Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

homosexuality obvious?

15 views
Skip to first unread message

Peter Brooks

unread,
May 18, 2012, 3:32:16 AM5/18/12
to
I'm a little surprised by this study that appears to be showing that
people recognise whether somebody is homosexual or heterosexual simply
from seeing their photograph:

http://www.telegraph.co.uk/science/science-news/9269298/Women-really-do-have-a-gaydar.html

http://www.upi.com/Health_News/2012/05/17/People-can-make-split-second-gay-judgement/UPI-33941337309646/?spt=hs&or=hn

Presumably this would mean that a neural network could be taught to do
this, so CCTV cameras could label people fairly accurately - not a
technology that would do much for human rights in Saudi Arabia.

It's a very small study - though it seems to have been widely
reported, so it may be nothing at all.

Dave Smith

unread,
May 18, 2012, 5:56:22 AM5/18/12
to
Also, perhaps the experiment did not divide the photos to reflect the base rate of homosexuality in the normal population?

Dave Smith

Peter Brooks

unread,
May 18, 2012, 6:44:13 PM5/18/12
to
On May 18, 11:56 am, Dave Smith <davidelliottsm...@btinternet.com>
wrote:
> On Friday, May 18, 2012 8:32:16 AM UTC+1, Peter Brooks wrote:
> > I'm a little surprised by this study that appears to be showing that
> > people recognise whether somebody is homosexual or heterosexual simply
> > from seeing their photograph:
>
> >http://www.telegraph.co.uk/science/science-news/9269298/Women-really-...
>
> >http://www.upi.com/Health_News/2012/05/17/People-can-make-split-secon...
>
> > Presumably this would mean that a neural network could be taught to do
> > this, so CCTV cameras could label people fairly accurately - not a
> > technology that would do much for human rights in Saudi Arabia.
>
> > It's a very small study - though it seems to have been widely
> > reported, so it may be nothing at all.
>
> Also, perhaps the experiment did not divide the photos to reflect the base rate of homosexuality in the normal population?
>
Well, yes, though your point isn't very PC, I understand it.

Dave Smith

unread,
May 18, 2012, 8:24:11 PM5/18/12
to
Here's the report of the study, for you to plough through if you have sufficient interest!

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0036671

Dave Smith

Peter Brooks

unread,
May 19, 2012, 3:15:22 AM5/19/12
to
On May 19, 2:24 am, Dave Smith <davidelliottsm...@btinternet.com>
> http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.00...
>
Thank you, yes, I found it. I didn't read it all, but it says that it
was already known that people could spot sexuality very quickly.

Philip

unread,
May 19, 2012, 3:49:39 PM5/19/12
to

"Dave Smith" <davidell...@btinternet.com> wrote in message
news:d7eafb1c-655f-441a...@googlegroups.com...
Thanks for finding that. It's an interesting paper but it seems to make an
implicit assumption that the photos people post of themselves on Facebook
represent unbiassed examples of what they look like. It seems to me at
least plausible that people who describe themselves as either straight or
gay might be biassed in their choices of photos of themselves towards those
that tend to match their self-descriptions, and given a choice of, say, a
dozen photos of each target, third parties might be just as successful at
selecting ones that signal arbitrarily chosen sexual orientations.

Dave Smith

unread,
May 19, 2012, 9:18:06 PM5/19/12
to
Yes, but your point would need to be considered in the context of the earlier studies referenced in the introduction. The two main hypotheses of the research concerned the type of facial stimuli used and the possibility of a difference between the judgements made about males and the judgements made about females:

"
Hypothesis 1.

Configural face processing contributes to accurate snap judgments of sexual orientation. Because sexual orientation is phenotypically ambiguous, we predicted that the deeper, more individuating type of face processing – configural face processing – would contribute to judgment accuracy. In practical terms, this means that judgment accuracy should be reduced when faces are presented upside-down (vs. upright).

Hypothesis 2.

