Translating Enhanced Conversational Recast to a Telepractice Setting

Abstract

Purpose:

This investigation adapted a well-studied language treatment method, Enhanced Conversational Recast, paired with auditory bombardment to a teletherapy format.

Method:

The study used a single case series approach (n = 7) to determine the feasibility of teletherapy with children ages 5 and 6 years of age. Treatment targeted grammatical errors in the context of dialogic reading and craft activities. Clinicians administered 24 doses in the form of focused conversational recasting, followed by 12 doses consisting of simple sentences containing the grammatical forms targeted for remediation. Children were treated for up to 26 sessions, with four children treated on consecutive weekdays and three treated twice a week. Treatment progress was operationalized as generalization of target grammatical forms to untreated linguistic contexts, as well as spontaneous use of the treated form. To control for nontreatment effects, generalization of an untreated form was also tracked throughout the treatment period.

Results:

Six of the seven children showed clinically meaningful gains in the use of the grammatical forms targeted for treatment within the treatment period. This was true for children enrolled in both treatment schedules. Learning for treated forms was retained after treatment was discontinued. In comparison, no change was seen for untreated forms for six of the seven children.

Conclusions:

The results suggest that this treatment method is feasible in a telepractice format, even with young children. The range of individual results is generally comparable to previous face-to-face versions of this treatment.

Developmental language disorder (DLD) is characterized by the presence of language deficits that emerge during childhood, in the absence of other handicapping conditions including intellectual disabilities, frank neurological impairments, and hearing impairments. DLD has also been referred to as specific language impairment or primary language disorder and is generally treated by a speech-language pathologist. During the COVID-19 pandemic, clinicians were often not able to meet with patients in-person, leading to a greater use of telepractice (American Speech-Language-Hearing Association, 2020a; Sylvan et al., 2020; Tohidast et al., 2020). Telepractice refers to the use of telecommunication devices to administer speech-language services (American Speech-Language-Hearing Association, 2020b).

The use of telepractice for adults and children for speech and language service delivery has generally been supported, with most studies showing positive results (Mashima & Doarn, 2008; Reynolds et al., 2009). In the past few decades, speech-language assessments and interventions for children delivered via telepractice have been examined in general terms (Fairweather et al., 2016; Gabel et al., 2013; Grogan-Johnson et al., 2010; Langbecker et al., 2019; Wales et al., 2017), with studies indicating that telepractice may be as effective with children as in-person services (Wales et al., 2017).

Past research concentrated on the benefits of increased access to speech-language services and was typically conducted in underserved rural school settings or caseload management (Boisvert & Hall, 2019). The teletherapy studies involved children with a wide range of speech and language goals, and often the language interventions used were not specified (e.g., Fairweather et al., 2016; Gabel et al., 2013; Grogan-Johnson et al., 2010), combining various types of interventions, the duration of intervention, populations, settings, and measures of treatment outcomes even within studies (Wales et al., 2017). In addition, the children treated within these studies typically represented a very wide range of ages (e.g., Fairweather et al., 2016; Gabel et al., 2013; Grogan-Johnson et al., 2010). It may well be that older, school-age children derived more benefit than preschool children simply due to their greater maturity and ability to comply with the demands of teletherapy. Younger children, when being treated in a telepractice modality, may have greater difficulty in maintaining attention to the treatment, due to limited attentional capacity and environmental distractions.

Given that the existing studies of teletherapy for children typically aggregate the results of multiple therapy methods and multiple treatment goals, the question of whether a specific type of intervention might be effective in a teletherapy format remains unclear. Therefore, it is important to test treatments already known to produce positive effects in face-to-face contexts in a telepractice modality. In this article, we seek to determine whether a well-established treatment method previously used in face-to-face settings is a good candidate for implementation in telepractice and how successful it may be when delivered in this modality. We specifically examined the feasibility of Enhanced Conversational Recast therapy paired with auditory bombardment. A meta-analysis of conversational recast studies indicates these treatment procedures are useful for children who make grammatical errors (Cleave et al., 2015). Enhanced Conversational Recast therapy is modeled after traditional conversational recast therapy and also has been shown repeatedly to be effective in treating morphological targets in preschool children (e.g., Eidsvåg et al., 2019; Encinas & Plante, 2016; Meyers-Denman & Plante, 2016; Plante et al., 2018, 2019).

Enhanced Conversational Recast treatment uses the same basic format as other recast treatments. One grammatical error is targeted for treatment at a time, as treatment that focuses on a single grammatical target tends to be more effective than treatment that targets any ungrammatical element in the child's speech (Cleave et al., 2015). The clinician waits for or encourages the child to attempt to use the grammatical form targeted for intervention. Immediately following the child's attempt, the clinician repeats or “recasts” the child's attempt, correcting any ungrammatical elements in the child's utterance, including the one targeted for treatment. Because the targeted form always occurs in clinician recast, children have the opportunity to notice the linguistic pattern represented by the targeted form and have opportunities to correct their own use of that form.

In addition to these traditional components, Enhanced Conversational Recast therapy includes high-variability linguistic input during recasts. A previous study (Plante et al., 2014) found that providing children with 24 recasts that targeted a single morpheme error but varied the verbs associated with that morpheme (e.g., jumped, looked, opened, connected) produced better learning than a treatment in which the same number of recasts was provided, but each verb was repeated twice in the recasts (e.g., jumped, jumped, looked, looked). In addition, clinicians are strongly encouraged to vary words in the noun phrase portion of the recast (e.g., “The horse jumped.” “Whinny chomped carrots.” “He kicked the dirt.” vs. “He jumped.” “He chomped carrots.” “He kicked the dirt.”). This noun phrase variability prevents elements not targeted for remediation (“he” in the example) from becoming more frequent in the clinician's input than the grammatical form to be trained (-ed in the example). This is consistent with experimental studies that show variability around the linguistic structure to be learned highlights the grammatical pattern for the learner (see Plante & Gómez, 2018, for a discussion of this principle).

