Data interpretation from the graph Consider the problem described which relates to graph as shown below

The members of which of the four groups had the least accurate perception of their body weight?

  
*
  
*Underweight  
  
*Normal weight   
  
*Moderately  overweight   
  
*Severely overweight   
  
*It cannot be determined from the information given in the graph.
  

Example taken from - 
Green,Sharon Weiner,M.A.; Wolf,Ira K.,Ph.D. (2015-02-01). GRE, 21st edition (Barron's Gre) (Kindle Locations 11285-11290). Barron's Educational Series. Kindle Edition. 
My solution which is wrong.
x and y axes are considered as depicted below,
|(y)        
|       
|          
|------ (x) 

If you look at the About right in each of the 4 categories on the x axis, About right is highest in Underweight category whereas About right is lowest in Severely overweight category.
Therefore Severely overweight members have the least accurate perception of their body weight. Hence (4).
However, the answer is (1).
And to be straightforward, I am unable to digest the solution. Please elaborate.

Almost all overweight females correctly considered themselves to be
  overweight; and more than half of all females of normal weight
  correctly considered themselves “about right.” But nearly 70% of
  underweight adult females inaccurately considered themselves “about
  right. Hence 1”

 A: I sympathise with you. I think it's a poorly posed problem. It may be that this is entirely deliberate and the challenge is to think through a lousy presentation. I will say no more except that as someone in university teaching for my entire career I never approve of "trick questions". 
The pedagogic or other motives aside, there are various problems with the presentation. 


*

*There isn't a match between actual and perceived categories. The only exact match between the two schemes is that "underweight" occurs in both. So, we have to juggle with the words. 

*The order of perceived categories on the graph is perverse. The order should go: underweight, about right, overweight. (The reverse would be better than what is given, but is not so good as that given a sequence from underweight up for the actual categories.) 

*The shading of categories is perverse. It doesn't match either the order used in the legend (as if that made sense) or any kind of natural order. 
I couldn't rest until I had made a better graph, meaning for me one I could think about with pain. Naturally several other designs are possible. I flag that although I am often happy with graded shading for ordinal (ordered) responses, it seems simpler to avoid shading altogether here. 
FWIW, I call this a two-way bar chart. The percents shown are just based on a quick by-eye guess. Someone who scanned the data or had access to the original data could produce better numbers. A small virtue of this design is that there is always room to show the percents (more generally, the amounts) as numeric text, even if the amounts are very small. Zeros are not labelled as such, although that is a convention, not a rule. So the display is hybrid graph and table. 
All that done, I agree (with qualifications) with the answer given. Only about 22% of those who were underweight admitted it, so they are the least accurate group. The qualifications are that for all the other categories the schemes do not match one to one, so the comparison is muddied, but the answer appears correct nevertheless. 
 
