Implicit Memory in Childhood:
Reassessing Developmental Invariance
Marianne E. Lloyd & Nora S. Newcombe
To appear in M.L. Courage & N. Cowan (Eds.),
The Development of Memory in Childhood, 2nd edition.
London: Psychology Press
It’s like riding a bike: once you’ve learned it, you’ll never forget how to do it.
Although it is uncertain whether this adage has been tested empirically, personal experience of the authors suggests that the passage of many years is indeed insufficient to destroy bike-riding ability. Recently, skills such as bike-riding have been termed implicit memory, and their retention has been
contrasted with what is seen with explicit memory tasks, in which participants are aware that they are
making a memory decision, as occurs when answering the question, ―Who taught you to ride a bike?‖.
Implicit memory has been described as encompassing various tasks, including not only acquisition of motor skill, but also priming (as when something comes quickly and easily to mind because it has been encountered previously) and classical conditioning. (For a typical graphic depiction of these distinctions, see Squire & Zola, 1996). There has been considerable interest in typologies of this kind as supporting investigations of the architecture of human memory that are informed by combinations of evidence from behavior and neuroscience, and that encompass research with a variety of human and non-human populations.
The terms explicit and implicit memory were proposed by Graf and Schacter (1985; also see Schacter, 1987). In the 30 years since the introduction of these terms, several debates have occurred over the necessity to posit multiple memory systems (e.g., Tulving, 2002) and the possibility that such dissociations are based on tasks rather than processes (Roediger & Blaxton, 1989; Toth & Hunt, 1999). However, the distinction seems to have withstood the test of time, and is in wide use today (e.g., in introductory psychology, physiological psychology, and cognitive psychology texts). Operationally, explicit memory refers to tasks in which people are asked to evaluate their memory, either by saying that they did or did not encounter something before (i.e., recognition test), or by producing a previously-
encountered stimulus or fact (e.g., autobiographical recollections, free recall, cued recall). In contrast, when the task does not refer to a study episode (e.g., naming tasks, priming measures, stem completion, sequence learning, conditioning), we refer to the results as a measure of implicit memory.
Developmentally, it has been claimed that implicit memory is present robustly from the start of life, and does not undergo the kind of age-related changes that are commonly seen in explicit memory (Reber, 1989). In fact, this theme was the dominant conclusion of the chapter by Alan Parkin (1998) covering implicit memory in the first edition of this volume. The claim has several attractive characteristics. First, explicit memory improves dramatically over the course of childhood (for a review see Kail, 1990), and many standard explicit tasks such as recall and recognition are simply too difficult to use with children younger than three years of age. Although there is increasing recognition that explicit memory is present in infancy (for a review, see Bauer, 2006), implicit memory measures still seem to offer a potentially more sensitive way to determine what information young children have encoded, while still employing measures that can also be used with adults. Second, implicit memory seems to offer a way of understanding a paradox in cognitive development, namely, the fact that early competence is often discerned using looking time measures while, in contrast, studies examining deliberate actions and explicit predictions often show protracted phases of immaturity (see Keen, 2003). That is, perhaps initial knowledge is implicit in nature, but considerable experience with the world is necessary to create explicitly accessible information that can be used to make judgments.
There is actually considerable uncertainty, however, concerning whether implicit memory is as developmentally robust as investigators believed at first. There are several reasons to rethink this position. First, it was based primarily on studies of perceptual priming, without much consideration of other types of implicit memory, or of conceptual priming (Roediger & Blaxton, 1989). Second, it became associated with other propositions that are not logically entailed by it, notably the idea that implicit memory is resistant to decay or interference (as in the bike riding example). Third, it was based on studies of children that began at quite an advanced age from the point of view of discussion of developmental origins of knowledge or invariance over the life span—rarely younger than 3 or 4 years of age.
