Introduction

The Five Minute Speech Sample (FMSS) has emerged as a key measure in recent parenting research (Sher-Censor 2015). To administer the FMSS, a researcher/clinician asks the respondent to speak for 5 min, with minimal interruptions, about their thoughts and feelings about the subject and how they get along together. This sample is audio-taped and coded. The FMSS was originally developed as a brief measure of expressed emotion (EE) in the families of individuals with a psychiatric illness (Magaña et al. 1986). Using the measure, high levels of EE (marked by high criticism or high emotional overinvolvement) in families has been associated with both short and long-term symptomology and increased risk of relapse for adults with schizophrenia (Magaña et al. 1986; Marom et al. 2005), and various other forms of adult psychopathology.

Associations between this index of EE and clinical outcomes in adult populations have led to the increasing use of the FMSS with child and adolescent populations. Parents’ FMSS scores have been associated with a wide range of parent and child correlates across a variety of domains, including intellectual disability, physical health, and psychopathology. For example, in a study involving mothers of children with intellectual impairment, high levels of EE were associated with higher levels of child behavior problems (Beck et al. 2004). Less positive and more negative parental relational schemas as rated from the FMSS have also been associated with higher levels of conduct problems for both clinic-referred and community samples of children (Pasalich et al. 2011a). FMSS ratings of maternal criticism have also been linked to the prospective development of depressive episodes in children, independent of preexisting levels of maternal and child depressive symptoms (Burkhouse et al. 2012). Furthermore, FMSS ratings have been shown to be linked to treatment outcomes across a range of child and adolescent mental or physical health conditions (e.g., Przeworski et al. 2012; Wamboldt et al. 1995).

Despite the widespread use of the FMSS in such research, surprisingly little is known about how FMSS scores relate to other methods for indexing parent–child dynamics. The gold standard for assessing parenting and parent–child dynamics is direct observation (Hawes et al. 2013). While self-report measures may be subject to socially desirable responding by parents, direct observation is thought to be less biased, with research suggesting that even when instructed to deliberately modify “bad” behaviors, participants in observed interactions are unable to do this (Johnson and Bolstad 1975). Furthermore, the parenting processes captured by direct observation may often be difficult for participants to self-evaluate, given that parenting often involves highly overlearned patterns of behavior which are enacted outside of conscious awareness (Gardner 2000; Hawes et al. 2013). However, the complexity and expense associated with direct observation can prove a limitation to its use in both clinical assessment and research. The FMSS may hold distinct advantages for use as a measure of parent–child dynamics; it has the potential to provide independent data on parent–child relationship domains that are free from biases associated with self-reports, and at the same time is significantly more convenient and affordable to administer than many forms of direct observation.

Developmental considerations have arisen regarding the appropriateness of using a measure originally intended for adult psychiatric patients to study parents and children. Furthermore, there is considerable expense associated with access and training for the original FMSS coding system. As a result of these issues, various modified FMSS coding systems have been created for families of young children and adolescents. A comprehensive review of these was provided by Sher-Censor (2015). The current review focuses on systems for which data from observed interactions is available. These systems are free and publicly available upon request. Detailed in Supplementary Table 1, they are the Preschool Five Minute Speech Sample (PFMSS; Daley 2001), the Family Affective Attitude Rating Scale (FAARS; Bullock and Dishion 2007), and a system developed by Caspi et al. (2004).The PFMSS does not aim to measure EE at all, but includes codes for initial statement, relationship, warmth, emotional overinvolvement (criteria altered from the original FMSS-EE system), critical comments, and positive comments. The FAARS (Bullock and Dishion 2007), which also does not meausure EE, focuses instead on the categories of Criticism and Warmth (additional codes available for Family Climate, Respondent Attributes, and Characteristics of Speech Sample). The Caspi et al. (2004) system focuses on the coding of positive and negative comments, and global ratings of Wamth and Negativity.

As highlighted by Sher-Censor (2015), use of the FMSS in child and family research has increased considerably in recent years. Much of the research in which the FMSS has been used as a measure of parent–child dynamics has followed the assumption that the verbal content of parents’ speech reflects behaviors or processes implicated in parent–child interactions. For instance, in a study by Burkhouse et al. (2012), the Criticism scale of the original FMSS is used to examine the effect of maternal criticism toward the child on the development of depression in children. Similarly, a study by Pasalich et al. (2011b), used the FMSS to explore the relationship between parental warmth and children’s problem behavior. However, it is conceivable that the thoughts and feelings expressed by a parent about a child may not necessarily translate into how they actually behave with the child. Indeed, this concern has previously been raised by researchers who have called for research into the assumption that thoughts and feelings verbalized by a parent about their child reflect their behavior during interactions with that child (Cruise et al. 2011; McCarty et al. 2004).

