I frequently review evaluations in which a child produces a fragmented, poorly organized, or difficult-to-follow narrative, yet the report concludes that narrative language is “within normal limits” because selected SALT measures fall within the expected range. The problem is not necessarily the SALT analysis. The problem is the conclusion drawn from it.
SALT, or Systematic Analysis of Language Transcripts, is a software system used to transcribe, code, analyze, and compare spoken-language samples. It provides valuable information about vocabulary, utterance length, syntax, grammatical accuracy, verbal facility, discourse features, and, when specifically scored, narrative macrostructure.
SALT is powerful, but its findings are only as broad as the task administered and the analyses completed. Expected performance on selected measures does not establish that a child can organize, connect, and communicate a coherent narrative.
SALT’s own guidance makes this clear. Miller, Andriacchi, and Nockerts (2019) caution that the Standard Measures Report may show no specific deficits even when narrative organization is problematic (p. 105). A child can produce adequate words and sentences and still struggle substantially at the discourse level.
What SALT Does Very Well
SALT addresses many of the practical problems that historically made language-sample analysis difficult to use consistently. It provides standardized transcription conventions, systematic analyses, reference databases, and efficient ways to examine multiple levels of spoken language. Its dynamic-norming approach also allows clinicians to compare a child’s performance with an appropriately selected reference group, and research supports the stability of these comparison samples when they are adequately sized (Tucci et al., 2022).
SALT is also not limited to sentence-level measures. Its Narrative Scoring Scheme (NSS) evaluates narrative macrostructure across seven areas: introduction, character development, mental states, referencing, conflict/resolution, cohesion, and conclusion (Heilmann et al., 2010; SALT Software, LLC, n.d.-b).
So the issue is not that SALT “only looks at syntax.” It does not. SALT can support a sophisticated narrative analysis. The quality of the final interpretation depends on what was elicited, what was coded, which analyses were selected, and whether the clinician understands narrative discourse well enough to interpret the results appropriately.
The Story Is More Than Its Sentences
A narrative is more than a series of grammatical sentences. It requires the speaker to establish characters and context, organize events, represent goals and motivations, connect actions to consequences, maintain reference, and bring the story to a meaningful resolution.
This distinction matters because children with developmental language disorder may show weaknesses in both microstructure and macrostructure. In a systematic review and meta-analysis, Winters et al. (2022) found a clear narrative disadvantage in children with DLD, with grammatical accuracy and story grammar among the most consistent areas of weakness.
SALT itself recognizes this distinction. Its Reference Book notes that word- and utterance-level language may appear adequate even when the narrative is poorly organized, with weak character introduction, missing plot information, and incomplete or misplaced conflicts and resolutions (Miller et al., 2019, p. 105).
A child can therefore produce acceptable sentences and still produce a poor narrative.
SALT’s Own Instructions Provide the Safeguards
One of the strongest points in SALT’s favor is that its own materials already describe the safeguards needed for sound interpretation. Problems arise when those safeguards are skipped.
1. Read the Narrative Before You Read the Numbers
Before interpreting a transcript, SALT recommends rereading it while listening to the recording, checking transcription decisions, and asking whether the sample actually reflects the referral concerns (Miller et al., 2019, p. 44).
For narrative samples, the clinician should also ask a simple question before looking at the scores: if you did not already know the story, could you follow what happened, why it happened, what the characters wanted, what they tried, and how the problem was resolved?
If not, that weakness matters. Expected microstructural scores should not override a narrative that is fragmented, incomplete, or difficult to follow.
2. After the Numbers, Return to the Narrative
SALT recommends returning to the sample after running the Standard Measures Report and asking whether the quantitative measures actually captured the difficulties heard in the child’s language. The Reference Book explicitly states that the SMR does not capture every aspect of language difficulty and identifies narrative organization and nonspecific referencing as areas that may require further analysis (Miller et al., 2019, pp. 94–95).
The Standard Measures Report should guide analysis, not end it. It can quantify selected language features, but it cannot determine by itself whether a narrative is coherent, complete, causally organized, or effective for a listener.
The clinician still has to return to the sample and decide whether the numbers accurately represent what the child actually produced.
