HIGH FALSE -m- LOW TRUE
figure 9.2. Relation between age and a corrected measure (A') of false recognition of the critical distractors of high and low Deese/Roediger-McDermott paradigm lists in Experiment 3 of Brainerd, Reyna, and Forrest (2002).
increased during childhood, but they continued to increase between age 11 and young adulthood, continuing to reinforce the theme of very gradual emergence of false memory in the DRM paradigm. Another finding that reinforces the same theme concerns false memories on "high" versus "low" lists. Specifically, a glance at the A' values for false memory reveals that the high-low list difference, which is so prominent in adults, did not emerge until after age 11.
We followed up these experiments with further experiments on children and adolescents, the novel feature of which was that an experimental paradigm and mathematical model were used that allowed investigators to tease apart and quantify the different retrieval processes that support and suppress false memory (Brainerd, Holliday, & Reyna, 2004). The objective was to determine whether the developmental increases were confined to a specific process or processes and, more particularly, to determine whether they would be confined to gist processing (as the earlier analysis implies). The paradigm that was used is called conjoint recognition, and it is an instructional variant on a standard old/new recognition design. Participants study a word list and then respond to a recognition test on which there are target probes, related distractors, and unrelated distractors, with the difference between the false-alarm rates for the two types of distractors being the false-memory index. However, participants respond to these probes under three types of instructions: verbatim (V: accept targets and reject all distractors);gist (G: accept related distractors while rejecting targets and unrelated distractors); and verbatim + gist (VG: accept targets and related distractors and reject unrelated distractors). When data from these conditions are analyzed with a mathematical model that was developed by Brainerd, Reyna, and Mojardin (1999), three distinct retrieval processes can be measured for false alarms to related distractors: (1) recollection rejection (suppresses recognition of related distractors by retrieving verbatim traces of targets, which neutralizes the familiarity of distractors' meanings); (2) similarity judgment (supports recognition of related distractors by retrieving gist traces of targets, which make distractors' meanings seem familiar); and (3) phantom recollection (supports recognition of related distractors by retrieving gist traces that are so strong that participants have illusory vivid mental experiences of distractors' prior "presentations"). An intriguing feature of false memory in the DRM paradigm, which was first detected by Roediger and McDermott (1995) with Tulving's (1985) remember/know procedure, is that critical distractors tend to provoke phantom recollective experiences (their prior "presentations" echo in the mind's ear or flash in the mind's eye). In contrast, false memories in other tasks are dominated by the weaker and less illusory experience of meaning familiarity (i.e., similarity judgment rather than phantom recollection). The fact that the DRM task provokes high levels of phantom recollection squares nicely, of course, with the high levels of false memory that this paradigm produces, but, more important for our purposes, it suggests that age increases in false memory may be primarily due to age increases in phantom recollection.
In the research that Brainerd et al. (2004) reported, participants from three age levels studied a total of nine DRM lists. Following the first three lists, the participants responded to a recognition test containing four types of probes: some targets from each of the lists, the critical distractor for each list (e.g., music), some other related distractors (e.g., drums, guitar), and some unrelated distractors. One-third of the participants at each age level responded under each of the three types of instructions (V, G, VG). The age increases in false memory that had been reported by Brainerd et al. (2002) were replicated in that the tendency to judge critical distractors as old under V instructions was much higher at older age levels than among the 7-year-olds. The more instructive findings, which are displayed in Figure 9.3, are for the underlying retrieval process. There are three results, in particular, that are of considerable theoretical interest. First, age increases in phantom recollection were driving the age increases in false memory. Of the two retrieval processes that support old responses to critical distractors, one, phantom recollection, increased steadily between the ages of 5 and 14, while the other, similarity judgment, remained invariant. Thus, the fact that the potency of the DRM illusion waxes with age is somehow tied to its ability to induce illusory vivid mental phenomenology. The second finding of theoretical interest is that the age increases in illusion potency occur despite the fact that the retrieval process that suppresses old responses to critical distractors also increases. Here, although recollection rejection remained relatively constant between the ages of 5 and 11, it jumped dramatically between the ages of 11 and 15. Taken together, the first two findings illustrate one of the most illuminating benefits of using mathematical models to separate and quantify underlying memory mechanisms: developmental variation in important types of memory performance (false memories in this instance) can be the net result of developmental variation in processes that work against each other—that is, of processes that produce age increases (phantom recollection in this instance) and processes that produce age decreases (recollection rejection in this case). The third finding of theoretical interest is that the patterns of developmental variation in underlying retrieval processes explained the detailed age trends in false memory performance. Concerning the latter trends, the false-alarm rate for critical distractors increased dramatically between the ages of 5 and 11, but it increased only slightly between the ages of 11 and 14. A glance at Figure 9.3 reveals the explanation. Phantom recollection increased substantially between the ages of 5 and 11—doubled, in fact—whereas recollection rejection did not increase; hence, age increases false memory. Phantom recollection increased again between the ages of 11 and 14, but recollection rejection also increased, allowing no net age change in false memory. Thus, the process-level analyses revealed that major developmental changes in underlying mechanisms can be underway during age ranges in which false memory would not appear to be changing if only raw memory performance were considered. This is yet another example of one of the fundamental lessons that has been learned from the application of mathematical models in the study of development (e.g., see Howe, Brainerd, & Kingma, 1985): that an absence of developmental variability
figure 9.3. Relationship between age and three retrieval processes that control false memory in Experiment 1 of Brainerd, Holliday, and Reyna (2004). Phantom = phantom recollection, similarity = similarity judgment, and recollection = recollection rejection.
in surface level performance can occur in the presence of variability, even massive variability, at the more important level of the processes that control surface performance.
