The second in a series introducing HMH Pulse, the learning science behind HMH Performance Suite®
In article 1 in this series, we introduced Pulse as HMH’s learning sciences intelligence layer and named the three connected components that make it work: an enterprise ontology, a proficiency model, and a cascading inference layer. This post zooms in on ontology where we will explain what an ontology is, why a learning sciences platform needs one, and why the ontology behind Pulse is structured the way it is.
What is ontology?
An ontology is a structured map of how the concepts in a particular domain relate to one another. It captures the entities that exist in that domain (the things), the relationships among them (how they connect), and the constraints that govern them (what’s allowed and what isn’t). Done well, an ontology is what allows a computer to reason about a domain much like a domain expert would.
In Pulse, the domain is K–12 teaching and learning. The entities include the standards published by a state, the skills those standards reference, and the assessment items and instructional resources used to measure and teach each skill. The relationships connect these elements, showing how skills, standards, assessments, and resources relate to one another. They show which skills are linked to specific standards, which skills rely on others, which items assess particular skills, and which resources support them. Constraints reflect what we know about how learning works: a sixth-grade algebra skill isn’t a prerequisite for kindergarten counting, and an item measuring whole-number multiplication doesn’t assess decoding, among others. In plain language, the ontology is the foundational map that Pulse is built upon.
Why a learning sciences layer needs an ontology
There is a temptation, in 2026, to imagine that a large language model is enough on its own. Drop in your data, write a prompt, and get a recommendation. The problem with that approach is what happens when you ask the system to justify why it gave that recommendation, what would change if it had different evidence about a student, or whether the same evidence would lead to the same conclusion next week. Without a structured representation of the domain, the answers to those questions are unstable.
The contemporary view in AI research, across both academic and applied work, is that ontologies are not optional scaffolding for credible AI systems. They are a precondition for reasoning that generalizes, that can be audited, and that connects naturally to the structures domain experts already use. If you want a system whose claims about learning hold up under scrutiny, you start by writing down what you know about learning in a form a machine can reason over.
Over the past few years, NWEA has committed itself to building the Pulse ontology, extending its tradition of investing in research-informed approaches to understanding and supporting student learning. The Pulse ontology draws on three foundations: knowledge representation from computer science, learning sciences research on how students develop understanding of concepts, and the practical expertise of HMH’s curriculum and assessment teams who have worked alongside educators for decades. The skill portion of the ontology, in particular, is built by PhD learning scientists with backgrounds in cognitive science and in either mathematics or English language arts education, all with prior classroom teaching experience. Every skill cluster design goes through internal review by at least two subject-matter experts before entering the production ontology. As an external check, we also run cognitive interviews with classroom teachers, who walk through portions of the ontology in real time and have so far judged the structure accurate, instructionally meaningful, and developmentally appropriate. The output of this work is a map that reflects how learning actually unfolds, which is what makes the reasoning on top of it trustworthy.
The ontology is the foundational map that Pulse is built upon.
What does Pulse ontology contain
As we introduced in article 1 in this series, Pulse’s ontology brings together three things every educator already works with: the standards adopted in their state, the skills those standards reference, and the instructional resources and assessment items used to teach and measure each skill.
The scale is substantial. 28,958 standards from all 50 U.S. states across grades K–12 (15,642 in mathematics; 13,316 in English language arts). As of today, there are 967 skills: 554 in mathematics, organized into eleven clusters spanning grades K–8, and 413 in ELA. Tens of thousands of assessment items, including MAP Growth and HMH curriculum-embedded assessments, are each tagged with the specific skills they measure through a structured mapping that clearly indicates which skills each item’s evidence supports.
What makes this a map rather than a stack of inventories is the connections among the parts. Equivalent standards across states are navigable from one to another while remaining distinguishable when a district needs them to be. Skills within a cluster are linked through prerequisite and conceptual relationships. Items and resources are tied to skills, and through those skills, back to the standards to which a school is held accountable. The result is a single navigable structure: a teacher in Ohio and a teacher in California can be working with different state standards, different curricula, and different assessments, and Pulse can reason about the underlying learning consistently across both.
What do we mean by “skill” in the Pulse ontology? Think about what it takes for a student to solve a problem like 29 minus 14, or to identify the main idea of a text. Two things are happening at the same time. The student is using a way of doing it: regrouping the ones, identifying the main idea across paragraphs. And the student is drawing on what they already know: that 29 is two tens and nine ones, how authors use details to develop a main idea. The doing depends on the knowing. They work together. A skill, in Pulse’s ontology, is what shows up in that moment: the ability to bring procedure and knowledge together to get a job done. Skills aren’t lists of facts a student has been taught. They aren’t topics a student has covered. They are what a student can actually do when it is time to do it. That is why, when the proficiency model claims that a student is proficient in a skill, the claim is about what the student can do, not which lessons they have sat through.
Built around how learning develops in each subject
A common question we get from district leaders is whether one ontology covers all subjects. The short answer is no, because the way learning unfolds in mathematics and in ELA is fundamentally different, and an ontology that papered over those differences would compromise both.
Math builds in conceptual progressions. Certain skills function as prerequisites for others. A student who hasn’t internalized place value will struggle with multi-digit operations. A student who hasn’t understood fractions as quantities will struggle with conceptualizing and using rational–numbers effectively. The ontology for mathematics has to capture these order-of-development relationships, because instructional recommendations depend on knowing what a student is ready for, which in turn depends on knowing what foundations are already in place. In practice, this means the math ontology has more chain-like structure: clusters of densely connected skills, with prerequisite relationships running both within and between them.
Reading and writing develop across multiple dimensions. Decoding, language comprehension, knowledge building, and writing all grow over time, often unevenly and in interaction, as students engage with increasingly complex texts and tasks. A student doesn’t finish learning to decode and then move on to comprehension. Both continue to develop, supporting and shaping one another as students read more challenging texts. The ontology for ELA has to capture this multidimensional development, which means it looks less like a chain and more like a web: multiple dimensions growing alongside one another, with the right instructional move depending on the combination of signals across them.
The point isn’t that one model is better than the other. The point is that the model has to match the discipline. A math recommendation rooted in a chain-like understanding of prerequisite skills and an ELA recommendation rooted in a web-like understanding of multidimensional development are both more credible than a recommendation produced by a system that treats every subject identically.
The Pulse series
This article is the second in a series. See the full list below:
- Post 3: How Pulse’s proficiency models turn assessment data into meaning teachers can act on.
- Post 4: How insights become specific, classroom-ready instructional recommendations.
- Post 5: The science behind mathematical understanding, and what makes math recommendations distinct.
- Post 6: The science behind reading and writing development, and what makes ELA recommendations distinct.
***
See how it all comes together. Explore HMH Performance Suite and discover how Pulse transforms assessment data into real-time insights that support smarter, more connected instruction.
Turn data into timely, targeted instruction for every student with our free guide.