Our aim in this chapter is to reconsider the development of implicit memory in a more differentiated way than was possible when Parkin’s chapter in the first edition was written. However, a
complete review of implicit memory development is beyond the scope of one chapter; indeed, an entire book on the topic has been written (Rovee-Collier, Hayne, & Colombo, 2000). We have chosen to restrict our scope in several ways. First, we will not discuss the development of explicit memory except when it is useful to make contrasts with implicit performance, because many other chapters in this book deal with aspects of explicit memory such as autobiographical or source memory. Second, we will concentrate on developments in childhood rather than address in detail the issue of what infant memory measures are implicit and which are explicit, which has a complicated history of its own (see chapter 2 in this book and also chapters 5-8 in Oakes and Bauer, 2007).
We first review what the support was for Parkin’s (1998) claim that ―procedural memory appears age invariant‖ (p. 124). Most of the evidence for the invariance claim came from studies in which implicit memory was tested using perceptual priming tasks. These findings of equivalent priming effects across childhood have largely been supported subsequently, but there have also been studies showing developmental change in conceptual priming tasks that have received much recent attention. We argue,
however, that this work entangles issues of priming with issues of conceptual growth and additions to semantic memory, so that the data do not actually challenge the invariance claim in a fundamental way. We then conclude the section on priming with a discussion of what is known about the interplay of implicit with explicit memory, where there does appear to be considerable, and rather late, developmental change. In the second major section, we review findings regarding another kind of implicit memory, different from priming, not covered in the first edition of the book, and for which there is clear evidence that memory does change developmentally, namely, sequence learning. We close the chapter with a brief review of some developments in neuroscience, implicit memory in clinical populations, and suggestions for future research.
One of the most striking aspects of the implicit memory construct is that it encompasses many compelling everyday experiences. When a familiar face pops out of a crowd, perceptual priming may be
at work. When a seemingly novel idea comes to mind, and we later realize that the idea was proposed by someone else in conversation a few days ago, a different kind of priming is at work—conceptual rather
than perceptual (as well as, sadly, an embarrassing rather than an adaptive aspect of priming). In this section, we consider the developmental invariance hypothesis for perceptual priming, followed by an examination of conceptual priming (including its effects on false memory), closing with consideration of how children come to realize that processing fluency, i.e., the speed and easy with which an item is perceived, may be a clue to making explicit memory judgments.
Perceptual priming is studied experimentally in situations in which we can be sure that stimuli have been encountered, but in which it is likely that they will not be explicitly remembered. We can then seek evidence that the stimuli are processed more quickly and easily as a function of prior exposure, despite the lack of recall, or even recognition. For example, when participants have recently studied the word ―tulip‖, they can read it faster when it appears on a screen, identify it at a higher level of perceptual masking, or respond to it more quickly on a lexical decision task, compared to a group that has not recently seen the item. Similarly, on a stem completion task, participants are more likely to solve the stem ―tu___‖ with ―tulip‖ than people who have not recently seen the word ―tulip.‖ These effects can occur independently of whether they recognize the word on a later recognition test, list it in a recall test, or can accurately remember that the word was viewed as a picture during a source monitoring test (for a review of the effects of encoding and retrieval variables on memory performance, see Richardson-Klavehn & Bjork, 1988; Roediger & McDermott, 1993). Further, unlike explicit tasks such as recall, these priming effects seem to be relatively stable throughout adulthood (e.g., Fleischman, Wilson, Gabrieli, Bienias, & Bennett, 2004).
