Beyond the Milestones: Selecting and Interpreting Curriculum-Based Assessments (VB-MAPP, ABLLS-R, and AFLS)

The Default Tool Problem: Why Most Behavior Analysts Reach for the Same Assessment

Dr. Jessica Osos opened with a picture of the field, not a feature list. She pointed to a 2020 survey study by Padilla that was sent out through the BACB. About 1,400 practitioners took part. They could check more than one tool. Most reported using the VB-MAPP (a map of early talking skills). Every other tool fell below 50%. The AFLS sat near 12%.

Two other numbers in that survey matter more. 71% of respondents said their direct supervisor trained them on assessments. Not the publisher. Not a set course. Not a competency check. Osos was quick to say this is not a knock on people. It is how most of us learned our craft. But fidelity (how closely you follow steps) can slip when nobody checks the steps. What a supervisor does, right or wrong, gets passed down.

And 58% said they chose their tool because of the research base. That is a field that wants to make evidence based choices. Yet the survey author concluded that these tools carry limited psychometric support (proof the test holds up). That gap is worth sitting with.

The cost is clinical, not academic. The tool sets the window you look through. If a repertoire (a set of related skills) is never sampled, it never becomes a goal.

you can't write a goal for a repertoire that you've never assessed

From the talk — Dr. Jessica Osos

What a Curriculum-Based Assessment Does and Does Not Do

All three tools in this talk are curriculum-based assessments (tests tied to teaching steps). A tool like this samples a set curriculum. It then compares what the learner did to a stated criterion (a set standard to meet). Can this learner do this thing, under these conditions, to the standard? That is the whole question.

There is no normative sample (a large peer comparison group). There is no ranking against other kids. Items turn straight into teaching targets. The tools are built for frequent retesting, so testing effects are less of a worry. That fits how payers think. They ask about six months of progress. A once a year tool cannot answer that. These tests are also meant to move when you teach.

Here is where teams get in trouble. These tools give no standard scores. No percentiles. No age equivalents. They do not decide eligibility. They do not support a diagnosis. They say nothing about where a learner stands among peers. And they do not replace the clinical judgment that picked them in the first place. Osos treats them as the engine for treatment planning, with other tools carrying the bigger picture.

no major decision is going to rest on one single test score

From the talk — Dr. Jessica Osos

VB-MAPP Structure: Five Parts and a 48-Month Ceiling

The VB-MAPP has five parts. Most teams use two. The milestones are the famous one. The barriers assessment holds 24 barriers, scored 0 to 4. It covers things like instructional control (the learner follows your lead) and prompt dependence (relying too much on hints). It also covers failure to generalize and weak motivation.

The transitions assessment is the part Osos called genuinely underused. She leans on it to support a move to less restrictive care or fewer hours. It also helps translate ABA (applied behavior analysis) skills into school terms.

The fourth part is task analysis and skills tracking (breaking a skill into small steps). That is where the curriculum lives. When a learner misses a milestone, go there to find the small steps and prerequisite skills behind it. The fifth part covers placement and IEP (special education plan) goals. Few behavior analysts write IEPs, so it mostly sits idle. It earns its keep when you join a school team.

Two design facts drive the rest. First, the sequencing is age banded. It tells you roughly where a learner sits in a progression, not just what is missing. Second, that detail stops at 48 months. That is the ceiling. A learner who is developmentally past 48 months is a poor match, whatever their age.

Sundberg, who created the VB-MAPP, is explicit about this in the manual. He calls it a criterion-referenced assessment (compared to a fixed standard, not to peers). It is also a curriculum guide and a skill tracking system. It is not a test of ability.

ABLLS-R and AFLS: What Each One Adds That the VB-MAPP Misses

The ABLLS-R (a fine grained early skills checklist) does not label domains by age. It labels them by letter, A through Z. A is cooperation. B is visual performance. The names look like VB-MAPP domains. But you cannot pull an age from them. Nothing tells you a skill shows up by 18 months.

