Maternal Linguistic Input to Infants Born to Mothers With and Without Postpartum Psychosis and Infant Language Skills: A Preliminary Study

Abstract

Purpose:

This preliminary study investigated maternal linguistic input (MLI) to infants born to mothers with and without postpartum psychosis (PPP), infant language skills, and the impact of MLI on infant language skills.

Method:

This is a cross-sectional case–control study from South India. The study included 14 mother–infant dyads, with seven dyads each in the PPP group, M_{age of infant}(SD) = 8.44 (1.93) months, and the no-PPP group, M_{age of infant}(SD) = 8.14 (1.8) months. MLI was assessed through video recordings of mother–infant interactions, focusing on both quantity (total number of words) and quality (type of words, type–token ratio, length of utterance, and types of maternal verbal responsiveness to infant vocalizations). Infant language skills were assessed between 6 and 12 months of infants' age using two parent interview measures: (a) Receptive–Expressive Emergent Language Test–Third Edition and (b) MacArthur–Bates Communicative Development Inventories, Words and Gestures. These measures were administered in languages spoken by the families, such as Kannada, Tamil, Malayalam, and Hindi.

Results:

Findings indicated that mothers with PPP provided significantly fewer total words and word types, shorter length of utterances, fewer verbal responsiveness to infant vocalizations, and a significantly higher number of no verbal responses to infant vocalizations when compared to mothers without PPP. No significant differences were observed in infant receptive and expressive language skills between the groups. Overall, strong positive correlations were observed between maternal responsiveness to infant vocalizations and expressive language skills as well as between no verbal responses to infant vocalizations and receptive vocabulary counts.

Conclusions:

The study highlights that there exist differences in the patterns of MLI in mothers with and without PPP. However, further research is needed with larger samples and longitudinal designs to explore the long-term impact of PPP on infant language development.

Mothers in the postpartum period are vulnerable to mental health concerns. Some mothers experience severe forms of mental illness. Postpartum psychosis (PPP), although rare, is a severe condition that affects one to two mothers per 1,000 deliveries worldwide (Meltzer-Brody et al., 2018). Its onset is rapid, typically occurs within 2 weeks after childbirth, and it often necessitates hospitalization. In India, however, there is delay in seeking help for such severe forms of mental illness, and patterns of help-seeking are reported to be around 11.3 weeks (Yadawad et al., 2021). The delay in seeking help could be due to (a) lack of resources; (b) the debilitating severity of clinical symptoms, leading to uncooperativeness; and (c) lack of awareness of the availability of appropriate care centers (Thippeswamy et al., 2018). Additionally, cultural perceptions of mental health may further contribute to barriers in seeking timely help from mental health professionals in India. For instance, Banerjee et al. (2021) reported such barriers as persistent perceived stigma, treatment side effects, misinterpretations, and time constraints from mental health professionals. Despite these barriers, mothers with severe mental illness express a need for early, direct communication and guidance regarding child care and parenting issues (Banerjee et al., 2021).

Symptoms of PPP include hallucinations, delusions (often centered on the infant), mania, depressive symptoms, and cognitive disturbances. These symptoms may disrupt mothers' overall perception of the infant and the quality of mother–infant (M-I) interaction, wherein mothers may demonstrate passivity, reduced responsiveness, and reduced emotional involvement or interest in the interaction with the infant (Snellen et al., 1999). For instance, Noorlander et al. (2008) found that mothers with PPP demonstrated reduced eye contact, physical touch, and vocal contact (dialogue exchanges between the mother and the infant) with the infant. These maternal behaviors indicate compromised quality of early input. Furthermore, the severity of PPP often leads to the hospitalization of mothers, leading to separation of the dyad. This early separation of the dyad may contribute to reduced input to the infant and subsequently affect infant development (Zhang et al., 2024). Notably, the onset of PPP coincides with the critical period of infant language development.

Talking to infants and verbally reciprocating their vocalizations are among the earliest forms of linguistic input provided by caregivers, such as mothers. As described by Rowe and Snow (2020), these early forms of high-quality linguistic input can be characterized into three related dimensions: linguistic, interactional, and conceptual. During infancy, the linguistic dimension of input includes repetitive utterances, the interactional dimension includes responding contingently to the infant, and the conceptual dimension includes talking about concrete and visible referents contingently. These linguistic, interactional, and conceptual dimensions of linguistic input help sustain the back-and-forth exchanges between caregivers and infants, which, in turn, play a crucial role in fostering infant language development (Bruner, 1981; Vygotsky, 1978; Warlaumont et al., 2014).