In addition to high linguistic variability, Enhanced Conversational Recast procedures include attentional cuing just prior to providing each recast. The clinician gives an attentional cue to ensure that the child is actually attending to the treatment dose (i.e., either a recast or an instance of auditory bombardment in this study). Maintaining a child's attention in telepractice is of particular concern, given that the child and clinicians are in separate locations. Additionally, children with DLD often struggle with sustained attention in general (e.g., Ebert & Kohnert, 2011; Kapa & Plante, 2015; Smolak et al., 2020; Spaulding et al., 2008). Therefore, given that obtaining a child's attention during in-person therapy can already be a challenge, it could be a major concern when administering Enhanced Conversational Recast therapy in a telepractice setting.

In addition to conversational recast therapy, auditory bombardment has also been shown to support language development in children with DLD (Courtright & Courtright, 1976; Kouri, 2005; Weismer & Murray-Branch, 1989). Auditory bombardment provides an opportunity for the child to recognize and process the appropriate use of the grammatical form targeted for treatment. Through this method, the clinician models the grammatical form targeted for remediation repeatedly in a series of short, linguistically unique sentences. When paired with recasting, this technique has been found to produce clinically meaningful change for more children when following Enhanced Conversational Recast treatment than when preceding this treatment (Plante et al., 2018). Following recasts with a period of bombardment may serve to promote consolidation of learning that accrues during the recast period. Given that auditory bombardment only requires clinician input, it should transfer readily to a telepractice format. In this study, we administered Enhanced Conversational Recast therapy through dialogic reading to children between the ages of 5 and 6 years of age. The Enhanced Conversational Recast therapy was followed immediately by auditory bombardment during a craft activity.

The purpose of this study is to address the feasibility of this treatment delivered through telepractice. We administered this treatment to seven children and gathered feedback from parents as consumers of this treatment. Using a close replication of previous studies of Enhanced Conversational Recast therapy, we administered the treatment daily to four children over a period of approximately 5 weeks. In addition, we treated another three children for a similar number of sessions, but scheduled twice a week for 15 weeks. These two schedules were selected because school-based clinicians often treat children biweekly but also have the option of scheduling their students in blocked time periods. We wished to determine whether both scheduling options were feasible in this initial study. However, the inclusion of two treatment schedules was not intended to determine which produced better outcomes, as this is beyond the scope of a feasibility study. We hypothesize that a teletherapy version of treatment will be effective with children when administered with either treatment schedule.

Method

Participants

Seven children (five boys, two girls) with DLD participated in this study and completed all of the treatments. All children had been originally recruited through preschool and day care programs. An initial four participants (B1–B4) were enrolled in the blocked condition, in which treatment was delivered daily. Only after that treatment was successfully concluded were the subsequent three children (S1–S3) enrolled in the spaced condition, in which sessions occurred twice per week. One additional participant, assigned to the spaced condition, was withdrawn from the study during treatment due to excessive absences. Children were enrolled according to the phase of the research (first blocked, then spaced), rather than randomly to the different conditions. The age range of children who completed treatment was 5;5 (years;months) to 6;5 (M = 5;10, SD = 0;4). The demographics of the participants are reported in Table 1.

Table 1.

Demographic information for children in the blocked (B1–B4) and spaced (S1–S3) conditions.

Participants	Sex	Age (years;months)	KABC-II	SPELT-P2	SPELT-3	GFTA-3
B1	M	5;11	108	67		70
B2	M	5;4	115	59		57
B3	M	5;5	108	53		40
B4	M	5;9	85	42		60
S1	F	6;5	108	64a	67	53
S2	M	6;0	92	73b	65	49
S3	F	6;0	81	84c	72	49
M		5;10	99.6	63.1	68.0	54.0
SD		0;4	12.3	13.6	2.9	8.8

Note. Test scores are standard scores (mean of 100 and an SD of 15). M = male; F = female; KABC-II = Kaufman Assessment Battery for Children–Second Edition, nonverbal scale; SPELT-P2 = Structured Photographic Expressive Language Test–Preschool: Second Edition (NB: S1–S3 were administered the SPELT-P2 at ages ^a5;0, ^b5;7, and ^c5;9 (years;months), respectively; SPELT-3 = Structured Photographic Expressive Language Test: Third Edition; GFTA-3 = Goldman-Fristoe Test of Articulation–Third Edition.

The children were diagnosed with DLD through testing by a certified speech-language pathologist. All children passed a pure-tone hearing screening at the time of diagnosis and had to score above the range for intellectual disabilities on a nonverbal intelligence test (70 + 1 SEM, on the Nonverbal Scale of the Kaufman Assessment Battery for Children–Second Edition; Kaufman & Kaufman, 2004). Although children ranged in age from 5;5 to 6;5 at the time of treatment, they were all originally recruited before their sixth birthday. Accordingly, they were originally tested using the Structured Photographic Expressive Language Test–Preschool: Second Edition (Dawson et al., 2005) to confirm language status. For this test, a score at or below a standardized score of 87 is consistent with a language disorder with a sensitivity of 90.6% and specificity of 100% (Greenslade et al., 2009). All children showed evidence of morphosyntax errors in their conversational speech, indicative of a language disorder at the time of their initial testing. Because some children had turned six before enrollment in treatment, these children were also administered the Structured Photographic Expressive Language Test: Third Edition (Dawson et al., 2003) to confirm the continued existence of morphosyntax errors. In addition, the Goldman-Fristoe Test of Articulation–Third Edition (Goldman & Fristoe, 2015) was used to assess articulation. According to this test, all children had a speech sound impairment, but this varied in degree. All test scores are reported in Table 1.

