Dear Terry,

There are two issues here. One is the empirical nature of the 
research. Like much psychological research, this work must be 
carefully bounded to achieve something that a science journal will 
publish. Therefore, it fills in minor gaps in knowledge. The meaning 
of the gaps is not always clear.

The issue of context keeps vanishing in this conversation. Even when 
we are speaking of mathematical averages for combining, compounding, 
or morphing, we're talking about processes that select among 
populations that fit reasonably well to the norms of what an "average 
model" looks like. This population is preselected out of the far 
larger population from which the image of an "average human being" 
might be fabricated. The very idea of a model at a model agency is 
predicated on the a priori notion of a model of human physical beauty 
or the physical characteristics that comprise a model human being. 
These are already tightly bounded and constrained, and they exclude 
vast numbers of human beings in any specific population that shares 
cultural norms of beauty.

If you allow to many outliers into the population, added factors are 
just as likely to skew the average in a direction that would not be 
considered beautiful. Consider someone who attempted to make an 
"average" chili from "outlier" chilies containing weird spices or the 
strange-flavor chilies that appear from time to time (cinnamon-raisin 
chili, peppermint chili, fish chili). Mixing tablespoons of these 
chilies into a standard beef chili or pork chili probably wouldn't 
win a chili cook-off.

For that matter, there are many different ways to make chili. Kind of 
meat? Beef or pork for standard chili, lamb or chicken less common, 
and dried, jerked beef for the kind of chili one might have made on 
an old trail expedition? Ways of preparing the beef or the pork in a 
standard chili? Cooked first before making the chili? Roasted first 
and shredded, roasted first and cubed, boiled first and cubed, boiled 
first and diced, fried first and cubed, fried first and diced? Meat 
placed directly in the chili? Ground, cubed or diced? Basic kind of 
chili -- chili rojo is most common but chili verde is also well 
known, especially for pork. Different kinds of chili peppers? Many to 
choose from. Different kinds of foundation sauces. Again, many to 
choose from.

Once we enter the empirical world, the reality of what we work with 
affects what we average from. One of the ideals of a "model chili" is 
a degree of self-consistency along with the subtlety of spices and 
undertones that add to the richness and complexity of flavor. Making 
a chili by sampling and averaging chilies made from different 
approaches to preparing the meat, or different kinds of chili (chili 
verde, chili rojo, etc.) might be as likely to create a bland or even 
a bad chili as to create a winning chili.

Think about wine. Averaged wines blend up to become box-wine plonk. A 
model wine requires depth and balance. A great wine is usually based 
on terroire.

I know these experiments -- some date back for years -- and I 
understand the concept. What I'm questioning here is the question of 
context and selection, and even then, I'm saying that the empirical 
facts only tell us what happened. One can always propose causal 
explanations, but there is not yet enough evidence to say which 
explanation tells us why.

Warm wishes,

Ken



>Dear Chris,
>Good point. As a matter of detail, in terms of the preferred aesthetics of
>faces (combined, compounded or morphed) it seems to be _exactly_ the
>mathematical average that is used (New Scientist 2 Oct 2004 and 22 Feb
>1992). There are claims that compounding helps increase 'beauty' by
>increasing complexity and that people have an  preference for increased
>complexity of facial detail. An empirical touchstone is model agencies
>apparently preferred photos of potential models whose faces had been
>digitally modified in this way.
>Cheers
>Terry
>
>
>>I don't think there is any suggestion that this "average" follows the rules
>of mathematical averages, I put it in quotes to indicate that it's a kind of
>rhetorical gadget.
>