Furthermore, these components of linguistic input can be broadly categorized into quantity and quality. Quantity is defined as the total number of words or utterances spoken to or directed at an infant. Quality encompasses (a) diversity, obtained by counting different types of words; (b) repetitiveness of utterances, represented by the ratio of different types of words to the total number of words; (c) complexity of linguistic utterances, measured as the length of utterance, that is, the total number of words in an utterance; and (d) verbal responsiveness. Both quantity and quality of linguistic input show the ability to promote infants' receptive (understanding) and expressive (spoken) language skills (see the meta-analysis by Anderson et al., 2021). These linguistic input parameters, when provided by a mother, are often referred to as maternal linguistic input (MLI) and are associated with infant language skills. For instance, a higher total number of maternal words (quantity) at 18 months of age links to advanced vocabularies at the ages of 24 months (Hurtado et al., 2008) and 26 months (Huttenlocher et al., 1991). Similarly, the maternal length of utterance (quality) demonstrates predictive abilities for later vocabulary sizes at 24 months of age (Hoff & Naigles, 2002).

The quantity and quality of MLI are crucial for language development in infants (see the meta-analysis by Anderson et al., 2021). However, in the context of PPP, most studies evaluate the overall quality of M-I interactions using rating scales. These rating scales use descriptive qualifiers to rate interactions, such as “sensitive/insensitive responses”; “synchronous/asynchronous”; and “infant cooperativeness, compliance, passivity,” which are often subjective. These qualifiers provide only broad insights into the quality of maternal utterances or infant behavior. They do not provide insights into finer details, such as the structural and interactional components of MLI. Analyzing these aspects requires transcribing M-I interactions, a process that is often labor intensive. To our knowledge, no studies have yet examined the structural and interactional components of MLI in the context of PPP. Thus, this study aims to address the gap in understanding both the structural and interactional components of MLI through video recordings and verbatim transcription of M-I interactions.

In the context of infant language skills, there exist mixed findings on the language skills of infants born to mothers with PPP. Earlier research (McNeil et al., 1988) suggested no developmental differences between infants born to mothers with PPP and those born to mothers without postpartum mental health concerns. However, more recent studies have indicated the opposite. For instance, Chen et al. (2021) reported a higher likelihood of attention-deficit/hyperactivity disorder later in life (M_age = 6.24 years) in infants of mothers with PPP. Similarly, Biaggi et al. (2024) found lower language domain scores in these infants by 12 months of age. Therefore, this study aims to build on the limited existing literature by exploring differences in the language skills of infants born to mothers with PPP compared to those born to mothers without PPP during the first year of life.

Specifically, two research questions have been addressed: (a) Are there differences in the quantity and quality of MLI among mothers with PPP when compared to mothers with no mental health concerns in the postpartum period (referred to as “no-PPP”)? (b) Are there any differences in the language skills of infants (measured concurrently when infants were between the ages of 6 and 12 months) born to mothers with PPP when compared to infants born to mothers with no-PPP? An additional exploratory analysis of the associations between MLI and infant language skills was undertaken.

Method

Participants

This study was part of a larger ongoing study conducted at a tertiary health care center (National Institute of Mental Health and Neurosciences in Bangalore, India) with specialized units for perinatal outpatient psychiatry services and mother–baby inpatient psychiatry services. A cross-sectional case–control study design was employed. Participants were recruited between March 2022 and June 2023. Fourteen M-I dyads were included in this study, with seven dyads each in the PPP and no-PPP groups. The M-I dyads in the PPP group were recruited from the specialized units at the tertiary health care center. The M-I dyads in the no-PPP group were recruited from primary health care and rural child care centers (e.g., anganawadis). Descriptive statistics of the maternal, infant, and family characteristics of both groups are presented in Table 1. Infants in both groups were age-matched and recruited between the ages of 6 and 12 months. Participants in the PPP group were recruited from Bangalore, and participants in the no-PPP group were recruited from Bangalore and Kottayam, cities in Southern India, where a variety of languages are spoken. Recruiting families with a single/similar language background to maintain homogeneity of the spoken language was difficult, more so impossible, given the linguistic plurality in the data collection sites. Thus, patterns of assimilation of languages (e.g., Kannada and Hindi, Tamil and Kannada) were observed at the data collection sites. There was equal distribution of monolingual (n = 4) and bi/multilingual (n = 3) families in the PPP and no-PPP groups. In bi/multilingual families, caregivers identified primary home language as the language comfortably understood and spoken by the family members for daily routines or transactions. See Table 1 for primary languages of the participating families. Additionally, caregivers reported that they “borrowed words” from English to communicate—a pattern commonly seen in urban India. Written informed consent was obtained from mothers on behalf of themselves and their infants. This study was approved by the institutional review board (NIMH/DO/BEH.Sc.Div./2021–22).

Table 1.

Demographic characteristics of the mothers and their infants in the postpartum psychosis (PPP) and no-PPP groups.