The education level of the children's mothers ranged from the junior year of high school to a graduate degree, with a modal educational achievement of 2 years of college or technical school. In addition, three children were White, one was American Indian, and two were biracial. The race of one child was not reported. Three children were Hispanic. By parent report, all children were native English speakers and three had been exposed to Spanish. For these children, the parents reported the child spoke little to no Spanish. In addition, the children were instructed in English in the preschool programs they attended. All parents provided informed consent for their children's participation, and the study procedures were approved by the University of Arizona institutional review board.

Materials and Procedure

General Treatment Context

Services were delivered using personal computers and Zoom for Health software. Children were located in their homes in a space that was generally quiet. Caregivers (four mothers, one father, one grandmother, and one older sibling) attended each session with the child. In most cases, these individuals were in close proximity to the child and helped with activities, but in some cases the caregiver was in the general area rather than actively engaged with the child. Caregivers were provided with a written schedule of the treatment days and activities, along with any expendable supplies to be used during the treatment.

The session began with Enhanced Conversational Recast therapy lasting about 20–25 min that targeted a single grammatical error (the target). In the second portion of the treatment, auditory bombardment of the target was conducted for 10–15 min. The exact time was determined by the time needed to deliver all doses in a particular session. For those in the blocked condition, children received five sessions a week over the course of 6 weeks. For the spaced condition, children were treated for two sessions a week spaced over 15 weeks. The number of sessions each individual child attended is reported in Table 2. Due to occasional absences, these numbers range from 24 to 26. Four clinicians in total administered the treatment; two for the blocked condition and two for the spaced condition. These clinicians were graduate students in the speech-language pathology program at the University of Arizona.

Table 2.

Treatment parameters for children in blocked (B1–4) and spaced (S1–3) conditions.

Participant	Target form	Control form	Treatment days	Unique nouns or verbs recast a	Unique nouns or verbs bombarded b
B1	Past tense -ed	She	26	297	92
B2	He	Auxiliary “is”	26	304	106
B3	Auxiliary “is”	She	26	285	102
B4	He	Auxiliary “is”	26	295	105
S1	She	“What” questions	17	202	73
	Negative “doesn't”	“What” questions	9	130	56
S2	Past tense -ed	Auxiliary “is”	24	242	68
S3	3rd person -s	Past tense -ed	24	258	92
		M	22.3	251.6	86.8
		SD	6.2	60.00	18.8

^a Unique nouns or verbs recast refer to the noun or verb used by the child that was recast by the clinician with the target form.

^b Unique nouns or verbs bombarded refers to the noun or verb the clinician used with the target form during the bombardment phase of treatment.

Each child received 36 treatment doses per session. A dose is defined as actions by the clinicians intended to remediate the child's use of the grammatical form targeted for treatment. Twenty-four of these treatment doses were administered through recasting the child's attempts (successful or unsuccessful) to use utterances containing the grammatical form targeted for treatment. The remaining 12 doses were administered through auditory bombardment. The cumulative intervention intensity (Dose Number × Dose Frequency/Day × Total Duration in Days) had a range of 576–624 (M = 610, SD = 23.4).

Data were collected by clinicians on a paper record form. Clinicians recorded the verb they used in the recast or bombardment, in order to assure that targets were heard in a different linguistic context with each dose. The clinicians also recorded whether the child's attempt at using the target was correct or incorrect. In addition, clinicians recorded whether the target was spontaneously produced by the child at any time during the session. Spontaneous use, by operational definition, was a correct production of the grammatical form with a verb that was not modeled by the clinician immediately beforehand.

Pretreatment Baseline

Immediately preceding the first day of treatment, probes of potential treatment targets were administered to the children over the course of 3 days. These probes were used for two purposes: (a) to identify forms that had low pretreatment use, making them suitable candidates for treatment; and (b) to create baselines of pretreatment use against which treatment performance could be compared. These pretreatment probes allowed for five to six grammatical forms to be probed during 45-min sessions. Each was probed 10 times per session by the clinician, typically by eliciting its use in a context that obligated the child to use that grammatical form in a way that was grammatically correct and pragmatically appropriate. Spontaneous use of the probed forms was also documented and included in the 10 child attempts required for each morphological structure. The clinician recorded if the attempt was correct, if the grammatical form was omitted from the response, if a different form was substituted (e.g., “not” for “doesn't”), or if no response was given. From the grammatical forms probed, two were chosen that were used correctly in less than 30% of obligatory contexts. One of these was selected for treatment (the target form) and the other was probed over the course of treatment, but not treated (the control form). Given the speech sound status of the children, we verified that the phonemes contained in the target and control forms were in the child's sound inventory. If multiple grammatical forms met these criteria, then specific forms were given preference if they had lower accuracy or contributed to balancing those forms selected as target or control forms across participants. The target and control forms selected for each child are displayed in Table 2.

Enhanced Conversational Recast

Doses consisting of clinician recasts were provided in the context of dialogic book reading by sharing the book pages through the Zoom platform. Dialogic book reading is the practice of interspersing text reading with discussions between the reader and child about the text and pictures. Dialogic book reading has been a component of all face-to-face versions of Enhanced Conversational Recast treatment (Eidsvåg et al., 2019; Encinas & Plante, 2016; Meyers-Denman & Plante, 2016; Plante et al., 2014, 2018, 2019). Given that books can be readily presented via computer, we anticipated that this would translate well to teletherapy as a treatment context.

The book reading activity was designed to elicit child use of the target form with 24 unique verbs for pronoun or verb inflection targets (see Table 2) and 24 different nouns for the negative “doesn't” form. This assured high linguistic variability in the subsequent recasts. Each teletherapy session used a different book. This allowed for a variety of linguistic contexts to occur naturally not only within a session, but across sessions as well. Clinicians were encouraged to use a variety of verb contexts to elicit use of the target form across days, and the number of different nouns or verbs used by each clinician during recasts is listed in Table 2.

