Tuesday, May 17, 2011

A reverse decline effect for RSVP?

Last week I attended the Vision Sciences Society annual meeting in Florida. Good times, good science. Although I don't use this blog for talking about my own research or field, I was struck by a talk from Molly Potter that was germane to this blog. (In full disclosure, Molly was the chair of my PhD committee, a personal hero of mine, and a large influence on my thinking).

In the 1960s and 1970s, Prof. Potter sought to study the temporal limits of complex visual processing. As our eyes move multiple times per second, the visual input we receive is constantly changing. To emulate this process, she developed the technique of rapid serial visual presentation (RSVP). In this method, one presents a participant with a stream of photographs, one after another, for a very brief time (half a second or less per picture). She found that when you give a participant a target scene (either by showing the picture or describing the picture), the participant can detect the presence or absence of this picture even when the pictures are presented for a tenth of a second each! Below is an example of one of these displays. Try to find a picture of the Dalai Lama wearing a cowboy hat.

Pretty cool, huh?

In her new research, Prof. Potter was trying to determine how much faster the visual system can be pushed by presenting RSVP steams that were only 50, 33 or even 13ms per picture. Here is a graph adapted from my notes at her talk:

Even at 13 ms per image, participants were performing at about 60% correct, and by 80ms per image, they were nearly perfect.

"Huh" I thought to myself during the talk, "this is really high performance. It seems even higher than performances for longer presentation times that were in the original papers".

So back in Boston, I looked up the original findings. Here is one of the graphs from 1975:


So, participants in 1975 needed 125ms per picture to reach the same level or performance that modern participants can perform with 33ms/picture.

I've complained a bit here about the so-called "decline effect", the phenomenon of effect sizes in research declining over time. The increased performance for RSVP displays can be seen as a kind of reverse decline effect.

Why?

In 1975, the only way to present pictures at a rapid rate was through the use of a tachistoscope. Today's research is done on computer monitors. Although the temporal properties of CRT monitors are well-worked out, perhaps these two methods are not fully equivalent. On the other hand, compared to 1975, our lives are full of fast movie-cuts, video games and other rapid stimuli, and so the new generation of participants may have faster visual systems.

3 comments:

  1. Hi,

    My monitor probably takes 33 ms to display the pictures and I can easily see the Dalai Lama. Had you alerted me to another picture if I would have seen that as easily. For example the kick boxing like image is relatively easy to spot but there is a green image that I cannot make out. So it seems to me that the experiment is testing some form of priming.

    My memory of tachistoscopic display is of seeing black and white line drawings, these would not be comparable to colour images.

    ReplyDelete
  2. sorry about the typo: "..I wonder if I would have seen that as easily."

    ReplyDelete
  3. Yes, they are surprisingly easy to find which is what makes the effect cool. There is also a fair amount of work done on detecting two targets in these streams - in general, you can detect 2 targets, but not when the second target is within 200-500 ms of the first. This is called the "attentional blink".

    And yes, most research with tachistoscopic displays did use line drawings, but the original Potter studies used slides of color photographs.

    ReplyDelete