Participant characteristic	PPP	No-PPP	Test statistic (U)	p
	(n = 7)	(n = 7)
	M (SD) Mdn (IQR) Range (min–max)	M (SD) Mdn (IQR) Range (min–max)
Maternal characteristics
Maternal age (years)	25.91 (3.4)	29.54 (5.12)	35	.201
	23.8 (5.65)	27.9 (2.85)
	22.5–30.2	24.6–40.2
Maternal education (years)	11.29 (1.7)	13.43 (1.81)	39	.056
	12 (2)	12 (3)
	9–14	12–16
Gestational age (weeks)	37.42 (1.9)	37.71 (1.60)	27	.792
	37 (1.5)	38 (1.5)
	34–40	37–40
Infant characteristics
Infant age	8.44 (1.93)	8.14 (1.8)	21.5	.749
	8.21 (2.87)	7.29 (2.3)
	5.62–10.71	5.91–11.04
Infant sex, n (%)
Female	4 (57.14)	5 (71.43)	—	1a
Male	3 (42.86)	2 (28.57)
Birth weight (kg)	2.74 (0.31)	3.28 (0.44)	40	.054
	2.75 (0.25)	3.50 (0.6)
	2.10–3.0	2.63–3.8
Family characteristics
Family type, n (%)
Joint	3 (42.86)	4 (57.14)	—	1a
Nuclear	4 (57.14)	3 (42.86)
Language exposure, n (%)
Monolingual	4 (57.14)	4 (57.14)	—	1a
Bi/multilingual	3 (42.86)	3 (42.86)
Language group based on video recordings, n (%)
Dravidian	6 (85.71)	6 (85.71)	—	1a
Indo-Aryan	1 (14.29)	1 (14.29)
Primary language of the family for assessments and video recordings, n (%)
Kannada	4 (57.14)	2 (28.57)	—
Tamil	2 (28.57)	—
Malayalam	—	4 (57.14)
Hindi	1 (14.28)	1 (14.28)
Socioeconomic status,b n (%)
Middle class – lower	2 (28.57)	4 (57.14)	—	.592a
Lower class – upper	5 (71.43)	3 (42.86)

Note. Em dashes indicate data not available. IQR = interquartile range; min = minimum; max = maximum.

^aFisher's exact test.

^bSocioeconomic status classification based on the revised socioeconomic scale (Ain et al., 2021): upper class, middle class (upper middle and lower middle), and lower class (upper lower and lower).

Selection Criteria for M-I Dyads in the PPP and No-PPP Groups

Mothers were recruited to the PPP group if they (a) had a clinical diagnosis of psychosis with a postpartum onset, as specified in the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10; World Health Organization, 2004) and as diagnosed by a psychiatrist; (b) had onset of a psychiatric illness within 24 weeks (6 months) following childbirth; (c) had a current illness severity score of 3 or below at the time of recruitment, as measured on the Clinical Global Impressions–Severity rating scale (Guy, 1976); (d) consented for audio–video recording; and (e) were comfortable understanding and speaking at least one or a combination of the following languages: Kannada, Tamil, Malayalam, Telugu, Hindi, and/or English. Mothers were excluded if they had (a) a diagnosis of intellectual disability based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (American Psychiatric Association, 2013), and (b) known sensory, motor, or neurological conditions interfering with providing language input (e.g., visual or hearing impairment). Infants born to these mothers were included if they were between the ages of 6 and 12 months. They were excluded if they (a) had known medical syndromic or neurological conditions, sensory impairment, and/or global developmental delay; (b) had extremely low birth weight (< 1,000 g) based on the guidelines (World Health Organization & United Nations Children's Fund, 2004); and (c) were extremely premature (i.e., gestational age of less than 32 weeks) based on the ICD-10 (World Health Organization, 2004).

Mothers in the no-PPP group were recruited if they had no mental health concerns, as reported in the Mini-International Neuropsychiatric Interview screening tool (Sheehan, 1998). Other inclusion criteria related to consent for recording, language, and exclusion criteria were like those for the PPP group. Infants born to mothers in this group were included if they were between the ages of 6 and 12 months. They were excluded if they had (a) a known medical diagnosis of any syndromic or neurological condition, sensory impairment, and/or global developmental delay; (b) birth weight less than 2,500 g; (c) gestational age less than 37 weeks; (d) significant prenatal history during the gestational period; (e) known or reported family history of developmental delay in first-degree relatives; and (f) greater than 2-month delay on any two developmental domains on the Communication DEALL Developmental Checklist (CDDC¹; Karanth et al., 2014), a criterion set for this study.

