واکنش پذیری کورتیزول، حساسیت مادر، و یادگیری در نوزادان 3 ماهه

Abstract This study investigated the effects of adrenocortical functioning on infant learning during an emotionally challenging event (brief separation from mother). We also explored possible relationships between maternal sensitivity and both infant and maternal cortisol reactivity during the learning/maternal separation episode. Sixty-three 3-month-olds and their mothers were videotaped for a 10 min normal interaction period, and mother–infant behavioral synchrony was measured using Isabella and Belsky's [Isabella, R. A., & Belsky, J. (1991). Interactional synchrony and the origins of infant-mother attachment: A replication study. Child Development, 62, 373–384] coding scheme. The percentage of synchronous behaviors served as a measure of maternal sensitivity. Learning and short-term memory involved relating the infant's mother's voice with a moving colored block in a preferential looking paradigm. Infants whose cortisol increased during the session showed no learning or memory, infants whose cortisol declined appeared to learn and remember the association, while infants whose cortisol did not change evidenced learning, but not memory for the voice/object correspondence. Sensitivity and cortisol reactivity were correlated for mothers, but not for infants. Infant and maternal cortisol values for the first sampling period were highly correlated, but their cortisol reactivity values were uncorrelated, supporting the notion that infants and mothers have coordinated adrenocortical functioning systems when physically together, but become uncoordinated during a separation/learning event.

Introduction Young infants are in possession of many cognitive abilities that aid them in acquiring an expressive vocabulary of 50 words by the time they are about 16 months old (Bates, Bretherton, & Snyder, 1988). For example, they have highly developed visual and auditory sensory systems (Aslin, 1987) for encoding the auditory and visual components of objects or events that are named by their caregivers, and an attention system that is particularly responsive to human speech. Even newborns are differentially responsive to listening to their mothers’ voices, compared to other female voices (DeCasper & Fifer, 1980), and by 2 months of age, can discriminate sentences differing by a single phoneme (Mandel, Jusczyk, & Kemler Nelson, 1994). Another prelinguistic ability needed to learn first words is visual-auditory integration. During stimulus encoding, very young infants can coordinate visual objects with sounds (Spelke, 1979) and faces and voices (Walker, 1982). Finally, evidence for rapid word learning in 13-month-old infants can be found in recognition measures (e.g., Woodward, Markman, & Fitzsimmons, 1994). Long-term recognition of lengthy speech passages spoken by infants’ mothers has even been demonstrated in 1- to 2-month-old infants (Spence, 1995). Importantly, in studies such as these, after removing data from infants that were fussy during the experimental event, who fell asleep, or who showed biases in their responses to a region of the stimulus field, a final data set remains that is comprised of averaged data demonstrating the ability in the group at large; however, contained in this averaged set are data for infants who clearly possess the cognitive ability under question, as well as infants that do not. Ascertaining the nature of individual differences in cognitive processing is a central focus of many child developmental researchers (e.g., Siegler, 1978 and Thompson, 1994). However, relatively little is known about the sources of individual variability potentially affecting infant learning. One widely varying factor is infants’ social-interactive environments. For example, infants’ primary caregivers differ in how often they speak to their infants, in the prosodic variation of their speech, and in terms of the contingent nature of their language input. Nine- and 13-month-old infants whose mothers responded contingently to their vocalizations and play activities outperformed infants of less responsive mothers on several indices of language development, including timing in the onset of their first words and in combinatorial speech productions (Tamis-LeMonda, Bornstein, & Baumwell, 2001). Our primary interest is in furthering knowledge about the psychophysiological parameters affecting infant learning in the experimental context. That is, what differences inherent in infants, such as temperament and psychophysiological responding, contribute to differences amongst infants in their ability to attend to, learn, and remember speech/visual representations during a novel learning event? One commonly studied psychophysiological measure is cortisol, an adrenal steroid hormone, which is secreted in response to physical and emotional stress and is also associated with high states of arousal. When the brain perceives a stress, corticotropin-releasing hormone is released from the hypothalamus, which then stimulates the pituitary gland to release adrenocorticotropic hormone, which, in turn, stimulates the release of cortisol from the adrenals (Sapolsky, 1992). Research on human adult populations shows that, up to a point, cortisol elevation