Children are often thought to be imaginative and fanciful, not only in their perception of the world but also in the veridicality of their memories. It may therefore be surprising that a robust method for eliciting false memories in adults is actually ineffective in children. In fact, children even tend to show better performance than adults in terms of false recall in the DRM paradigm, a task in which a list of words must be remembered where every word in each list is strongly associated with a word that is not presented in each list. Adults tend to show a robust tendency to incorrectly recall these “critical lures,” but in general, kids don’t!
Examined more closely, however, previous work has shown some conflicting results in young children: while most work seems to demonstrate fewer false memories among children than older adults in the DRM paradigm (putatively related to underdeveloped “gist extraction” or “gist processing” abilities), yet other work has actually shown either no age related trends in false recall/recognition or even inverse trends, such that preschool-aged children have a higher proportion of falsely to correctly recalled words relative to 7-year-olds and adults (putatively related to poor “source monitoring” among young children).
In the July/August issue of Child Development, Carneiro et al report new findings from an age-appropriate version of the DRM paradigm. In the past, the lists provided to children have not always been words that the children clearly know, nor have the critical unpresented words necessarily been those which a young child would consider a strong semantic associate of the other words. In addition, studies have varied with respect to the interval over which subjects had to remember the words, the length of the lists themselves, and the rate at which each word is presented to the subjects. All of these facts complicate any interpretation of the DRM paradigm in young children.
To remedy these problems, Carneiro et al gave age-appropriate word lists to 384 subjects of varying ages. In their first experiment, the authors played several recorded lists consisting of 8, 10, 12 and 15 words presented every 2 seconds to subjects aged 3-4, 7-8, and 11-12 and adults (respectively; preschoolers were tested only on four lists total, whereas every other age group was tested on eight lists, and all list orders were counterbalanced). Subjects were given two minutes to freely recall any words they remembered after each list. After completing all the lists, subjects completed a recognition task by answering “yes” or “no” to each of 48 words, each of which was either presented or not presented during the preceeding task.
Unsurprisingly, Carneiro et al’s results showed that correct recall and recognition increases with age (incidentally, preschoolers did not show a strong primacy effect – in other words, they didn’t show the normal effect where words presented early in a list are recalled with greater probability than those in the middle of the list.)
Slightly more surprising is the finding that preschoolers showed lower absolute levels of false recall than adults. In other words, they performed better than adults even when the lists were designed to maximize their false recall! Results from a recognition test showed similar trends, where older subjects also showed higher false recognition of critical semantic lures.
The preschoolers were equally likely to falsely recall the critical semantic associates of each list as they were to incorrectly recall studied items from preceding lists (aka, an “intrusion” error). In contrast, all older age groups showed the opposite tendency; they were more likely to falsely recall the semantic associates than to show intrusion errors. Furthermore, the discrepancy between these error types increased with age.
However, it turns out that these age differences in false recall are less clear when false recall is examined in relation to the total number of words recalled. That is, preschoolers don’t show lower false recall than adults when taking into account their much lower level of correct recall. And if we count the number of intrusion errors in our estimate of false memory, Carneiro et al. demonstrate that preschoolers show even greater false memory proportionate to their correct recall than do adults.
A second experiment demonstrated that older children both correctly recall more studied words and falsely recall more semantic lures than preschoolers. Other results from the second experiment demonstrated fewer intrusion errors than false recall errors (in contrast to the first experiment) and increasing correct recognition with age. List length (8 vs 12 items) did not seem to affect false recall.
The authors conclude that the construction of age-appropriate lists is an important variable for eliciting false memories in preschoolers, and that list length is not particularly important. The authors also suggest that age differences may be more apparent on measures of recall than recognition. Furthermore, in the right experimental situation, preschoolers can indeed show evidence of false semantic memories, perhaps indicative of gist extraction/processing.
The authors further speculate that underdeveloped semantic networks in preschoolers may contribute to their lower levels of false recall/recognition than older groups. Alternatively, it could be that such differences in semantic development are secondary to the ability for kids to retrieve or reactivate previously presented words. Those who can retrieve or reactivate more previous words will also be more likely to reactivate the critical lures. This is consistent with the absence of age differences in false semantic recall when that value is normalized by total recall, indicating that preschoolers are showing an overall lower level of recall. It is however inconsistent with observed trends in adults, where correct and false recall are inversely correlated.
Finally, the authors conclude that at least two processes may contribute to false recall in the DRM paradigm. The first consists of strategic mechanisms such as elaboration and rehearsal, whereas the second is the phenomenon of spreading activation through the nodes of semantic networks which may be aided by more static or crystallized knowledge.