So far, we have seen that, contrary to the commonsense view of memory development, the DRM illusion increases dramatically between early childhood and young adulthood, when it is measured by both intrusions during recall and false alarms during recognition. Because this outcome is so counterintuitive, before proceeding any further we ought to consider a possible artifactual explanation—specifically, age differences in association norms. Deese (1959) constructed his lists by selecting associates of critical distractors from the Russell and Jenkins (1954) word-association norms. Those norms were gathered from University of Minnesota students. Suppose that young children's associations to these same critical distractors are markedly different. If so, one would not expect them to have false memories for music, rough, and so on, the way older children and adults do, because they are not being exposed to the correct DRM lists—correct for them, that is. Metzger et al. (2004) investigated this potential artifact and ruled it out. They reported a study in which new association norms for some of Deese's critical distractors were obtained from a sample of children. In agreement with prior developmental research on word association (e.g., Bjorklund & Jacobs, 1985), Metzger et al. found a strong correlation (r = 0.75) between the associates that were given by children and the corresponding data in the Russell and Jenkins norms. They also reported a developmental study in which these child-generated lists were administered to children of various ages and to adults, using a recall-followed-by-recognition design as in the third experiment of Brainerd et al. (2002). They found that these "child appropriate" DRM lists also yielded age increases in false recall and false recognition of critical distrac-tors. Another crucial finding was that child appropriateness made no difference at all in the performance of the young children. They showed the same low levels of false memory, regardless of whether they were exposed to standard DRM lists or child-appropriate lists.
Next, Howe, Cicchetti, Toth, and Cerrito (2004) reported a developmental DRM study that is highly relevant to the aims of this volume, as it was concerned with true and false memories in maltreated children. Howe et al. were particularly interested in whether the effects of the chronic stress that is associated with maltreatment might influence the basic processes that control false memory. The authors used a 3 (age level: 5 to 7, 8 to 9, or 10 to 12 years old) x 3 (middle socioeconomic status (SES), low SES nonmaltreated, low SES maltreated) x 2 ("high" or "low" DRM list) design in which the effects of each of these factors on children's false memory were separately assessed. Children were exposed to 12 DRM lists in a recall-followed-by-recognition design. A key aspect of the design is that related distractors other than the critical distractor were included on the recognition test. The developmental increases in false memory that had been observed in prior studies were again detected. Both recognition and recall of critical distractors increased with age, the age increases were greater for recall than for recognition (a finding that Brainerd et al.  also obtained), and there were age increases in false recognition of related distractors as well as false recognition of critical distractors. The most informative new finding is that developmental increases in false memory were observed, although in varying degrees, for all three types of child populations—middle SES, low-SES nonmaltreated, and low-SES maltreated. This is informative because although middle-SES children would be comparable to the children in prior developmental studies, low-SES nonmaltreated and low-SES maltreated would not. As mentioned, age trends in false memory did vary somewhat as a function of which population children were drawn from, but the central hypothesis under investigation—that levels of false memory might be different for maltreated children owing to their history of chronic stress—was not borne out.
very recent studies
Since the above research appeared, the developmental literature on the DRM paradigm has been growing apace, which is surely connected to fact that an age trend has been identified that, though surprising, has proven to be easily replicable and highly robust (but see Ghetti, Qin, & Goodman, 2002). Surprising data stimulate investigators to formulate and test possible theoretical accounts, though this is one of those rare instances in which the data were predicted on theoretical grounds long before they were obtained. Most of the recent work on the DRM illusion is in this vein in the sense that it includes new manipulations or new measures that are intended to generate data on a specific theoretical hypothesis or to pit different hypotheses against each other. Considering that theoretical understanding is the surest way to make progress, this is a welcome development. In the remainder of this section, we consider six articles that are representative of this hypothesis-testing orientation: Dewhurst and Robinson (2004); Howe (2005); Holliday and Weekes (in press); Howe (2006); Brainerd, Forrest, Karibian, and Reyna (2006); and Brainerd and Reyna (2007).