is adaptive, but if the response is prolonged or severe, there appear to be negative effects on health, behavior, and learning and memory processes (e.g., Flinn & England, 2003). Maternal trauma experienced during pregnancy has even been shown to affect infant adrenocortical functioning in utero ( Yehuda et al., 2005). In most infant cortisol studies, the objective has been to understand how situational, emotional, and/or temperamental factors are associated with different patterns of adrenocortical functioning. As an example of situational variables, infant cortisol has been found to be higher at home than when tested at the same time of day in the laboratory following a car ride (Gunnar, Mangelsdorf, Larson, & Hertsgaard, 1989b), and is higher in the afternoon for preschoolers in poorer quality child care settings than in higher quality settings (Dettling, Parker, Lane, Sebanc, & Gunnar, 2000). Regarding behavioral emotional reactions, Lewis and Ramsay (2005) recently reported a study relating anger and sadness to cortisol response in 4- and 6-month-old infants. For both age groups, a goal blockage paradigm was used to activate the emotional response system. Four-month-old infants learned a contingency between an arm movement they initiated and a pleasant event, and subsequently learned that their behavior no longer reinstated the event. The researchers employed a still-face paradigm (Tronick, 2003) to test the 6-month-old infants, in which mother displayed a still face instead of a communicative response to her infant. Infants’ facial and vocal affective displays were coded to determine the degree to which they felt anger, sadness, joy, and other emotions. Results showed that in both age groups, greater incidence of infants’ expression of sadness was related to increasing cortisol response, but anger was not. In some infant cortisol studies, a specific event occurred, and the adrenocortical response to the stressor was assessed and compared to a subsequent session. Any difference in the pattern of cortisol responding across sessions implied that infants remembered something about the original event. For example, across two sessions separated by 24 h, healthy newborn infants exhibited significantly reduced cortisol response to physical exam, and significantly greater cortisol response to heelstick (Gunnar, Connors, & Isensee, 1989a; Gunnar, Hertsgaard, Larson, & Rigatuso, 1992a). Cortisol levels also decreased in 6.5- to 13-month-old infants during two mother–infant swim classes (Hertsgaard, Gunnar, Larson, Brodersen, & Lehman, 1992), which was interpreted as an example of a novel event producing positive emotions, being remembered as pleasant, resulting in decreased cortisol levels over time (Gunnar & Donzella, 2002). However, in these studies, no measurements of infant cognitive abilities were reported. To our knowledge, there is only one reported study attempting to investigate relations between cortisol responsiveness and learning in the context of a relatively stress-free experimental paradigm. Haley, Weinberg, and Grunau (2005) used a conjugate reinforcement mobile task (Rovee-Collier, Sullivan, Enright, Lucas, & Fagan, 1980) to assess the relationship between adrenocortical reactivity and contingency learning in preterm and full-term 3-month-old infants. Infants were tested at various times of the day in their homes, in the presence of both the experimenter and the infants’ mothers, at a time of day when the infant should have been alert and active. The procedure involves attaching a ribbon to the infant's foot, attaching the ribbon to the mobile, and measuring the infant's rate of kicking to cause the mobile to move during several phases, including baseline, learning, and extinction. Researchers operationalized infants’ rate of kicking in specific ways as evidence for learning (Day 1), recognition memory (Day 1), and long-term retention (Day 2). Saliva sampling was undertaken two times both immediately before and 20 min after the introduction of the mobile task, and on 2 consecutive days. Two groups of infants were compared, those whose cortisol levels increased (±0.01 μg/dl) and those whose cortisol levels decreased. Results showed no difference in recognition memory for cortisol increasers compared to cortisol decreasers. However, infants whose cortisol levels increased .15 and .30 μg/dl across sampling periods on Day 1 (n = 5) showed strong evidence for long-term retention of the kicking/mobile movement contingency on Day 2, while decreasers and infants whose cortisol either increased minimally or more than .30 μg/dl showed no evidence for long-term retention. These results are consistent with the findings and perspectives of some researchers who argue that moderately increasing cortisol reactivity is associated with better memory ( Abercrombie, Kalin, Thurow, Rosenkranz, & Davidson, 2003; Buchanan & Lovallo, 2001; de Kloet, Oitzl, & Joels, 1999). van Bakel and Riksen-Walraven (2004) conducted a study on cortisol reactivity to a stressful event (a stranger/robot episode) and cognitive competence (measured using the Bayley Scales of Infant Development) in 15-month-old infants. They also assessed infant attachment security and anger proneness. Anger-prone infants, and