افسردگی مادران و آموزش ترویج اثرات هدایت شده گفتار و نوزاد: نقش حساسیت مادر، تشخیص افسردگی و نشانه های صوتی سخنرانی

Abstract The hypothesis that the associative learning-promoting effects of infant-directed speech (IDS) depend on infants’ social experience was tested in a conditioned-attention paradigm with a cumulative sample of 4- to 14-month-old infants. Following six forward pairings of a brief IDS segment and a photographic slide of a smiling female face, infants of clinically depressed mothers exhibited evidence of having acquired significantly weaker voice–face associations than infants of non-depressed mothers. Regression analyses revealed that maternal depression was significantly related to infant learning even after demographic correlates of depression, antidepressant medication use, and extent of pitch modulation in maternal IDS had been taken into account. However, after maternal depression had been accounted for, maternal emotional availability, coded by blind raters from separate play interactions, accounted for significant further increments in the proportion of variance accounted for in infant learning scores. Both maternal depression and maternal insensitivity negatively, and additively, predicted poor learning.

. Introduction Parents increase the salience of their speech to infants by slowing its rate, exaggerating changes in pitch, hyper-articulating vowels, and repeating key words and phrases (Fernald, 1984). For young infants, infant-directed speech (IDS) is more effective than adult-directed speech (ADS) at altering infant state, eliciting infant responding, and promoting infant attention and rudimentary cognitive-linguistic processes (Cooper and Aslin, 1990, Fernald, 1984 and Ma et al., 2011). Importantly, IDS may be particularly effective at inducing social preferences: relative to a novel face, infants prefer to look at the face of a woman they had seen and heard talking in IDS intonation (Schachner & Hannon, 2011). In contrast, infants prefer a novel face to that of a woman they had seen and heard talking in ADS intonation. Thus, IDS is a powerful promoter of infant attention and learning, and stimuli correlated with it may acquire greater interest. However, depressed mothers produce IDS that is deficient in perceptual salience, infant focus, and degree of contingency on infant behavior (Bettes, 1988, Kaplan et al., 1999 and Murray et al., 1993). Social factors may be particularly influential in cognitive and language development. According to Kuhl (2007) “social gating” hypothesis, parents use of IDS, and the arousal and attention generated by contingent parental responding, increases the robustness and durability of stimulus–stimulus (e.g., word-object) associations acquired during joint attentional states (see also Ma et al., 2011). Various aspects of rudimentary language development can be facilitated by contingent interaction with live partners, above and beyond mere exposure video and audio stimulation (Kuhl et al., 2003 and Roseberry et al., 2009). Research has revealed that, in comparison to IDS produced by non-depressed mothers, IDS produced by depressed mothers is less effective at promoting a basic kind of learning in infants (Kaplan et al., 1999). But given that deficits in the salience of IDS are highly correlated with deficits in its degree of contingent delivery (Bettes, 1988), and further given evidence that maternal depression and maternal insensitivity each adversely affect the kinds of joint mother–infant attention that promotes infant learning (Raver & Leadbeater, 1995), questions remain about relative roles of low perceptual salience and low contingency as determinants of learning deficits in infants of depressed mothers. The purpose of the present research was to evaluate these alternative explanations for effects of IDS on infant learning in a conditioned-attention paradigm using a cumulative sample of infants of non-depressed and depressed mothers. Originally motivated by evidence that IDS is particularly effective at increasing an infant's state of arousal (i.e., sensitization and dishabituation effects; Kaplan, Goldstein, Huckeby, & Cooper, 1995), Kaplan and colleagues performed a series of experiments to test whether IDS can facilitate the acquisition of an association with a visual stimulus that soon follows it (Kaplan, Jung, Ryther, & Zarlengo-Strouse, 