The major aim of this review was to systematically identify and evaluate the evidence that can now be brought to bear on this assumption. This evidence has grown significantly in recent years, as the FMSS has been used in a growing number of observational studies involving diverse populations. Furthermore, findings that FMSS ratings are associated with child and adolescent psychopathology suggest that it may have clinical utility as a brief alternative to time-consuming observational paradigms in clinical settings. In a previous review, Sher-Censor (2015) provided a broad overview of the FMSS, examining its history and a range of general issues arising from its use in parent–child relationship research. The author also described the various systems that have emerged to code the FMSS, and key findings from the use of the FMSS in developmental research. In contrast, the current review provides a systematic review and in-depth examination of one specific aspect of the current evidence base for the FMSS, namely, evidence regarding associations between FMSS data and observational indices of parenting. Key questions addressed were: (1) In which age groups and populations have associations between the FMSS and parent–child interactions been demonstrated?; (2) Do these associations differ between mothers and fathers?; (3) How do these associations vary across different FMSS coding systems, and across the different scales within these coding systems?; and (4) What are the associations between FMSS systems and observational indices of parenting and parent–child interaction at different ages?

Method

Overview of Search Strategy

A comprehensive literature search was performed on the following electronic databases in November 2015: PSYCinfo (1806-present); Medline (1950-present); Web of Knowledge (1864-present). A first group of searches aimed to identify studies involving the FMSS (as indicated in title or abstract, indexing it as a measure, citing the original paper which developed the FMSS, or mentioning variables measured by the FMSS in title or abstract; for instance—EE). A second group of searches identified studies that contained parenting or parent–child interaction variables by using related search terms (e.g., parenting; parent–child interaction; family interaction; discipline; parenting practices; parent–child relationship). Exact search terms for each database are displayed in Supplementary Table 2. These two search groups were then combined. A total of 618 studies (after removing duplicates) were retrieved using this method. Studies were initially screened by title and abstract in order to identify relevant articles, resulting in the exclusion of 435 studies (127 studies were excluded due to not using a parent and child/adolescent population, the remaining 308 studies did not include a parenting or parent–child interaction variable). The remaining 183 studies were screened at the full text level. One hundred and fifty-eight papers were excluded at this stage (156 due to not investigating the relationship between a parenting/parent–child interaction variable and the FMSS, and two due to modifying the administration or coding of the FMSS so much as to render it incomparable with other studies), resulting in a final set of 25 studies. Search strategy, along with reasons for exclusion at each stage, is displayed in Fig. 1.

Fig. 1
figure 1

Overview of search strategy, including exclusion criteria at each stage

Full size image

Inclusion/Exclusion Criteria

The current review considered all studies that investigated associations between the FMSS and parenting or parent–child dynamics variables as coded from parent–child observations. Associations were defined quite broadly, to include any statistical evidence of a relationship between the FMSS and observed interactions, for instance via correlation, path analysis, or mean differences in behavioral measures between groups differentiated by FMSS scores. Only studies published in English language and in peer-reviewed journals were included. No limitations were placed on sample characteristics (e.g., child/adolescent or caregiver psychopathology or physical health status). Included studies had to use the FMSS and an observed parent–child interaction, and to examine associations between these two measures. Observed parent–child interactions had to be coded for variables associated with parenting, or parent–child dynamics. Observed interactions coded only for child variables (e.g., only child behavior, or only child attachment) were not included. Studies using various procedures for administering and coding the FMSS were included, unless the nature of these variations significantly altered the fundamental nature of the FMSS as a measure of parents’ thoughts and feelings about their children. For instance, St Jonn-Seed and Weiss (2002) had respondents give speech samples about their infant while the infant was present, and expanded the coding system to include behaviors toward their infant during administration.

Quality Assessment

Studies were assessed for quality using a modified version of the Downs and Black (1998) checklist for measuring study quality (Ferro and Speechley 2009), which includes 15 items assessing domains of study reporting, internal and external validity, and power. Indicators of quality considered most important to the current review, were the description of participants (both children/adolescents and parents), sample size, and perhaps most crucially, the validity and reliability of measures. In order to assess the latter domain, we considered whether both observational measures and the FMSS were coded by several raters who were properly trained, blinded to participant information (when relevant), and had acceptable inter-rater agreement and internal consistency.

Key Definitions

Observational indices of parenting and parent–child interaction were categorized as either positive or negative, in keeping with recent reviews and meta-analyses of parenting behavior (e.g., Thomas and Zimmer-Gembeck 2007; Wilson and Durbin 2010). Definitions of positive and negative parenting were adapted from Wilson and Durbin (2010), such that parenting behaviors were classified as “positive” if they involved, for instance, warm, affectionate, sensitive, engaged, positive, accepting, and supportive behaviors and/or interactions. Alternatively, parenting behaviors were classified as “negative” if they involved, for example, hostile, coercive, intrusive, restrictive, controlling, negative, critical, and dysfunctional behaviors and/or interactions. If an observational variable did not clearly fit into either of these categories, for instance “separate play,” it was not included. However, this was a very rare occurrence and did not result in the exclusion of any identified studies from the review.