A recast is defined as a clinician utterance that immediately follows a child attempt to use the target grammatical form. The clinician's recast follows the general form of the child's utterance, correcting any grammatical errors that occur (e.g., Child A: “Boy not jump,” Clinician A: “The boy doesn't jump.”; Child B: “She open the door.” Clinician B: “She opened the door.”). Recasts were administered by the clinicians following both correct and incorrect utterances elicited from the child.

To elicit child attempts to be recast, the clinicians read the book text and then used the illustrations as a visual reference, in screen share mode, to prompt use of the target form. To prompt the child's use of the target form, the clinician used the verb they were attempting to elicit up to three times in either uninflected form or a form other than that targeted by the treatment (e.g., Clinician: “Peter likes to chew carrots. See him chew here? He's chewing. What did Peter do to this carrot?” Child: “He chewed it.”). This avoided modeling the target form prior to eliciting its use. Clinicians prepared a list of words to recast within the context of the book but would also recast a child's response if the child attempted use of their target without a clinician elicitation, as long as the child's utterance had not been recast previously in the same session. The majority of recasts, however, followed clinician-elicited utterances by the child. Past research has indicated that there is no substantial difference in effectiveness between recasting a spontaneous response and a child attempt elicited by the clinician (Hassink & Leonard, 2010). Given that the clinician elicited most utterances in ways that obligated use of these verbs and the content of child utterances was further constrained by the book content, clinicians did not have difficulty recasting child utterances despite speech sound errors.

In addition to the incorporation of high linguistic variability in the recasts, Enhanced Conversational Recast Therapy has two additional components that are not always standard in other versions of recast treatment. The first component is that one grammatical form is targeted for treatment at a time, opposed to targeting any error that the child makes in the moment (i.e., broad recasting). The second component is ensuring the use of attentional cues prior to the recast. In the teletherapy format, the clinicians cued the child to look at them before the recast was delivered (e.g., “Look!” “Eyes on me!”), and children were verbally reinforced for this behavior after the recast was delivered. We note that during book reading, both the book and the clinician's face were present on the screen, facilitating visual attention to the screen. The clinician would exit screen share mode to speaker mode just prior to the recast so that their face occupied the full screen at the time the recast dose was delivered. The clinicians also pointed to their own mouth when they produced the target form during the recast in order to focus the child's attention on the relevant portion of the recast. The clinicians' gestural cue toward their mouth during production of the target form was intended to focus the child's attention on that portion of the recast that contained the target form.

Auditory Bombardment

Auditory bombardment was accomplished by having the clinician produce 12 short but grammatical utterances that contained the target (e.g., “I glued.” “You folded it.”). These doses were administered while completing a simple craft activity with the child. The craft activity was related conceptually to the book content used during that session. For example, if the clinician used the book Just Me and My Dad (Mayer, 2003), which features the two title characters fishing and a bear stealing the fish, the children made and decorated a construction paper fish during the craft portion of the session. The clinician provided an example of the finished product and then modeled the craft step by step through Zoom. The child performed each step either by themselves or with the assistance of a nearby adult or the older sibling. All necessary materials for the crafts were supplied to the families beforehand. For this reason, crafts and books were preplanned, and all children, regardless of their treatment schedule, were read to from the same books and made the same crafts over the course of treatment.

Clinician utterances used for auditory bombardment each contained a subject and verb and would also potentially contain articles, a negative, or an adverb additionally (e.g., Clinician A: “She doesn't cut.” Clinician B: “We colored”). Different activities were used throughout treatment so that every session included a variety of nouns and verbs. The specific verbs were tracked to ensure that a variety of linguistic contexts were used. The number of unique verbs that each child heard in association with their target form is reported in Table 2. Each utterance was presented by the clinician after ensuring they had the attention of the child, as indicated through eye contact. Attentional cues were used as needed, as described above.

Probe Sessions

Probe sessions were used to measure the generalization of the child's use of their target form to untrained verbs. A list of 20 verbs was reserved for use solely during probe sessions so that generalization to untrained linguistic contexts could be measured. Noun use was conditioned on the materials included in each probe kit, and these were also limited to probe sessions. Seven probe kits consisted of cut-out pictures and felt board materials based around a specific theme (e.g., farm life, race cars, and marine animals). Each probe kit was different to enhance the child's interest with the activity and allow for a variety of contexts to elicit probe verbs (e.g., the verb “pick” could be used for using a tool in hay, choosing a race car driver, or plucking a flower). The clinician recorded which kit was used on each probe session. Clinicians modeled actions using the materials to elicit verbs reserved for the probes per session.

Probe sessions occurred seven times throughout treatment. For the blocked condition, probes were conducted 1 time per week for the first 4 weeks and 3 times (Monday, Wednesday, and Friday) in the final week. For the spaced condition, the probes were conducted once just before every fifth session and before each of the three final sessions. Six weeks after the final treatment session, children were given a final probe session to test long-term retention. The probe sessions lasted approximately 10 min and occurred just before treatment to ensure that the retention of the target form was being measured rather than measuring a byproduct of the treatment that had just occurred.

During the probes, the clinician elicited 10 child attempts for each of the target and control forms. The clinicians could choose any 10 of the probe nouns or verbs for each probe session, and clinicians were encouraged to use all of the probe vocabulary across probe sessions. The target and control forms could be probed in the same or separate utterances (e.g., a pronoun and verb tense could be elicited in the same utterance or in separate utterances). Clinicians could elicit the target and control grammatical forms in any way they wished as long as they presented obligatory contexts. Variability in the ways in which the words were prompted decreased the likelihood that children would have automatic responses and allowed for a more accurate measure of generalization. If spontaneous use of a probe word occurred that had not been used previously in the probe session, then that utterance would count as one of the 10 probe responses.

