Percent grammatical utterances: Meet your new go-to LSA measure

8.png

We know that few SLPs use language sample analysis. And, real talk, we get it—transcribing and analyzing language samples takes forever, and sometimes you end up with a whole bunch of numbers and no idea what they mean. To help with that, this study gives us a little more guidance for analyzing narrative language samples using percent grammatical utterances (PGU).

The authors used data from 4- to 9-year-old children who took the Edmonton Narrative Norms Instrument (ENNI). As part of the ENNI, children generated stories for six picture sequences, which were transcribed and coded. PGU coding is pretty straightforward. You:

  1. Divide the sample into C-units

  2. Decide whether each C-unit has a verb

  3. Mark each C-unit with a verb as grammatical or ungrammatical

  4. Divide the number of grammatical C-units by total eligible C-units (those with a verb) to get PGU

That’s it—no complex coding, no lengthy rubrics, just a yes/no decision for each utterance (see the article and supplemental material for more examples and guidance). And as easy as it is, PGU is also a good measure. The authors found that PGU was reliable, valid, and able to distinguish children with and without developmental language disorder (DLD) with acceptable diagnostic accuracy. Using the data in the article, you can supplement diagnostic decisions (Table 5) or track progress (Supplemental File S4). Note that you should use the same language sample context that they did; luckily, the ENNI pictures are freely available. And for an even faster measure to use with 3-year-olds, check out these researchers’ previous work on percent grammatical responses (PGR).  

*Note that we shouldn’t use PGU to score samples from speakers of non-mainstream dialects of American English because the scoring rules don’t (yet) account for dialect differences.

 

Guo, L., Eisenberg, S., Schneider, P., & Spencer, L. Percent grammatical utterances between 4 and 9 years of age for the Edmonton Narrative Norms Instrument: Reference data and psychometric properties. American Journal of Speech–Language Pathology. doi:10.1044/2019_AJSLP-18-0228

Shifting and switching from Spanish to English

5.png

In the US, children who speak Spanish at home often begin learning English when they start school, and their dominant language shifts from Spanish to English over time. To get a better idea of how this happens, the authors of this study looked at the change in grammatical accuracy (percent grammatical utterances or PGU*) in Spanish and English narrative retells from kindergarten to second grade.  

As expected, children’s PGU in English went up over time, while PGU in Spanish went down. The researchers compared children in bilingual (English–Spanish) vs. English-only classrooms. For children in bilingual classrooms, the decrease in Spanish PGU was slower, but the increase in English PGU was slightly slower also.  

The researchers also looked at a subgroup of the children who had lower PGU in Spanish at the outset. They called this group “low grammaticality” because they didn’t have enough measures to confidently diagnose developmental language disorder (DLD). Children in this group showed a different pattern, with Spanish PGU holding steady for those in bilingual classrooms, suggesting that they benefited from bilingual teaching.

For a brief time (around age 8), English and Spanish PGU scores for the low grammaticality group looked similar to the rest of the children, which means that if we assessed them at this point, we might not be able to tell who does and doesn’t have DLD. The authors encourage us to assess children in their home language early on, before this shift happens.

So as if assessing English language learners wasn’t hard enough, we also need to consider the type of instruction children are getting and their skills in each language over time.  Ideally, we’d assess children in their home language right when they start school. When that’s not possible, dynamic assessment might help us to differentiate language disorders from normal language dominance shifting during the early school years. For other resources on diagnosing DLD in English language learners, see reviews here, here, and here.

 

*Remember that higher PGU means more accurate use of grammar.

Castilla-Earls, A., Francis, D., Iglesias, A., & Davidson, K. (2019). The impact of the Spanish-to-English proficiency shift on the grammaticality of English learners. Journal of Speech, Language, and Hearing Research. doi: 10.1044/2018_JSLHR-L-18-0324.

Tutorial Throwback (2014): Am I doing this right? Some under-the-hood tips to structuring language therapy

SLPs have to make LOTS of decisions about how to structure our therapy:

How many minutes do you see each student?