Results

Data abstracted from the identified studies included sample size, sample characteristics, study design, participating parent (mother/father), interaction task and parenting/parent–child interaction variables, FMSS administration and coding system, and findings regarding associations between FMSS ratings and observed parent–child interactions (see Table 1 for an overview, and Table 2 for detailed descriptions of associations including statistical results). These studies spanned a 26 year period from 1989 to 2015. Most studies (22) examined associations between the FMSS and cross-sectional data on observed parent–child interactions, two studies examined these associations both cross-sectionally and longitudinally, one study examined only a longitudinal association between observed parenting, and FMSS data two years later. Given the nature of the review question, all studies except two, were not intervention studies but examined the relationship between predictor variables and child or family outcomes. In most studies (21), either all or most participants in the FMSS and observed interactions were mothers, with several other studies including both fathers and mothers (3), and one study involving fathers only. It should be noted that five studies (Smith et al. 2013, 2015; Waller et al. 2012, 2014, 2015) included in the review were not independent but made use of the same sample, and this is also true of a further two studies (Pasalich et al. 2011a, b), which shared the same sample.

Table 1 Characteristics and results of included studies
Full size table
Table 2 Observed parenting and parent–child interaction variables associated with FMSS by sample age
Full size table

Study Quality

Quality index scores for all studies are displayed in Table 1. Scores varied from 8–12 out of a possible total of 15. However, given the focus of the review, the overall score was not considered the most important index of quality. Rather, key issues of quality in the papers reviewed were small sample size (n < 40) in 6 (24 %) of the 25 studies (Hermanns et al. 1989; Kubicek et al. 2013; Marshall et al. 1990; Schobinger et al. 1992; Scheeringa et al. 2015; Narayan et al. 2012) and questionable reliability or validity of FMSS or observational indices (or inadequate reporting of this information) in 7 (28 %) of the 25 studies (Challacombe and Salkovskis 2009; Kubicek et al. 2013; Psychogiou et al. 2008; Scheeringa et al. 2015; Smith et al. 2013; Waller et al. 2012, 2014). Given that exclusion of the 11 studies identified above would result in substantially reducing the number of total studies reviewed, we decided to not to exclude them in order to provide a review of all the available evidence. However, issues of quality were considered in reporting results of the review. Furthermore, the quality assessment findings highlight the need for future studies to use larger sample sizes, and to ensure and report clearly the reliability and validity of FMSS and observational measures.

FMSS and Observed Parent–Child Interactions

In 21 of the 25 included studies (84 %), significant associations were seen between FMSS ratings and observational indices of parent–child interactions. Although not all studies used correlation analyses to demonstrate these associations, when correlations were reported, they ranged in magnitude from small to large (r = .12–.65). These studies were diverse in their sample type and characteristics, parent participant, observed parent–child interaction variables, and FMSS administration and coding. Four studies (Kubicek et al. 2013; Pasalich et al. 2011b; Schobinger et al 1992; Scheeringa et al. 2015) did not find any association between the FMSS and observed parent–child interaction, however, three of these studies, as identified above, suffer from small sample sizes, which may contribute to the lack of significant findings.

Child Sample Characteristics

Child samples in the identified studies spanned a broad range of ages, from infants with a mean age of 17 months, to adolescents with a mean age of 17 years. Child participants were sampled from diverse populations. Fifteen papers used largely homogeneous samples from the following populations: children in the community (1), children with asthma (2), infants with special needs (1), children with externalizing problems (3), low-income families (1), homeless families (2), traumatized children (1), children at risk for either externalizing problems, family problems, socioeconomic problems (4). The remaining ten papers used mixed samples, including: children with OCD or panic, and controls (1), children with internalizing and externalizing problems (1), children of depressed mothers and children of non-depressed mothers (2), children with, and without asthma (1), children with, and without ADHD (2), adolescents with high, and low levels of antisocial behavior (1), children with, and without externalizing problems (1), and youths at high, medium, and low, risk of antisocial behavior (1). There were two populations for which associations between the FMSS and observational indices were not found: traumatized children (Scheeringa et al. 2015), and infants with special needs (Kubicek et al. 2013). Again, it should be noted that both of these studies were hampered by a small sample size, and furthermore, the Kubicek et al (2013) study suffered from poor inter-rater agreement for FMSS ratings.

Parent Sample Characteristics

Among the 21 studies in which most or all parents were mothers, 18 (86 %) studies demonstrated significant associations between FMSS ratings and observational indices. Of the three studies that included both fathers and mothers, 2 (67 %) found evidence for some relationship between FMSS and observed parent–child interactions. The only study in which all participants in the FMSS and observed interactions were fathers (Schobinger et al. 1992) did not find any significant association between the two measures, however small sample size was an issue for this paper.

FMSS Coding Systems

FMSS Coded Using Original Magaña et al. (1986) Method

Nine studies administered the FMSS using the original system devised by Magaña et al. (1986), and an additional three studies used modified versions of it. Among the four studies that indexed High-EE, 2 found associations between this rating and observed parent–child interaction. A revised version of High-EE, developed by Wamboldt et al. (2000), was also associated with parent–child interaction in this study. Of the nine studies that indexed Criticism, 7 found associations between this rating and observed parent–child interaction. Of the four studies that indexed EOI, none found it to be related to observed parent–child interactions. However, one of the revised versions of EOI (but not the other), developed by Wamboldt et al. (2000), did demonstrate such a relationship in that study. Overall, of the 12 studies that have used the original FMSS coding method or modified versions of it, 10 (83 %) demonstrated at least some association between the FMSS and observed parent–child interactions, with correlations across studies ranging from small to large (r = .19–.65).