Procedural Fidelity

Before clinicians began treatment, they were trained on procedures for treatment sessions, probe sessions, and auditory bombardment. Clinicians were required to read prior publications on recast treatment by our group and others (i.e., Hassink & Leonard, 2010; Plante et al., 2014). The procedures of Enhanced Conversational Recast and the format of Probe sessions were explained by the supervising speech-language pathologist, and how these would be done in the teletherapy context. Clinicians had an opportunity to practice the procedures on their own. All had used Zoom software for other purposes and were familiar with it, although neither the clinicians nor the supervising speech-language pathologist had used it for treatment prior to this study. During all of the sessions, they were monitored by a certified speech-language pathologist who attended the teletherapy session with the camera off and audio muted. The supervising clinician was also responsible for ensuring the correct administration of the procedures for treatment and provided feedback typically just after the treatment session. Fidelity was measured during live sessions by an observer in treatment and probe sessions. Fourteen (20%) of the probe sessions and 42 (23.6%) of the treatment sessions were coded. Coding for fidelity was spaced evenly throughout the course of treatment for each child. The fidelity codes included descriptions of the activities, the amount of unique and total platform words used during probes and treatment sessions, and whether probe words or materials were used when they weren't supposed to be.

Several dimensions of the treatment were assessed for fidelity. Coders confirmed that the clinician did not model the target before the child's attempt (99.6% [98.6%–100%]), that the recast occurred immediately following the child's attempt (99.1% [97.9%–100%], that the clinician cued the child's attention during the recasts and during the bombardments (95.6% [91.7%–100%]), that no words reserved for probes were recasted during treatment (99.9% [99.3%–100%]), that each verb recast in a session was unique (99.9% [99.5%–100%]), and that 24 recasts and 12 instances of bombardment were delivered during the session (100%).

Probe fidelity was measured by whether the clinician used the probe materials (100%), whether the target or control forms were obligated by the clinician (99.6% [98%–100%]), that these forms were not modeled by the clinician preceding the child's attempt (98.8% [96%–100%]), that no feedback was given to the child by the clinician indicating correctness (99.6% [98%–100%]), that 10 target and 10 control items were probed (100%), and that only probe words were used (100%). The calculations were based on how many times these procedures were followed during the session divided by the total number of times they were required.

Scoring Reliability

Scoring reliability for probe sessions was conducted through video recordings. Fourteen (20%) of the probe sessions were coded, and these were evenly distributed throughout treatment period for each child. Scoring was based on how often the clinician and reliability coder agreed on whether or not the child correctly produced the target and control forms during probe sessions. Coders did not consider whether other words in the child's utterance were also grammatical. For example, if the target for the child was the pronoun she and the child said, “She kick,” then that would be scored as a correct production of the target, even though the verb tense was in error. Point-to-point reliability averaged 96.4% with a range of 92.5%–100%. We also calculated the agreement between clinicians and coders for the number of spontaneous productions of the child's target form per session. The range of these values was very restricted, ranging from 0 to 3 per session, with a Spearman rho = .77, p = .00.

Parent Feedback

After treatment was concluded, parents were interviewed concerning their reactions to the treatment. This interview was conducted by the supervising speech-language pathologist, using the questions contained in Table 3. The supervising speech-language pathologist typed notes on parent responses during the call. These questions covered whether caregivers did anything to reinforce learning in treatment, as well as whether they noticed any changes in their child as a result of treatment. In addition, we gathered information concerning their own and their child's attitudes toward treatment.

Table 3.

Questions asked of parents posttreatment.

Have you changed the way you talk to or interact with your child as a result of your participation in the program?

Have you noticed any changes in your child that you think may be related to his participation in the program?

How do you think your child felt about the program?

How did you feel about the program?

Results

The data were treated as a series of single case studies. Data were analyzed visually (see Figure 1). In-treatment performance tends to track generalization probe performance for Enhanced Conversational Recast treatment (Hall & Plante, 2020), and this was also the case here. We focus on probe data as the metric of treatment progress as this reflects generalization to untrained linguistic contexts, rather than forms trained during treatment sessions. The single subject treatment effect size d was calculated to represent the degree of change from the pretreatment baseline performance. This particular effect size was calculated as in previous studies of Enhanced Conversational Recast (Eidsvåg et al., 2019; Meyers-Denman & Plante, 2016; Plante et al., 2014, 2018, 2019), allowing for direct comparisons with these studies. The treatment effect size d is calculated as the mean for the final three generalization probe data points minus the mean of the three pretreatment baseline data points, with this difference divided by the standard deviation of the final three generalization probe data points. If this standard deviation was zero (e.g., the child was at 100% for all three final sessions), then the minimum nonzero standard deviation (i.e., a difference of 10% for one of three final data points) was used to avoid dividing by zero. Because these data were taken from the generalization probes, the effect sizes reflected generalization of the grammatical form to untrained linguistic contexts. This and other treatment results are included in Table 4.

Figure 1.

Results for children treated in blocked and spaced conditions using Enhanced Conversational Recast treatment followed by auditory bombardment. Use with untreated vocabulary words was probed once per week in the first 4 weeks of treatment and 3 times per week (every other day) in the blocked condition. Probes occurred every fifth session for the first 4 weeks of treatment, and before each of the final three treatment sessions in the spaced condition. Ages are reported in years;months. PreTx = Pretreatment session.

Table 4.

Group and individual results for blocked (B1–B4) and spaced (S1–S3) conditions.

