SALT vs Free Alternatives: A 2026 Comparison for SLPs

Every school SLP eventually has to answer one budget question: should we buy a SALT licence, or can we get away with the free tools? The honest answer depends on caseload size, district expectations, and how much of the workflow is bottlenecked at the scoring step versus the transcription step. The marketing pages from SALT Software, the free-tools landing pages on educational SLP sites, and the AI vendors all tell different versions of the same story — and almost none of them put the four options side by side with prices, feature parity, and scoring rules in one place.

This guide does that. It walks through SALT (Systematic Analysis of Language Transcripts; Miller & Iglesias, 2008) as the long-standing professional benchmark, the free browser-based LSA calculators on this site (MLU, DSS, IPSyn, NDW, PGU, and the Brown’s stages and SUGAR norms lookups), CHILDES/CLAN as the academic-research alternative, and ConductSpeech as the AI option that automates the slow step. By the end you should know which tool fits your caseload, your budget, and your district’s expectations — and what the failure modes of each option look like in real clinical use.

A note on neutrality: the free calculators on this site and ConductSpeech are both built by the same team that wrote this guide. Where the comparison is in our favour we have tried to include the SALT counter-argument; where SALT is the better pick we say so. The whole guide is checked against the most recent published SALT documentation and the SUGAR (Pavelko & Owens, 2017, 2019) and Heilmann et al. (2010) reliability literature so that the recommendations are anchored in something other than vendor copy.

SALT — Systematic Analysis of Language Transcripts — is a desktop language sample analysis package developed by Jon Miller and Aquiles Iglesias and distributed by SALT Software, LLC out of Madison, Wisconsin. The first version shipped in the 1980s, written in Pascal; the current version (SALT 22, released 2024) runs on Windows and macOS, uses a proprietary transcription syntax with C-unit and T-unit segmentation, and ships with a 700-child reference database that lets clinicians compare a child’s transcript against age-banded percentiles drawn from typically-developing American English speakers.

The transcription syntax is the part of SALT that takes a half-day to learn. Utterances are broken with tab-separated speaker codes, mazes are wrapped in parentheses, omissions are marked with asterisks, abandoned utterances end with `>`, and bound morphemes are tagged with explicit slashes (`walk/ed`, `dog/s`). Once a transcript is entered, SALT scores MLU in morphemes and words, total number of utterances, NDW, TTR, MATTR, percent grammatical utterances, percent intelligible utterances, mazes per hundred words, and roughly twenty other secondary measures, and then maps the child’s scores onto the matching percentile band from the reference database.

The reference database is the part of SALT that genuinely has no free analogue. It is the only commercial LSA package in 2026 that lets a clinician say "this child’s MLU-w of 3.4 at age 4;6 is at the 21st percentile relative to a national sample of 1,200 typically-developing children". The free SUGAR norms lookup on this site covers the SUGAR (2017, 2019) reference data — which is itself a 360-child sample — but the SALT reference database is broader, deeper, and ships with the percentile interpolation built in.

The free LSA toolkit on conductscience.com is a suite of single-purpose calculators built around the same Brown’s rules and SUGAR scoring conventions that SALT uses. Each tool runs entirely client-side — the child’s transcript never leaves the browser — and each one accepts plain-text input rather than a proprietary transcription syntax. The trade-off is that the free tools do not have a reference database, do not produce a single combined report, and do not store transcripts between sessions.

The MLU Calculator implements Brown (1973) morpheme rules in code: compounds count as 1 morpheme, irregular pasts count as 1 morpheme, and "gonna"/"wanna"/"hafta" each count as 2. It returns MLU-m, MLU-w, total morphemes, total utterances, and the matching Brown’s stage in one pass. The Lexical Diversity Calculator returns NDW, TTR, MATTR, and vocd-D from a typed transcript. The DSS Calculator implements the eight Lee (1974) categories with the original numerical weights. The IPSyn Calculator implements the 60-item Index of Productive Syntax in code. The PGU Calculator returns the percent of utterances that are grammatically complete using the standard SUGAR rule (no maze, no omission, no abandonment).