Procedure

In-person caregiver interview sessions were conducted to gather information on sociodemographic variables, maternal mental health, and infant language development. The responses from both mothers and alternate caregivers were considered to ensure that adequate and accurate information was collected. Sociodemographic information was collected using a study-specific questionnaire. The overall severity of the psychiatric symptoms of the mother's illness for the PPP group was measured using the Brief Psychiatric Rating Scale (BPRS; Overall & Gorham, 1962), a U.S.-based tool with 18 symptom domains rated on a 7-point Likert scale (1 = symptom present to 7 = extremely severe). Although there are no cutoff scores for the BPRS validated for mothers in India, this tool is widely used in clinical settings in India (Innamuri et al., 2022) and has been previously used in mothers with postpartum psychiatric illness at the study site (Yadawad et al., 2021). Infant receptive and expressive language skills were measured on the Receptive–Expressive Emergent Language Test–Third Edition (REEL-3; Bzoch et al., 2002). This is a standardized, norm-referenced (United States) parent interview measure. It assesses language skills across receptive and expressive domains in infants and toddlers from birth to 3 years of age. The REEL is used by speech-language pathologists in India to document receptive and expressive language skills and to explain to caregivers how their child's current skills align with those expected for their chronological age. However, as this tool was developed in English, prior to data collection, the REEL-3 tool was forward- and backward-translated from English to the languages spoken by the families in this study (Kannada, Hindi, Malayalam, and Tamil). Vocabulary counts for words understood (receptive) and words understood and spoken (expressive) were measured on the MacArthur–Bates Communicative Development Inventories (CDI), Words and Gestures (CDI:WG; Fenson et al., 2007). The CDI:WG is a standardized, norm-referenced (United States) parent report questionnaire for infants between 8 and 18 months of age. The CDI has been used in India previously in other populations (Rahman et al., 2016). As adaptations and validation (including normative data) were not yet available for the languages spoken by the families in this study, we opted to use the U.S.-based English version of the CDI. Caregivers were provided with English vocabulary words alongside translated equivalents in the languages spoken by the family. Assessments in bi/multilingual families were conducted in the primary language that the caregivers identified. Although the researcher asked questions in the family's selected primary language, an infant was given credit for a skill if they demonstrated it in any language spoken at home. Since norms were unavailable for most of the languages in our sample, we made an a priori decision to use raw scores in our analyses. Following these in-person assessments, 8-min audio–video recordings of a semistructured M-I play interaction were conducted. After completion of the recordings, families were briefed on the findings of the assessment sessions.

Recording of M-I Play Interaction

A total of 8 min of semistructured M-I play interactions were recorded when the mother and the infant were alert and active and the infants were well fed. Only the mother and the infant participated in the recording session to reduce spontaneous participation and linguistic input from other caregivers. Mothers were provided with a set of toys (e.g., books, windup toys, balls) that were culturally and developmentally appropriate. These toys were developed in consultation with five multidisciplinary experts in the field as part of a larger ongoing study. Mothers were briefed on the functioning of each of these toys and were provided time to understand how to use these toys with their infants. Mothers were instructed to play with their infants on a floor mat using these toys, as they naturally would at home. As mothers were recruited from outpatient departments, primary health care centers, and rural centers, a recording room within these organizations with minimal distractions and noise was chosen. Recordings were initiated when the mothers indicated that they were comfortable to start.

Coding Measures

MLI was defined as any transcribable verbal utterance spoken by the mother directed to the infant during the 8-min M-I interaction. Verbal utterances directed toward the researcher were eliminated from the coding procedure. MLI was measured in terms of quantity and quality. The quantity of input included the total number of words or token of words (TW). The quality of input included (a) the type of words (TyW); (b) type–token ratio (TTR); (c) the length of utterance (LU); and (d) the types of maternal verbal responsiveness to infant vocalizations (VR), further categorized as maternal verbal responsiveness to infant vocalizations (RIV) and no verbal responses to infant vocalizations (NR). The total number of infant vocalizations (IV) was also coded. The definitions of these parameters are listed in Table 2.

Table 2.

Abbreviations and definitions of maternal linguistic input (MLI) and infant vocalization parameters used in the study.

MLI parameter	Definitions
Token of words (TW)	Total number of words spoken by the mother
Type of words (TyW)	Total number of different words spoken by the mother
Type–token ratio (TTR)	Total number of different words (TyW) by total number of words spoken by the mother (TW)
Length of utterance (LU)	Total count of words spoken by the mother in a single utterance
Verbal responsiveness (VR)	Total count of a mother's prompt and contingent verbal responses to her infant's vocalizations. This was further classified as listed below:
Responsiveness to infant vocalizations (RIV)	Verbal maternal responses within 3 s (Fagan & Doveikis, 2019) of an infant's vocalizations (prompt) that are contingent to the infant's vocalization. This was calculated as the number of verbal responses by the total number of infant vocalizations.
No verbal responses to infant vocalizations (NR)	Instances when the mother does not produce any verbal response following an infant's vocalization. This was calculated as the number of no verbal responses from the mother by the total number of infant vocalizations.

Coding Process

A codebook was developed as part of a larger ongoing study to systematically annotate the M-I interactions (https://osf.io/wc79n/?view_only=9cf5be35aa10478587d28098c5e47801). All components of MLI from the M-I video interactions were uploaded onto the ELAN (EUDICO Linguistic Annotator) software (Version 6.0; The Language Archive, 2020). Figure 1 illustrates the coding process. Although families were bi/multilingual (i.e., spoke more than one language), mothers used their primary language in combination with borrowed words from the English language in the video recording. If the mothers used the same word in their primary language and in English (e.g., “elephant” in English and /a:ne:/ in Kannada [meaning “elephant”]), it was counted as two different words for token and as a single type of word, as both words are synonyms. The videos for the current study were coded by two experienced, independent, and reliable raters, where the second rater was blind to the group status. Both researchers were fluent in the languages spoken by the mothers in the video recordings. The primary rater manually segmented and transcribed the verbal utterances spoken by the mother. The second rater assessed the segmentation and transcription accuracy and resolved disagreements by obtaining consensus with the primary rater prior to coding MLI parameters. Interrater reliability was analyzed using the Wilcoxon signed-ranks sum test for all MLI parameters. No significant differences were observed between both raters for any of the MLI parameters (TW: p = .104, TyW: p = .525, TTR: p = .688, LU: p = .177, RIV: p = 1, NR: p = 1).