Much initial work on implicit memory development focused on applying the paradigms typically used with adults to work with children. Some of this work has been verbal (Billingsley, Smith, &
McAndres, 2002; Komatsu, Naito, & Fuke, 1996; Perez, Peynircioglu, & Blaxton, 1998). For example, participants may be asked to complete word stems or fragments after studying a list of items. Some of these may be completed with studied items while others would be completed with novel items. Priming is
measured by an increased likelihood in completing items with studied words. Such techniques are not suitable for children who cannot read, however. To modify the methods, studies with children as young as three have used pictorial stimuli (Billingsley, et al., 2002; Cycowicz, Friedman, Snodgrass, 2000; Hayes & Hennessey, 1996; Parkin & Streete, 1988; Perez, et al., 1998). For example, Billingsley, et al. (2002) demonstrated that children and adults perform similarly on two implicit tests: both verbal (category generation) and pictorial (identification) tasks. Working only with pictures, Parkin and Streete (1988) showed common objects to children aged 3, 5, and 7 and later had the children identify distorted pictures that were slowly made clearer. In such a paradigm, priming is represented by faster naming of pictures that have been recently presented. In general, particularly when levels of explicit memory performance were made equivalent (e.g., by reducing the length of the study list for younger children), perceptual priming effects were equivalent across age.
Along the same lines, Drummey and Newcombe (1995) showed pictures to 3- and 5- year old children as well as adults and later tested their ability to recognize the objects at various degrees of perceptual degradation. The degree of priming (higher masking level for identification for studied versus unstudied pictures) was fairly consistent for the three groups of participants. The participants were also asked for recognition judgments, and this measure of explicit memory showed expected age-related improvement. Similar results were reported by Hayes and Hennessey (1996) who also used implicit (picture identification) and explicit (recognition) memory tasks. In their study, children who were 4, 5, and 10 years of age studied a list of pictures and returned 48 hours later for a memory test. During the test, participants first attempted to identify pictures at varying degrees of blurriness. After identifying a picture, participants made a recognition judgment. There was no difference in the degree of priming among the 4-, 5-, and 10- year-olds. That is, the advantage of having studied a picture on subsequent identification tasks was similar for all three ages. However, recognition memory improved from age 4 to 10. Notably, in both studies, the implicit measure of picture identification was not correlated with the explicit measure of recognition; when there are such correlations, developmental invariance may not be found because explicit memory improvements may create what appear to be implicit memory effects (Parkin, 1998).
As we have seen, evidence that has been accumulating since the first edition of this book continues to suggest that perceptual priming is relatively stable throughout development. One exception is that Cycowicz, Freidman, Snodgrass, and Rothstein (2000) have reported evidence for developmental improvements between the ages of five and nine on a picture identification task. However, the task was harder than that used in other studies – participants were under time pressure, and the investigators
acknowledge that differences in retrieval speed may have caused the apparent improvement in priming. Further, the youngest children still demonstrated significant priming for the repeated objects.
Findings of equivalent priming rates across age have been taken as evidence that the implicit memory system is older evolutionarily than the explicit system (Reber, 1989) and that it is an independent memory system (e.g., Schacter & Graf, 1985). The idea that perceptual priming is equivalent across the
lifespan has not, however, been tested at ages younger than 3 years or so. Although recent data suggest that priming effects exist in infancy (Myers, Clifton & Clarkson, 1987; Perris, Myers & Clifton, 1990; Rovee-Collier, 1997; Snyder, 2007; Webb & Nelson, 2001), their magnitude has not been cleanly compared to effects shown by older children, and in fact, such an enterprise is fraught with methodological and conceptual difficulty. For example, work by Clifton and colleagues have shown that infants maintain some memory of motor responses for actions completed in infancy for long periods of time. However, studies with older children and adult rely primarily on verbal or pictorial measures of priming. It is certainly possible that new and more sensitive techniques may show some early changes in the magnitude of perceptual priming. Given the current state of the evidence, it appears that Parkin was correct in his assessment in this book’s first edition: perceptual priming effects are similar from the
preschool years on, and remain stable throughout normal aging. However, true developmental invariance beginning with infancy has yet to be assessed.