How much time should you spend on a single goal?

What type of activities should you use to target a specific goal?

 The list goes on…

The answer to these questions isn’t always clear (even when you consult the available research), so when an article comes along that tackles these issues, we get pretty excited! This 2014 article by Eisenberg reviewed the research to break down what SLPs need to know about treatment dosage, dose form (type of treatment the student receives), and treatment procedures. Note that the author decided to focus on specific aspects of language therapy rather than any particular packaged approach, which helps us think critically about what’s making a difference in children’s success.

Language and education researchers have borrowed the concept of dose from our friends in medicine. While we might not typically talk about language services in terms of dosage, this analogy actually works really well in helping us think about our therapy and its “active ingredients.”  Imagine a kindergarten student with a goal of using progressive forms who is seen weekly for 30 minutes. In this case, dose would be the number of targets given per session (“Spiderman is swinging” and “Minnie is laughing” would be two), and frequency would be the amount of time the dose is given (30-minute weekly sessions).

We know that children with developmental language disorders (DLD) need a higher dosage than their typically-developing peers to learn new words and new grammatical forms, but just how much more? The article provides some numbers for comparison:

Screen Shot 2019-06-14 at 12.03.46 PM.png

For frequency, it seems that spaced exposure (compared to mass exposure) leads to better learning. So a student receiving a consecutive 60-minute session (mass exposure) may not be benefiting as much as a peer who receives two 30-minute sessions (spaced exposure).

So now we can talk about what the dose (language input) should actually look like. Here are some general guidelines for teaching specific grammatical targets (e.g., passives, bound morphemes). You should:

Vary the surrounding vocabulary

  • Children learn forms better when they’re combined with different vocabulary

  • e.g. The boy was scolded. This cat was licked.

Keep the target consistent across trials

  • Children acquire rules better when exceptions are limited 

  • e.g. She walks. / He runs……….. Instead of: She walks. / I walk.

Speak in grammatical (not telegraphic) sentences

  • There’s no compelling evidence that telegraphic speech improves comprehension

  • Children need grammatical models to produce grammatical forms

  • Children use grammatical knowledge to understand sentences and learn new words

  • e.g. The boy helps his mama…….. Instead of: Help mama.

Finally, for treatment procedures, therapy activities range from least natural (e.g., drills) to most natural (e.g., literacy based interventions or conversation activity). There seems to be a trade-off, with imitation training giving you quicker short-term gains, but activities like recasts being more important for spontaneous productions.

Screen Shot 2019-06-14 at 4.02.07 PM.png

Lots to think about, right? Now, while this information certainly doesn’t answer all our questions, it’s a good framework that can help SLPs think about why something in therapy may (or may not) be working and making meaningful improvements to our sessions.

Looking for more about dose and treatment intensity? Check out our blog post on that topic from 2018.

Eisenberg, S. (2014). What works in therapy: Further thoughts on improving clinical practice for children with language disorders. Language, Speech, and Hearing Services in Schools. doi:10.1044/2014_LSHSS-14-0021

Mistakes preschoolers make in multi-symbol utterances using AAC

14.png

Preschoolers learning to communicate via AAC systems typically start by using one symbol at a time. Many go on to construct 2–3 symbol utterances, but make mistakes along the way. In this study, the researchers looked back at data from a prior study to explore 10 three–four-year-olds’ errors when producing multi-symbol utterances. In total, they made errors on 45% of their utterances!

We’ll get to those errors, but first, some background. All the preschoolers had a speech sound disorder/delay diagnosis, although one child also had a secondary ASD diagnosis and another was diagnosed with cerebral palsy. During the intervention sessions, the kids used an AAC device (an iPad with Proloquo2Go) to describe videos, then the clinician briefly modeled the targets and facilitated play-based therapy for 20 minutes. Keep in mind that the preschoolers didn’t have access to the device outside of the study, and 8 of the 10 participants actually had no prior experience with AAC before the intervention.  