Preschool Five Minute Speech Sample (Daley 2001)

In two studies the FMSS was administered and coded according to the system published by Daley et al. (2003). The scales of Initial statement, Warmth, and Positive comments, each demonstrated an association with observed parent–child interaction in one of the two studies. In both studies, Relationship and Critical comments were associated with parent–child interactions. In a study by Scheeringa et al. (2015), the PFMSS (Daley et al. 2003) was administered, however five of its scales (with the exception of EOI) were combined to create a scale intended to represent emotional sensitivity. This study did not find any association between the emotional sensitivity rating derived from PFMSS scales and observed parenting. In summary, for both of the studies that used the PFMSS method in both administration and coding, parental speech samples demonstrated some associations with observational indices, with correlations ranging from small-medium to medium (r = .22–.44), yet in the study which modified the coding system to reflect “emotional sensitivity,” no associations were found.

FMSS Coded Using the FAARS (Bullock and Dishion 2004)

Eight studies used the FAARS system to code the FMSS. Of the seven studies that included Warmth in the FMSS, 6 demonstrated associations between the FMSS and observed parent–child interactions. Of the five studies that indexed Criticism with the FMSS, all showed an association with observational indices. In total, seven of the eight studies (88 %) in which the FMSS was administered and coded using the FAARS system demonstrated associations with observational indices, with correlations, where reported, ranging from small to medium-large (r = .12–.51).

FMSS Coded Using Caspi et al. (2004) Method

Three studies (two already mentioned, also used the FMSS-EE system) used Caspi et al.’s (2004) method for coding the FMSS. Both studies that indexed Negativity demonstrated associations with observed interactions, and one of the two studies that included Warmth found associations between Warmth and observed parent–child interactions. One study found associations between Positive comments and observed parenting behavior, but no association between Negative comments and observational indices. In summary, all three studies using Caspi et al.’s (2004) method demonstrated associations between the FMSS and observational measures of parent–child interaction, with correlations ranging from small to large (r = −.17–.65).

Observed Parenting and Parent–Child Interaction Variables

Positive Parenting

Seventeen studies, spanning all coding systems including in this review, examined an observational measure of positive parenting behaviors (e.g., warmth, problem-solving). Of these studies, fourteen (82 %) found some association between the FMSS and positive parenting variables. Where reported, correlations between positive parenting variables and FMSS scales ranged from small to large (r = .12–.65)

Negative Parenting

Eighteen studies, spanning all coding systems included in the review, examined an observational measure of negative parenting behaviors (e.g., observed criticism, harshness). Thirteen of these studies (72 %) found some association between FMSS ratings and negative parenting variables, with correlations, where reported, ranging from small to large (r = −.13–.60).

Negative Parent–Child Interactions

Four studies included observational indices that were focused on the dynamics of interaction between parent and child, as opposed to focusing only on parenting behaviors (e.g., mother-adolescent coercion). All of these parent–child interaction variables were classified as negative. All four studies demonstrated associations between the FMSS and observed interaction variables. Only one of these studies used correlations to analyze these associations, and found the association to be medium in size (r = −.36).

Associations Between FMSS and Observational Variables by Age

In order to examine the evidence for the use of the FMSS as a measure of parents’ interactions with their children at different ages, studies were categorized based on the mean age of the sample, or where this was not provided, on the median of the age range. Table 2 displays the statistical evidence for associations between observational indices and FMSS ratings, grouped by sample age. Parenting and parent–child interaction variables that were associated with FMSS ratings for each age group are summarized below.

Infancy

Two studies used the original FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993) system for coding the FMSS. The study by Kubicek et al. (2013) did not find any association between EE, Criticism, or EOI and observed positive parenting variables. Kaugars et al. (2007), used the FMSS-EE criticism scale, and several other scales developed for the purpose of the study. Observed responsiveness and acceptance were associated with several of the scales developed by Kaugars et al. (2007), yet neither were associated with FMSS-EE criticism. Correlations for significant associations were small (r = .19–.24).

Preschool Age

Seven studies included a sample of preschool aged children. Two of these studies (Daley et al. 2003; Psychogiou et al. 2008) used the PFMSS (Daley 2001), and both studies demonstrated associations between the FMSS and observed parenting. Among these studies, observed praise and observed affection demonstrated associations with PFMSS scales, as did observed criticism. Five studies (Smith et al. 2013, 2015; Waller et al. 2012, 2014, 2015), which also used the same sample (or subsets of this sample) used the FAARS (Bullock and Dishion 2004), and in all studies the FMSS was related to observed parenting or parent–child interactions. Specifically, observed positive behavior support and observed warmth were associated with FAARS scales, as was observed harsh parenting. In summary, all seven studies that used the FMSS with preschool children and their parents found associations between the FMSS and observational indices. Where reported, correlations ranged from small to medium-large (r = .12–.45).

