CVNet - summary responses to visual memory query

CVNet (cvnet@skivs.ski.org)
Sun, 18 Aug 96 11:08:16 PDT

From: bnglaser@tohu0.weizmann.ac.il (Daniel Glaser)
To: cvnet@skivs.ski.org
Cc: bnglaser@tohu0
Subject: SELECTION of responses to visual memory q. (LONG)

Hi,

In the second of two mails I include some of the replies I received.

For the sake of brevity, I have not included mails which contain just
the references listed in my first mail, and I have compressed the
references themselves. The responses have also been gently reformatted.

Daniel Glaser.

* - * - * - * - * - * - * - * - * - * - * - * - *
* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Thu, 25 Jul 96 17:11:57 EDT
From: Mary C. Potter <molly@psyche.mit.edu>

Dear Dr. Glaser,

I suspect that you are conflating several different experiments with
picture memory: three studies in the 1960's or early '70's by,
respectively, Shepard, Nickerson, and Standing, each of which showed
that people have an excellent recognition memory for pictures, even
those shown for only a second or two--and that that memory can be
quite long-lasting (only Standing's study, I believe)--and some
studies of mine with rapidly presented pictures, which showed that at
still higher rates of presentation (in the range of normal eye
fixations and faster) people can see and understand the pictures
rather well if they are simply looking for a generally specified
target (e.g., "a boat") (Potter, 1975, 1976) but they forget many or
even most of them immediately (Potter & Levy, 1969). Additional work
on this topic was carried out by H. Intraub in the late 1970's and
1980's.

...

I'll be interested to see what other references you come up with.

Sincerely,

Molly Potter

* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Thu, 25 Jul 1996 18:04:42 -0500 (EST)
From: ROBERT TEGHTSOONIAN <RTEX@smith.smith.edu>

... The parameters aren't quite the ones you cite: Shepard used 612
pictures and tested at various delays, ranging up to 120 days. But
others have used more pictures and longer delays. You can take the
quotes off your 'fact' I think. Hope this helps.

*******************************************************************
Robert Teghtsoonian Telephone: (413)585-3902
Psychology Department Fax: (413)585-3786
Smith College e-mail: RTEX@SMITH.EDU
Northampton, MA 01060
USA
*******************************************************************

* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Thu, 25 Jul 1996 19:07:34 -0400 (EDT)
From: JOSH WALLMAN <wallman@scisun.sci.ccny.cuny.edu>

Try the 1995 ARVO meeting, abstract 1778.

Josh Wallman

* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Thu, 25 Jul 96 19:42:44 PDT
From: cwt@skivs.ski.org (Chris Tyler)

The key studies were by Lionel Standing and Norman Haber, (Perception
and Psychophysics, in the '70s), who measured performance for 100,
1000 and 10000 pictures, as I recall. They extrapolated the declining
percent recall function to conclude that recognition memory for
pictures had a limit of 1 million pictures (35 mm slides, maximally
distinct). The 10000-picture series took a week to view!

Christopher Tyler

* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Fri, 26 Jul 1996 00:21:24 -0500
To: bnglaser@tohu0.weizmann.ac.il

Daniel-

...

The Nickerson et al. paper examined memory for 10,000 pictures.
Shepard used about 700. Observers initially studied a large set of
images, but at test they were shown substantially smaller sets of old
and new images and recognition rates for the small set were used to
estimate representation of the complete set of studied images. Let me
know if I can be of any more help.

-Dan

-------------------------------------------------------------------
Daniel J. Simons djs5@cornell.edu
Department of Psychology, Uris Hall
Cornell University W: (607) 255-6398
Ithaca, NY 14853 H: (607) 273-3156
-------------------------------------------------------------------

* - * - * - * - * - * - * - * - * - * - * - * - *
Date: Fri, 26 Jul 96 8:46:03 BST
From: Professor Vicki Bruce <vb1@stir.ac.uk>

Try Standing, Conezio and Haber (1970) though I can't off-hand
remember if they used an old-new or forced choice recognition memory
test.

Anyway there were a number of similar papers in the late 60's early 70s -
Roger Shepard (1967), Nickerson, R S (1965).

After the early 70's, people started to focus on memory for more
homogeneous items, so that they could explore more closely *what*
people were remembering. Goldstein and Chance (1971) used faces,
inkblots and snowflake patterns, and a much tougher recognition test.

Cheers

Vicki Bruce

-- 
Professor Vicki Bruce                    
Department of Psychology             
University of Stirling                
Stirling                              
Scotland                             
FK9 4LA                                
                                        
Phone (+44) (0)1786 467638 or 467640
Fax                467641
E-mail vb1@forth.stir.ac.uk   

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Fri, 26 Jul 96 8:54:26 +12000 From: "johan wagemans" <Johan.Wagemans@psy.kuleuven.ac.be>

I know of one study by Standing et al. (1970) presenting 2560 images, each during 10 seconds, leading to 93% correct recognition.

...

Cheers,

Johan Wagemans University of Leuven Department of Psychology Tiensestraat 102 B-3000 Leuven Belgium e-mail: johan.wagemans@psy.kuleuven.ac.be

* - * - * - * - * - * - * - * - * - * - * - * - * From: "Jeffrey C. Liter" <jliter@yam.mpik-tueb.mpg.de> Date: Fri, 26 Jul 1996 10:12:42 +0000

...