Perhaps because the issue of developmental invariance of perceptual priming, at least from 3 years on, seems relatively settled, much recent research in implicit memory development has focused on the contrast between perceptual priming and conceptual priming. Conceptual priming often shows marked improvements with age. For example, although children of different ages show equivalent priming rates
for recognizing a picture of a bear that has been presented earlier, young children are often less likely than older children to subsequently complete the stem b________ with the word ―bear‖ (e.g., Komatsu et al.,
1996). Similarly, when asked to list words that fit the category ―vegetable‖ older children and adults are more likely to respond with words that were recently presented, relative to younger children (e.g., Billingsley et al., 2002; Perez et al., 1998). Whereas perceptual priming is based on the physical overlap between prior and subsequent presentation, conceptual priming is assumed to be based upon the activation of related concepts in memory. Using the example of the category generation task mentioned above, a person who has recently been presented with the words ―eggplant‖ and ―corn,‖ should be more likely to include these two items in a task asking for a list of vegetables.
The evidence for developmental invariance of conceptual priming has been mixed. Some work in conceptual priming suggests that it is age invariant while other work suggests developmental changes. A direct comparison between perceptual and conceptual priming was conducted by Perez et al. (1998), who also studied perceptual and conceptual memory of an explicit nature. In their study, preschool and elementary school age children as well as adults participated in four memory tests (perceptual explicit: pictorial cued recall; conceptual explicit: category cued recall; perceptual implicit: picture identification; conceptual implicit: category production). The advantage to the design of this study was having all four measures available for each participant. Perez et al. found that neither the perceptual nor the conceptual implicit tests showed improvement with age. Similarly, Billingsley et al. (2002) found equivalent priming across age with a category generation task. Unfortunately, the youngest group of participants was 8-10 years of age. It may be that younger children would have shown lower performance on the task. However, other researchers do report differences in conceptual priming during development. For example, Perruchet, Frazier, and Lautrey (1995) showed that children improved in measures of conceptual priming when the exemplars were atypical but not when they were typical members of the category. This finding has since been replicated several times (e.g., Mecklenbrauker, Hupbach, & Wippich, 2003; Murphey, Mckone, & Slee, 2003).
Although these changes have been argued to be evidence for conceptual priming development, we would argue that perhaps this is an unfair test of priming. That is, it is not priming that is changing,
but rather the concepts that these paradigms attempt to test. For example, it has been well documented that the categorical structure of concepts such as animals grows with age. Thus, preschool children may show conceptual and perceptual priming invariance, if tested with suitable materials. Further, we suspect that children may even be able to show better conceptual priming than adults when the stimuli are made
to be relevant to their lives (e.g., for many toddlers today the popularity of The Wiggles probably makes
Dorothy a better prime for ―Dinosaur‖ than for ―Wizard‖). Although many of the papers that find age
differences are willing to attribute them to differences in memory, we believe that it is unfair to consider these tasks assessments of conceptual priming capacity.
One line of research that is relevant to developmental changes in category structure is work by Sloutsky and colleagues (Fisher & Sloutsky, 2005; Sloutsky & Fisher, 2004). In this work, an incidental encoding task is used, in which many exemplars of the same animal are presented (e.g., pictures of cats). During a surprise recognition memory test, children outperform adults. That is, they are better able to recognize the pictures of cats that were presented in the study list relative to pictures of other cats that were not presented. The explanation for this pattern is that adults naturally categorize the pictures as "cats" while the children maintain more of the individual features of each item in memory. Contrasts of this kind suggest that changes in conceptual priming ability may be better understood by changes in categorization and semantic organization.
Our point about the role of knowledge in age differences in conceptual priming was also made by Murphey et al. (2003), who argue that conceptual priming reflects changes in knowledge accessibility. Murphey et al. also suggest that priming may be a way to test knowledge levels in children (p. 159).While this is true, an equally important question in studying implicit memory development should be the capacity for perceptual and conceptual priming across childhood. That is, is it possible to test conceptual priming in a way that is fair to both adults and children independent of pre-experimental knowledge? We advocate for attention to the basic mechanisms supporting priming as well as to the outcome of priming
tasks. Given the vast age-related changes in semantic structures and general knowledge, future work aimed at investigating differences in conceptual priming needs to be first concerned with finding measures for which children of different ages have equivalent knowledge for the concept.
