A faster way to measure grammar skills


We’ve previously pointed you to research supporting “Percent Grammatical Utterances” (that’s PGU for the acronym-inclined) as a good language sample analysis to help diagnose developmental language disorder (DLD). While great practice, the procedure for computing PGU can be, in reality, pretty time-consuming.

In this study, the researchers that brought us PGU have given us a faster way to accomplish pretty much the same thing. Yay, science! They want to find a good method to measure and monitor growth in grammar skills — there really isn’t anything like that right now — so the process needs to be efficient enough to do multiple times a year for each kiddo. Enter Percent Grammatical Responses (you guessed it… PGR for short).

So how does PGR work? Kids between 3;0 and 3;11 saw a series of 15 pictures (described here). For each, the adult asked “What is happening in the picture?” The whole response was scored as either grammatical or ungrammatical. Take the number of grammatical responses, divide by 15, and, voila! PGR. No dividing responses up into C-Units… woo!

That’s too easy. It can’t be valid! It can! PGR appears to actually measure grammatical ability, since it correlates closely with SPELT-P 2 scores, while not being correlated with a measure of vocabulary. It also correlates with PGU, which has proven validity. As diagnostic tools for DLD, PGR and PGU agreed 92% of the time on “passes” and 94% of the time on “fails,” with given cutoff scores for each.

Awesome, right? But remember: this initial validation study only established a cutoff score for 3-year-olds, so we don’t have enough information to substitute PGR in for PGU with older kids. Also, hold off for now on using PGR to monitor progress, in addition to adding it to your diagnostic toolkit. More study is needed to determine if PGR is sensitive enough to reflect skill growth over time.

Cultural/Linguistic Diversity Note: The sample “ungrammatical” responses in the paper are constructions that are perfectly good in Non-Mainstream American English (or African American English). The kids in this study spoke “mainstream English,” but as always, be mindful of dialect differences in assessment.

Eisenberg, S. L. & Guo, L. (2017). Percent Grammatical Responses as a General Outcome Measure: Initial Validity. Language, Speech, and Hearing Services in Schools. Advance online publication. doi: 10.1044/2017_LSHSS-16-0070.