The ENNI for narrative language samples

This page draws on the ENNI manual and on Schneider, Hayward and Dubé’s Storytelling from pictures using the Edmonton Narrative Norms Instrument, published in 2006.

Which stories do I use?

The ENNI has a training story and two sets of three test stories. Each set has the same characters throughout and grows in length and complexity:

The simplest stories have one episode and two characters; the longest have three episodes and four characters. The stories were written to fit the story grammar model, a description of the information a good story contains, so that children’s stories could be scored against it. ENNI Manual: Introduction

Different measures use different stories. Story Grammar is scored on A1 and A3 only. First Mentions and the language measures use all six. Giving both sets, as in the normative study, provides what every measure needs. Pictures, Analyses & Comprehension

What is specific about administering the ENNI?

The child tells the story without hearing it first. This is story generation rather than retell. In the developers’ research, children included more story information when they retold a story they had heard. ENNI norms therefore describe stories children compose from the pictures, and a retell is a different task. Schneider, Hayward & Dubé, 2006, p. 225

The examiner cannot see the pictures. Hold the book up facing the child. The instructions tell the child that you won’t be able to see the pictures, so they need to tell the story really well. The child cannot rely on pointing, so the story has to be told in words.

The child previews each story first. Go through all the pages, return to the beginning, and then turn each page when the child seems to have finished with it. Either set can come first, but the stories within a set are given in order. The stories are audio-recorded.

The training story allows more help than the test stories. For the training story you can supply a beginning, such as "One day there was a boy who…". For the test stories, the manual lists a small set of prompts, such as "How would you start your story?" and "What happens in the story?", and asks you to use only those, so that the stories are collected as they were for the norms. Beyond those prompts, ask no questions and give only neutral responses such as "uh-huh" or "okay". Administration of the Instrument

How are the stories transcribed?

The ENNI divides stories into communication units (C-units): an independent clause with any clauses that depend on it, or a sentence fragment. Clauses joined by and, but or then begin new C-units. Clauses introduced by because, when or although stay with their main clause. Utterance boundaries are set on these grammatical grounds, not by intonation.

The manual gives conventions for both CHAT and SALT, including how to mark mazes and contractions. For comparison with the word-count norms, the transcript needs to match the normative sample’s. Grammatical endings are marked so that bouncing and bounced count as one different word, gonna and wanna are written out, and a short list of words is spelled as one word or two. Transcribing the Stories

Story Grammar and First Mentions can be scored by listening to the recording, without a full transcript.

How are the stories scored?

The ENNI scores the content of the story, how characters are introduced, and the language used to tell it:

Story Grammar credits what a listener could understand without seeing the pictures. The scoring sheets list typical acceptable responses for each element of A1 and A3. The Story Grammar page works through four scored examples: two stories each from a typically developing four-year-old and a four-year-old with a language impairment. They can be used to check your own scoring.

MLCU excludes mazes, responses to questions, utterances that are interrupted or broken off, comments unrelated to the story, and story endings such as "the end". The MLCU and Complexity Index instructions give the full rules.

How do I use the ENNI’s norms?

The norm tables convert each raw score to a standard score (mean 10, standard deviation 3) for the child’s age in years, from 4 to 9. Story Grammar and First Mentions have tables. The language measures have a spreadsheet that calculates the standard score when you enter the raw value. All are linked from the analysis pages.

The norms come from 377 children in Edmonton, Alberta: 50 typically developing children per year of age and 10 to 17 children per age group with a previously identified language impairment. The impairment group was weighted to about 7% of each age group when the norms were calculated. The manual reports the sample’s socioeconomic status and compares its ethnic make-up with the 2001 census for Edmonton and Canada. Description of the Normative Study All the children spoke English as their first language at home. Schneider, Hayward & Dubé, 2006, p. 227

In the 2006 study, Story Grammar scores from A1 and A3 increased with age to about 7 or 8, then leveled off. At ages 4 to 8 they correctly classified 81% of children as having typical language or a language impairment: 84% of the typically developing children and 70% of those with an impairment. At age 9 the two groups’ scores did not differ, and at age 7 the difference on the simple story fell short of significance. The authors conclude that Story Grammar alone is not sufficient to identify a language impairment at any age, and they recommend caution at age 9. Schneider, Hayward & Dubé, 2006

The Syntactic Complexity page notes that MLCU and the Complexity Index do not separate the two groups well at age 6, so scores at that age are used with particular caution. Syntactic Complexity In their 2008 update, the developers recommend using several ENNI measures rather than just one or two. ENNI Update, 2008

These comparisons apply to stories collected with the ENNI pictures and procedure. A story from another picture book, or a retell, is a different task.

How can the ENNI help identify treatment or IEP goals?

Because the measures are scored separately, the results show which part of storytelling a child finds difficult:

These are examples of how results can point toward target areas, not a separate ENNI scoring system. The ENNI also has comprehension questions for A1 and A3. They ask about story events, the problem and its resolution, and which parts of the story matter most. The developers describe the answers as showing what a child can provide with support, alongside what the child included unprompted when telling the story. ENNI Update, 2008

What additional analyses can I use with the sample?

ENNI stories are language samples, so they can be analyzed for grammatical accuracy, verb morphology or vocabulary, with each measure’s own rules. Later studies from the ENNI normative data report reference data for percent grammatical utterances, finite verb morphology and clausal density. The measures differ in how well they separated children with and without language impairment. Percent grammatical utterances did so acceptably or well at every age from 4 to 9. Finite verb morphology did well to age 7 but poorly at 8 and 9. Clausal density did poorly, and its authors recommend it for describing complex syntax and following change over time, not for identification.

SALT includes an ENNI reference database built from the typically developing children in the normative sample. It compares a transcript with age-matched ENNI stories inside SALT, for samples transcribed in SALT’s format. SALT ENNI database

Transcripts for 361 of the 377 children, in CHAT format, are available through CHILDES.

Useful ENNI materials

• Pictures, Analyses & Comprehension: the training story, Sets A and B, comprehension questions for A1 and A3, and links to each measure’s scoring sheets and norm tables.
• ENNI Manual: rationale, the normative study, administration with the full prompt list, and transcription in CHAT and SALT.
• Schneider, Hayward & Dubé (2006): the instrument’s development and the Story Grammar results by age and group.
• Publications & Presentations: later papers and conference materials, including a clinician scoring-reliability study and a comparison with two other narrative tests.