1996). The rationale was that the increase in central arousal elicited by IDS would persist after the presentation of the visual stimulus, and promote the formation of a stimulus–stimulus association. Analogous effects have been documented in non-human animals. To test this hypothesis, the researchers developed a laboratory model of associative learning, a conditioned-attention paradigm, diagrammed in Fig. 1. In the pairing phase, a 10-s segment of IDS (the nominal conditioned stimulus or CS) preceded the presentation of a 10-s face reinforcer (the nominal unconditioned stimulus or UCS; forward pairings). In a control condition, a 10-s presentation of a face reinforcer preceded the presentation of a 10-s segment of IDS, such that the IDS did not predict the visual stimulus (backward pairings). To test whether the IDS segment acquired the ability to control infant attention, all infants were tested with 4 10-s presentations of a novel 4 × 4 checkerboard pattern. The IDS segments from the pairing phase were played simultaneously with the first and fourth checkerboard presentations, whereas the second and third checkerboard presentations occurred in the absence of sound. The extent to which the IDS stimuli increased looking at the checkerboard pattern (“positive summation”) in a forward-pairing condition, above that in a backward-pairing condition (and in other studies, random and “no CS” control conditions, Kaplan, Fox, & Huckeby, 1992), comprised the measure of associative learning. Associative learning is important because it is a phylogenetically old and ubiquitous mechanism for infants to learn “what goes with what” in the environment (Rovee-Collier, 1986). Results showed that when an unfamiliar non-depressed mother's IDS signaled the occurrence of a face (forward pairing condition), it acquired the ability to significantly increase looking at the checkerboard pattern in 4-month-old infants of non-depressed mothers. In contrast, when an unfamiliar non-depressed mother's IDS occurred after the face (backward pairing condition), it had no effect on looking at the checkerboard. Similarly, an unfamiliar non-depressed mothers ADS had no effect on looking at the checkerboard pattern following either forward or backward pairings with a face (Kaplan et al., 1996). These findings showed that IDS is more effective than ADS at promoting this form of learning, possibly because of differences in IDS vs. ADS stimulus-induced sensitization (Kaplan, Zarlengo-Strouse, Kirk, & Angel, 1997). Schematic diagram of the conditioned-attention paradigm. Fig. 1. Schematic diagram of the conditioned-attention paradigm. Figure options A conditioned-attention paradigm was used in subsequent experiments to assess the learning-promoting properties of IDS produced by mothers with symptoms of depression. Although all mothers increase the mean fundamental frequency (F0) of their IDS, clinically depressed mothers and mothers diagnosed with major depression in partial remission exhibit a significant restriction in fundamental frequency (F0) range in their IDS (F0 max-F0 min, or ΔF0), in comparison with non-depressed mothers and with mothers diagnosed with minor depression, anxiety disorders, and major depression in full remission ( Porritt, Zinser, Bachorowski, & Kaplan, 2014). Because of the demonstrated importance of F0 modulation in eliciting infant responding ( Cooper and Aslin, 1990, Fernald and Kuhl, 1987 and Kaplan and Owren, 1994), it was predicted that IDS produced by depressed in comparison with non-depressed mothers would less effectively promote infant learning in this paradigm. Indeed, groups of 4-month-old infants of non-depressed mothers for whom a segment of IDS produced by an unfamiliar, depressed mother signaled a smiling face showed no evidence of associative learning on the post-conditioning test, whereas groups of 4-month-old infants of non-depressed mothers tested with IDS produced by an unfamiliar non-depressed mother showed evidence of significant learning on the post-conditioning test (Kaplan et al., 1999). In preliminary support of the perceptual salience hypothesis, ΔF0 in the maternal IDS stimuli was significantly correlated with the group mean learning scores for groups of infants who were tested with those stimuli. This initial study assessed the learning-promoting effects of unfamiliar depressed