3. Score the Narrative Dimensions, Not Just the Composite
The Narrative Scoring Scheme (NSS) is SALT’s macrostructure measure. It evaluates seven areas: introduction, character development, mental states, referencing, conflict/resolution, cohesion, and conclusion.
The NSS is useful because it requires the clinician to examine how the story is organized, but it still depends on trained human judgment. SALT instructs clinicians to read the narrative as a whole, look beyond the transcription codes, and score narrative features across the full story. Proficiency in NSS scoring develops with experience (SALT Software, LLC, n.d.-b).
The composite score should not be interpreted in isolation. SALT’s own Reference Book provides an example in which the overall NSS composite was only 0.94 SD below the database mean, while Mental States, Referencing, and Cohesion were substantially weaker (Miller et al., 2019, p. 106).
The individual narrative domains are often where the clinically important information is found.
4. Do Not Confuse Statistical Range With Clinical Normality
SALT uses plus or minus one standard deviation on the Standard Measures Report to help clinicians identify areas of relative strength and weakness that may warrant closer examination (Miller et al., 2019, p. 77). That is a descriptive convention, not a diagnostic cutoff.
A score within that range means only that performance on that specific measure, under that specific sampling condition, was not statistically unusual relative to the comparison group. It does not prove that the underlying narrative skill is intact.
This matters because a child can show expected performance on several microstructural measures while still demonstrating significant weakness in story grammar, cohesion, causal organization, referencing, or narrative completeness.
“Within normal limits” should describe the measure, not become a global conclusion about the child’s narrative ability.
5. Extra Prompting Can Change the Sample
SALT database comparisons are most meaningful when the elicitation protocol is followed. In the Narrative Story Retell protocol, examiners are instructed to use minimal, open-ended prompts and avoid specific WH questions, vocabulary cues, or prompts that supply story content (SALT Software, LLC, 2023).
This matters because targeted prompting can make a weak narrative appear stronger than it is. Questions such as “Why did he do that?” or “How did she feel?” may provide the very causal structure, vocabulary, or perspective-taking information the child failed to generate independently.
Once the examiner begins helping construct the narrative, the sample no longer reflects the same level of independent performance represented in the reference protocol.
Protocol fidelity therefore protects more than standardization. It protects the validity of the interpretation.
Picture Support Changes the Narrative Task
The SALT Narrative Story Retell is intentionally structured. The child first hears a model story while following the pictures. For students in Grades 2 through 6, the child then retells the story with the written text covered while the pictures remain available (SALT Software, LLC, 2023).
This is a useful standardized task, but it measures performance under a specific condition: an immediately modeled, picture-supported retell.
The pictures and model reduce some of the demands of independent narrative generation by supplying the characters, event sequence, much of the content, and a visual scaffold for temporal progression. Research confirms that elicitation conditions affect language-sample performance (Spencer et al., 2023).
Retell is not inferior to generation, but the two are not interchangeable. Adequate performance on a picture-supported retell does not establish that the child can independently plan, organize, and communicate a coherent narrative without that support.
How Serious Narrative Deficits Can Still Produce Reassuring SALT Results
Consider an 8-year-old who performs reasonably well on a structured, picture-supported SALT retell but is then asked to tell a previously heard story without the book or pictures available. The child says:
“There was a boy and his frog got out. He looked for him. Then his dog fell. There were bees and then they went somewhere. A deer came and then he found the frog. That’s it.”
The sentences themselves are not strikingly abnormal. They are understandable and largely grammatical, and the child communicates several major events. Yet once the visual sequence is removed, the weakness in narrative organization becomes much more apparent.
The child provides little orientation for the listener, incompletely establishes the central problem, omits important causal relationships, provides minimal information about goals and motivations, and moves from event to event without clearly connecting attempts to consequences. The listener receives a list of remembered events rather than a well-organized representation of the story.
This discrepancy is clinically important. A child may perform adequately on selected SALT measures during a structured, picture-supported retell while showing substantially greater difficulty when required to organize and communicate the story without that scaffold. The stronger supported performance does not invalidate the weaker performance. The difference between the two conditions is itself meaningful information about the child’s narrative abilities.