Early-Mid Childhood

Six studies included samples of children in early-middle childhood. Three studies (Narayan et al. 2012, 2015; Wamboldt et al. 2000) used Criticism coded from the FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993), one also used EOI from the FMSS-EE, and these three studies also used additional codes from other methods or systems (refer to Tables 1 and 2 for further detail). Positive parenting variables associated with FMSS-EE Criticism were: observed positive involvement, overall effective parenting, skill encouragement, and problem-solving. Negative observed parenting variables associated with FMSS-EE Criticism were: inept coercive discipline, negative affect, and harshness/hostility. Several of the observational indices were also related to other FMSS scales used in these studies, namely Warmth and Negativity, coded using Caspi et al.’s (2004) method. Two studies (Pasalich et al. 2011a, b; these studies were not independent as they made use of the same sample) coded the FMSS using the FAARS (Bullock and Dishion 2007), and one of these demonstrated associations between observed parenting and FMSS ratings. In this study (Pasalich et al. 2011a), observed warmth/responsiveness, and engagement were associated with FAARS scales for both mothers and fathers, but praise was only associated with FAARS scales for fathers. One study, by Scheeringa et al. (2015) used the PFMSS (Daley 2001) to score speech samples, yet combined the ratings to form a novel “emotional sensitivity” scale. However, this study only examined an association between this scale and observed parenting longitudinally (observed parenting at time 1, speech samples 2 years later), and did not find a relationship between them. In total, four of the six studies (75 %) that used the FMSS with early-mid childhood samples found associations between the FMSS and observed parenting or interaction variables, with correlations ranging from small to large (r = .18 to −.65).

Mid-Late Childhood

Seven studies (Challacombe and Salkovskis 2009; Cruise et al. 2011;Hermanns et al. 1989; Marshall et al. 1990; McCarty et al. 2004; Park et al. 2008; Schobinger et al. 1992) included samples of children in mid-late childhood, and all of these studies used the FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993) coding system. Both observed positive relations and observed negativity were associated with both positive and negative remarks made during the FMSS. Observed autonomy-granting and warmth were associated with High-EE, whilst observed warmth was associated with Criticism. Negative parenting and interaction variables associated with FMSS-EE Criticism were: observed critical behavior and criticism, sequence length of negative verbal exchange between dyads, negative affective style, coercive verbal processes, antagonism, negativity, disgust, and harshness. In summary, six of the seven studies (86 %) that used the FMSS with families of children in mid-late childhood demonstrated associations between FMSS ratings and observational indices. Where reported, correlations between the FMSS and observational measures in these studies ranged from small-medium (r = −.14–.27).

Adolescence

Four studies (Wamboldt et al. 2000; Tompson et al. 2012; Bullock and Dishion 2007; Scott et al. 2011) used the FMSS in relation to adolescents. Two studies used FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993) coding, with one study also using additional revised EE and EOI scales (Wamboldt et al. 2000). FMSS-EE Criticism was associated with interpersonal attunement, observed problem-solving and observed positivity. Observed positive affect, maintenance of interpersonal boundaries, and interpersonal attunement were also related to additional/revised scales. One study (Bullock and Dishion 2007) used the FAARS (Bullock and Dishion 2004) in addition to positive, also critical remarks as rated from the FMSS-EE system. Suggesting ideas, acknowledging/responding, mother coercion, mother-adolescent coercion, and observed criticism were associated with FMSS-EE comments and FAARS scales. One study used Caspi et al.’s (2004) method to code positive and negative comments (Scott et al. 2011) and found that both observed positivity and negativity were associated with positive comments. In summary, all studies that used the FMSS with parents of adolescents demosntrated associations between FMSS ratings and observational indices, with correlations ranging from small to medium-large (r = −.17–.51).

Longitudinal Associations

Three studies examined longitudinal associations between the FMSS and observational indices of parent–child interaction. In two of these studies, FAARS-rated Criticism, was found to be related to trajectories of parent–child coercion over time (Smith et al. 2013, 2015). The third longitudinal study examined retrospective associations between parenting behavior at time 1 and an “emotional sensitivity” scale based on PFMSS ratings 2 years later, yet no evidence of a link was found (Scheeringa et al. 2015). It must be noted, however, that this final study was limited by a small sample size. In summary, 2 of the 3 (67 %) studies to examine longitudinal associations demonstrated a link between FMSS ratings and observed parent–child interactions (β = .28–.30).

Discussion

Use of the FMSS in parenting research has become increasingly widespread in recent years, raising important questions regarding whether and how it maps onto observational indices of parent–child interactions. Among the 25 studies that have provided data on such associations to date, significant associations between these respective measures have been reported in the vast majority (20). This indicates good support on the whole for the use of the FMSS to index parent–child interactions. At the same time, the available evidence highlights a number of important considerations with respect to key issues, as follows.

In all age groups represented in the identified studies (samples spanning 17 months to 17 years of age), significant associations were demonstrated between the FMSS and observational indices of parent–child interactions. However, as will be detailed below, the data for use of the FMSS with parents of infants is limited. Likewise, FMSS ratings were consistently associated with parent–child interactions in all of the various populations represented in these studies, with the exception of families of young children with special needs (Kubicek et al. 2013), and traumatized children (Scheeringa et al. 2015). Further research is needed to clarify the applicability of the FMSS to these populations, given the current availability of only one study using the FMSS with observational data for each of these populations. Overall, findings suggest that the FMSS is a promising measure of parent–child dynamics across a diverse range of developmental stages and populations.