Figure 1.

Schematic representation of the coding process used to code the maternal linguistic input parameters on ELAN (EUDICO Linguistic Annotator).

Analytical Plan

Analyses were conducted using the RStudio statistical software (Version 4.2.3; R Foundation for Statistical Computing, 2023). As the data did not follow normal distribution, nonparametric tests were performed. Descriptive statistics were used to describe the data in terms of means, standard deviations, medians, and interquartile ranges (Quartile 3 – Quartile 1) for continuous variables and in terms of frequencies and percentages for categorical variables. In addition, 95% confidence intervals are reported. Fisher's exact test was used to assess the associations between categorical variables. To understand the group differences in MLI, Quade's analysis of covariance (Q-ANCOVA) was conducted, with maternal years of education as a covariate. For infant language skills, raw scores from the assessments were considered, as the norms from these tools were not standardized for infants in the Indian context. To understand the group differences in infant language skills, Q-ANCOVA was conducted, with infant age and birth weight as covariates. Only a subset of infants (n = 11) aged 8 months and above had CDI scores. The effect sizes of the findings were measured using partial eta-squared and were interpreted as small (0.01–0.05), medium (0.06–0.13), and large (˃ 0.14; Cohen, 1988). Spearman's correlation was utilized to understand the association between the severity of a psychiatric illness (as measured on the BPRS) and the parameters of MLI. Finally, an exploratory analysis was conducted to investigate the overall association between MLI and infant language skills. Owing to nonnormal data, these associations were explored using Spearman's partial correlation, while controlling for infant age and birth weight, for the receptive and expressive language skills on the REEL-3 and receptive vocabulary counts on the CDI. The strength of the associations was interpreted as very weak (ρ_s = .00–.19), weak (ρ_s = .20–.39), moderate (ρ_s = .40–.59), strong (ρ_s = .60–.79), and very strong (ρ_s = .80–1.0; Dancey & Reidy, 2007).

Results

Descriptive Characteristics

A total of seven M-I dyads each were included in the PPP and no-PPP groups. Among the maternal characteristics, maternal education approached statistical significance (p = .056) between the two groups, where the mean number of years of maternal education was lower in the PPP group (M = 11.29, SD = 1.7). Among the infant characteristics, infant birth weight approached statistical significance (p = .054), where the mean birth weight was lower for infants in the PPP group (M = 2.74, SD = 0.31). There were no statistically significant differences between the groups for other family demographic variables, such as family type, language, and socioeconomic status.

Mothers in the PPP group had a mean onset of psychiatric symptoms at 9 weeks postpartum (SD = 5.01), with 6.26 as the mean duration of illness (range: 1.84–10.17 months). The mean score for the overall severity of psychiatric symptoms, as measured on the BPRS, was 21.6 (SD = 3.91, range: 18–28), indicating minimal illness at the time of recruitment. The symptoms reported by the mothers on the BPRS profile included emotional withdrawal (n = 6), anxiety and tension (n = 3), blunted affect (n = 2), depressive symptoms (n = 2), and motor retardation (n = 1).

MLI in Mothers With PPP and No-PPP

Statistically significant differences in MLI were observed for the quantity of input, that is, TW (F = 6.63, p = .024, η_p² = .36), and for the four parameters of quality of input, that is, TyW (F = 6.68, p = .024, η_p² = .36), LU (F = 9.43, p = .01, η_p² = .44), RIV (F = 5.78, p = .033, η_p² = .33), and NR (F = 5.27, p = .041, η_p² = .08), as outlined in Table 3. Lower means were observed for all parameters, except NR, in the PPP group. Both RIV and NR are indexed to IV, which did not statistically differ (p = .10) between the infants in the PPP (M = 22.29, SD = 24.12) and no-PPP (M = 19.14, SD = 20.46) groups. Effect sizes, as measured in terms of partial eta-squared, were large for TW (η_p² = .36), TyW (η_p² = .36), LU (η_p² = .44), and RIV (η_p² = .33) and medium for NR (η_p² = .08). No significant correlations were observed for any of the MLI parameters and the severity of illness, as measured on the BPRS scale (TW: ρ = .270, p = .558; TyW: ρ = .126, p = .788; TTR: ρ = −.252, p = .585; LU: ρ = .126, p = .788; RIV: ρ = 0, p = 1; NR: ρ = −.545, p = .210).