The process of reading sexual orientation from faces may differ as a function of whether the stimulus person (face) is male or female. In the present experiments, participants judged both men’s and women’s faces, allowing for direct comparisons of judgments as a function of target gender. This hypothesis is exploratory in nature and does not carry a directional prediction.
"


I'm not sure how accurate the judgements of sexual orientation were. The authors tend to slide from 'better than chance accuracy' to 'accurate' when discussing the results, but it seems clear that many of the judgements made were wrong or just lucky given the even chance of getting the right answer. Also, in an everyday situation most people will be heterosexual, so the 50:50 split imposed in the experiments was artificial.

The authors acknowledge that their work tells us little about everyday life:

"Although the present experiments deal primarily with whether above-chance accuracy in snap judgments of sexual orientation from faces can occur and how faces are processed to give rise to such judgments, it does so in an experimental setting wherein individuals are instructed to make forced-choice judgments of sexual orientation. Recent work, e.g., [41], [42] suggests that inferences of sexual orientation need not depend on the explicit instructions to judge faces as gay or straight. Nonetheless, a relatively unexplored question that is ripe for future research involves the external validity of these effects – do snap judgments of sexual orientation from faces occur in real-life settings? Additionally, what are the downstream consequences of snap judgments of sexual orientation, for example, on the perceiver’s feelings, thoughts, and behaviors towards the target?"

It seems to me that the external validity of the research is further limited by the exclusion of possible behavioural 'clues' concerning sexual orientation which might normally influence judgements. I'm left with a so-what feeling!

Dave Smith

Philip

unread,
May 20, 2012, 7:25:27 PM5/20/12
to

"Dave Smith" <davidell...@btinternet.com> wrote in message
news:148f0037-2f26-4a6d...@googlegroups.com...
[...]
>I'm not sure how accurate the judgements of sexual orientation were. The
>authors tend to slide from 'better >than chance accuracy' to 'accurate'
>when discussing the results, but it seems clear that many of the judgements
> >made were wrong or just lucky given the even chance of getting the right
>answer. Also, in an everyday >situation most people will be heterosexual,
>so the 50:50 split imposed in the experiments was artificial.

I don't uderstand why that's important. What difference does it make?

Lance

unread,
May 20, 2012, 8:08:21 PM5/20/12
to
On May 21, 1:25 am, "Philip" <pp...@tiscali.co.uk> wrote:
> "Dave Smith" <davidelliottsm...@btinternet.com> wrote in message
Well think of the problem of identifying a fake bank note when you
expect to be handed one versus when you don't. In real life fake bank
notes tend to be rare in comparison with real ones so we don't
specially scrutinise them to see if they are real unless we have some
special reason do so - we just stuff them in our wallets and get on
with our business. Similarly if homosexuals make up about 4% of the
population (that is what surveys suggest) then even if we can identify
homosexual faces we are unlikely to carefully scrutinise faces for
signs of homosexuality unless we have some reason to expect them to be
present because they will be fairly rare.

Lance

Dave Smith

unread,
May 21, 2012, 6:31:26 PM5/21/12
to
Well, if the split in everyday life is more like 5:95, then the best strategy might be to judge that everyone is straight. That will result in a 95 per cent accuracy rate, which might be hard to improve upon.

Another consideration is that in a situation where homosexuality is fairly rare, attempts to pick out the homosexuals are likely to give rise to quite a few false positives (people judged to be gay who are in fact straight). Say, there is a rate of homosexuality of 5 per cent and the measure used identifies 80 per cent of cases correctly. Applying the measure to 100 cases might correctly pick up 4 out of 5 gays (four fifths of five), but incorrectly identify 19 out of 95 straights (one fifth of 95).

I'm sure this is rather crude and simplistic, but my point is only that the incidence of homosexuality in the general population needs to be taken into account when considering the implications of the experiments for everyday life situations.

Dave Smith
0 new messages