For feedback on earlier versions of this manuscript, we thank R. Barr and N. Zmyj. We thank all the infants and parents who participated in this research. We also thank Harlene Hayne for providing the puppet stimuli.
Ethics approval: This research was approved by the Faculty of Psychology ethics review board at the Ruhr University Bochum. Informed parental consent was obtained before participation. Open access funding enabled and organized by Projekt DEAL,
Why do infants remember some things and not others? Human infants frequently cycle through different states such as calm attentiveness, wakeful activity, and crying. Given that cognitive processes do not occur in isolation, such fluctuations in internal state might influence memory processing. In the present experiment, declarative memory in 9‐month‐old infants (N = 96) was heavily state dependent. Infants exhibited excellent retention of a deferred imitation task after a 15‐min delay if their state at encoding was identical to their state at retrieval (e.g., calm). Infants failed to exhibit retention if their state at encoding was different from their state at retrieval (e.g., calm vs. animated). Infant memory processing depends on internal cues.
Human infants can appear unpredictable. Compared to adults, they fluctuate more rapidly between different internal states, such as sleep, calm attentiveness, wakeful activity, and crying (Clifton & Nelson, 1976). These fluctuations bring with them obvious behavioral changes, making them a trademark of infancy. The potential relevance of state for early memory was evident in some earlier studies using an operant conditioning procedure (Fagen, Ohr, Fleckenstein, & Ribner, 1985; Fagen, Ohr, Singer, & Klein, 1989). In these studies, 3‐ to 4‐month‐old infants who cried in response to a shift in the reinforcer during training and who were tested in a calm state exhibited disruptions in their long‐term memory, interpreted as a retrieval failure by the authors. The reason for the retrieval failure might have been be a mismatch in infant state at encoding and retrieval in that the physiological and affective states associated with crying became part of infants’ memory of the learning experience (Fagen et al., 1989).
This idea shares similarities with the phenomenon of state‐dependent memory (Bower, 1981). Research with adult populations has shown that, for example, being in the same mood at encoding and retrieval can facilitate memory performance compared to being in a different mood at encoding versus retrieval (Eich, Macaulay, & Ryan, 1994). Similarly, being either drunk or either sober at both encoding and retrieval can help memory, compared to being once drunk and once sober (Goodwin, Powell, Bremer, Hoine, & Stern, 1969). Associative network theory can explain state‐dependent memory by conceptualizing aspects of internal state, such as emotions, as nodes in a semantic network (Bower, 1981). These nodes are linked with other nodes in the network that describe personally experienced events during which a particular internal state was activated. If a particular internal state gets aroused again, activation spreads from the representing node to the connected nodes containing information about events that happened when this state was experienced previously. This increases the chance of recalling these events.
One characteristic of early memories is their high degree of specificity. To access a particular memory, human infants require an almost perfect match between external cues present at encoding and external cues present at retrieval (e.g., Borovsky & Rovee‐Collier, 1990; Hayne, Boniface, & Barr, 2000; Hayne, Greco, Earley, Griesler, & Rovee‐Collier, 1986; Robinson & Pascalis, 2004; Rovee‐Collier, Schechter, Shyi, & Shields, 1992). For example, 6‐month‐old infants fail to exhibit memory for an action sequence after a 24‐hr delay if the encoding and testing locations differ (Hayne et al., 2000). They show excellent retention after this delay if the location of retrieval is identical to that of encoding (Barr, Dowden, & Hayne, 1996; Collie & Hayne, 1999). While such findings show that infant memories are impressively detailed and a far cry from a “blooming, buzzing confusion” (James, 1890), they also suggest boundaries in the efficiency of knowledge acquisition. As adults, humans flexibly adapt to novel situations. We can operate a new coffee machine, solve an unfamiliar math problem, or drive a rental car because we retrieve related knowledge from memory and build upon it in a new situation. Infants, in contrast, might need to re‐learn certain things even though they have potentially useful information stored in memory. Flexibility of memory retrieval increases throughout infancy and beyond (Herbert & Hayne, 2000). These changes gradually enable infants to generalize previously learned information to a wider range of circumstances. Nevertheless, even older infants struggle to retrieve recently encoded memories if any essential aspect of a learning experience change, such as the social context or aspects of originally encountered stimuli (Hayne, MacDonald, & Barr, 1997; Learmonth, Lamberth, & Rovee‐Collier, 2005). Given this profound reliance on matching external cues, infant memory might be highly susceptible to variations in internal state.
To test this idea, we presented infants with a memory task and randomly assigned them to either receive induction of the same state or of diverging states immediately before encoding and immediately before retrieval. State induction occurred by evoking either calm or animated play with their caregivers for 5 min. We hypothesized impaired memory retrieval if infant state at encoding was different from state at retrieval. Hence, our main analyses were relatively confirmatory in nature.