Most researchers believe that preverbal infants possess only a primitive, implicit memory system which processes information automatically and without awareness. They hold that the explicit memory system, which mediates conscious recollection, does not mature until late in the 1st year. This widespread belief, however, is based solely on memory studies with aging amnesics--not on studies with infants! This article reviews new evidence that even the youngest infants exhibit the same memory dissociations as normal adults on recognition (explicit) and priming (implicit) tests, revealing that the roots of both memory systems can be traced to earliest infancy.
Since the time of Freud, scientists have sought the roots of adult memory in the infancy period. This search has led to the view that adults possess two independent memory systems that emerge hierarchically during the 1st year of life--a primitive system that is present early in infancy and a late-maturing system that appears at the end of the 1st year.
The notion of two memory systems originated with clinical observations that amnesics are impaired relative to normal adults on one kind of memory test but not on another. Amnesics, for example, perform poorly on recognition tests when asked to choose which of four words was on a list they had just studied minutes earlier, but they perform as well as normal adults on priming tests when asked to complete word fragments with the first words that come to mind. Despite being unable to recognize them, amnesics typically complete the fragments with words from the prior study list. Such dissociations in memory performance led researchers to assume that recognition and priming tests tap different underlying memory systems--one that is impaired in amnesia and one that is not.
Applying the Jacksonian principle of the hierarchical development and dissolution of function to memory--a first in/last out account, researchers assumed that the memory system that is spared in amnesia--the "last to go"--develops first, whereas the memory system that fails in amnesia matures last.
Over the years, many multiple memory systems have been proposed. The most widely accepted today are nondeclarative/declarative memory, semantic/episodic memory, and implicit/explicit memory. Table 1 lists their distinguishing characteristics. Although "conscious awareness" was a late addition to this list, it has become the defining characteristic of explicit memory.
The fact that preverbal infants cannot say whether they consciously recollect experiencing a particular event before has contributed to the assumption that they lack explicit memory. This assumption, however, has never been studied directly in infants but has only been inferred from studies of aging amnesics. For many years, my laboratory has studied long-term memory in 2- to 6-month-olds. Our data suggest that if there are two independent memory systems, then they develop in parallel and not hierarchically from early in life. In approaching this issue, we have not appealed to the notion of conscious awareness; rather, we have simply asked if preverbal infants show the same dissociations as adults on memory tasks analogous to those that were the basis for positing two different memory systems in the first place.
First, let me describe our procedures. Because infants lack a verbal response to say what they recognize, we teach them a motoric one--a foot-kick that moves a particular crib mobile. During the retention test, we show them a mobile that is either the same or different in some way, and infants "say" whether or not they recognize it by whether or not they produce the motoric response. If they recognize it, they kick above baseline ("yes"); otherwise, they do not kick above baseline ("no").
We teach infants to kick by stringing a ribbon from one ankle to the overhead hook that suspends the mobile. Infants learn rapidly that kicking moves the mobile and usually double or triple their kick rate within a few minutes (see Figure 1a). Before and after training, we measure baseline and immediate retention, respectively, by attaching the ribbon to an "empty" hook so that infants can see the mobile, but kicks cannot move it (see Figure 1b). Long-term retention is measured under these same conditions.
After training, infants are subjected to one of two different memory tasks (see Figure 2). In the delayed recognition task, we simply hang the test mobile over the infant's head and ask if he/she recognizes it. In the reactivation task, we briefly expose infants to a memory prime (a reactivation treatment) at some point before the long-term test (see Figure 3). The prime, a component of the original training event (e.g., the original mobile or context), reactivates the memory, increasing its accessibility. Later, the prime's effectiveness is assessed in a delayed recognition test.
As an example of how priming works, consider that at 3 months of age, delayed recognition gradually declines as a function of the time since training. After 6-8 days, infants exhibit no evidence of remembering the mobile game. Yet, if they are exposed to a brief mobile prime on Day 13, it restores their retention to the same level seen immediately after training, and then the reactivated memory is again forgotten gradually, at the same rate as it was before! Thus, the effect of the prime on the accessibility of a memory is not fleeting but can persist for many days.
The delayed recognition and reactivation tasks used with infants correspond to the recognition and priming tasks, respectively, that are used in studies of implicit and explicit memory with adults. For both adults and infants, priming is thought to reactivate or increase the accessibility of a memory representation. Also for both, priming is item-specific and appears to initiate an automatic perceptual identification process that results in memory retrieval. Likewise, the delayed recognition task used with infants corresponds to yes/no, same/different recognition tests used with adults.
A number of independent variables affect memory performance on priming and recognition tests differently in adults, who presumably have two memory systems. These same variables, however, have the same effect on memory performance in reactivation and delayed recognition tests with preverbal infants, who presumably have only one memory system. The evidence for each independent variable is briefly reviewed below.
Retention interval. In adults, the magnitude of recognition decreases with the interval between study and recognition testing, but performance on priming tasks is all-or-none and stable over the same delay. Similarly, the magnitude of recognition at 3 and 6 months decreases with the interval between training and the delayed recognition test, but the magnitude of reactivation is all-or-none, unaffected by the interval between training and priming.
