Schacter, D.L., Israel, L. & Racine, C. (1999). Supressing false recognition in younger and older adults: The dinstinctiveness heuristic. Journal of Memory and Language, 40, 1-24.

Background:

• This study concerns false recognition in the DRM paradigm.  Reminder for everyone in this paradigm people are presented with lists of words that are all semantically related to a single nonpresented word (i.e. critical lure).
• Earlier work by Schacter's lab has found that false recognition is lower for pictures than it is for words.
• The authors argue that this is because subjects rely on a distintiveness heuristic. It basically comes down to this: Subjects expect to be able to remember distinctive details of individual items for pictures but don't necessarily expect to for words.
• Another way to think about this is that subjects decide whether something is a real memory or not by comparing how detailed it is compared to their other memories for the items they studied.  For words, all the subject's memories are pretty sparse, and so their memories for the critical lure don't seem that much less detailed.  For pictures, the subjects memories for the presented items (i.e. targets) seem pretty detailed, but their memories for the critical lures don't seem as detailed by comparison.

Overview: Two experiments are reported.  First they simply show that the results of Israel and Schacter (1997) can be replicated with older adults.  Second, they show that if you mix it up, show some lists pictorially and some lists with words that there is no supression of false memories in the picture condition.  Why? because distinctiveness becomes a less reliable cue.

Experiment 1

Method

• Two groups of subjects, college age and elderly.  The data from the college aged subjects was actually the same as that reported in Israel and Schacter (1997) Experiment 2, although the data was analyzed more extensively
• Used 12 item DRM lists, but only included items that could be pictorially encoded
• A total of 14 such lists were presented.  In the picture condition subjects saw the picture and heard the word.  In the word condition subjects saw the word and heard the word.
• The recognition test included four types of items.   Subjects made remember/know decisions on these items
   

Results

I'm going to break the results into two parts.  First I'm going to cut to the chase and give the overall conclusions of the results section.  Then I'm going to talk a little bit about all the different types of analyses they did and why.

Cutting to the Chase

For recognition of targets people recognized more pictures than words and younger folks recognized somewhat more targets than older folks.  In addition in the picture encoding condition people recognized more targets when the test included pictures

For false memories the finding was that there were more false memories for words than for pictures and that the elderly had somewhat more false memories than younger people.

What's the Deal with all these different types of analyses!?!  Or as a student said to me once, "What's up with this results section, I signed up for Psychology not calculus?!?"

The authors describe a number of different analyses.  Some of these concern how you correct for response bias, or the tendency to just say "old" to anything.  The corrections they report are a signal detection analysis using A' and corrected recognition scores like they reported in Israel and Schacter (1997) and that have also been used by other false memory researchers (Payne, Lampinen & Cordero, 1996; Seamon, Luo & Gallo, 1998).

The important point is that no matter what correction procedure they used the above results pretty much held up.  Its also important in our own research to think about how one should correct for response bias.

The other analyses they report are analysis of the Remember/Know judgments.

The interesting thing here is that they report a correction to the estimate of recollection and familiarity based on the assumption that recollection and familiarity co-occur and that Know judgments therefore underestimate the role of familiarity.

This correction is one developed by Yonelinas et al (1998).  Yonelinas among other things has produced versions of Jacoby's Process Dissociation Procedure that handles ROC results better.

Here's the logic behind it.  Assume there are two processes, recollection (which is measured by remember judgments) and familiarity which is partly measured by know judgments.

Like PDP recollection and familiarity are statistically independent.  This means that familiarity is underestimated by "know" judgments.

Remember judgments are high threshold judgments that are basically just on or off.  Know judgments are driven by a signal detection process.  To figure out the signal detection parameters though you need the probability of old and new items being recognized if there was no recollection.  This is given by the equations:

Based on these estimates the signal detection parameters d' (strength) and b (response bias) are calculated.

YAY!  It basically just combines Process Dissociation with Signal Detection and its not hard to understand if you understand those two things.

Method

The method was basically identical to Experiment 1 except the mixed lists were used.  Some lists were presented as pictures and some lists were presented as words. The distinctiveness heuristic should not occur here according to the authors because now true memories as a whole will not be substantially more distintinctive than false memories Comment: Its not obvious why this would be true.  That is, its certainly possible that people are aware of how distinctive each list should be.  One possibility for why this doesn't happen is that people use local distinctiveness information on the test, "How distincitive is this item compared to other items on this test, especially items I've recently seen on this test."  This possibility would allow the false recognition reversal phenomenon to merge with the distinctiveness heuristic.

Results
 

• As they predicted when some lists were presented pictorially and some lists were presented as words to the same subject there was no decrease in false memories for the picture lists relative to the word lists.
• Older adults had a lower hit rate and more false memories than younger adults.

General Discussion

• The distinctiveness heuristic is a meta-memorial strategy in which you determine a recognition criterion relative to other items on the list.
• This raises interesting questions about which other items you compare the item to.  The authors talk about at least three levels: A global level, a list specific level, and an item specific level.
• Findings of Experiment 1 support the idea that the global distincitiveness of items on the entire list might be used to set response criterion.  Findings of Experiment 2 support the idea that one does not use list by list distinnctiveness.  However, findings of False Recognition Reversal support the idea that one can sometimes use item level distinctiveness.
• Although the authors don't discuss this, the idea of a distinctiveness heuristic are potentially important in the debate over recovered childhood memories.  The question is, when a recovered memory is being evaluated how distincitive and detailed does it need to be.  Against the backdrop of which other memories is it being compared?  If the answer is other childhood memories, then the criterion might be pretty low.  However it may be other memories for atypical events, in which case it might be pretty high (see Pezdek, Finger & Hodge, 1997).