There is now considerable evidence that mothers’ FMSS ratings are associated with parent–child interactions, however the evidence regarding fathers is far less clear. This reflects the overwhelming focus on mothers within the parenting literature (Cabrera et al. 2000), despite evidence that fathers play a unique and important role in children’s development. The only study in which participants were all fathers did not find associations between the FMSS and observed interactions (Schobinger et al. 1992). This study was limited by small sample size (n = 24), however, another similarly sized study that included only mothers did demonstrate a significant association (Hermanns et al. 1989). We identified two studies that examined the relationship between FMSS and observed interactions separately for mothers and fathers. One of these studies did not find associations between FMSS and observational indices for either mothers or fathers (Pasalich et al. 2011b), whereas the other study found differences in how mothers’ vs. fathers’ FMSS ratings were associated with observed interactions (Pasalich et al. 2011a). Specifically, observed praise was associated with both FAARS warmth and criticism for fathers, whereas it did not demonstrate any significant association for mothers. Secondly, all positive parenting behaviors demonstrated significant negative associations with FAARS criticism for fathers, whilst none did for mothers. The reasons for such a finding are unclear, however it appears to suggest that when fathers’ speak more negatively about their child and their relationship with their child, this is more likely to be reflected in a lack of warm parenting behavior. For mothers, however, more negative speech about their child may not actually reflect less warmth in parent–child interactions. Perhaps this is due to mothers tending to talk more overall, both positively and negatively, during speech samples. This interpretation, however, is largely speculative. In any case, differences between mothers and fathers in associations between FMSS ratings and observed behavior are potentially important, and represent an important area for future research. This is particularly relevant in the context of previous research showing that mothers and fathers may differ in aspects of parenting (Simons and Conger 2007), and that mothers’ and fathers’ behaviors may be differentially associated with child outcomes (Gryczkowski et al. 2010).

All four systems for administering and coding the FMSS have demonstrated associations with observed parent–child interactions. For studies using the FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993), the most supported scale was Criticism. The evidence for High-EE was less clear: it demonstrated associations in two of the four studies in which it was examined. Perhaps this is due to the broad definition of High-EE, which can be coded when parents are high in either criticism or emotional overinvolvement (the latter which does not appear to be associated with parent–child interaction). Finally, EOI failed to demonstrate any associations with observed parent–child interaction across all studies that examined it, with the exception of a revised EOI code used by Wamboldt et al. (2000). McCarty et al. (2004) offer two explanations for a lack of association between EOI and observational indices. One explanation is that parent behaviors associated with EOI may not be possible to code from parent–child interactions (McCarty et al. 2004). However, this seems unlikely, as the current review demonstrated that behaviors associated with EOI (such as autonomy-granting, and maintenance of interpersonal boundaries) have been reliably coded from observed interactions, and have also demonstrated associations with FMSS scales (Challacombe and Salkovskis 2009; Wamboldt et al. 2000). An alternative explanation for the lack of associations between the original EOI scale and observational measures, is that key features of EOI may indeed represent emotional overinvolvement for relatives of adults (for whom the FMSS was originally devised), but not for children. For instance, FMSS scoring for the EOI scale includes expression of love, giving detail about the target’s past, and “excessive” praise. Whilst these speech sample features may indeed correspond to intrusive and guilt-inducing behavior when they are expressed about an adult, they may actually be developmentally appropriate when directed at a child. This explanation is supported by the consistent lack of evidence for an association between FMSS-EE EOI and observed parent–child interaction across several studies examining a range of parent–child interaction variables. As such, the original EOI code appears to have limited utility in capturing qualities of the parent–child relationship.

The PFMSS (Daley 2001), designed specifically for use with preschool children, has been used alongside observational indices in three studies, and demonstrated associations with observed parent–child interaction in two of these. All scales showed some association with interactions in at least one of these studies, except for EOI. In a third study (Scheeringa et al. 2015), the PFMSS was used with children in early-mid childhood, however in this study, the authors combined the ratings from PFMSS scales (excepting EOI) to form a novel “emotional sensitivity” scale. In this study, no relationship was found between the emotional sensitivity scale and observed interactions, however it must also be noted that this study examined longitudinal associations between observed parenting, and PFMSS-rated emotional sensitivity 2 years later, and also that this study was limited by a small sample size. There is thus preliminary evidence that the PFMSS coding system captures meaningful variance in parent–child interactions, however ongoing research is needed to examine this.

There was evidence that the FAARS (Bullock and Dishion 2004) system for coding the FMSS was associated with observed parent–child interaction, in regards to both Warmth (demonstrated associations in all but one study) and Criticism (demonstrated associations in all studies) scales. Finally, Caspi et al.’s (2004) system for coding warmth and negativity from the FMSS has been used in two studies, and a third study coded for positive and negative comments using this method. From these few studies, negativity was associated with observed interactions in two studies, warmth was in one study, and positive comments was in one study. There is thus some preliminary evidence that Caspi et al.’s (2004) system relates to parent–child interactions, yet further research is needed to clarify this.

