8 as appropriate for the short forms. Using a guide of only prorating if fewer than 10% of items are missing gives this.
Prorate up to four missing for the full BSQ, leaving smallest measure, minimum k (number of items) therefore 30.
Prorate up to two missing for the 16 item short forms, smallest k = 14.
Only one missing allowed for the 8 item short forms, minimum k = 7.
As ever, the key thing is to declare how you handled missing items, and their numbers, in the methods section of a paper whenever you do use prorating, it’s horribly easy to forget to do this and weakens our literature when we forget (yes, I have at times!)
Scoring the short forms
The scoring of the short forms is based on the same principle: add up the scores on the items. Very roughly, you can convert a score on a 16 item version to what its equivalent is on the full BSQ by multiplying the score on the 16 item version by 34/16. By the same principle you can convert scores on any 8 item version to BSQ equivalent score by multiplying by 34/8. Bear in mind that this IS approximate: because different items will have different probabilities of being scored positively at the same level of body shape preoccupation a score on one item is not equivalent to a score on another item and a so such rescaling is always only a guide. There are ways to get better rescaling rules based on empirical data using the measures, or just based on looking at the scores on the items in the shorter forms when embedded in the full form. I’m not aware of work like that having been published for the BSQ though. Do please contact me if you seen it reported for any of the short forms.
Cutting points for classifying scores
People also ask about cutting points between “normal” and “abnormal”. As far as I know, there are none. Given the complex cultural and other determinants of general body image concerns and of frank clinical eating disorders, I believe that any cutting points should be checked very carefully and not assumed to generalise across cultures.
Melanie (Bash, née Taylor) has the following in her thesis but not in the original paper about the BSQ:
|Full BSQ score||Classification|
|less than 80||no concern with shape|
|80 to 110||mild concern with shape|
|111 to 140||moderate concern with shape|
|over 140||marked concern with shape|
For UK English samples, it is currently reasonable to convert those cutting points on the full BSQ to get cutting points for the 16 item and 8 item versions by multiplying the BSQ cutting points by 16/34 and by 8/34 respectively. That gives these cutting points for the 16 item short forms.
|16 item score||Classification|
|less than 38||no concern with shape|
|38 to 51||mild concern with shape|
|52 to 66||moderate concern with shape|
|over 66||marked concern with shape|
and for the 8 item versions:
|8 item score||Classificationi|
|less than 19||no concern with shape|
|19 to 25||mild concern with shape|
|26 to 33||moderate concern with shape|
|over 33||marked concern with shape|
As noted above, that’s not a great way of working out new cutting points for the short forms as the different items can have rather different mean scores (in clinical and in non-clinical samples) so multiplying by the numbers of items is not going to guarantee the best possible map from one version to another. Even for the UK this way of multiplying the BSQ category cutting points by the reduced numbers of items in the 16 and 8 item short forms is really only a sensible guide pending other empirical data from large clinical and non-clinical samples emerging. If you have done such work or know of such work, do please contact me and I’ll edit this to point to the work..
Given the complex cultural and other determinants of general body image concerns and of frank clinical eating disorders, I believe that for other cultures and for translated versions, any cutting points should be based on local data not transferred from those UK guidelines assuming generalisability across cultures and languages.
Psychometric properties of the BSQ
I haven’t worked in body image or eating disorders for over a decade now but I do continue to work on the psychometrics and other instruments (mostly CORE-OM and shortened forms and PSYCHLOPS) and have been surprised to be seeing rather little work on the psychometrics of the BSQ though I certainly haven’t been following this exhaustively. If you have a moderately large dataset (n ≥ 500) of data from the full BSQ to recheck the shortening specification, or of any of the BSQ forms to look at their psychometric properties in your sample, and and you need psychometric help and would trade for authorship, do contact me.
Cautionary note about psychometric properties of the BSQ and short forms
I get a number of contacts asking me for “the reliability and validity of the [BSQ and/or short form(s)]” Please don’t ask this!
a) Because the answer for Cronbach’s reliability in the development sample data are in Melanie’s paper and thesis and for the short forms are in the paper I and Dr. Dolan did to develop them (and I’m happy to send you a copy of that paper if you ask me).
b) However, those are simply internal reliabilities and there are other indices of reliability, principally test-retest reliability and these are no indication of validity. There is good work addressing validity in Melanie’s thesis but validity is a many faceted idea.
c) Even these internal reliability values are from the measure in English, from specific samples that were collected over twenty years ago now. There is no guarantee that they are still good guesstimates of the population reliability of the measures in UK non-help-seeking and help-seeking samples any more: attitudes to body image change. Even more importantly, attitudes to body image are substantially determined by cultural issues so might be very different in other countries and cultures. It’s not for the BSQ but for a dramatic example, see the work I did around the time of doing the BSQ short form work: Evans, C., Dolan, B., & Toriola, A. (1997). Detection of intra- and cross-cultural non-equivalence by simple methods in cross-cultural research: Evidence from a study of eating attitudes in Nigeria and Britain. Eating and Weight Disorders, 2, 67–78. https://doi.org/10.1007/BF03397154. Contact me if you want a copy of that but make sure you say that’s what you want: the messages I get from the contact form don’t say what page they came from.
d) Translation of any measure is never perfect, even when the cultural location is the same full “measurement invariance” is rare if samples are large enough to catch the likely changes which may sometimes be substantial. For examples exploring this, not for the BSQ, from my own work (again, contact me if you want copies):
- Paz, C., Evans, C., Valdiviezo-Oña, J., & Osejo-Taco, G. (2020). Acceptability and Psychometric Properties of the Spanish Translation of the Schwartz Outcome Scale-10 (SOS-10-E) Outside the United States: A Replication and Extension in a Latin American Context. Journal of Personality Assessment, 1–10. https://doi.org/10.1080/00223891.2020.1825963
- Paz, C., Hermosa-Bosano, C., & Evans, C. (2021). What Happens When Individuals Answer Questionnaires in Two Different Languages. Frontiers in Psychology, 12, 688397.
- Paz, C., Mascialino, G., & Evans, C. (2020). Exploration of the psychometric properties of the Clinical Outcomes in Routine Evaluation-Outcome Measure in Ecuador. BMC Psychology, 8(1), 94–105.
So please: no more requests for “the reliability and validity of the [BSQ and/or short form(s)]” If I do bother to waste time answering it will be just to point you here.
Downloading the measures and translations
Comments or questions?
If you still have questions or suggestions to improve this age do contact me.
Page created 5.i.19 from page on the old psyctc.org site from 2003, last updated 8.xi.23. As with most pages on this site, the content is freely available under a Attribution-ShareAlike Creative Commons Licence so you can quote as much or as little of it as you like but you must make whatever you do available on the same licence and give the attribution to me with a link back to here.