Posts by Stephen Hill
Last ←Newer Page 1 2 3 4 5 Older→ First
-
@ Evan Yates
More random research findings: people's self-ratings of their face recognition abilities have no relationship to their actual performance on lab face recognition tasks.
Does this mean people who think they are good at face-recognition are actually not so good? Or people who think they are crap are actually better than they think they are? Or both?The research I've seen shows that the correlation between people's judgements of their own face recognition ability and their performance on lab tests is nearly 0 (-.05 for those who care). For a correlation to be near 0 'both' is likely to be the answer to your question. That makes good sense to me - a lot of people I meet reckon their memories are awful but do okay on lab tests. On the other hand there are always a bunch of people who think they're flasher than they really are.
I realise this thread is getting old but I hope you can answer. Does that test at Face Blindness Testing have any relevance to the true nature of someone's actual ability? (especially the Unknown Faces test)
The low correlation might occur because lab tests don't tell us about our face recognition performance in 'the real world'. I suspect that this is not the case and that a good score on a face memory test is a reasonable guide to how well we remember faces day-to-day. It is also probably the case that some people ARE good judges of their face memory ability (but most of us aren't).
-
@ Sacha
most research shows that women are better at recognising women's faces than men but no better at recognising men's faces
How does that work? - please say more, StephenI don't think anyone really knows - I certainly haven't seen a convincing explanation in the literature. Some speculate that women pay more attention to women's faces than men's but why that happens (if it does) is unclear. One consequence of this phenomenon is that many tests of face memory only use male faces (like the ones on the face blindness site) ... and they make them bald so as to avoid providing 'hair cues' that can lead to good performance even in the absence of normal face memory (prosopagnosics are famous for using such cues to navigate the world of people ... a la The Man who Mistook his Wife for a Hat).
-
Since I research this stuff in my day job can I ask those people who say they have 'bad memories for faces' whether they mean
1. The face is familiar but you can't put a name to it
2. The face is unfamiliar (looks like a stranger)Number 1 is extraordinarily common, Number 2 not so much.
Although I love the idea of calling #1 'Brown Syndrome' I suspect we'd have to diagnose about 75% of population as being being positioned somewhere on the 'Brown Syndrome' spectrum.
One prosaic reason why lots of people like Russell forget stuff is because they have lots they have to remember - they often forget about same percentage of material as everyone else but that turns out to be rather a lot in absolute terms.
And on the male/female thing - most research shows that women are better at recognising women's faces than men but no better at recognising men's faces.
More random research findings: people's self-ratings of their face recognition abilities have no relationship to their actual performance on lab face recognition tasks.
-
That was my point exactly - the formulae I used are those used by the research companies. They basically shouldn't be used with small sample sizes. The reason some of the confidence limits (lower and upper percentages) extend past 0 and 100 is that the 'margin of error' calculations are phenomenonally large when using tiny samples.
There's nothing wrong with the maths - it's the research design that's the problem.
-
Just for fun I have calculated the confidence levels for the figures Russell quotes from the Herald. There's a 99% chance that the actual percentage of New Zealand First supporters who wish NZ First to go into coalition with National is between 64.5% and 115% (!). For ACT and United Future its between 42% and 156%, and for the Maori Party between -2.6% and 92.5%.
These statistics are based on the extremely dubious assumption that the people surveyed provide a representative snapshot of the range of opinions of the supporters of the four parties. The stats are, of course, meaningless as Russell and others point out. It's not simply a case of OVERinterpretation here - rather, it's a case of basic innumeracy. You can't conclude anything at all from the survey about coalition support.
BTW - for the purposes of arriving at the ACT and United Future figures I had to use the figure of 99% support for coalescing with National rather than 100% - the stats can't be calculated using 100%.