infants with higher Bayley scores, showed greater cortisol reactivity to the stressful event than infants who were less prone to anger and who had less cognitive competence. Further, among those infants showing an insecure attachment with their mother, greater cortisol reactions occurred in the high cognitive competence group than in the low cognitive competence group, while no such relation existed in the secure attachment group. The association between mother–infant behavior and infant psychophysiological responding has been extensively studied. Gunnar and her colleagues explored the relationship between infant stress reactivity in cortisol and attachment security (e.g., Gunnar, Brodersen, Nachmias, Buss, & Rigatuso, 1996). They found that pre-test cortisol levels in 2, 4-, and 6-month-old infants were not associated with mother's greater responsiveness during a physical exam, but were associated with later attachment security. Specifically, higher cortisol levels in 2- to 6-month-olds were associated with later insecure attachment classifications assessed when the infants were 18 months old. In a sample of 9-month-old infants, Gunnar, Larson, Hertsgaard, Harris, and Brodersen (1992b) found that providing infants with a sensitive and responsive substitute caregiver completely prevented elevations in cortisol normally observed in a 30 min maternal separation period; importantly, elevations across this interval were witnessed when these substitute caregivers ignored the infants. Similar findings were reported in studies of 18-month-old infants (Nachmias, Gunnar, Mangelsdorf, Parritz, & Buss, 1996), and younger infants (Spangler, Schieche, IIg, Maier, & Ackerman, 1994). In their recent literature review, Gunnar and Donzella (2002) concluded that, in general, highly fearful, insecurely attached and disorganized infants had higher cortisol reactivity than those more securely attached to their mothers. However, some studies have reported a lack of a relationship between maternal sensitivity and infant cortisol response (e.g., Fleming, Steiner, & Corter, 1997; Haley & Stansbury, 2003). Haley and Stansbury (2003) investigated 5- and 6-month-old infants’ behavioral and psychophysiological responses using a modified version of the still-face paradigm that included an additional still-face reunion sequence. Parent responsiveness was coded from videotaped interactions, wherein parental behavior that was contingent upon infant facial affect or vocalizations contributed to the percentage of observed responsive behaviors. Measures of infant psychophysiological response included heart rate and cortisol change between pre-test and 30 min after the start of testing, while the percentage of time infants looked at the parent's face and the frequency of negative affective responses comprised the infant behavioral response measures. Results showed that parental responsiveness was associated with greater regulation of heart rate and lowered negative affect during the still-face procedure, but that infant cortisol response remained unaffected by parental responsiveness. In summary, the picture of infant psychophysiological, and in particular, cortisol, response is complex, and is affected by intrinsic factors, the social-interactive environment, and situational factors. Our empirical paradigm was designed so that we could focus on these questions: (a) How are different patterns of infant cortisol response associated with demonstrations of infant learning in an experimental context? and (b) How is maternal sensitivity associated both with infants’ initial cortisol level at the start of a learning event and with infants’ cortisol response to the learning situation? Given the complex, and as yet, sparse picture of infant cortisol response to learning situations, we deemed it important to outline the theoretical groundwork and specific design parameters of our investigation that are relevant to our predictions. We are guided in part by some recent findings that, in toddlers, the control of both emotion and cognitive processes resides in common brain areas (Posner & Rothbart, 2000). If this is also true earlier in development, consistent with this finding, our general expectation is that infants who can manage their emotional response to a moderately stressful situation would be more likely to show learning during an experiment than infants who manage their emotional response less well. Further, in line with other researchers, we assume that cortisol elevations reflect a failure of the infant's behavioral coping system (e.g., Gunnar et al., 1996; Levine & Wiener, 1988; Spangler & Grossman, 1993), and further, that a maternal separation episode constitutes a challenge to the 3-month-old infant's emotional response system (e.g., Ahnert, Gunnar, Lamb, & Barthel, 2004). Lastly, empirical work is supportive of the phenomenon that prolonged or severe stressors interfere with encoding and memory processes (e.g., Heffelfinger & Newcomer, 2001). Taken together, past theoretical and empirical work implies that if the infant is left to cope without the physical presence of his/her mother while simultaneously experiencing a novel experimental learning task, failures in emotion coping may be associated with reductions in