Item counts show the real focus. Basic learner skills hold 381 items. That is 71% of the whole tool. One source lists 382, which changes nothing in the clinic. Scoring ranges vary by item, and there are four score columns: a baseline plus three updates. The form is built as a long record of one learner. The big differentiator is content. The ABLLS-R covers academic and self-help work. The VB-MAPP has early math and reading, then stops. Sections U through X hold self-help and daily living items.

The AFLS (a daily life skills tool) comes in six separate protocols. You almost never need all six for one learner. Vocational is the largest, with 16 areas. The structure is the point. Items are grouped by function and setting. There are no levels and no age bands. You cannot place a learner developmentally in it, and that is on purpose. For these skills, Osos said an age band would be unhelpful and arguably disrespectful.

Sequencing is where the three tools truly split. The VB-MAPP orders items by when skills usually emerge. The ABLLS-R sequences inside each domain, from early to harder. The AFLS sequences by context. Each choice costs you something. The AFLS can show that a learner cannot finish a store task. It will not tell you that a missing discrimination (telling two things apart) is why.

The purpose of this assessment is to identify deficits and also areas of strengths.

From the talk — Dr. Jessica Osos

Criterion-Referenced vs Norm-Referenced: Sentences You Can and Cannot Write

Norm referenced means comparison to same age, typically developing peers. That needs a standardization sample. You get back standard scores, deviations from the mean (how far off average), and age equivalents. The Vineland works that way.

Criterion referenced means comparison to a stated criterion, item by item. So, can this learner do this thing under these conditions to the standard? And there's no standardized sample. You're always using that same child as their own comparison in the future. So when someone calls the VB-MAPP a standardized assessment, they are half right at best. The way you give it is standardized. The scores are not standardized against a population.

That rules out a whole set of sentences. You cannot say a learner scored at the 12 month level. The levels are printed, but no normative sample backs that number. You cannot say a learner is in the third percentile for language. You cannot say the ABLLS-R shows a 14 month delay. Be careful with claims that a point gain proves the treatment worked, since too many things move at once.

What you can do is name skills. Report which items are absent and which are present. Use the tool to flag deficits and strengths. If your sentence needs a peer comparison, you need a different instrument, not better wording.

both types are standardized administration. But only one is standardized against a population.

From the talk — Dr. Jessica Osos

VB-MAPP Reliability and Validity Evidence: Read the Domain Scores With Care

The VB-MAPP was first published in 2008. The first reliability data did not arrive until 2019. That is eleven years. In that work, raters scored in pairs across 32 autistic children, using interclass correlation (a measure of rater agreement). The milestones total score held up well. The barriers total was only moderate.

The domain level is the part that should change your practice. Few of us report the total. We read tacting (naming what you see) or intraverbal (answering what someone says). Those domain profiles are less reliable than the total. The authors say it plainly: high reliability for an individual domain should not be assumed. Osos's rule follows from that. Do not hang a decision on a one point domain difference. If a domain score drives a major call, get a second score.

Padilla and Acres collected validity evidence in 2021. Thirteen subject matter experts rated the tool. All were BCBAs (board certified behavior analysts) with at least seven years of experience. They scored every milestone and barrier item on relevance, developmental fit, measurement method, and domain representation. Relevance was reassuring: the items belong. Only 31% reached statistical significance for relevance under their criterion, which does not make the rest bad items. Representation was the soft spot. The content is right. There may not be enough of it. Their advice was to use the VB-MAPP alongside other sources.

Two more findings sit at the level of design. A factor analysis (a test of item grouping) found that items clustered by skill complexity. They did not split cleanly into separate operants (types of verbal behavior, like asking or naming). Dixon and colleagues predicted that in 2005. PEAK, another skills assessment for verbal behavior, offers a more expansive option.

Training matters too. Staff who only read the manual ran the protocol under two thirds correct. After behavior skills training (teach, model, practice, feedback), accuracy climbed sharply.

most milestone domains reached only a moderate reliability. And most barrier domains were poor.