Vulnerability. In adults, explicit memory is vulnerable to interference, but implicit memory is not. Similarly, 6-month-olds show excellent recognition in the training context 1 day later, but exposure to a novel context between training and testing eliminates recognition in the original context--a classic retroactive interference effect. Nonetheless, the magnitude of reactivation is invulnerable to the same retroactive interference procedure: Infants show excellent recognition in the original context whether they are exposed to a novel one between priming and testing or not.
Age. The magnitude of recognition increases from childhood to middle age, but the magnitude of priming remains constant over this period. Likewise, the magnitude of delayed recognition after
1 week increases between 2 and 6 months, but the magnitude of reactivation after 3 weeks is the same at all ages.
Number of study trials. Increasing the number of study sessions (equivalent to study trials) for both adults and infants protracts delayed recognition but has no effect on reactivation: At 3 months, each additional session prolongs recognition by 1 week, but the magnitude of reactivation after 3 weeks is the same whether infants receive two sessions or three.
Amount of study time. As for adults, increasing the study time protracts infants' delayed recognition. With 9 minutes of study time, 3-month-olds show no retention 1 week later; with 12 minutes of study time, they remember for 1 week, and with 18 minutes, they remember for 2 weeks. (We have no reactivation data for infants.)
Number of studied items. Infants and adults show the same dissociation for the number of studied items. When the number of items was defined by the number of different types of blocks on the training mobile, the magnitude of recognition after 1 day was better when the study list was shorter at both 2 and 3 months, but the magnitude of reactivation after 3 weeks was the same.
Affect. In adults, recognition is sensitive to affect, but performance on priming tests is not. When 3-month-olds being trained with a 10-object mobile were suddenly switched to a 2-object mobile, 50% of them cried. One week later, noncriers exhibited excellent recognition, but criers exhibited none. Infants exhibited the same magnitude of reactivation 3 weeks later, however, whether they had cried or not.
Serial position. Adults who learn a serial list display a primacy effect on recognition tests after longer delays not on priming tests. Likewise, 3- and 6-month-olds who were trained with a serial list composed of three different mobiles displayed a primacy effect on a delayed recognition test 1 day later. When primed with a mobile from the immediately preceding serial position just before the 24-hour test, however, they recognized all mobiles equally well, irrespective of serial position.
Memory load. For both adults and infants, increasing the memory load impairs performance on recognition tests but not on priming tests. Increasing the length of the serial list from three mobiles to five eliminated the primacy effect, and 6-month-olds recognized all mobiles from the list, irrespective of serial order. They did not recognize a novel mobile, however, indicating that they did not respond indiscriminately to all test mobiles when the list was longer; rather, they recognized only mobiles from the original list--but they no longer remembered their order.
Level of processing. This is an encoding variable that affects adults' performance on recognition but not on priming tests. We studied this variable with 3-month-olds by training them with a mobile containing a single "L" block (the target) amidst six "+" blocks (the distractors). When infants are tested with this mobile, the single "L" apparently pops out and captures their attention, because they behave as if the mobile contains only "Ls." We exploited this phenomenon as a means of enhancing infants' attention to the target and increasing its depth of processing during encoding. In fact, infants trained with one "L" amidst six "+s" remembered a test mobile displaying seven "Ls" longer than infants trained with seven "Ls" in the first place. The same result was found when infants were trained with one "+" amidst six "Ls" and tested with seven "+s."
Studied size. For adults, changing object size impairs performance on recognition but not on priming tests. When 3-month-olds were trained with "+s" of one size and tested with "+s" that were 33% larger or smaller, their delayed recognition 1 day later was impaired, but the magnitude of reactivation 2 weeks later was the same whether the size of the "+s" was changed or not.
The preceding data demonstrate that preverbal infants exhibit exactly the same memory dissociations as adults in response to exactly the same independent variables. Although some might argue that infants' memory dissociations are only dissociations within a single, primitive memory system, it is illogical to use these same dissociations as evidence for two memory systems in adults and for only one memory system in infants. In addition to exhibiting the same memory dissociations as adults, however, young infants also exhibit other characteristics of explicit memory (see Table 1). Three examples are described below.
First, because the implicit memory system does not encode information about specific events, young infants, with only that system, are thought to be unable to encode and retrieve information about a specific prior episode--but they can. In one study, we trained infants with a series of yellow-block mobiles displaying "A's" in different colors and then briefly showed them a stainless-steel and colored-glass wind chimes either immediately after training or 4 days later. One day after this exposure, infants kicked vigorously during a delayed recognition test with the wind chimes, apparently attempting to move it with the same response they had previously used to move the yellow-block mobiles. Moreover, the magnitude of delayed recognition was greater when the wind chimes was exposed immediately after training than when its exposure was delayed by 4 days, but the magnitude of reactivation 2 weeks after the wind chimes prime was the same whether its exposure was immediate or delayed--the familiar explicit/implicit dissociation. Clearly, infants' recognition of the wind chimes could only have been based on their memory of the single, specific occasion that they were shown it. Recall that they never were reinforced for kicking to move it, and they never practiced kicking to move it--they only looked at it on one, brief occasion.