infants’ capacities for learning and remembering aspects of an experienced event. There are three notable differences between this, and the single other (Haley et al., 2005), published investigation into relations between infant cortisol reactivity and learning in young infants. First, we include measurements of maternal cortisol. Studies investigating associations between human maternal behavior and cortisol have found that mothers with higher levels of cortisol show the most affectionate behaviors (Fleming, Steiner, & Anderson, 1987), and are better able to identify their own infants’ odors when tested 2 days after birth (Fleming et al., 1997). Spangler (1991) studied cortisol levels in newborn and several-month-old infants and observed that mean infant cortisol levels were significantly related to mean maternal cortisol for newborn and older infants, which could be due to shared genetics, shared environment, or the interaction. Our study is the first to explore potential links between maternal and infant cortisol reactivity, and between maternal cortisol and maternal behavior, in the context of an infant learning experiment. Second, in the Haley et al. (2005) experiment, infants demonstrated contingency learning by making kicking movements. Studies on children show a direct relationship between cortisol levels and physical activity (e.g., Cielslak, Frost, & Klentrou, 2003). Our design used an experimental paradigm where infants were confined to an infant seat, allowing much less opportunity for body movements, thereby minimizing the contributing influence of physical exertion in infant cortisol response. Third, we provide a measure of maternal sensitivity to assess whether or not sensitivity is associated both with infants’ cortisol levels at the start of the learning task, and infants’ changing cortisol levels resulting from their experiences during the learning task. 1.1. Assessment of mother–infant synchronous and asynchronous co-occurrences The quality of early mother–infant interaction plays a significant role in the formation of a secure attachment bond (Ainsworth, Blehar, Waters, & Wall, 1978; Bowlby, 1969). When mothers notice their infant's signals and respond in a sensitive manner throughout the first year, a secure attachment bond is a more likely outcome than an insecure bond (Atkinson et al., 2000; de Wolff & van Ijzendoorn, 1997). Recognizing that maternal sensitivity is made manifest both through the initiation of a sensitive behavior, and by a sensitive response to an infant signal, Isabella, Belsky, and von Eye (1989) developed a bidirectional system for coding mother–infant interaction that is used extensively in research (e.g., Isabella & Belsky, 1991; Wendland-Carro, Piccinini, & Millar, 1999). They used 13 mother–infant interaction categories to define both synchronous and asynchronous behavioral exchanges. A synchronous exchange such as infant vocalizes—mother vocalizes/smiles is an indicator of caregiver sensitivity, while an asynchronous exchange, such as infant vocalizes/cries—mother unresponsive is an indicator of caregiver insensitivity. As in Isabella et al. (1989), we determined the incidence of both synchronous and asynchronous exchanges between mother and infant in a 10 min videotaped free play session, and calculated the percentage of total codable behaviors that fit their synchronous and asynchronous behavioral exchange categories. The percentage of synchronous behavioral exchanges thus served as a measure of maternal sensitivity in our study, and was assumed to be indicative of the interactive experiences typically shared by each mother–infant pair. 1.2. The present study Learning in 3-month-old infants was tested using a two-monitor standard preferential looking procedure (Kemler Nelson et al., 1995), a common tool used in investigating infants’ speech (Hirsh-Pasek et al., 1987) and visual (e.g., Thompson, Madrid, Westbrook, & Johnston, 2001) processing abilities. The procedure we employed most closely resembled one used by Courage and Howe (1998) in their study of novelty and familiarity preferences in 3-month-olds. Acquisition and retention constituted two phases of the procedure that are indicative of progressive steps in the learning process. During acquisition, a consistent auditory stimulus (a recording of the infant's mother) co-occurred with the presentation of a moving colored block (either red or yellow), with the trial length contingent upon the infant's head turning behavior. The other color block was never paired with mother's voice. All infants accumulated 90 s of total looking time during this phase. We compared infants’ looking times to the monitor with the voice/color block correspondence against looking times to the other monitor displaying the color block that was not paired with mother's voice. Learning was operationalized as a significant preference in looking times for the voice/color block pairing, compared to the colored block that was not paired with mother's voice. A retention phase immediately followed the acquisition phase, during which the red and yellow block stimuli were again presented on the two side monitors, without the voice stimulus. Evidence for short-term memory was operationalized as a significant difference