Recent work, however, suggests that memory can be much worse if more similar stimuli are used:

Franken, R. E. & Rowland, G. L. (1979).

I look forward to seeing the other replies to your query.

Jeff

---------

Jeffrey C. Liter Max-Planck-Institute for Biological Cybernetics Tuebingen, Germany

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Fri, 26 Jul 1996 11:00:17 +0100 (BST) From: P Thompson <pt2@unix.york.ac.uk>

I seem to remember being a subject in such an experiment when I was an undergraduate. This would have been in around 1970 and the author would have been Alan Allport, now of Oxford University. Hope this helps.

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Fri, 26 Jul 96 09:23:00 EDT From: madden@grip.cis.upenn.edu (Brian Madden)

You are possibly referring to a study by Shepard (JVLVB 1967) where immediate recognition was nearly perfect and after a week was still 90%. After 4 months, however, performance fell a considerable amount. Ralph Haber did a study with thousands of images (Psychonom. Sci. 1970) that found 90% recognition after 4 days. There was also some earlier work by Hochberg.

Regards, Brian Madden

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | Brian Madden, Ph.D office: 328C | GRASP Laboratory phone: (215) 898-0337 | University of Pennsylvania fax: (215) 573-2048 | 3401 Walnut Street internet: madden@grip.cis.upenn.edu | Philadelphia, PA 19104 | USA http://www.cis.upenn.edu/~madden/home.html ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Fri, 26 Jul 1996 10:30:57 -0500 (CDT) From: jbartlet@utdallas.edu

Prof. Glaser --

A colleague who knows of my interest in visual long-term memory forwarded your query about picture recognition to me. The classic reference for the picture memory experiment you have heard about is Shepard, R.N., (1967). I might note that Roger Shepard used all kinds of pictures, not just scenic pictures -- I believe his goal was to have as diverse a set as possible, and this actually is critical to get such high levels of performance. I might also note that he did not use 1000 pictures, but only about 700 (still alot, obviously). Somewhat later, Standing Conezio and Haber (1970) replicated the basic result using 2500 pictures. Then, Standing (1973) published a paper on recognition memory for 10,000 pictures. High performance levels depend on use of a heterogenous set of pictures and the use of dissimilar distractors. For example, Standing et al. showed relatively poor performance at distinguishing left-right reversals from identical copies of previously viewed pictures, a finding my colleagues and I explored some years ago, using only scenic photographs (see,e.g., Bartlett, Gernsbacher & Till, (1987) and Frederickson and Bartlett (1987).

This is probably more information than you want or need, but I hope it helps.

Jim Bartlett Human Development and Communication Sciences University of Texas at Dallas Box 830688 Richardson, TX 75083-0688

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Fri, 26 Jul 1996 18:35:40 -0400 (EDT) From: Peter Tse <ptse@wjh.harvard.edu>

Hi, Richard Herrnstein at Harvard did similar reseach with pigeons in the 70s I believe. Pigeons were remarkably good at recognizing scenes that they had seen. And they were also very good at recognizing a person, even when she was shown in an unfamiliar position or place. Good luck, Peter Tse

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Wed, 31 Jul 1996 10:08:56 -0400 (EDT) From: Richard A Roemer <roemer@astro.ocis.temple.edu>

Hi, Brenda Milner in Montreal did something like that in the late 1960's. Another person to check out is Matt Erdelyi who was at NY City College. His work focused on hypermnesia and somewhat parallels the model you describe ---------------------------------------------------------------------- Richard A. Roemer, Ph.D., D.M.S. (215) 707-7283 Professor of Psychiatry and Neurology (215) 707-4086 Fax Temple University School of Medicine (215) 552-6568 Beeper 3401 N. Broad Street. Philadelphia, PA 19140 e-mail roemer@astro.ocis.temple.edu ----------------------------------------------------------------------

* - * - * - * - * - * - * - * - * - * - * - * - * Date: Wed, 7 Aug 96 09:58:07 PDT From: dwh@st-andrews.ac.uk

Dear Daniel

Standing, L. (1973), Standing, L., Conezio, J. & Haber R.N. (1970).

Its a long time since I read either of these papers but I seem to recall that the main methodological difficulty is that no proper estimates of false alarm rates were taken.

I am going to "do" this experiment with our postgraduate students this coming semester to demonstrate the application of non-parametric signal detection theory methods (rating scale tasks). I am using a PC and software written in Delphi/Object Pascal. Training session is 200 pictures each for 500 msecs. Test session is 200 pictures self paced - 100 new 100 old - response is 5 point rating of confidence-that-had-been-seen-before. Plot ROC curve and calculate P(A). Seems to work o.k. task is actually very easy.

best wishes

dave

------------------------------------- Dr. David W. Heeley Vision Science Laboratory School of Psychology University of St. Andrews St. Andrews, Fife KY16 9JU U.K.

E-mail: dwh@st-andrews.ac.uk Date: 07/08/96 Time: 09:58:07

* - * - * - * - * - * - * - * - * - * - * - * - * * - * - * - * - * - * - * - * - * - * - * - * - * Daniel Glaser bnglaser@tohu0.weizmann.ac.il Department of Neurobiology http://www.weizmann.ac.il/~bnglaser Weizmann Institute of Science Phone: + 972 (0)8 934 3080 Rehovot 76100 ISRAEL Fax: + 972 (0)8 934 4129