Table 3.

Differences in the quantity and quality of maternal linguistic input (MLI) for semistructured play between the postpartum psychosis (PPP) and no-PPP groups while controlling for maternal education.

MLI parameter	PPP (n = 7)		No-PPP (n = 7)		F	p	Effect size (ηp2)
	M (SD) Mdn (IQR) Range (min–max)	95% CI [LL, UL]	M (SD) Mdn (IQR) Range (min–max)	95% CI [LL, UL]
TW	150.57 (114.65)	[44.54, 256.61]	522.28 (196.15)	[340.88, 703.69]	6.63	.024 *	.36
	138 (63.5)		637 (303)
	0–375		226–711
TyW	43.43 (25.48)	[19.86, 66.99]	89.42 (18.78)	[72.05, 106.8]	6.684	.024 *	.36
	50 (30.5)		91 (11.5)
	0–75		54–116
TTR	0.28 (0.18)	[0.11, 0.45]	0.18 (0.04)	[0.14, 0.23]	0.596	.455	.05
	0.26 (0.2)		0.16 (0.08)
	0–0.56		0.13–0.24
LU	1.20 (0.92)	[1.15, 2.85]	5.77 (3.28)	[2.27, 8.81]	9.427	.01 *	.44
	2.30 (0.46)		5.06 (3.98)
	0–2.75		2.43–11.05
RIV	0.22 (0.25)	[−0.01, 0.45]	0.68 (0.19)	[0.50, 0.86]	5.782	.033*	.33
	0.14 (0.19)		0.71 (0.26)
	0–0.71		0.33–0.88
NR	0.46 (0.34)	[0.15, 0.79]	0.05 (0.07)	[0.01, 0.12]	5.269	.041 *	.08
	0.47 (0.53)		0.04 (0.09)
	0.53–0.87		0–0.17

Note. Effect sizes reported are measured using partial eta-squared. IQR = interquartile range; min = minimum; max = maximum; CI = confidence interval; LL = lower limit; UL = upper limit; TW = token of words; TyW = type of words; TTR = type–token ratio; LU = length of utterance; RIV = verbal responsiveness to infant vocalizations; NR = no verbal responses to infant vocalizations.

^*p < .05.

Infant Language Skills Between Infants Born to Mothers With PPP and No-PPP

On the REEL-3, no statistically significant group differences were observed for both receptive (F = 1.31, p = .275) and expressive (F = 0.37, p = .555) language skills between infants in the PPP and no-PPP groups (see Table 4). The CDI was administered to a subsample of infants (n = 6 in the no-PPP group, n = 5 in the PPP group) aged 8 months and above. Given the age of the infants, there were no entries to the “words understood and spoken” category for infants in both the PPP and no-PPP groups. Hence, only findings related to “words understood,” that is, receptive vocabulary counts, have been reported. No significant group differences were seen for “words understood” (F = 0.3, p = .58) between infants in the PPP and no-PPP groups (see Table 4).

Table 4.

Language skills of infants on born to mothers with and without postpartum psychosis (PPP) based on Receptive–Expressive Emergent Language Test–Third Edition (REEL-3) and MacArthur–Bates Communicative Development Inventories (CDI) assessment tools.

Assessment domain	PPP		No-PPP		F	p	Effect size (ηp2)
	M (SD) Mdn (IQR) Range (min–max)	95% CI [LL, UL]	M (SD) Mdn (IQR) Range (min–max)	95% CI [LL, UL]
REEL-3
	n = 7		n = 7
Receptive language skills	18.86 (6.28)	[13.05, 24.67]	22.57 (6.97)	[16.12, 29.02]	1.31	.275	.01
	19 (11.5)		23 (10.5)
	11–25		15–27
Expressive language skills	16 (7.12)	[9.42, 22.58]	18.57 (6.78)	[12.30, 24.84]	0.37	.555	.03
	14 (13)		20 (10)
	8–24		10–28
CDI
	n = 5		n = 6
Words understood (receptive vocabulary)	35.8 (38.6)	[−12.12, 83.72]	12 (9.3)	[2.25, 21.75]	0.32	.58	.026
	31 (23)		8 (10)
	4–101		4–28

Note. Quade's analysis-of-covariance results are presented, with infant age and birth weight as covariates. Effect sizes reported are measured using partial eta-squared. IQR = interquartile range; min = minimum; max = maximum; CI = confidence interval; LL = lower limit; UL = upper limit.

Association of MLI to Infant Language Skills

Overall, strong positive associations were observed between (a) MLI-RIV and expressive language skills on the REEL-3 (ρ_s = .663, p = .02) and (b) MLI-NR and words understood (receptive vocabulary) on the CDI (ρ_s = .706, p = .03), as reported in Table 5.

Table 5.

Overall Spearman's partial correlations between maternal linguistic input parameters and infant language skills.