in looking times toward the color block previously associated with mother's voice, compared to the block that was not. This procedure is not informative of the degree of learning or the degree of retention exhibited by infants; rather, significant differences between the two categories of stimuli are indicative of whether the infant group in question learned and/or remembered the visual/object correspondences. The time course of events in the experiment was important for establishing what aspects of psychophysiological functioning are reflected in cortisol values. Table 1 presents the sequence of activities during the testing session. Testing occurred at a time of day that was most convenient for the mother, and when the baby was likely to be alert and awake—typically, in the morning hours. Most mothers drove to the laboratory. Upon entering the laboratory, mothers chatted with the experimenter, settling in for several minutes before participating in a 10 min videotaped session. Basal saliva sampling took place roughly 20 min after entering the lab for both infant and mother. Then the learning experiment occurred, when the infant was put in the seat and the mother disappeared out of view from the infant. The duration of the learning experiment varied, depending on the infants’ looking behavior, but never exceeded 25 min. Finally, a second saliva sampling event occurred between 20 and 30 min after the beginning of the learning experiment. Given that peak cortisol response is typically 20–30 min following an event (e.g., de Weerth & van Geert, 2002) we assume that the cortisol level for the basal sampling period (T1) primarily reflects infants’ (and mothers’) psychophysiological response to the novel lab context as experienced in each other's physical presence, and that this level reflects a certain degree of physiological, and perhaps emotional, arousal due to the physical activities in getting to the lab and to the novelty of the laboratory environment. We further assume that the difference in cortisol levels between sampling periods (ΔT1 − T2) is reflective of infants’ reactions to both maternal separation and the novel learning event. And finally, we believe it is important to categorize infants as “increasers” or “decreasers” based on T1 and T2 cortisol values that are different enough to be indicative of true change. To this end, we used a stringent criterion of requiring one-half standard deviation difference in cortisol levels between sampling periods as evidence for either increasing or decreasing cortisol reactivity. This amount of change was approximately the same as the interassay coefficient of variation in our sample. Table 1. Sequence of events during an experimental session Event Duration Events reflected in cortisol Car ride to lab Variable Enter lab Settling in 10–15 min Videotaped observation Exactly 10 min Total time before T1 20–25 min T1 saliva sampling Response to new lab context in presence of mother Acquisition phase Variable Short break 2–3 min Retention phase 2–3 min Wait Variable Total time T1 to T2 20–30 min (M = 24.8 min) T2 saliva sampling New learning situation and mother separation Table options 1.3. Hypotheses We hypothesize that infants who experience a decline in cortisol levels during the experimental session will show learning and memory for the object associated with their mothers’ voices, whereas infants who experience an increase in cortisol levels will not learn, nor consequently remember, the object associated with their mothers’ voices. Based on our literature review, we also predict an association between maternal sensitivity and infant cortisol values; this hypothesis would receive support in correlations between behavioral synchrony scores assessed in the videotaped 10 min mother–infant interaction period and both infant T1 scores and infant ΔT1 − T2 scores. Specifically, lower absolute values of infant T1 scores and less reactivity are predicted to be positively correlated with our maternal sensitivity measure. Several additional, more exploratory hypotheses will be entertained. This is the first study to explore whether or not there is a relationship between maternal sensitivity and maternal cortisol response. While past research has shown a relationship between positive attitudes toward mothering and cortisol response ( Fleming et al., 1987), it is unknown whether a similar relation exists between maternal sensitivity and maternal cortisol. Finally, because mothers and infants were in close physical proximity to each other prior to T1 sampling, and consistent with past research, we predict that infant and maternal values will be significantly positively correlated at T1 ( Spangler, 1991). However, research has not yet addressed whether maternal and infant cortisol reactivity are correlated during a separation/learning event. A lack of a correlation between mothers and babies in ΔT1 − T2 could occur given that they could plausibly have differing emotional reactions to the separation experience and the learning experiment. On the other hand, it might also be expected that mothers would have an empathic psychophysiological response to their infants during the separation/learning event, which would be reflected in a positive correlation.