and non-depressed mothers’ IDS in infants without a history of potentially disordered interactions with a depressed primary caregiver. To test whether infants of depressed mothers would exhibit a similar pattern of findings, and further to determine if their ability to form voice–face associations might have been degraded as the result of disordered mother–infant interactions, a follow-up study compared voice–face associative learning in 4-month-old infants of clinically depressed versus non-depressed mothers. Results showed that 4-month-old infants of depressed mothers did not acquire associations in response to their own or an unfamiliar clinically depressed mother's IDS, but did exhibit significant learning in response to an unfamiliar non-depressed mother's IDS (Kaplan, Bachorowski, Smoski, & Hudenko, 2002). Thus, 4-month-old infants of depressed mothers did not differentiate between their own versus an unfamiliar depressed mother's IDS, but the fact that they did exhibit an ability to learn in this paradigm in response “normal” IDS produced by an unfamiliar non-depressed mother suggested that, at least at this age, their general associative learning abilities were intact. Perhaps depressed mothers’ IDS sounded similar to ADS to these infants, or lacked the positive affect that typically characterizes IDS (Kitamura & Burnham, 1998). As outlined above, an alternative hypothesis was that a history of prior non-reinforcement in social interactions with depressed mothers reduced the “associability” of the infant's own mother's IDS, and this “tuning-out” accounted for the poor voice–face associative learning during the conditioning test. Such an effect would be analogous to “latent inhibition” or “learned irrelevance” effects in classical conditioning studies: pre-conditioning exposure to isolated presentations of the to-be CS, or to random presentations of the to-be CS and the UCS, produce a transfer effect or “retardation” of conditioning when the pre-exposed stimulus is subsequently used to positively predict the UCS in a forward-pairing conditioning arrangement (Linden, Savage, & Overmeier, 1997). Under this view, social interactions leading up to the laboratory visit, more than the low perceptual salience of “depressed” IDS, were the proximal causes of poor infant learning in the laboratory. Here, we refer to this experience-based tuning-out of an infant's own mother's IDS “specific learned irrelevance.” Despite the preliminary support for it, several lines of evidence now cast doubt on the perceptual salience hypothesis, and are at least consistent with the learned irrelevance account. First, although there was a significant negative correlation between ΔF0 in maternal IDS samples and mean learning scores for groups of infants of non-depressed mothers who were tested with those samples ( Kaplan et al., 1999), several subsequent small sample studies have failed to show significant correlations between ΔF0 in IDS speech samples and individual infant learning in infants of depressed and non-depressed mothers ( Kaplan, Burgess, Sliter, & Moreno, 2009). Second, although the learning failures in response to an infant's own depressed mother's IDS have been observed consistently in infants ranging in age from 4 to 13 months (Kaplan et al., 1999, Kaplan et al., 2004 and Kaplan et al., 2009), learning outcomes for these infants in response to “normal” IDS produced by unfamiliar non-depressed mothers have varied with the chronicity and timing of the infant's own mother's depression. In one study, the strength of the associative learning among infants of depressed mothers in response to non-depressed mothers’ IDS was inversely proportion to the postpartum duration of the infant's own mother's depression (Kaplan et al., 2004). In another study, individual infants of depressed mothers were found to exhibit significant learning in response to an unfamiliar non-depressed mother's IDS at 4 months, but not at 12 months (Kaplan, Danko, Kalinka, & Cejka, 2012). Finally, 12-month-old infants of currently depressed mothers with perinatal depression onset did not exhibit significant voice–face associative learning in response to an unfamiliar non-depressed mother's IDS, whereas 12-month-old infants of currently depressed mothers with later depression onset (mean duration of 4 months) did (Kaplan, Danko, Diaz, & Kalinka, 