Both studies that used the FMSS alongside observational measures in families with infants used the original FMSS-EE coding system (Magaña et al. 1986; Magaña-Amato 1993). Although one of these studies found some associations between observed positive parenting variables and additional codes created for the study, neither study found any association between FMSS-EE ratings and observed behavior. There is therefore currently no behavioral evidence to support the use of the FMSS-EE system as a measure of parent–child dynamics with infants, although further research is needed to clarify this. In particular, the two studies with infant populations identified by this review both indexed only positive observed parenting variables, whereas it may be expected that negative parenting variables would be more closely associated with the FMSS-EE coding categories, as they reflect largely negative constructs. Future research should also include observational measures of negative parenting and parent–child interaction variables. Furthermore, additional research should examine whether other FMSS coding systems may capture meaningful differences in parent–infant interactions.

Research in which the FMSS has been used with preschool-aged children has used two coding systems; the PFMSS (Daley 2001), and the FAARS (Bullock and Dishion 2007). Each of these systems have demonstrated links to observed interactions between parents and preschool-aged children, for both positive and negative observed parenting. As such, the available evidence supports the use of these two coding systems with preschool-aged children.

Coding systems used to rate the FMSS of parents with children in early-mid childhood include the original FMSS-EE system (Magaña et al. 1986; Magaña-Amato 1993), the FAARS (Bullock and Dishion 2007), the PFMSS (Daley 2001), and Caspi et al.’s (2004) system. The three studies using the FMSS-EE provide some support for the utility of the Criticism scale in corresponding to observed interactions between parents and children of this age, however only one study also used the FMSS-EE codes of EOI and EE, neither which demonstrated associations to observational indices. Two studies also provided preliminary support for the Caspi et al. (2004) system with this age group. Of the two studies using the FAARS system, only one study found associations between FAARS ratings and positive observed parenting variables (but not negative parenting variables). The other study to use the FAARS with this age group did not find any association between FAARS warmth and coercive parenting but did not include FAARS criticism, or an observational measure of positive parenting, thus further research is needed to evaluate the behavioral evidence for the FAARS with this age group. Finally, the PFMSS system did not demonstrate an association with observed parenting in this age group- however, as noted previously, this study modified the coding system of the PFMSS to combine several codes into a novel scale, examined associations only longitudinally, and also was limited by a small sample size. Therefore, further research is needed to clarify whether the PFMSS may be reflective of parent–child interactions beyond the ranges of its intended preschool-aged population.

The only coding system for which behavioral data is available when the FMSS is used for parents of children in mid-late childhood, is the FMSS-EE (Magaña et al. 1986; Magaña-Amato 1993). Six of seven studies using samples of this age provide support for the use of this system in mid-late childhood, however some scales were better supported than others. Specifically, whilst Criticism is most consistently related to both positive and negative parenting and there is some evidence for High-EE being linked to observational indices, there is no evidence that EOI represents meaningful differences in parent–child behavior. Further research is needed to establish whether other coding systems may be useful for children in this age group.

The FMSS-EE system (Magaña et al. 1986; Magaña-Amato 1993), FAARS (Bullock and Dishion 2007), and Caspi et al.’s (2004) system, have been used alongside observational indices for adolescent populations. Two studies suggest that FMSS-EE Criticism may relate meaningfully to observed positive parenting behaviors, yet these studies did not provide any evidence that the FMSS-EE may also reflect negative parent–child interactions. The one study to use the FAARS with parents of adolescents provided preliminary evidence for the utility of this measure to capture both positive and negative parent–child interactions, although this requires replication. Finally, positive comments as coded by the Caspi et al. (2004) system were linked to both positive and negative parenting, yet further research is needed to investigate whether the other scales in this system (warmth and negativity) relate to observed parenting variables for this population.

Considerable evidence is now available to support the use of the FMSS as a measure of parent–child dynamics, particularly in terms of the affective quality of the parent–child relationship. This is encouraging given that the FMSS may represent a desirable alternative to self-report measures and assessments based on the direct observation of parent–child interactions. The unstructured, free-flowing quality of speech samples may not only be more naturalistic than data collection using questionnaires, but may allow for a greater breadth of information to be collected regarding quality of parenting and the parent–child relationship. At the same time, the evidence identified in this review highlights a number of considerations that should be taken into account when applying the measure.

First, that although there is some evidence for each of four FMSS coding systems, some scales within these systems may not be useful in capturing meaningful differences in parent–child interactions (for instance, EOI from the original FMSS-EE system, and the novel composite code of “emotional sensitivity” created in the Scheeringa et al. 2015 study). Second, that the evidence for different coding systems differs by the age of the sample (for instance, the FMSS-EE is the only coding system for which evidence exists for children in mid-late childhood, but there is no evidence to support its use for preschool aged children). Third, the evidence regarding the use of the FMSS-EE with fathers is inconclusive and requires further attention from future research. Finally, the current review also highlights the lack of evidence regarding the use of the FMSS to index parent–child dynamics across a range of cultures, a concern also identified in the recent review by Sher-Censor (2015). Cultural norms regarding parenting practices and beliefs, and expressions about relationships, may affect the meaning of the constructs assessed by the FMSS (Sher-Censor 2015). Western, white families dominate the samples of studies included in the current review. Three studies used samples in which African American caregivers formed the majority of respondents (Scheeringa et al. 2015; Narayan et al. 2012, 2015), with two of these finding associations between the FMSS and parenting behavior. Future research should investigate whether and how the FMSS reflects parent–child interaction in other cultural contexts.