Because AAC literature has focused heavily on inversions (word order reversals), the researchers checked for other error patterns.

 
The Complexity Approach for Grammar_.png
 

The main takeaway? Inversions and omissions were more common than substitutions and additions overall, but there were differences across targets and—business as usual for AAC studies—there were differences across the children in the study.

So, what can we do with this information? For starters, when you collect data, think beyond the number of words in the utterance. Instead, try classifying the types of errors the students are making. Are they inversions, substitutions, omissions, or additions? We’d wager that you’d approach instruction just a little bit differently depending on the error type—and that type of modification to your instruction could make all the difference!

FYI: As we mentioned, data analyzed here was originally collected for a different study. Although the data set isn’t perfect, this is the first study we have that conceptualizes the errors these children might be making. Also, keep in mind these targets were chosen specifically for speakers of English, which has a Subject + Verb  + Object syntax structure.

 

Binger, C., Richter, K., Taylor, A., Williams, E. K., & Willman, A. (2019). Error patterns and revisions in the graphic symbol utterances of 3- and 4- year old children who need augmentative and alternative communication. Augmentative and Alternative Communication. doi:10.1080/07434618.2019.1576224

Mixing language with science to target "because" and "so" in preschoolers with DLD

As SLPs, we all love the intricacies of grammar… right? No?

Well, we at least love ourselves some good ol’ adverbial clauses, right? Anyone? Just me?

Ok, so maybe we don’t all share the same nerdy love of all things grammar, but we can probably all agree that complex sentences are essential for both conversation and academics, and that children with developmental language disorder (DLD)—who struggle to use and understand these sentences—need effective language intervention to learn them. Also, even for us grammar enthusiasts, complex syntactic constructions can be difficult to teach. So what do we do?

That’s where this single-case study* from Curran and Van Horne comes in. They hypothesized that recast strategies—which have been researched extensively for teaching morphology—would improve preschool children’s use of causal adverbials, specifically because (“I ate because I was hungry”) and so clauses (I was hungry, so I ate). Critically, the authors distinguish between kids being able to use the word “because” in their speech (which happens pretty early) and actually acquiring multi-clause sentences that express a cause–effect relationship; and we’re interested in that second, more complex skill.  

What’s great about this study’s approach is that it looked at causal adverbials in the context of science instruction, which relies heavily on understanding of cause and effect. After some baseline probes and some standard science lessons, the researchers provided 20 sessions of science instruction combined with language intervention, using visuals, recasts, and prompting for those so and because clauses. A typical recast might sound like this: 

Child: “The kite goed up. Wind pushed it.”

Adult: “The kite went up because the wind pushed it.” 

Wondering how they got the kids to produce these structures in the first place? They used prompts like this one for because: “She pushed air in. The plunger popped out. Why did the plunger pop out? Start with ‘The plunger...’”

So did it work? Well, six of the seven children improved in their use of because clauses, showing strong positive trends during the intervention phase, compared to control structures. So clauses did not improve significantly, maybe because they were less frequent or as the result of a possible “competition effect” between because and so. Finally, while the kids did learn the science content over time, it didn’t seem to be the result of language skills gained.

The authors sum it up nicely: “Multiclause adverbials can be effectively addressed in clinical intervention, even for children who do not yet possess significant written language or metalinguistic skills.” Larger studies could help to clarify some of these findings, but using recasts as a way to teach complex syntactic structures is a promising strategy for children with DLD.

 

*Single case designs have their own special place in research and are valuable tools for treatment studies. They can highlight individual differences (because group designs only look at mean differences), and because it’s pretty comparable to what SLPs are doing in the real world, they have high social validity. We still would love to see similar results come from a larger study design, but these smaller studies teach us a lot in the meantime.

 

Curran, M., & Van Horne, A. O. (2019). Use of Recast Intervention to Teach Causal Adverbials to Young Children With Developmental Language Disorder Within a Science Curriculum: A Single Case Design Study. American Journal of Speech-Language Pathology. doi:10.1044/2018_AJSLP-17-0164