One way to achieve this goal would be to look at novel category learning. This would control the level of knowledge for each category for both children and adults. Additionally, this would allow for a paradigm that can test both conceptual and perceptual implicit memory in the same children while controlling exposure duration. To our knowledge, research has not yet been conducted with such a design. Rather, work with novel categories has focused on generalization and reasoning (e.g., Sloutsky, Kloos, & Fisher, in press) That is, instead of testing for priming effects, these studies have focused on the way that young children learn about classification of novel items. These stimuli should prove very useful in research on priming as well.
Age Differences in False Memory
Increases in conceptual priming with age are not an unmixed blessing. In the false memory illusion (e.g., Deese, 1959; Roediger & McDermott, 1995; for a review see Gallo, 2006), related lists of words are presented that center around a topic (e.g., bed, rest, dream). The critical theme word that ties the list together is never presented (e.g., sleep). Adults are very likely to falsely recognize or to recall the word "sleep" on a memory test, but children do so to a much lesser degree (e.g., Holliday & Weekes, 2006; Howe, 2006). As in Sloutsky’s work on categorization in young children, the false memory effect is
an example of children outperforming adults on a memory task. It is assumed that one explanation for this pattern of results is that children do not have the semantic organization that would cause the word "sleep" to be more familiar after presentation of related words (e.g., Brainerd, Forrest, Karibian, 2006).
Another difference between children and adults in false memory may involve a shift from phonological to semantic processing across childhood. In work by Dewhurst and Robinson (2004), children (ages 5, 8, and 11) studied lists of categorized words and were then given a free recall task. The youngest children made significantly more phonological than semantic errors (e.g., were more likely to recall ―head‖ after studying ―bed‖ than to recall ―sleep‖ after studying ―bed,‖ ―rest,‖ and ―dream.‖) The pattern was the opposite in 11-year olds: they were more likely to recall the semantically related words. Semantic and phonological errors were balanced for 8-year-olds. These results suggest that the way in which children are storing information shifts from a phonological to a semantic base over time.
Further evidence for links between categorization and false memory has been reported by Holliday et al. (2006). In these studies, participants receive lists of categorized words similar to those used in the DRM paradigm described above. Rather than taking a standard recognition test, in which the goal would be to respond ―no‖ to a critical item such as ―sleep,‖ participants are asked to take recognition tests that ask them not only to respond positively to items that were presented during the study phase but also to items that are semantically related to those that were studied. Between the ages of 5 and 14 there are dramatic increases in the likelihood of responding positively to words that are highly related (e.g., sleep). This suggests that children are better able to determine how items are related to one and other along semantic dimensions with increases in reasoning skills and better semantic knowledge. The finding also supports the possible reason for the equivalent category generation in children between the ages of 8-10 that was reported by Billingsly et al. (2002). That is, by middle childhood, there is a better semantic network from which to achieve conceptual priming.
The previous examples using the DRM paradigm highlights measures of memory change in which participants actively generate responses. Such situations are generally conceptualized as belonging to the class of explicit memory, although it is possible that related associates are automatically activated (Underwood & Riechardt, 1977). To date, only one study has looked at memory for critical lures using implicit memory techniques (Diliberto-Macaluso, 2005). In this work, fourth and fifth grade children performed stem completion tasks under either explicit or implicit instructions. That is, participants were either encouraged to complete the lists with words that had been presented earlier, or were simply told to list words that came to mind. The children showed high levels of priming for the critical words in both the explicit and implicit conditions. This result fits nicely with the results of Billingsly et al. (2002), who found intact conceptual priming in 8-10 year old children, and suggests that by age 10, semantic associations are better formed. More work along these lines with younger children would be welcome. The Interaction of Priming and Explicit Memory
Although implicit and explicit memory may be functionally distinct (e.g., Rauch et al., 1997; Grafton, Hazeltine, & Ivry, 1995), in adults there is interplay between the two types of memory. For example, participants may use the experience of priming as a clue for explicit memory decisions. That is,