Results This research concerns how infant learning and short-term memory are affected by changes in cortisol concentration from Time 1 (prior to the looking time experiment; T1) to Time 2 (post-experiment; T2), termed “cortisol reactivity”, and how maternal sensitivity assessed prior to the learning experience may be related to infant adrenocortical functioning during a learning experience. We explored possible additional relationships between mothers’ and infants’ cortisol levels at T1, between maternal and infant cortisol reactivity, and between behavioral synchrony and mothers’ cortisol reactivity. We first conducted Pearson R tests for relations between study variables (preferential looking during acquisition, preferential looking at retention, degree of synchrony, log10 cortisol values at T1 and T2) and potentially confounding variables (gender, mother's age, ethnicity, education, and income). None of these correlations were significant (all p's > .05), with one exception: gender was correlated with cortisol values at Time 2 (r(62) = .32, p = .01), indicating that boys’ T2 cortisol values tended to be lower than girls’ values. The cortisol data from the entire set of participants with obtained cortisol data was examined for outliers, defined as ±3 S.D. from the mean for each sample (Gunnar & White, 2001). In the data for the 63 complete and final mother–infant sets, there were no scores that fit this criterion. The raw cortisol values at T1 and at T2 were not normally distributed, so the raw scores were transformed to log10 scores (Larson, White, Cochran, Donzella, & Gunnar, 1998). All subsequent statistics were conducted on the log10 values. A value of 0.5 S.D. from the mean of the raw score sample was used in establishing the cortisol change category for the difference in cortisol between T1 and T2, or ΔT1 − T2 cortisol (M. Gunnar, personal communication, May 30, 2006). Thus, if an individual's ΔT1 − T2 cortisol was positive and changed 0.5 S.D. or more, this was categorized as a decrease in cortisol (infant n = 16; mother n = 8). If ΔT1 − T2 was negative and greater than 0.5 S.D., this was categorized as an increase (infant n = 14; mother n = 2). No change from T1 to T2 was categorized as such (infant n = 33; mother n = 53). Table 2 presents descriptive statistics for the cortisol data for the final 63 mother–infant pairs. Table 2. Means, standard deviations and ranges of the cortisol values of mothers and infants Measure n (%) Raw mean (S.D.) Raw range Log-transformed mean (S.D.) Infants T1 cortisol in μg/dl 63 (100) .26 (.21) .04–1.0 −.70 (.31) T2 cortisol in μg/dl 63 (100) .23 (.13) .05–.66 −.69 (.23) ΔT1 − T2 cortisol (μg/dl) 63 (100) −.02 (.20) −.74–.54 −.01 (.28) ΔT1 − T2 cortisol increased 14 (22.2) ΔT1 − T2 cortisol decreased 16 (25.4) ΔT1 − T2 cortisol no change 33 (52.4) Mothers T1 cortisol in μg/dl 63 (100) .22 (.12) .08–.80 −.72 (.22) T2 cortisol in μg/dl 63 (100) .18 (.13) .05–.97 −.80 (.23) ΔT1 − T2 cortisol (μg/dl) 63 (100) −.03 (.07) −.30–.16 .09 (.12) ΔT1 − T2 cortisol increased 2 (3.2) ΔT1 − T2 cortisol decreased 8 (12.7) ΔT1 − T2 cortisol no change 53 (84.1) Note: Cortisol values are presented as raw values and log-transformed to base 10. Statistical analyses were conducted with the transformed data. Table options Fig. 1 graphically displays the infant T1, T2 and ΔT1 − T2 data for all three cortisol change categories. A post-hoc analysis showed infants in the decreasing cortisol category had T1 values that were significantly higher than infants in the increasing cortisol category, t(28) = 6.14, p < .0001. As in other recent studies (e.g., Haley et al., 2005; Ramsay & Lewis, 2003) time of day for saliva collection and ΔT1 − T2 cortisol were not correlated in either the mother's data or the infant's data (p's > .05), and was not considered in further analyses. Mean log10 cortisol values at T1, T2, and ΔT1−T2 values for infants in all three ... Fig. 1. Mean log10 cortisol values at T1, T2, and ΔT1 − T2 values for infants in all three cortisol change categories (bars = 1 S.E.). Figure options To test the hypothesis that infants with decreased ΔT1 − T2 cortisol levels would both learn and remember object fields associated with mother's voices, while infants with increased ΔT1 − T2 cortisol would not, separate t-tests were conducted on the infants’ preferential looking data at acquisition and retention phases who fell into these two categories, as well as the infants with no ΔT1 − T2 cortisol change. Data were averaged across trials within each phase. In the decreased ΔT1 − T2 cortisol group, infants’ looking times were significantly greater for the color block associated with their mothers’ voices both during the acquisition phase (t(15) = 2.44, p = .028, d = .61), and during the retention phase (t(15) = 2.25, p = .04, d = .56), looking longer at the monitor with the color block that also presented their mothers’ voices, and showing that they expected to hear their mothers’ voices again associated with the same color block during the retention trials. In contrast, in the