Dewhurst and Robinson (2004)
These authors conducted a simple developmental recall study of DRM lists in which they added a measurement wrinkle that bears on FTT's hypothesis that young children's limitations in connecting the gist across words is a major factor in developmental increases in intrusions. Children from three age levels (5-, 8-, and 11-year-olds) studied and recalled a total of five DRM lists. Remember in this connection that Brainerd et al. (2002) found that young children's intrusions were qualitatively different than that of older children and adults in that they were not dominated by unpresented words that were semantically related to lists. Dewhurst and Robinson hypothesized that there is a developmental shift in the informational basis of false memory and, specifically, that young children's false memories may be dominated by a reliance on perceptual similarity (e.g., auditory or visual resemblance) that is later supplanted by a reliance on semantic similarity. To test this hypothesis, they classified the intrusions that children made into three categories: phonologically related (unpresented words that sounded like one of the list words when spoken), semantically related, and unrelated. When the mean numbers of the different types of errors were plotted for the different age levels, a dramatic Age x Type of Relatedness crossover was detected. On the one hand, as predicted by FTT, the number of semantically related intrusions increased from 0.7 to 1.4 to 2.7. On the other hand, the number of phonologically related intrusions decreased from 1.2 at the two youngest age levels to 0.2 at the oldest age level. Thus these data provide preliminary evidence of a perceptual-to-semantic shift in the basis for false memory, a theme that is echoed in two other recent studies (Brainerd & Reyna, 2007; Holliday & Weekes, in press).
Howe employed a directed-forgetting design that had been introduced in some adult DRM studies by Kimball and Bjork (2002). The directed-forgetting procedure consists of simply telling participants to forget (or remember) previously studied information, with "forget" or "remember"
cues being given either immediately after each item is presented (the item method) or after a series of items has been presented (the list method). Kimball and Bjork used the list method to test a prediction from FTT: directed forgetting should suppress true memory for DRM list words more than it should suppress false memory for critical distractors, because the verbatim traces that support true recall/recognition are more sensitive to post-presentation interference than the gist traces that support false recall/recognition. Their results were as predicted in that directed forgetting had no effect on false recall, though it suppressed true recall. Howe used this basic design in a DRM experiment with 5-, 7-, and 11-year-old children. Some children simply studied and recalled individual DRM lists in the usual way. Other children studied DRM lists in pairs, receiving a "forget" instruction following one of the lists and a "remember" instruction following the other. Like adults, the true memories of children of all age levels were suppressed by directed-forgetting instructions. Unlike adults, however, children's false memories were also suppressed by directed-forgetting instructions. Howe concluded that adults' DRM false memories are so automatic that they are not amenable to conscious control via instructions, but that children's are effortful and deliberate, which leaves them open to conscious control. Considering that adults are known to be better than children at virtually all forms of conscious control of memory, this is a counterintuitive hypothesis, but then the developmental pattern that needs to be explained is equally counterintuitive.
Holliday and Weekes (in press)
These investigators followed up Dewhurst and Robinson's perceptual ^ semantic shift hypothesis and provided more differentiated data on this shift, as a test of FTT's analysis of false memory. Holliday and Weekes proposed that FTT would expect such a shift on the ground that although children's ability to connect the gist across multiple words increases with age, they are able to detect and transfer phonological resemblance at a very early age. To evaluate this proposal, they studied false recognition of related distractors in 8-, 11-, and 13-year-old children with two types of lists. First, DRM lists, as we know, consist of semantic associates of critical distractors. Half the children at each age level studied such lists, followed by the usual recognition test with targets, semantically related distrac-tors, and unrelated distractors. The other half of the children studied lists that were phonologically related to each other. Lists of this sort were developed by Sommers and Lewis (1999) using procedures that are reminiscent of Deese's (1959) method of list generation. First, Sommers and Lewis selected 24 stimulus items, each of which was a familiar three-letter word (e.g., cat, hit). Next, English lexical databases were used to select the 15 words that were most phonologically related to each stimulus word. To take a concrete example, the first 15 phonological associates of cat are fat, that, cab, caught, cot, sat, cut, hat, kit, vat, cap, mat, bat, cad, and chat.
The key feature of Holliday and Weekes's (in press) data is that, in line with the perceptual ^ semantic shift hypothesis, developmental trends for semantic associates of lists and phonological associates of lists were mirror images. This can be seen in Figure 9.4, where false-alarm rates for DRM critical distractors and unpresented phonological associates are plotted by age level. The two developmental trends are virtually mirror images of each other. The tendency to recognize unpresented semantic associates as old increased from a little over 60% to a little over 80%, whereas the tendency to recognize unpresented phonological associates as old decreased by almost exactly the same amount, from a little below 40% to a little below 20%. Also, it can be seen that perceptual and semantic illusions were equally seductive to 8-year-olds, whereas semantic illusions were more seductive to 11- and 13-year-olds. A prediction that would follow from the shape of the trends in Figure 9.3 is that if this design
Was this article helpful?
Parenting is a challenging task. As a single parent, how can you juggle work, parenting, and possibly college studies single handedly and still manage to be an ideal parent for your child? Read the 65-page eBook Single Parenting Becoming The Best Parent For Your Child to find out how. Loaded with tips, it can inspire, empower, and instruct you to successfully face the challenges of parenthood.