Although the concerns in the conclusions outlined above largely reflect gaps in the available evidence, this review is not without its own limitations. Firstly, although great effort was taken to conduct a comprehensive search that would identify all papers meeting inclusion criteria for the review, it is possible that papers exist which have not been located. Secondly, whilst the evidence reviewed is informative in suggesting that the FMSS does indeed reflect observed interactions between parents and children, the current review is somewhat limited in that it does not provide conclusive statements regarding the strength of associations between the FMSS and observed parent–child interaction. Given the broad exploratory focus of this review, the diverse range of positive and negative interaction variables encompassed in the reviewed papers, and the broad range of statistical methods used in reviewed papers, a systematic review approach was considered the most appropriate method to address the study aims. However, as more studies make use of the range of FMSS coding systems to index parent–child dynamics, a meta-analytic review (with coding system as a moderator) may shed further light on the strength of the associations between the FMSS and observed parent–child interactions.

The evidence reviewed here suggests important implications for parenting research. Despite growing findings that FMSS ratings are associated with the development of a range of child and adolescent psychological (and health-related) conditions, the mechanisms underlying these associations are currently unclear. Ongoing research is needed to examine the relationship between the FMSS and other factors that contribute to parenting processes. For instance, recent research has indicated that emotional flooding and cognitive appraisals of child affect contribute to hostile discipline in families of toddlers with disruptive behavior problems (Mence et al. 2014). It seems then, that parents’ thoughts and feelings about their children are likely to be shaped by their own interpretations of a child’s affect, and their own emotional responses to this perceived affect. The FMSS may inform research into the cognitive-affective underpinnings of parenting. Furthermore, recent research on the development of parenting practices and family interactions suggests a complex interaction between patterns of behavior and neurobiology (Feldman 2012). In this context, future research would benefit from investigating the neurobiological factors associated with parenting processes indexed by the FMSS.

Longitudinal research using the FMSS has been particularly limited and does not currently allow for firm conclusions to be drawn about stability or change in associations between FMSS ratings and observational indices over time. Interestingly, in the longitudinal study recently conducted by Smith et al. (2015), negative relational schemas were not found to be concurrently associated with coercive parent–child dynamics, but were associated with less reduction in coercive dynamics over time. This finding, in addition to a similar finding in a related study (Smith et al. 2013) provides preliminary evidence that the FMSS may be useful in predicting and characterizing the trajectory of parent–child relationships. However, the one study to examine retrospective associations between observed parenting behavior and the FMSS measured 2 years later, did not find any association (Scheeringa et al. 2015). Taken together, these findings may suggest that the FMSS is more predictive in terms of prospective parent–child interaction, and less associated with previous parent–child interaction. However, only one study has examined these retrospective associations, and it was limited by a small sample size. Future research is therefore needed to fully explore associations between FMSS ratings and observational indices of parent–child interaction over time, and to test for developmental effects regarding the precise ways in which FMSS ratings map onto parent–child interactions at different ages.

A final consideration concerns the utility of the FMSS as a tool for clinical assessment. Research has demonstrated that FMSS ratings differ between parents of children with and without clinical levels of psychopathology (Asarnow et al. 1994). Furthermore, intervention studies of children with internalizing and externalizing problems suggest that parents’ thoughts and feelings about their children, as captured by the FMSS, differ significantly from pre-treatment to post-treatment (Gar and Hudson 2009; Smith et al. 2013). The evidence that is now available from observational research further supports the potential clinical utility of the FMSS. It would be of much value to investigate the application of the FMSS for the purpose of identifying and evaluating treatment targets in family-based interventions.

There is now much evidence that the FMSS captures meaningful variance in parent–child dynamics, as exhibited during observational assessments of parent–child interactions. This has been demonstrated across a broad range of developmental periods and populations, however further research is needed to explore the utility of the measure in certain populations for which the evidence is currently limited (e.g., families of infants with special needs, traumatized children; Kubicek et al. 2013; Scheeringa et al. 2015). Four systems for administering and coding the FMSS (FMSS-EE; Magaña et al. 1986; Magaña-Amato 1993; PFMSS; Daley 2001; FAARS; Bullock and Dishion 2004; system used by Caspi et al. 2004) have been associated with observational indices. However, some FMSS scales have been more robustly associated with observational indices than others. In particular, EOI, as indexed by the original FMSS-EE system (Magaña et al. 1986; Magaña-Amato 1993), is not supported by any evidence from observed interactions, and may therefore not be a helpful scale with which to measure parent–child dynamics. Research into clinical applications of the FMSS is supported by the current evidence and represents a key priority for future research.