Measuring Social Communication in the Community: Novel Tools for Advancing Family Participation

Learning opportunities for young children occur in the family’s natural contexts and encompass myriad routines, locations, contextual features, and occupations. When children engage in community activities, they have the opportunity to practice skills that have cascading effects on development (Dunst et al., 2001), ultimately contributing to positive child outcomes over time (Law, 2002; Segal, 1999). Research has suggested that interventions are most effective when they are embedded in a family’s natural environment and routines (Bernheimer & Weisner, 2007). Although the importance of participation in natural contexts and child communication opportunities has been well documented (Anaby et al., 2013; Piškur et al., 2012), little empirical evidence exists on how these opportunities may differ across families’ complex daily routines and occupations. Therefore, the purpose of this study was to examine the use of a novel combination of objective instruments to measure child–parent social communication (using Language ENvironment Analysis [LENA^®]; LENA, Boulder, CO) and locations a family visits (using the Qstarz^® BT-Q1000XT; Qstarz, Taipei, Taiwan) to inform future intervention efforts.

Community activities in which children with and without disabilities participate can be categorized as instrumental (e.g., going to the grocery store), unstructured (e.g., going to the park), or structured (e.g., attending Boy or Girl Scouts; Little et al., 2014). Within these activities, the language adults direct at young children likely enhances the benefits of the children’s participation in the community context. For example, increased adult words at the park (e.g., “Time to go down the slide”) may help to enrich the child’s experience of the community. In the community context, similar to other contexts such as the home (e.g., Hart & Risley, 1992) and school (e.g., Girolametto et al., 2000), adult talk during routines is arguably a primary mechanism linked to developmental outcomes. As such, researchers have reported that adult language influences not only social communication (Warren & Yoder, 2004) but also other essential outcomes for children, such as social competence (Phillips et al., 1987), language ability (Girolametto et al., 2000), and cognition (Hart & Risley, 1992). In addition, it appears that many studies have examined adult and child language in home (e.g., Greenwood et al., 2011; Hart & Risley, 1995) and school (e.g., Dykstra et al., 2013; Irvin et al., 2013) contexts, but studies examining adult and child talk in community settings among children with and without disabilities are limited.

The LENA is a wearable sensor worn by the child that captures child and parent language, or speech–talk, in the natural environment. It can be particularly useful for researchers aiming to measure talk in community settings. The tool has allowed researchers to better understand the language environments that children with disabilities experience in the home and school (e.g., Irvin et al., 2013; Thiemann-Bourque et al., 2014; Warren et al., 2010), and it is beginning to be used in the community.

The results of one available study in the community using the LENA indicated that adult and child vocalizations vary by context (e.g., more words from children during public outings, such as to the supermarket, than during travel time; see Soderstrom & Wittebolle, 2013). Although the Soderstrom and Wittebolle (2013) study is informative, it is limited in that it relied on coders to listen to and infer context and specific community settings. Moreover, because this approach both is time consuming and requires reliable coders, it is unlikely to benefit occupational therapy practitioners working directly with families. Thus, the language environments children experience in different community settings remain largely unknown.

Measurement of family participation in the community is also limited. Parent report measures remain the popular approach to assessing children’s engagement within community settings (e.g., Young Children’s Participation and Environment Measure [YC–PEM]; Khetani et al., 2013). This approach, however, is limited in that it does not fully capture parent–child talk and the exchanges (i.e., conversational turns) that are the foundation of family routines. Similarly, researchers have relied on parent time diaries to identify the locations that families visit (Kuo et al., 2013; Mazurek & Wenstrup, 2013; Orsmond & Kuo, 2011). However, research has suggested that this method has a number of limitations, such as participant burden and respondent error (Anderson et al., 1985; Chen et al., 2015). For example, one study found that parent time diaries misclassified location in 48% of time–location responses versus a global positioning system (GPS) worn by typically developing children ages 3–5 yr (Elgethun et al., 2007). Combining LENA and GPS could address the known limitations of parent report and advance the measurement of adult and child language in community settings. If occupational therapy practitioners are able to understand the places where adult–child language is limited, interventions can better address such gaps. For example, if a parent is not talking with his or her child at the grocery store, an occupational therapist may emphasize the language opportunities at that community location.

If occupational therapy practitioners better understand the language environments across differing community contexts and activities for typically developing children, they can better develop and refine contextually relevant family-centered interventions that meaningfully support the participation of children with disabilities. This pilot investigation used a case-study design to examine the language environments in community settings experienced by one child during the family’s naturally occurring routines using precise, objective measures: the LENA system and Qstarz BT-Q1000XT (a GPS logger). This study addressed the following research questions:

Method

Procedure

We collected language and location data on 1 typically developing boy (age 5 yr, 2 mo) over an approximately 5-hr period (10:00 a.m.–3:15 p.m.) using LENA and the Qstarz GPS logger (Figure 1). The primary caregiver (a White man whose highest level of education was a high school diploma) was instructed in how to turn the devices on and off as well as where to place them on the child’s clothing. The LENA was placed in a specially designed t-shirt sold by the LENA Foundation, and the Qstarz logger was attached to a belt loop on the child’s right side. Using MATLAB (MathWorks, Natick, MA), a program was developed to place the LENA speech estimates (i.e., adult words, child vocalizations, and adult–child turns) processed via the tool’s Advanced Data Extractor at a rate of 1 s. The 5-s GPS estimates from the Qstarz logger were also transformed to 1 s and, in turn, tied to actual locations (e.g., park, grocery store) via Quantum Geographic Information System (QGIS; https://www.qgis.org/en/site/), a free, open-source desktop geographic information software (GIS). With the speech and location estimates on the same metric, we were able merge the two types of data. The caregiver was asked to keep a parent diary of time spent in individual locations to verify the QGIS analysis.