The Brown’s Stages Lookup, the SUGAR Norms Lookup, and the Speech Sound Development Chart cover the normative-data side: type in an MLU and read off the matching Brown stage; type in a child’s age and read off the SUGAR or Iowa-Nebraska norms. None of these is a substitute for a 1,200-child SALT reference database, but for the 80% of clinical questions that come down to "is this child’s MLU within typical limits for their age?", the published norm tables are good enough — and they are the same norms SALT references in its own documentation.

CHILDES (the Child Language Data Exchange System; MacWhinney, 2000) is the open-source academic alternative to SALT. The CHILDES project hosts thousands of transcribed child-language corpora in the CHAT transcription format, and the companion CLAN (Computerized Language ANalysis) program is the free analysis engine that operates on those transcripts. Like SALT, CLAN is desktop software with its own transcription syntax. Unlike SALT, the entire stack — corpora, software, tutorials — is free.

CLAN computes MLU, NDW, TTR, MATTR, MOR-tagged morphological analysis, KIDEVAL developmental scores, and dozens of secondary measures via command-line tools. The MOR tagger and KIDEVAL profile are arguably more rigorous than the SALT scoring engine for research purposes, and the CHAT format is the de facto standard in academic child-language journals. The trade-off is the learning curve: CHAT is the steepest transcription syntax of any of the four options in this guide, the analysis tools are command-line, and there is no clinical report template — the output is data files that you process yourself.

In practice, CLAN is the right choice for academic researchers, doctoral students, and clinicians publishing case reports. It is rarely the right choice for a school SLP running 30 IEPs per year because the front-end cost in transcription time and command-line literacy is too high relative to the time saved at the scoring step. The free browser-based tools on this site are designed precisely to be the clinical equivalent of CLAN — same scoring rules, dramatically lower learning curve.

The free browser tools on this site, SALT, and CLAN all assume you already have a typed transcript. For most school clinicians, that assumption is the entire problem: transcribing a 50-utterance language sample by hand takes 45 to 90 minutes, and that time bottleneck is the reason most caseloads run on standardised tests rather than language samples. ConductSpeech is the AI alternative built specifically for that bottleneck. It accepts the raw audio recording from the elicitation session, runs the transcription with speaker diarisation, segments the child’s utterances using the Brown rules, computes MLU-m, MLU-w, NDW, percent grammatical utterances, and morpheme accuracy in obligatory contexts, and drafts a present-levels paragraph for the IEP.

The right way to think about ConductSpeech is that it automates the slow step — transcription — so the clinician spends their time on the fast step (interpretation and goal writing). It does not replace SALT’s reference database; it does not replace the clinician’s judgement; it does not work as a black-box "compliance" tool. The output is editable at every step: the transcript, the utterance segmentation, the morpheme tags, and the present-levels draft are all visible and modifiable. ConductSpeech pays for itself on caseloads above 20-30 students because that is the point where the saved transcription time exceeds the licence cost.

The honest comparison is this: SALT gives you a percentile against a reference sample and a battery of secondary measures; CLAN gives you research-grade morphological tagging and the academic standard format; the free browser tools on this site give you the core SUGAR/Brown scoring without a learning curve; ConductSpeech gives you the audio-to-IEP-draft pipeline without the transcription bottleneck. They are not substitutes for each other — they solve different parts of the same workflow.

The matrix below compares SALT, CLAN/CHILDES, the free browser tools on this site, and ConductSpeech across the features that actually drive a clinical decision: scoring engine, normative data, transcription format, automation level, deployment target, and licence cost. None of these tools is strictly dominant — each one wins on a different axis — and the right pick depends on which axis matters most for your workflow.