The question is therefore not simply whether the child can produce grammatical sentences or recall major events. It is whether the child can organize those events into a coherent narrative when the structure is not being supplied visually.
How Narrative Deficits Can Be Missed
False reassurance usually develops through several small errors rather than one obvious mistake.
- The task does not match the referral concern. A picture-supported retell cannot answer every question about independent narrative generation, classroom explanation, personal narrative, or writing.
- The child is overprompted. Specific questions or cues can supply information, vocabulary, or organization the child did not generate independently.
- The transcript is not reviewed as discourse. A narrative may contain adequate sentences while remaining fragmented, vague, or causally weak.
- The Standard Measures Report is treated as the narrative analysis. Useful microstructural measures do not replace direct analysis of story organization.
- Macrostructure is not examined. Without NSS or another defensible macrostructure analysis, important narrative weaknesses may never be identified.
- Only the NSS composite is reported. A broadly expected total score can mask substantial weaknesses in individual domains.
- Statistical range is treated as a diagnostic cutoff. “Within one standard deviation” does not mean that narrative ability is intact.
- Production is assumed to reflect comprehension. A child may retell visible or recently modeled events without fully understanding causal relationships, motives, or inferences.
- One context is generalized to all contexts. Retell, generation, personal narrative, academic discourse, and written language place different demands on the child.
- The software replaces clinical reasoning. SALT should document and clarify the communication problem, not determine whether a problem exists.
The common error is overinterpretation. SALT describes the language that was elicited and analyzed. The clinician is responsible for deciding whether the task, the sample, and the conclusions actually answer the referral question.
Report Only What You Actually Measured
One of the easiest ways to prevent overinterpretation is to keep the conclusion matched to the task, measure, and sampling condition.
| Overreaching conclusion | More defensible interpretation |
|---|---|
| “SALT showed age-appropriate narrative skills.” | “On this picture-supported retell, selected SALT measures fell within the comparison range. Narrative organization was examined separately.” |
| “The NSS composite showed adequate narrative ability.” | “The NSS composite fell within the comparison range, but individual narrative domains were also reviewed.” |
| “Narrative comprehension was age appropriate.” | “Comprehension responses were interpreted descriptively because the SALT Narrative Story Retell database does not provide normative comprehension comparisons.” |
| “The language sample showed intact narrative functioning.” | “The sample describes performance under one elicitation condition. Additional contexts are needed to characterize broader narrative functioning.” |
The point is simple: a picture-supported retell tells us how the child performed on a picture-supported retell. An MLU describes utterance length in that sample. An NSS composite summarizes performance across NSS domains.
A conclusion should not be broader than the evidence that supports it.
A Better Way to Use SALT in Narrative Assessment
The solution is not to abandon SALT. It is to use it as one part of a broader narrative assessment that actually answers the referral question.
- Start with the referral question. Decide whether you need to assess retell, independent generation, personal narrative, comprehension, academic discourse, written narrative, or several of these.
- Use the correct SALT protocol when database comparison is intended. Preserve the elicitation conditions and avoid adding support that changes the task.
- Read and listen to the sample before focusing on scores. Determine whether the narrative is coherent, complete, and understandable to a listener.
- Use SALT to quantify the features it measures well. Examine relevant microstructural and discourse measures such as syntax, morphology, lexical diversity, errors, utterance length, and verbal facility.
- Analyze macrostructure directly. Use the NSS when appropriate and review the individual domains, not just the composite.
- Examine causal and inferential organization. Determine whether the child explains goals, motivations, causal relationships, attempts, and consequences rather than simply listing events.
- Assess comprehension separately when relevant. Do not infer comprehension from retell production alone.
- Add a less scaffolded narrative condition when needed. If independent organization is the concern, obtain a generated or personal narrative in addition to a picture-supported retell.
- Examine transfer to academic language and literacy. Consider how the child performs in reading comprehension, written narrative, classroom explanation, and expository language when those are part of the referral concern.
- Treat discrepancies as findings. If microstructural measures look reassuring but the narrative is fragmented or poorly organized, the discrepancy itself is clinically meaningful.
A strong narrative assessment does not depend on one score or one sample. It integrates the task, the transcript, the quantitative measures, the narrative structure, and the child’s functional language demands.