OPEN IN VIEWER

Figure 1.

(A) LENA digital language processor and (B) Qstarz GPS data logger.

Measures

LENA System.

LENA consists of a digital language processor and speech recognition software. The device records up to 16 hr of a natural audio environment. The accompanying software then processes the recording to provide adult word counts, child vocalizations (e.g., words, babbles, and prespeech communication), and adult–child turns (i.e., back-and-forth exchanges between the child wearing the LENA and adults; see Gilkerson & Richards, 2008, for additional information on LENA metrics). The system differentiates between meaningful and distant speech and silence; noise (e.g., overlapping speech, crying) and other environmental sounds (e.g., television) are categorized as nonmeaningful speech (Yoshinaga-Itano & Gilkerson, 2010). Considerable testing has been conducted to determine LENA’s reliability (see Warren et al., 2010; Xu et al., 2009). A recent synthesis showed that more than 50 studies from 2009 to 2017 relied on the LENA to capture child talk, adult talk, or both; the device has been used extensively with children at risk for and with disabilities, and its use in contexts outside the home (e.g., school and community) has expanded since the tool’s initial development (Greenwood et al., 2018).

Qstarz BT-Q1000XT.

The Qstarz BT-Q100XT is a GPS data logger that uses mapping technologies to provide real-time location estimates every 5 s (updates to the device now allow for location estimates to be captured on a 1-s interval). The tool has ample evidence of concurrent validity in use with young children in community settings (i.e., parent reports of locations visited showed 91.9%–99.9% agreement with κ statistics, ranging from 0.49 to 0.99; O’Connor et al., 2013).

Parent Reports (Demographic Form and Time Diary).

Parent reports were completed by the primary caregiver. The demographic form was completed at the onset of participation and included questions related to family and child demographic variables such as gender, family income, race/ethnicity, and child diagnostic and medical status. The parent time diary was a two-column document on which the parent indicated entry and exit times and the locations visited.

Data Analysis

The Qstarz GPS logger generated a new data point with associated latitude, longitude, speed, time, and date every 5 s. For this study, data generated from the Qstarz were downloaded from the device into Qstarz software to produce a comma separated values (CSV) file. The Qstarz CSV file was brought into QGIS. Each data point was then plotted using the latitude and longitude coordinates and validated in QGIS by comparing plotted data points with county GIS zoning data and satellite imagery. Data points in roadways were assigned to a travel group and further verified by confirming speed of travel of those data points. Using a QGIS vector analysis tool (i.e., Points in Polygon), we then associated the 6,102 location data points with the adult and child talk estimates that were processed via our MATLAB program. We relied on descriptive statistics to determine the amount of parent and child language as well as adult–child turns by location. The time diaries were used to help us ensure that there were no unforeseen issues with the child wearing the Qstarz logger (i.e., locations listed in the parent diary that did not emerge in the QGIS program).

Results

Research Question 1

Results for the child wearing the LENA and GPS showed four clusters of community activity and language: the family’s home, a fast-food restaurant, a big-box store, and a gas station. Data were also calculated for travel between locations. The 5-hr and 15-min speech–location recordings indicated that the three settings visited outside of the home resulted in 2,052 vocalizations from the child (mean [M] = 25 per minute), 2,684 words from the adult (M = 38 per minute), and 665 conversational turns (M = 10 per minute). The average adult word, child vocalization, and adult–child turn rate was calculated by dividing the total number of these speech counts by the time spent in specific locations; see Table 1 and Figure 2 for complete results. The caregiver reported no challenges in using the devices during data collection, and the locations in the time diaries aligned with those from the QGIS program.

OPEN IN VIEWER

Table 1.

Adult and Child Talk by Location

Location	No. Adult Words	No. Child Vocalizations	No. Adult–Child Turns	Min in Location	M Adult Words/min	M Child Vocalizations/min	M Turns/min
Residence	2,095	925	455	121	17.31	7.64	3.76
Restaurant	265	750	86	155	1.71	4.84	0.55
Big-box store	229	315	80	43	5.33	7.33	1.86
Gas station	42	6	6	4	10.50	1.50	1.50
Travel	53	56	38	17	3.12	3.29	2.24
Total	2,684	2,052	665	340	37.97	24.60	9.91

Note. M = mean.

OPEN IN VIEWER

Figure 2.

Parent and child talk and conversational turns per minute in home and community settings.

Research Question 2

Our results indicated that the child and parent vocalized most in the home, and this was also where the highest number of adult–child turns took place. The gas station was the location with the least amount of adult and child speech. The restaurant and big-box store were settings in which the rate of child speech was higher than adult words per minute. The rate of adult–child turns per minute at the gas station resembled that at the big-box store, even though time spent at the gas station was brief. Although a larger amount of time was spent in the home relative to the big-box store, the rates of child vocalizations were similar. The rate of adult–child turns were high during travel and were only exceeded by that in the home.