Participants	Target forms						Control forms
	Pretreatment use (%)	End-treatment use (%)	Effect size (d) a	Follow-up use (%)	Total spontaneous productions	Unique spontaneous productions	Pretreatment use (%)	End-treatment use (%)	Effect size (d) a
B1	0	100	17.3	80	30	21	10	0	−1.7
B2	6.6	86.7	5.3	100	73	61	3.3	0	−0.6
B3	0	20.0	0.8	40	14	12	0	0	0
B4	3.3	0	−0.6	0	6	4	0	0	0
M	2.5	51.7	5.7	55.0	30.8	24.5	3.3	0	−0.6
SD	3.2	49.1	8.1	44.3	29.9	25.3	4.1	0	0.8
Spaced condition
S1 b	3.3	100	17.3	100	1	1	0	0	0
	0	100	17.0	100	19	15	0	0	0
S2	6.7	90	8.3	90	15	12	3.3	0	−0.6
S3	23.3	57	2.9	70	5	5	10	40	3.0
M	10.5	82.3	9.5	86.7	9.3	8.5	4.4	13.3	0.8
SD	11.3	22.5	7.2	15.3	5.1	4.9	5.1	23.1	1.9

^a Treatment-related change (d) is calculated as (mean end-treatment use – mean pretreatment use) / the end-treatment standard deviation.

^b S1 was treated for two target forms within the treatment period (see Table 2). The two values for S1 were averaged to obtain means and standard deviations for the spaced condition.

Blocked Condition

B1

B1 is a male who was 5 years, 11 months at the time of treatment. After the pretreatment baseline, he showed relatively steady progress after the first week of treatment (5 consecutive treatment days) toward full acquisition of the target form (past tense -ed) by the end of the treatment period. His treatment effect size d (17.54) reflected the maximum possible effect, as he progressed from a pretreatment baseline of zero to a consistent end-treatment performance of 100%. This gain was largely maintained at the follow-up probe session, with a performance of 80%. He also used his target form spontaneously fairly frequently in treatment sessions, with 30 occurrences, 21 of which represented linguistically unique uses (i.e., not simply repetitions of previous uses). The number of linguistically unique uses demonstrated that he was not using the target form in a stereotypic way. In contrast, he showed no gains on his control form (pronoun “she”), either by the end of treatment (d = −1.75) or at follow-up (0% use). We note that it is not uncommon for performance on the control form to fall relative to baseline during treatment as the child is rarely hearing it in the treatment sessions.

B2

B2 is a boy who was 5;4 at the time of treatment. He began to show progress toward generalization of his target form (the pronoun “he”) after 2 weeks of treatment (10 days) and showed steady progress from that point on, with gains of about 20% accuracy per week, ending at 100% accuracy (d = 5.23). He maintained this accuracy level at follow-up. He showed the highest spontaneous use of the target form of all the children with 73 occurrences during the treatment period. Sixty-one of these were linguistically unique. In comparison, his control form (auxiliary “is”) use never exceeded baseline during treatment or at follow-up (d = −0.58).

B3

B3 is a boy who was 5;5 at the time of treatment. He showed a 2-week (10 treatment days) delay before generalizing his target form (auxiliary “is”) to an untreated linguistic context but made this change abruptly (going from 0% use to 80% use). However, use of his target form declined over the next 3 weeks, back to 0% and 10% use at the end of treatment. His clinician noted that he seemed to become confused concerning what he was supposed to be attending to in the clinician's recasts and how to change his own output during that time. However, he did use his target form spontaneously 14 times over the course of treatment, with 12 of these uses being linguistically unique. His final treatment d for the target form was 0.75 compared with a d of 0 for the control form (the pronoun “she”). However, some learning clearly accrued, as his follow-up probe performance was at 40% compared to a pretreatment baseline of 0%.

B4

B4 is a boy, age 5;9 at the time of treatment. He is the only child who received blocked treatment to be a frank treatment nonresponder. Use of neither his target form (pronoun “he”) nor his control form (auxiliary “is”) exceeded baseline at any time during the treatment period (target d = −0.58, control d = 0.00). His spontaneous use of the target form was more modest than the other children in this treatment condition, with six spontaneous uses, four of which were linguistically unique. His follow-up performance was at 0% for both target and control forms

Spaced Condition

S1

S1 is a girl, age 6;5 at the time of treatment. She showed a rapid and substantial gains in generalization of her target form (pronoun “she”) by the first probe session, 2.5 weeks (five sessions) after the onset of treatment. Her performance went from an average pretreatment use of 3.33%–100% use during this time, and this was maintained for three successive probe sessions (d = 16.96). At this point, a second target form was treated (negative “doesn't”), which showed the same jump from 0% to 100% correct use within the five treatment sessions (2.5 weeks) and was sustained at this level until the end of treatment (d = 17.54). Although performance on the initial target form (“she”) dropped to a low of 40% after treatment for the second target form began, performance rebounded and she was at 100% on both target forms by the end of treatment. This was maintained at follow-up. She used her target, “she,” fairly frequently (19 times in total, with 15 of these being linguistically unique) but had only one spontaneous production of the second-trained grammatical form “doesn't.” We note here that she had a more limited number of sessions over which to produce the treated “doesn't” form compared to children who were treated for grammatical error throughout the treatment period. Her control form (the “wh” question form, “What does…”) showed no progress throughout the treatment period (d = 0.00).

S2

S2 is a boy, age 6;0 at the time of treatment. S2's sessions tended to run the longest because of frequent interruptions needed to manage off-task behavior. Despite this, he began to generalize his target form within the first five sessions (2.5 weeks) and showed largely steady progress from that point on, with a small decrease in performance on the final session. His treatment effect was strong (d = 8.33), and his performance at follow-up was consistent with his performance in the final five sessions (2.5 weeks) of treatment. He used his target form spontaneously 15 times during treatment, 12 of which were linguistically unique productions. He showed no gains on his control form during the treatment period (d = −0.58) but did show 60% use of this form at follow-up. This was still lower than the 90% use of his target form at follow-up.

