Odegard, T.N., Holliday, R.E., Brainerd, C.J., & Reyna, V.F. (2008). Attention to global gist processing eliminates age effects in false memories. Journal of Experimental Child Psychology, 99, 96-113.
Background
This study concerns developmental differences in false memories. In particular, the study involves a procedure for producing false memories call the Deese-Roediger-McDermott (DRM) paradigm. In the DRM paradigm participants are presented with lists of words (e.g., bed, rest, nap, snooze, etc.) that are all associates of a non-presented critical lure (e.g., sleep). When people are presented with lists such as these it is often the case that they will falsely recall or recognize the critical lure, and will often do so with very high degrees of confidence. What’s great about the DRM procedure is that it is very quick and easy way of producing large numbers of confidently held false memories in the laboratory under tightly controlled conditions.
Let’s Get Fuzzy. The authors of the present paper explain the false memory effect by appealing to a theory of cognitive development called fuzzy trace theory. According to fuzzy trace theory, when presented with a list of items at study people will tend to encode two types of representations in memory. A verbatim trace is an exact record of the item that was presented including information about the surface details associated with the item. A gist trace is not an exact record of what was presented. Rather, a gist trace records information about the overall sense or meaning of the items that were presented.
So you’ve got these verbatim traces and gist traces stored away in memory. Later on when you are tested, you look through your memory and see what you can recover. Imagine you are presented with one of the actually studied items (e.g., nap). If you retrieve a verbatim trace, you will correctly say that it was presented to you previously and you’ll do so with high confidence. Such judgments are called Identity judgments, because the item on the test is identical to the item you retrieve from memory. If you fail to retrieve a verbatim memory trace, you may still retrieve a gist memory trace. Under those circumstances, you won’t be certain that you were presented with the word nap, but you will know that there were a lot of sleep related words on the study list. So you may decide to accept the word nap on the memory test. Such judgments are called Similarity judgments, because the test item is similar to the gist of the list.
Imagine that instead of being presented with the studied item on the test, you are instead presented with the unstudied critical lure (e.g., sleep) on the memory test. If you retrieve a verbatim trace, you may end up rejecting the critical lure (i.e., it wasn’t sleep it was nap). Such a judgment is called recollection rejection. If you fail to retrieve a verbatim trace, you may instead retrieve a gist trace. If you do you may end up falsely accepting the critical lure, because the critical lure (sleep) is very highly related to the gist of the presented list. Such a judgment is also called a Similarity judgment. Finally, it has been argued that sometimes gist memories occur in a highly convincing form. This process is called Phantom Recollection. A phantom recollection is a false memory that seems like an Identity judgment.
Brainerd and colleagues developed a mathematical model that can be used to estimate these different processes. To estimate these processes, subjects are given one of three different sets of instructions and then model fitting program is used to provide estimates of the underlying processes. We won’t go into detail about how this is done, but we will look at the estimates that Odegard gives us.
Developmental Trends in False Memories. Extensive research has now shown that the false memory effect in the DRM procedure is weaker for young children than it is for older children or adults. One way of explaining this is that young children are less likely to notice the fact that all the list items are related to each other (e.g., global gist). For instance, adults are more likely to have false memories when lists are blocked by theme (making the global gist obvious) than when they are randomly intermixed (making the global gist more difficult to notice). This effect of blocking doesn’t effect first graders. Older children have some ability to notice global gist, but it isn’t as well developed as it is in adults.
Odegard decided to examine this age group and compare them to college aged adults. Participants were presented with modified DRM lists. Rather than presenting the list items in isolation, each item on the list was paired with another word which biased the interpretation of the list words. For some lists, the items were paired with items that biased the interpretation of the words in the direction of the critical lure. For other lists, the items were paired with items that biased the interpretation of the words away from the critical lure. For instance, for the critical lure WINDOW the associate SHADE was presented in the context of either the word DRAPES or the word TREE.
Methods
Odegard and crew tested 11-year-olds and college students. Students studied six DRM lists. Each list contained 11 items -- out of 15 from the Stadler norms. For three of the lists, each list item was paired with an associate that biased the meaning of the items in the direction of the critical lure. For the other three lists, each list item was paired with an associate that biased the meaning of the item away from the critical lure. Word pairs were presented visually for three seconds a piece and 15 seconds between each list.
The recognition memory test included two studied items from each list (Targets), the critical lure from each list, non-presented list items (Related Lures), and targets, critical lures, and related lures from non-presented lists which served as controls.
Results
To control for response bias, results were presented in terms of A’. A’ is basically a correction that controls for how often people say “Yes” to unrelated items. The most important finding was that recognition of critical lures was very similar for adults and 11-year-old in the Context Toward condition. But when the context shifted the interpretation away from the critical lure, it had little effect on the adults in the study, but greatly decreased false recognition of critical lures for the 11-year-olds (note that a A’ of .50 is chance performance). Recognition of unrelated lures was not influenced by whether the context toward or context away condition was used.
The Conjoint Recognition Model was used to get estimates of the use of the different memory processes proposed by fuzzy trace theory. These can be thought of as the probability that the different memory processes will occur when given that kind of item. The main effect observed was for similarity judgments in response to critical lures and related lures. For young adults, critical lures produced more similarity judgments that related lures in both the context toward and context away conditions. In fact, the context away vs. context toward manipulation didn’t have a huge effect for young adults. For the 11-year-olds in the context away condition, similarity judgments were about as likely for related lures as for unrelated lures. For the context toward condition, there were more similarity judgments for critical lures than for unrelated lures.
Discussion
So what does all this mean? Basically, adults are very good at noticing the themes of the list and they notice those themes (the global gist) regardless of the context provided by these additional cues. Eleven-year-olds, although they can process the global gist, don’t do so as readily. So they are helped when the context cues them to the global gist, and they are hurt when the context diverts them from the global gist.
Another nice feature of this article is the use of mathematical modeling to provide estimates of the underlying memory processes. In particular, fuzzy trace theory predicted that by changing the context the words were presented in, you would change the ability of the participants to extract the global gist of the lists. On a recognition test, gist results in similarity judgments, and in fact the estimates provided by the math model showed that the manipulation of context specifically influenced estimates of similarity judgments. The author of this study must have received especially good training during his graduate career.