2011). Thus, consistent with a role for an infant's experience with a depressed primary caregiver, infants of chronically depressed mothers not only failed to learn in response to their own mothers’ IDS in this paradigm, but also eventually exhibited poor learning in response to normal IDS. However, infants of chronically depressed mothers exhibit stronger-than-normal learning in this paradigm in response to an unfamiliar father's IDS (Kaplan et al., 2004 and Kaplan et al., 2010). Taken together, these findings were consistent with a progressive “tuning-out” of non-depressed mothers’ IDS in infants of chronically depressed mothers, and inconsistent with the perceptual salience hypothesis. The researchers hypothesized that learned irrelevance of an infant's own mother's IDS (“specific learned irrelevance”) might eventually generalize to other similar stimuli, such as non-depressed mother's IDS (“generalized learned irrelevance”), but not to less similar stimuli, such as an unfamiliar non-depressed father's IDS. Third, the “specific learned irrelevance” hypothesis has been supported in recent research in which the current quality of mother–infant interactions was used as a proxy measure for the general degree of contingent mother–infant interactions. Researchers (Kaplan et al., 2009) trained 5–13-month-olds on a forward pairing arrangement in which their own mother's IDS signaled a smiling female face, and separately assessed the quality of mother–infant interactions during a 10-m semi-structured play session. Videotapes of play sessions coded by blind raters using the Emotional Availability Scales (EAS; Biringen et al., 1993 and Emde, 1980) showed that, relative to non-depressed controls, depressed mothers were significantly lower in sensitivity and significantly higher in (covert) hostility. Infants of depressed mothers were rated as significantly lower in responsiveness to the mother relative to infants of non-depressed mothers, consistent with the hypothesized “tuning-out.” However, only maternal sensitivity correlated with infant learning in response to maternal IDS. A hierarchical linear regression revealed that maternal depression was not significantly related to infant learning after demographic risk factors and antidepressant medication use had been taken into account, but that maternal sensitivity did significantly predict infant learning after all other variables, including ΔF0 in IDS, had been entered into the regression equation. This finding suggested that the quality of mother–infant interactions was more important than either maternal depression per se or speech acoustic variables in predicting infant learning in this paradigm. The purpose of the present study was to further investigate the roles of the perceptual salience of IDS, maternal depression, and maternal sensitivity in accounting for infant learning in response to their own mother's IDS in a conditioned attention paradigm. The prior study included only a small sample of infants. The present report includes analyses of a cumulative data set, including the 55 mothers and infants in the Kaplan et al. (2009) study, plus an additional 81 mothers and infants whose data are reported here for the first time. Based on previous work, we predicted that, consistent with a specific learned irrelevance account, ΔF0 in maternal IDS would not predict individual infant learning. Further, we predicted that any effects attributed to maternal depression would be mediated by maternal sensitivity.

. Results 3.1. Demographic and diagnostic information Table 1 provides a summary of demographic and diagnostic information for the 136 mother–infant dyads that completed all testing. Depressed mothers were more likely than non-depressed mothers to be from an underrepresented group, and had significantly lower levels of formal education. Fifty-nine mothers (43.4%) had elevated self-report scores on the BDI-II, and 20 (14.7%) received a current DSM-IV Axis-I diagnosis of major depressive disorder (MDD; n = 16) or depressive disorder not otherwise specified (DDNOS; n = 4). In addition, 17 mothers (12.5%) were diagnosed with a depressive episode in partial remission (PR), and 21 mothers (15.4%) were diagnosed with a depressive episode in full remission (FR). In comparisons between NDEP and DEP conditions below, FR and PR mothers were included in the NDEP group, unless otherwise noted. 