Assessment domain	TW	TyW	TTR	LU	RIV	NR
Receptive–Expressive Emergent Language Test–Third Edition (n = 14)
Receptive language skills	.215	−.63	−.318	.011	.264	−.162
Expressive language skills	.382	.243	.138	.327	.663 *	.359
MacArthur–Bates Communicative Development Inventories (n = 11)
Words understood (receptive vocabulary)	.265	.011	−.350	−.034	−.195	.706 *

Note. TW = token of words; TyW = type of words; TTR = type–token ratio; LU = length of utterance; RIV = verbal responsiveness to infant vocalizations; NR = no verbal responses to infant vocalizations.

^*p < .05.

Discussion

The present study is the first to understand (a) the structural and interactional components of MLI in terms of differences in the quantity and quality of MLI among mothers with PPP when compared to mothers with no mental illness in the postpartum period (no-PPP) and (b) the differences in the receptive and expressive language skills of infants born to mothers with and without PPP. The components of quantity and quality of MLI were systematically coded using 8-min semistructured M-I play videos. Infant receptive and expressive language skills and vocabulary were assessed through caregiver interviews. The findings of this study indicate that there are differences in the quantity and quality of MLI between mothers in the PPP and no-PPP groups. Infants did not demonstrate differences in receptive and expressive language skills based on caregiver reports. Overall, correlations between MLI parameters and infant language skills revealed strong positive correlations between RIV and expressive language skills on the REEL-3 and between NR and receptive vocabulary on the CDI.

We found significantly lower means for TW, TyW, and LU in the PPP group compared to those in the no-PPP group. Although there is no literature directly measuring MLI parameters in the context of PPP, a few studies have examined maternal verbal output in relation to individual symptoms that a person with psychosis can experience, such as emotional withdrawal, depressive mood, anxiety, and tension (also reflected in the BPRS profile in our sample). For instance, a systematic review by Scheiber et al. (2022) reported that four out of six studies observed a decrease in the total number and type of maternal words in mothers with depression. Similarly, two out of six studies reported a decrease in LU spoken by mothers with depression when compared to mothers without depression. Thus, findings of lower TW and TyW in the PPP group could be attributed to the influence of underlying psychiatric symptoms.

Findings of lower means for RIV and higher means for NR in the PPP group indicated reduced attention or a lack of response to infants' vocalizations. These findings suggest that, as seen in other postpartum psychiatric disorders (e.g., postpartum depression), mothers with PPP may miss cues from their infants (Ramsauer & Achtergarde, 2018), thereby possibly limiting opportunities to facilitate their infants' language development. The TTR was similar between the two groups, which could indicate similar patterns of repetitiveness utilized by mothers in both groups. However, the content of repetitiveness (i.e., the number of repetitions and the class of words repeated) should be explored in future studies.

The second aim of this study was to identify the differences in infants' language skills. No significant differences were observed in both receptive and expressive language skills on the REEL-3. The CDI was administered on a subsample of infants, and no differences were found for receptive vocabulary measures between infants in both groups. These findings contrast with the only other study conducted by Biaggi et al. (2024), which included a large sample of infants (n = 103) conducted in the United Kingdom. First, the infants in the current study were younger (M_age = 8 months) compared to those in the study conducted by Biaggi et al. (M_age = 12 months). Identifying differences at this young age may be challenging, as these differences may not have emerged this early. Thus, future studies must consider longitudinal study designs with infant language skills measured within both the first and second years of an infant's life. Second, unlike the study by Biaggi et al., where infant language skills were observed through tasks involving interactions with the infant, that is, a direct observation measure, the current study recorded caregiver report for infant language skills. An additional component of direct observation measures for infant language skills in this study could have provided a comprehensive overview, including information on the consistency of words or sound repertoires or communication patterns used by the infant upon interaction with a new individual (researcher). Thus, future studies must include assessment tools that measure infant language skills using a combination of direct observation measures and caregiver reports or interviews. Third, the current study included mothers who had mild psychiatric symptoms with a shorter duration of exposure, compared to those in Biaggi et al. In fact, a longer duration of symptoms has been related to adverse child outcomes (Murray et al., 2010). Last, these differences could be explained by cultural differences in the family structure. In the United Kingdom (the setting of Biaggi et al., 2024), the family structure predominantly includes two biological parents (Jarvis et al., 2023). In such a scenario, if maternal mental health is affected, the infant may receive reduced quantity and quality of linguistic input. In contrast, in India (the setting of the current study), the family structure is more variable and often involves extended family members, such as paternal or maternal grandparents. The presence of extended family members may buffer the effects of maternal mental health and provide an environment that facilitates the quantity and quality of linguistic input to the infant, thereby subsequently supporting the development of infants' language skills. Thus, future studies, particularly in collectivist society in India, should focus on accounting for the influence of linguistic input from other family members.