SALT 22 single-user licences run roughly $295 in the US for a perpetual desktop licence (no annual renewal required, but updates ship as paid upgrades roughly every two to three years). Site licences scale with user count: a 5-user district licence is in the $1,000-1,500 range, and university teaching licences are negotiated separately. The 30-day trial available at saltsoftware.com is fully functional for the analysis engine but does not include reference database access. SALT does not ship a cloud version; the data lives on the desktop the clinician installs it on, which is a privacy advantage and a workflow disadvantage in equal measure.

CLAN and the entire CHILDES stack are free for academic and clinical use. The download lives at talkbank.org/clan/. There is no licence cost, no per-user pricing, and no cloud component. The cost is in the learning curve and in the lack of a clinical report template.

The free browser-based LSA tools on this site are free with no account required and no subscription. They are designed to cover the 80% of clinical scoring questions that come down to MLU, Brown’s stages, and SUGAR norms; they are not a substitute for the SALT reference database. ConductSpeech is a paid web product priced per-clinician on a monthly subscription, with a free trial for new accounts. The honest break-even versus a SALT-only workflow lands around 20-30 students on caseload — below that, the free calculators on this site cover almost everything you need; above that, the time saved on transcription justifies the subscription.

The single most useful framing for picking between SALT, CLAN, the free browser tools, and ConductSpeech is caseload size and the proportion of that caseload that needs a language sample. A clinician with 8-12 students on caseload, most of whom need standardised testing rather than LSA, has a different optimal stack from a clinician with 60+ students and a district that requires LSA-anchored present levels at every IEP.

For caseloads under 15 students, the free browser tools on this site cover almost everything: the MLU Calculator and Brown’s Stages Lookup for the 3-5 year range, the DSS or IPSyn Calculator for school-age syntax complexity, the SUGAR Norms Lookup for the percentile question. SALT is over-engineered for this volume and the licence cost is hard to justify against the budget cycle. For caseloads of 15-30 students, the free calculators are still enough for scoring — the question becomes whether transcription time is the bottleneck. If it is, ConductSpeech is the right add. If transcription is not the bottleneck (most of the work is cued-elicitation tasks already pre-segmented), the free tools alone are still enough.

For caseloads of 30+ students, the free tools alone start to break down because the transcription overhead compounds. ConductSpeech becomes the highest-leverage add at this scale because every saved transcription hour buys back direct service time. SALT enters the picture if your district report template specifically requires "percentile relative to a national normative sample" — which is the one feature SALT genuinely has no free analogue for. Many large-caseload clinicians end up with a hybrid stack: ConductSpeech for the audio-to-draft pipeline and SALT for the percentile lookup on the IEPs that need it.

The four most common objections clinicians raise to free LSA tools and to AI-based LSA are worth addressing directly rather than letting them sit as background uncertainty. They come up at every district training and they are reasonable questions — the answers in 2026 are just clearer than they were five years ago.

The workflow recommendation that comes out of this comparison is hybrid and pragmatic: pick the tool whose strengths line up with the part of the LSA pipeline that is currently slowing you down, and accept that the right answer for a 12-student caseload is different from the right answer for a 60-student caseload. The four options compete on different axes; the right combination compounds savings rather than trading one for another.

For most school SLPs, the recommended starter workflow is: use the free Language Sample Worksheet to elicit the sample in the clinic room, transcribe by hand (or via ConductSpeech if caseload is large), paste into the MLU Calculator and the appropriate metric calculator (DSS, IPSyn, NDW, PGU), look up the matching norm in the SUGAR Norms Lookup or the Brown’s Stages Lookup, and use the IEP Goal Generator to draft the goal. The whole workflow can be completed in 20-40 minutes per child on a 50-utterance sample, end to end, without any paid software.

For larger caseloads, the recommended hybrid is: ConductSpeech for the audio-to-draft pipeline, the free calculators on this site as a sanity-check QA layer, and a SALT licence on a single shared machine in the SLP office for the IEPs that specifically need a national-norm percentile. This stack is not the cheapest possible setup, but it is the one that actually fits the workflow of a high-caseload school clinician in 2026 and it is the configuration that several large districts have settled on after running the math.