3.2. Emotional availability scales Mean ratings from the EAS were analyzed as a function of mother's diagnosis. A 2 (NDEP vs. DEP) × 5 (EAS subscale) MANOVA carried out on the subset of infants who also completed the conditioned-attention test (n = 136) showed no significant multivariate effect of EAS subscales, F(5, 130) = 1.15, p = .34, and no significant effect of depression diagnosis on any of the scales, F(1, 134) = 1.95, p = .17 for Sensitivity, F(1, 134) = 3.00, p = .09 for Hostility, and F(1, 134) = 2.20, p = .14 for Child Responsiveness. 2 Similarly non-significant effects were obtained when a 4-level maternal depression variable (NDEP, FR, PR, DEP) was used. 3.3. Speech acoustics Mean ΔF0 calculated across each mother's 3 “pet the gorilla” utterances did not differ for NDEP vs. DEP mothers, Ms = 139 vs. 130 Hz, respectively, F(1, 134) = .39. ΔF0 correlated positively with maternal education, r = .19, p = .05. When depression was entered as a 4-level variable (NDEP, FR, PR, DEP), there was a significant overall effect, F(3, 131) = 3.49, p = .02, η2 = .077, with significant differences in ΔF0 between the FR (M = 170 Hz) and each of the other 3 conditions (never depressed, M = 137 Hz, PR, M = 117 Hz, currently depressed, M = 130 Hz). 3.4. Conditioned-attention tests Pairing phase: Table 2 shows mean looking durations on the first and last presentations of the IDS speech segments (during which the projection screen was uniformly illuminated) and on the first and last face presentations in the pairing phase. As in all prior conditioned attention studies, there were no significant differences as a function of maternal depression in mean response levels during IDS segments or face stimuli, and no significant changes in responding to either stimulus across trials. The same findings were obtained here, and for the sake of brevity, those aspects of the data are not discussed further. Table 2. Mean looking times in response to voice and face during pairing and summation test phases. Phase Stimulus DEP NDEP Pairing Trial 1 Voice 4.96 (3.17) 4.57 (3.67) Trial 6 Voice 4.68 (3.30) 4.36 (3.35) Trial 1 Face 6.12 (3.77) 6.60 (2.98) Trial 6 Face 6.04 (2.60) 5.65 (3.18) Summation test Trial 7 Voice + checkerboard 5.85 (3.13) 6.82 (2.88) Trial 8 Checkerboard 6.03 (2.62) 5.02 (3.05) Trial 9 Checkerboard 4.34 (2.98) 4.68 (2.76) Trial 10 Voice + checkerboard 4.60 (3.62) 5.47 (3.22) Difference score 0.04 (1.68) 1.30 (2.32) Table options Summation test phase: During the post-learning test phase, speech segments were presented simultaneously with the first and fourth checkerboards (Trials 7 and 10), whereas the other two checkerboard presentations occurred alone. As seen in Table 2, averaged across the two checkerboard-alone test trials, there were no differences in duration of looking as a function of maternal diagnosis, F(1, 134) = .40, consistent with equivalent baseline response levels in those groups. A 2 (mother NDEP or DEP) × 4 (test trials) split-plot repeated measures ANOVA on looking times across all 4 test trials (with Greenhouse–Geisser corrected df) revealed no significant effect of diagnostic condition, but a significant effect of trials, F(3, 375) = 6.62, p = .001, η2 = .05. And although the maternal diagnosis x trials interaction was not significant, F(3, 375) = 2.30, p = .08, there was a significant quadratic trend in the diagnosis x trials interaction term, F(1, 134) = 4.74, p = .05, η2 = .035. Infants of NDEP mothers showed a more pronounced U-shaped response function across test trials than did infants of DEP mothers. When the test trial data were reanalyzed using the 4-level depression variable rather than the binary depression variable, there were no significant effects. For each infant, a difference score was calculated by subtracting the mean looking time on the two checkerboard only test trials from the mean looking on the two voice-plus-checkerboard test trials. This difference in looking reflects the extent of “positive summation,” a standard indirect assessment of associative learning (Rescorla, 1971). Because previous research has shown significant positive summation after forward but not backward pairings of IDS segments and faces, it has been interpreted to reflect the formation of a voice–face association. Mean difference scores, shown in Table 2, were