To explore the associations between MLI and infant language skills, all infants were grouped together. Significant strong positive associations were obtained between MLI RIV and REEL-3 expressive language skills. These findings are in line with McGillion et al. (2017) and highlight the importance of relevant responses and expressive language skills even at this young age. Significant strong positive associations were also observed between MLI NR and CDI receptive vocabulary counts. A possible explanation for this correlation is that NR reflects silence from the mother following an infant's vocalization. However, it does not preclude the possibility of mothers engaging in actions (e.g., actions related to the toys used in the M-I interaction). Thus, although there were no verbal responses to infant vocalizations from the mother, there could have been actions related to the toys (e.g., pointing to the book to direct the infant's attention) that influenced the receptive vocabulary of the infant—a line of thought somewhat comparable to that in Fagan and Doveikis (2019), which emphasized the frequent occurrences of nonverbal actions related to objects following maternal silences. However, these additional components of behaviors following maternal response categories need to be explored in future studies to explain these associations better.

The current study, despite its small sample size, has strengths. To the best of our knowledge, this is the first study to systematically examine the structural and interactional linguistic features of maternal input among mothers with PPP. A thorough methodology has been incorporated throughout the study. For instance, (a) age-matched infants were recruited; (b) culturally and developmentally appropriate toys were selected and reviewed by experts to elicit linguistic input in these M-I dyads; (c) a codebook was developed to systematically code the linguistic input; (d) responses to infant language measures were collected from both the mothers and the accompanying caregivers, indicating some reliability of the information; and (e) two researchers coded the videos, enhancing the reliability of the parameters of MLI.

The study also comes with limitations. First, the sample size in this study is small. Second, only structural and interactional aspects of MLI were evaluated. We did not examine the prosodic aspects related to MLI. Future studies need to adopt the trifocal view of MLI (i.e., linguistic, interactional, and conceptual aspects) and explore both verbal and nonverbal components of maternal behaviors, as suggested by Rowe and Snow (2020), to get a comprehensive view of the early linguistic environment. Owing to the limited sample size, we did not explore the mediating role of infant vocalizations on maternal verbal responsiveness and, thereby, on infant language skills. Due to ethical regulations at the study site, we could not include mothers with varying degrees of severity of the illness. Mothers included in this study had recovered and were on psychotropic medications. We did not explore the effect of psychotropic medications on mothers' linguistic utterances. Future studies must evaluate the linguistic utterances of mothers with varying degrees of severity of the illness overtime (i.e., at admission, at discharge, and during a follow-up visit). Next, although the mothers used a single language (monolingual) in the video recording, many infants were growing up in bi/multilingual environments. Future studies must investigate the influence of multiple languages spoken by family members on the infant's language skills. The linguistic plurality and recruitment from two data collection sites reduced the language homogeneity within the PPP and no-PPP groups. Last, we used U.S.-based versions of assessment tools to measure infant language skills. These tools were not normed on infants from India.

Future studies should (a) employ longitudinal study designs to understand the nuances of linguistic input and its influence on infants' later language development; (b) look into the comprehensive view of linguistic input, as suggested by Rowe and Snow (2020); (c) account for the totality of linguistic input provided to the infant, that is, input throughout the day and from other family members caring for the infant; and (d) explore the influence of bi/multilingual environments on infant language skills. These need to be carefully explored especially in countries such as India, where infants could be part of joint families (e.g., living with uncles, aunts, and grandparents).

Conclusions

In summary, findings from the present preliminary study indicate that the patterns of both quantity and quality of input are different among mothers with PPP when compared to mothers who have no mental illness in the postpartum period. Infants born to mothers with PPP did not demonstrate any differences in receptive and expressive language skills at 6–12 months of age. Significant strong positive correlations were observed between overall RIV and expressive language skills as well as between NR and receptive vocabulary counts. However, these findings need to be explored in a larger sample with a longitudinal study design.

A potential implication for clinical practice includes providing speech-language stimulation strategies to caregivers of infants younger than 12 months of age to help structure their utterances to enhance a rich language environment. Specifically, speech-language stimulation strategies could focus on (a) increasing the overall total number, variety, and length of utterances; (b) emphasizing repetitiveness, that is, repeating new words that are introduced; and (c) aiding the mother in identifying her infant's vocalizations while meaningfully expanding them. This could be done using video modeling or video feedback sessions, where caregivers and the clinician jointly view the interaction and identify areas of opportunities to promote input to the infant (e.g., Nevill et al., 2018; Whitehouse et al., 2019). Another potential implication for clinical practice includes involving speech-language pathologists in the multidisciplinary care of these M-I dyads. There is a need for collaboration between speech-language pathologists and mental health professionals to initiate screening of infant language development and support early caregiver–infant interactions in the context of PPP.

Data Availability Statement

Data will be made available from the corresponding author upon request.

Funding Statement

We are thankful to India Alliance for funding the salaries of two authors (Reny Raju and Malavi Srikar) through a grant from the Department of Biotechnology, Government of India, and the Wellcome Trust (IA/CPHE/18/1/503968; awarded to Shoba S. Meera). The funders had no role in study design, data collection, analysis, data interpretation, or writing of the report.