S3

S3 is a female, age 6 years, 0 months at the time of treatment. She showed the most variable treatment of all of the children, both for the target (third person singular -s) and control forms (past tense -ed). Her generalization probe performance was not consistently above baseline until well into treatment (by 20 treatment sessions, or 10 weeks). Her final treatment effect size d was 2.89, which was not substantially different than the d for the control form of 3.00. The slightly larger value for the control d versus the target d reflects the lower pretreatment performance for the former versus the latter grammatical form, even though the average use of the target form was higher at the end of treatment. She used her target form spontaneously only five times, but each was a linguistically unique utterance. At follow-up, the target form use was at 70% versus 20% for the control form.

Parent Feedback

Caregivers (parents, and in one case a grandparent) had a range of responses to the question concerning whether they changed their own behavior as a result of seeing the treatment. Three caregivers said that their behavior did not change (e.g., “Not really.” “We thought we weren't supposed to change anything because of the research…”). Two expressed changes in very general terms (e.g., “I'm trying to do better with him.” “I sometimes asked him to re-say incorrect utterances”). One talked about discussing elements of stories (characters, settings, problems, solutions) when reading books. This caregiver had explicitly asked questions about storybooks and dialogic reading during the course of treatment. Two parents cited specific things they did to improve their child's speech (e.g., encouraged use of the target). These were the parents of one child who did not show a treatment response (B4) and one who did (S1).

Caregivers cited many positive changes they associated with their child's participation. Many of these reflected general benefits around language and behavior, rather than those specific to the morphosyntactic goals of the treatment sessions. Positive comments included increased talking in general, “giving more thought to what he is saying,” willingness to say “more difficult words,” and “doing better with sounding out words.” One parent noted that their child had “a much better understanding of stories” as a result of the treatment. One specifically noted increased use of the target form. Parents cited non-linguistic behavioral gains, including “he is more patient with us when we can't understand him,” “he was enjoying his learning,” and “the preparation for each session (i.e., checking the schedule and taking out the materials for the day) modeled good planning and organizational skills for him.” Finally, three parents specifically noted that the experience with distanced online learning helped to prepare their child for school, which was expected to be done remotely due to the COVID-19 pandemic. One noted that their child “seems to have more confidence…at school.” Another noted overgeneralization of the past tense in home speech, as a concern for them.

A majority of parents expressed positive opinions about the program. These included comments concerning perceived benefit to their child, the convenience of not having to transport their child for treatment, and the organization of the program. Two parents favorably compared the one-on-one format to their child's previous online group treatment sessions. One noted higher benefit in individual treatment versus previous group treatment. Another noted that “group therapy tends to get too competitive,” and that their child often “yells at the computer to be heard” among the other children vying to speak. One parent's opinion was more neutral, saying, “It was different. Everything is different online.”

Parents also universally felt their child liked the treatment sessions overall. Four noted that their child asked about when their treatment would happen again between sessions. Others noted their child's positive engagement during sessions, or noted that their child liked the books or craft activities in particular. One parent did state that her child (B3) got restless and the child had asked “Why do I have to copy so much?” We believe this referred to having him use a particular verb produced by the clinician when attempting to elicit an attempt at the target form. Alternately, it could have referred to the clinician's recast as copying his own utterance.

Discussion

This study addressed the feasibility of implementing Enhanced Conversational Recast therapy paired with auditory bombardment with preschool-aged children diagnosed with DLD in a telepractice setting. The children participating in this study presented with a range of language severities and, like many children with impaired language, also showed evidence of a comorbid speech sound disorder (Shriberg et al., 1999). The majority of the children generalized their target form over pretreatment levels. The effect size of the target form compared to the control form was very large for two of the four participants in the blocked condition, more moderate for a third, and negligible for a fourth. Two of the three participants in the spaced condition had a strong improvement, with one child (S1) showing improvement for two targets within the 26-session treatment period. The third child showed more modest and more variable progress during treatment. Three of the four participants in the blocked condition and all three participants in the spaced condition had a higher follow-up use of their target form compared to their pretreatment baseline and use of their control form at follow-up. Treatment delivery in both the spaced and the blocked conditions was generally effective, with both treatment schedules being associated with largely positive outcomes. These data indicate that, when implemented in telepractice, this treatment method has the potential to produce change, regardless of whether the treatment is administered over consecutive or non-consecutive days.

Comparison to Face-to-Face Treatment

The children in this study, as a set, achieved the full range of treatment effects. This included children who showed the maximum treatment effect possible to one child who showed no effect at all. Most children showed gains that would be considered clinically meaningful. Moreover, these gains were sustained at levels well above baseline even after treatment stopped. This is consistent with previous face-to-face versions of Enhanced Conversational Recast (Eidsvåg et al., 2019; Meyers-Denman & Plante, 2016; Plante et al., 2014, 2018), for which a majority of children also show clinically meaningful and sustained changes.

In face-to-face treatment, initial test scores are a predictor of treatment outcomes (Justice et al., 2017; Kapa et al., 2020). In this study, most children had morphosyntactic test scores below –2 SD from the mean. This included both the top performers and the child who showed no treatment response at all. Therefore, there was likely insufficient variability in test scores among our participants for a trend similar to previous reports to emerge. Previous work has also demonstrated that use of the target from within the first 10 treatment sessions is a good prognostic indicator that the child will eventually be successful with Enhanced Conversational Recast treatment (Hall & Plante, 2020). This was also the case in this study, with all treatment responders showing initial grammatical form use within this window. In addition, children in face-to-face versions of Enhanced Conversational Recast tend to continue to use their target form after treatment ends at about the same level as they did at the end of treatment. They do not “close the gap” between their end-treatment use and full accuracy on their own, even when their end treatment use is relatively high (Hall & Plante, 2020). In this study, children also sustained their learning and, in some cases, returned with slightly better performance. Only two children had reached 100% accuracy over several treatment sessions by the end of treatment, and one of these achieved this accuracy at follow-up, with the other achieving 80% accurate use.