significantly higher for infants of NDEP than DEP mothers, F(1, 134) = 7.05, p = .01, η2 = .049. Mean difference scores were also significantly higher for infants of mothers with non-elevated in comparison with elevated self-report BDI-II scores, F(1, 134) = 4.28, p = .05, η2 = .031. Mean difference scores for infants of the never depressed, FR, PR, and currently depressed mothers were 1.22 s (SE = .27, n = 78), 1.26 s (SE = .59, n = 21), 1.79 s (SE = .52, n = 17, and −.12 s (SE = .38, n = 20), respectively. Difference scores did not correlate significantly with infant age, r(136) = −.02, or, for infants of depressed mothers, with the duration of the mother's current depressive episode, r(20) = .38, p = .10. 3.5. Relations between demographics, diagnostics, EAS and infant learning As shown in Table 3, mean difference scores were significantly positively correlated with maternal sensitivity and significantly negatively correlated with maternal hostility and maternal depression (1 = currently clinically depressed; −1 = not currently clinically depressed). To follow up on these correlations and to determine the relative contributions of diagnostic, demographic, speech acoustic, and EAS variables to infant learning as measured by difference scores, a hierarchical linear regression was performed (Table 4). Significant demographic correlates of clinical depression (ethnicity and maternal education) were entered into the analysis in step 1, followed by mean ΔF0 in IDS segments, antidepressant medication use (1 = current use, −1 = no current use), depression diagnosis (1 = current clinical depression; −1 = no current clinical depression), maternal hostility and maternal sensitivity in steps 2–6. Table 3. Correlations between measures of infant learning, demographic variables, depression diagnosis, and EAS ratings. 1 2 3 4 5 6 7 8 1. Learning (Diff) score – −.04 .08 −.10 −.22* −.05 .29* −.24* 2. Minority status – – −.38* −.20* .22* −.05 −.07 .12 3. Maternal education – – – .19* −.27* −.08 .14 −.07 4. ΔF0 – – – – −.05 .16* .06 −.12 5. Depression diagnosis – – – – – .16* −.12 .15 6. Antidepressants – – – – – – .02 −.04 7. Maternal sensitivity – – – – – – – −.46* 8. Maternal hostility – – – – – – – – * p = .05. Table options Table 4. Effects of demographic, speech acoustic, depression, medication, and EAS variables on infant difference scores from summation tests. Step Variable R2 change F change df1 df2 Sig. F change 1 Minority status/education .006 .43 2 133 .65 2 Mean ΔF0 .014 1.92 1 132 .17 3 Antidepressant meds .010 1.36 1 131 .25 4 Depression diagnosis .038 5.26 1 130 .02 5 Maternal hostility .051 7.42 1 129 .01 6 Maternal sensitivity .038 5.76 1 128 .02 Table options Table 4 shows that maternal education and ethnicity, medication use, and ΔF0 did not account for significant proportions of the variance in infant learning scores. However, after effects of those variables had been controlled, current depression diagnosis accounted for a significant proportion of the variance in infant learning (ΔR2 change = .038). After effects of maternal depression had been accounted for, there was a significant effect of maternal hostility, ΔR2 = .051, F(1, 129) = 7.42, p = .01. Finally, there was a further significant effect of maternal sensitivity, ΔR2 = .038 F(1, 128) = 5.76, p = .02, and the effects of maternal hostility was no longer significant (β = −.131, t = 1.42, p = .16) but the effect of maternal depression was still significant (β = −.170, t = 2.00, p = .05). Fig. 2 presents the mean difference scores as a function of depression diagnosis and maternal sensitivity category (created by a median split of the sensitivity scores). A 2 (NEP vs. DEP) × 2 (high vs. low sensitivity) ANOVA yielded a significant main defect of depression, F(1, 132) = 7.83, p = .01, η2 = .056, and a significant main effect of sensitivity category, F(1, 132) = 4.73, p = .05, η2 = .035, but no significant depression × sensitivity interaction, F(1, 132) = .62. Mean difference scores (mean looking at checkerboard plus voice, i.e., test ... Fig. 2. Mean difference scores (mean looking at checkerboard plus voice, i.e., test trials 1 and 4, minus mean looking at checkerboard alone, i.e., test trials 2 and 3), calculated from summation test phase data as a function of the infant's mother's clinical diagnosis and her EAS-based sensitivity category.