From the talk — Dr. Jessica Osos

What the Evidence Shows for the ABLLS-R and the AFLS

Padilla and colleagues reviewed reliability and validity evidence for ABA based assessments in 2023. The AFLS and Essentials for Living were not in it. Osos stressed why. Nothing met the reviewers' inclusion criteria inside the search window. The reviewers did not overlook these tools, and a missing study is not proof that nothing exists.

Still, the gap has teeth. As of that review, there was no published reliability or validity evidence for the AFLS. The tool may serve your learners well. But if you tell a payer you used a validated tool, that claim is weak. Osos mentioned a 2026 article that she believed appeared in Behavioral Interventions. It reported a psychometric evaluation of the ABLLS-R and the AFLS. It was co-authored by an AFLS developer. Osos had not read it in full, so she did not describe its findings.

Several ABLLS-R papers do exist. Read the author lists first. A 2018 study reported strong numbers, and Osos did not think they were inflated. But look at what the ICC (a score agreement number) measured. It measured agreement among expert panelists scoring the same recorded protocol. That is far easier than two clinicians assessing the same child on their own. Two of the three authors were the publisher. Another study found strong test-retest results (same test given again later), but with neurotypical children, not autistic learners. A third also used neurotypical children, and the authors called the evidence preliminary.

So combine on purpose. Osos pairs a curriculum-based tool with a standardized one, often a Vineland caregiver interview. Plan the overlap in advance and avoid redundant scope. Do not run a second tool just because the first score was not what you wanted. And when you switch instruments, run both during one period, then drop the old one. You cannot average scores across tools or compare one tool's score to another's.

we want to be careful with that sentence because a null result tells you nothing.

From the talk — Dr. Jessica Osos

A Five-Question Framework, Run Through Three Learner Profiles

Ask the five questions in order. One: what decision am I making? Some decisions no curriculum-based tool can serve, such as eligibility or a diagnosis. Two: where does this learner sit relative to the floor and ceiling? Three: which repertoire does this decision hinge on? Name it before you open a protocol. Four: what does this tool not measure that I still need? Plan the supplement now, not later. Five: what will I say when someone asks how I know? Think criterion, source, and limit.

Take a learner who is two years and ten months old, new to EIBI (early intensive ABA for toddlers). He has about 15 words, and nearly all are mands (requests for something). The decision is initial program design. There is no ceiling risk at that age. The repertoires at stake are manding, echoics (copying a sound you hear), and listener responding. Motor imitation and the barriers that block learning matter too. The blind spot is adaptive skills (daily self-care and home skills), since the VB-MAPP has almost none. Add ABLLS-R sections U through X or a caregiver interview. Report milestone counts by domain, not an age level, and flag domain scores as provisional.

Now take a seven year old in general education who speaks in full sentences. She has scored just over 160 on the VB-MAPP for three straight reauthorizations. The team keeps writing that she maintained gains. That is a ceiling effect, not a plateau in her learning. Two decisions follow: diagnose the stall and defend continued services. A flat score invites a denial. The repertoires in question are intraverbals, classroom routines, independent work, and peer social behavior. The ABLLS-R has finer grained items for those, and PEAK is worth a look. Then write plainly that the flat score reflects the instrument's range, not her progress.

Last, take a 17 year old who uses a speech generating device (a device that speaks for him). He mands and tacts fluently through it. He ages out of school services in ten months. The clinic's default assessment is the VB-MAPP, and he still places inside it. That does not make it the right fit. Osos would not want to explain a 48 month range to a parent or a vocational counselor. The same goes for explaining it to the young man himself. The target repertoires are money, meal preparation, work behavior, self-management, health, and safety. The AFLS fits. Its blind spot is why a task fails, so pair it with a barriers assessment or a real task analysis. The output travels: mastery by routine and setting, in words adult providers already use.

So I'm not going to automatically say what assessment am I going to use? I'm going to say what information do I need from an assessment.

From the talk — Dr. Jessica Osos