increased ΔT1 − T2 cortisol group, looking times across conditions were not significantly different both for acquisition (t(13) = 0.21, p = .840, d = .06) and retention (t(13) = −0.877, p = .396, d = .23) phases, showing that they did not encode the color/voice association, nor did they demonstrate a memory or expectation of the association. For the no change in ΔT1 − T2 cortisol infant group, looking times were significantly greater for associated voice/color trials at the acquisition phase (t(32) = 3.40, p = .002, d = .59), but not at the retention phase (t(32) = −0.98, p = .334, d = .17), revealing a looking time preference for block color/mother's voice combination with voice presentations, but no expectation of her voice during later silent moving block trials. These data are shown in Fig. 2 and Fig. 3 for acquisition and retention phases, respectively. Average looking times to blocks associated with mother's voice compared to ... Fig. 2. Average looking times to blocks associated with mother's voice compared to blocks not associated with mother's voice, for all three cortisol change categories of infants, during the acquisition phase (bars = 1 S.E.). Figure options Average looking times to blocks associated with mother's voice compared to ... Fig. 3. Average looking times to blocks associated with mother's voice compared to blocks not associated with mother's voice, for all three cortisol change categories of infants, during the retention phase (bars = 1 S.E.). Figure options In investigating the relationships between mother–infant behavioral synchrony and infant and maternal cortisol response, preliminary tests were conducted to determine if covariates should be included in subsequent tests. Cortisol responses are subject to the Law of Initial Values (LIV; Wilder, 1956). Greater ΔT1 − T2 increase should occur in a system that is at a lower level of activity at T1 than when the system is at a higher level of activity at T1. As explained by Gunnar and White (2001), two conditions must be met before a researcher should be concerned about the LIV. Namely: (a) there must be significant difference between T1 cortisol and T2 cortisol values, and (b) T1 and T2 values must be significantly positively correlated. Looking at infant cortisol first, T1 and T2 cortisol values were positively correlated (r(62) = .47, p < .0001), however, a paired samples t-test showed no significant change between T1 and T2 values (t(62) = −0.21, p > .05). For mother, T1 and T2 values were also positively correlated (r(62) = .86, p < .0001), and there was a significant decrease in cortisol from T1 to T2 (t(62) = 5.54, p < .0001). Thus, in the infant sample, no steps needed to be taken to remove the variance from change scores that might be associated with T1 scores. However, for correlations involving mothers’ ΔT1 − T2 values, T1 scores were included as a covariate due to LIV considerations. To investigate whether or not maternal sensitivity was associated with infant cortisol response, partial correlations were conducted on the set of mother–infant behavioral synchrony scores and infant ΔT1 − T2 values, controlling for maternal T1 values. The resulting correlation was nonsignificant (p > .05). The next test of a correlation between mothers’ cortisol change scores and behavioral synchrony revealed a marginally significant positive correlation, r(58) = .25, p = .056, showing that mothers with greater decreasing cortisol reactivity (on this scale, higher scores) showed higher degrees of sensitivity to their infants in the 10 min videotaped normal interaction period. Finally, as Table 3 shows, synchrony and asynchrony were significantly negatively correlated (r(58) = −.29, p < .05), indicating that those mothers exhibiting a greater percentage of synchronous behaviors with their infants also exhibited a lower percentage of asynchronous behaviors. Table 3. Partial correlations among cortisol reactivity in mother and infant, behavioral synchrony, and behavioral asynchronya Variables 1 2 3 4 ΔT1 − T2 infant – ΔT1 − T2 mother −.04 – Behavioral synchrony .12 .25b – Behavioral asynchrony −.05 −.13 −.29* – Note: Cortisol values were log-transformed to base 10. a Controlled for mothers’ log10 T1 values (d.f. = 58 for all values). b p = .056. * p < .05. Table options Furthermore, a bivariate correlation conducted on infant and maternal log10 T1 cortisol values yielded the predicted significant positive correlation, r(62) = .381, p < .005; and further, mothers’ and infants’ ΔT1 − T2 values were uncorrelated, p > .05. Taken together, the correlation analyses show that the degree of behavioral synchrony, or sensitivity, mothers display in their interactions with their infants is significantly related to their own, but not with their infants’, cortisol reactivity to a separation/learning event. Furthermore, mothers’ and infants’ cortisol levels were significantly correlated with each other when they began the experiment, but their patterns of cortisol reactivity in response to the separation episode/learning event were uncorrelated.