The present results support feasibility of the treatment delivered either daily or twice a week. Single subject designs are not intended to directly test the relative efficacy of different treatment manipulations (Fey & Finestack, 2009) but can sometimes suggest trends to be explored in future work. In this study, there was no obvious trend related to treatment delivery schedule (see Frizelle et al., 2021, for a review of this issue). This outcome contrasts with previous face-to-face treatments. Smith-Lock et al. (2013) found superior results for a morphosyntax treatment that was delivered weekly for 8 weeks compared to daily for 8 days. However, it is important to note that the number of treatment doses was not controlled across treatment schedules in that study. Justice and colleagues have suggested that the amount of treatment per session interacts with the frequency of treatment sessions (Justice et al., 2017; Schmitt et al., 2017). In these studies, the dose was defined as time spent on any treatment goal, rather than a specific number or particular types of therapeutic actions. Stronger outcomes were associated either with less frequent sessions paired with more time on treatment goals within the session or more frequent sessions and less time on treatment goals (Schmitt et al., 2017). In this study, dose was defined as the number of therapeutic actions, which was held constant across children, as was session length and number. The time needed to deliver these doses would have been considered on the high end compared to the range of time clinicians spent on treatment goals in Justice et al. (2017) and Schmitt et al. (2017). However, in this study, more frequent sessions (i.e., blocked sessions) resulted in faster learning, simply due to the greater number of sessions per week, compared with less frequent sessions that occurred in the spaced condition. Indeed, progress appeared to be more strongly linked to the number of sessions accrued, rather than to how these sessions were spaced.

Teletherapy Format

The Enhanced Conversational Recast treatment adapted well to a teletherapy format overall. Others have noted that this therapy format can come with technological challenges, such as unreliable Internet access, poor sound quality, and other technical difficulties, but solutions to many of these are often available (McGill & Fiddler, 2021). Although we also occasionally experienced technical difficulties, they did not negatively impact children's progress overall. In addition, the physical separation of the clinician and child was not a large impediment, as children did learn the recast treatment format. Dialogic book reading, also used in face-to-face treatment formats, translated readily to an online format. Use of the screen share and speaker view settings during reading allowed clinicians to switch reasonably easily between views of book pages and a full-screen image of their face during the recast. However, not all children's books available in an electronic format were well suited for recast treatment. The best books have a limited amount of text per page, and large-scale, engaging artwork that depicts varied actions. Limited text allowed for time for the clinician and child to read the story and discuss the pictures as a way to set up clinician elicitations of the target forms. Artwork that depicted varied actions allowed clinicians to elicit the target form in highly varied verb contexts, a feature of Enhanced Conversational Recast treatment known to enhance outcomes. At least one parent felt their child benefited from the exposure to books specifically. Craft activities, which were paired with clinician models containing the target form, were also noted by caregivers as a positive of the treatment. These crafts were simple and quick to complete and used low-cost and readily-available materials. However, crafts did require advanced planning to provide materials to the families. However, caregivers noted that availability of a schedule and materials was a positive aspect of the treatment.

One notable limitation to online delivery of doses involved attentional cuing. Children do not appear to derive benefit from recasts that they do not attend to (Eidsvåg et al., 2019). In face-to-face treatment, clinicians have numerous methods for cuing children to attend during each recast, including physical proximity, waiting for the child to orient to the clinician before delivering the recast, touch, moving their face into the child's line of sight, and auditory cuing. In the teletherapy format, cues were limited to the auditory modality (e.g., “Look here!” “Eyes!”). In particular, simply waiting for the child to orient to the clinician before recasting, which often works in face-to-face formats, was not effective online. Even a short time delay before recasting often resulted in the child going off-task all together. Although children did learn to look during recasts over the course of the treatment, clinicians had fewer options for shaping this behavior.

Another potential limitation of the teletherapy format for children of this age was the need for caregiver supervision. The amount of supervision needed varied greatly among children. For S1, a 6-year-old, and the strongest responder, supervision consisted of a parent being in the general area in case help was needed. The parent of the single non-responding child (B4), a 5-year-old, sat next to the child daily. For other children, direct caregiver supervision still did not remove the need for the clinician to actively manage the child's behavior during treatment. This need for local supervision adds to the cost of teletherapy, in that it requires the time of an adult in addition to the time of the clinician. However, this cost may be offset in some situations. For children typically treated in outpatient hospital settings, private practices, or university clinics, the cost to the parents in time might be offset by the convenience of home treatment.

It is worth noting that caregivers' observations of the child's treatment did not necessarily translate into parent carry-over of treatment methods at home. When asked, a majority of parents did not report incorporating techniques used in treatment into their interactions with their child. Indeed, a majority did not note changes related to the child's target form as a benefit of treatment, even though most children did show changes. This suggests that caregiver-administered home treatment as an adjunct to clinician treatment would require more explicit training than simply observing treatment sessions. On the other hand, several parents did note that exposure to the teletherapy format had a secondary benefit for preparing their young children for formal schooling, also expected to be delivered online during the COVID-19 pandemic.

Conclusions

This single case series study of Enhanced Conversational Recast showed that this treatment can be delivered with success in an online format. Most children showed progress toward their treatment goals, as measured by generalization of a grammatical form targeted for treatment to untreated linguistic contexts. In addition, improved use of the targeted grammatical form during generalization probes was typically accompanied by spontaneous production of the target form. Caregivers were generally happy with the treatment, although their comments concerning children's gains were generally unrelated to the specific goal of the treatment. Caregivers also felt that their children largely enjoyed the treatment as well. This study demonstrates that a treatment that has been well established in a face-to-face format can be translated to an online format, even for relatively young children (ages 5–6 years).

Funding Statement

This work was supported by the National Institute of Deafness and Other Communication Disorders Grant R01 DC015642, (Co-Principal Investigators: E. Plante and M. Alt).