What to Use SALT For, and Where to Stop
| Focus on | Do not overinterpret |
|---|---|
| Whether the elicitation task matches the referral question | A picture-supported retell as a global measure of narrative ability |
| The full transcript and recording | A numerical report viewed in isolation |
| Patterns across microstructure and macrostructure | One expected-range measure as proof that discourse is intact |
| Individual NSS domains and the narrative as a whole | The NSS composite as the sole narrative conclusion |
| Protocol fidelity and examiner support | A heavily prompted sample as independent performance |
| Causal coherence, goals, motivations, attempts, and consequences | A chronological list of pictured events as a coherent narrative |
| Separate evidence of comprehension | Adequate retell production as proof of comprehension |
| Performance across relevant discourse contexts | One sampling condition as proof of broader functional language ability |
| Clinical integration | SALT output as an automated diagnostic verdict |
The principle is straightforward: use SALT to describe and quantify the language that was actually sampled. Do not ask the results to establish abilities that the task did not directly examine.
The Bottom Line
SALT is not the problem. The problem begins when a sophisticated language-sample analysis system is treated as a diagnostic verdict.
SALT’s own developers repeatedly instruct clinicians to examine the transcript, preserve the elicitation context, look beyond the Standard Measures Report, analyze narrative organization, and apply clinical judgment. Their guidance is direct: clinicians should “trust your clinical judgment” and use SALT to document the communication problems they observe (Miller et al., 2019, p. 114).
If a child produces adequate sentences but cannot organize a coherent story, that discrepancy is the finding. If a picture-supported retell is stronger than independent generation, the difference between those tasks is the finding. If the NSS composite is broadly expected while cohesion, mental states, or referencing are weak, the uneven profile is the finding.
Good language-sample analysis requires both measurement and expertise. SALT provides powerful measurement tools. The clinician must still know what to elicit, what to examine, how to interpret the pattern, and what the available data cannot establish.
A reassuring SALT result should never erase a meaningful narrative weakness. Measure precisely, interpret narrowly, and keep the conclusion tied to the narrative the child actually produced.
References
Heilmann, J., Miller, J. F., Nockerts, A., & Dunaway, C. (2010). Properties of the Narrative Scoring Scheme using narrative retells in young school-age children. American Journal of Speech-Language Pathology, 19(2), 154–166. https://doi.org/10.1044/1058-0360(2009/08-0024)
Miller, J. F., Andriacchi, K., & Nockerts, A. (Eds.). (2019). Assessing language production using SALT software: A clinician’s guide to language sample analysis (3rd ed.). SALT Software, LLC. https://www.saltsoftware.com/ssdownload/Book3PDF/SALTReferenceBook3.pdf
SALT Software, LLC. (2023). Narrative Story Retell: Elicitation protocol (Rev. 8/2023). https://www.saltsoftware.com/media/wysiwyg/reference_database/NarStoryRetellElicProtocol.pdf
SALT Software, LLC. (n.d.-b). Comprehension questions. https://www.saltsoftware.com/products/elicitation-materials/comp-questions
SALT Software, LLC. (n.d.). NSS scoring guide. https://www.saltsoftware.com/coursefiles/1502/NSS_Scoring_Guide.pdf
Spencer, T. D., Tolentino, T. J., & Foster, M. E. (2023). Impact of discourse type and elicitation task on language sampling outcomes. American Journal of Speech-Language Pathology, 32(6), 2827–2845. https://doi.org/10.1044/2023_AJSLP-22-00365
Tucci, A., Plante, E., Heilmann, J. J., & Miller, J. F. (2022). Dynamic norming for Systematic Analysis of Language Transcripts. Journal of Speech, Language, and Hearing Research, 65(1), 320–333. https://doi.org/10.1044/2021_JSLHR-21-00227
Winters, K. L., Jasso, J., Pustejovsky, J. E., & Byrd, C. T. (2022). Investigating narrative performance in children with developmental language disorder: A systematic review and meta-analysis. Journal of Speech, Language, and Hearing Research, 65(10), 3908–3929. https://doi.org/10.1044/2022_JSLHR-22-00017