Second, implicit memory is characterized as nonassociative, but even 6-month-olds, who presumably possess only implicit memory, have exhibited evidence of an associative network. In one study, infants learned to move a mobile or turn on a music box by either arm-pulling or foot-kicking, and then they learned the other cue - response pair in the same distinctive context. This procedure will be recognized as an analog of the traditional paired - associate task. Three days later, when infants received a delayed recognition test with the mobile, they produced only the mobile-appropriate response, be it kicking or arm-pulling, and not the music box response, even if they were trained with the music box last. This result shows that each training memory was specific to its cue and independent of the other memory.
Three weeks later, new training groups were again tested with the mobile, but first they were primed with either the mobile or the music box. As expected, infants who were primed and tested with the mobile exhibited the mobile-appropriate response, but infants primed with the music box and tested with the mobile also exhibited the mobile-appropriate response--and no other. In other words, priming with the music box "brought to mind" the memory of the mobile task as well, enabling infants to produce the mobile-appropriate response. Because these infants were not primed with the mobile, and the mobile task was forgotten when they were exposed to the music box prime, it must have indirectly primed or reactivated the memory of the mobile task, probably via a mechanism akin to spreading activation. Thus, infants' independent memories of the music box and mobile were associatively linked, possibly by their common training context, in a mnemonic network.
Third, although implicit memory is context-independent, young infants' memory of training is highly context-dependent. Three-month-olds, for example, were trained in their bedroom in a porta-crib that was draped on all sides with a distinctively colored and patterned cloth. Later, some infants were exposed to a memory prime in the bedroom, and others were exposed to it in the kitchen--a familiar room but different from where they were trained. Even though both groups were primed in the same porta-crib draped with the same distinctively colored-and-patterned cloth, the prime successfully reactivated the memory only in the room where infants were trained. Apparently, the details of the kitchen that were visible above the porta-crib prevented the prime from being effective in a different place. Similarly, when 6-month-olds were trained in one room in their home and tested in another just 1 day later, they failed to recognize the original training cue in a different place!
Figure 4 shows infants ranging from 2 to 18 months of age. Notice the vast difference between the youngest and oldest infants. The maximum duration of their delayed recognition is summarized in Figure 5. There are three important points to notice. First, at 6 months of age, memory performance in the mobile and train tasks is identical. Second, there is no indication that memory changes abruptly at the end of the 1st year, when the late-maturing memory system presumably emerges. And third, there is no indication that memory changes in the 2nd year with the appearance of language. In short, memory is remarkable for its continuity over significant periods of development.
The preceding body of evidence reveals that the roots of adult memory systems can be traced to earliest infancy, and that these systems develop in parallel--not hierarchically--thereafter. Let me caution, however, that uncritically accepting the notion of dichotomous memory systems is an easy trap to fall into. Although simple dichotomies--like 2-way interactions--are easier to undfirstand than numerous systems or sub-systems involving 16-way interactions or more, a this-or-that model is unlikely to capture the variety and richness of what each of us remembers on different occasions. Although perceptual identification and recognition appear to be different processes, for example, per-ceptual identification may be a precursor of recognition rather than part of a memory system entirely distinct from it. Hopefully, in the future, we will be able to progress beyond descriptive labels and find out how memory processing actually works!
|Figure 1. A 3-month-old in the mobile task (a) during training and (b) during baseline and all retention tests. During acquisition, the infant moves the mobile by means of an ankle ribbon connected to the mobile hook.|
Figure 2. (a) The delayed-recognition task, showing training and the long-term retention test. (b) The reactivation task, showing training and the brief reactivation (priming) treatment prior to the long-term retention test. The test cue in (a) is the memory prime in (b).
Figure 3. A reactivation treatment with a 3-month-old. The loose end of the ribbon is held by the experimenter at the side of the crib.
Figure 4. Infants of the ages tested in the mobile and train tasks, from left to right, they are 2, 3, 6, 9, 12, 15, and 18 months of age. Note the dramatic physical and behavioral differences between the younger and older infants.
Figure 5. The maximum duration of retention (in weeks) over the first 18 months of life. Independent groups of infants were studied in the mobile (2-6 months) and the train (6-18 months) tasks. Six-month-olds were trained and tested in both tasks.
article is condensed from the Psi Chi Distinguished Lecture presented
at the annual meeting of the Midwestern Psychological Association, May
1998, Chicago, IL. It is based on a report by the author, "Dissociations
in Infant Memory: Rethinking the Development of Implicit and Explicit
Memory," that appeared in Psychological Review, 1997, 104, 467-498.
References to the specific studies that underlie the general findings
described in the present article can be found there.
this article and the research described in it was supported by grants
R37-MH32307 and K05-MH00902 from the National Institute of Mental
Health. All correspondence pertaining to this article should be directed
to Carolyn Rovee-Collier, Department of Psychology, Busch Campus,
Rutgers University, 152 Frelinghuysen Road, Piscataway, NJ 08854-8020. Electronic mail can be sent to firstname.lastname@example.org.