Calibrated responses
Each grade-calibrated booklet presents passages and questions engineered to load specific cognitive operations, so that responses carry signal about the process, not just accuracy.
The science
Measuring the process, not the performance.
A reading test tells you whether a student got the answer. ARIA tells you how they arrived at it, and where, if anywhere, their understanding gave way. That difference is the whole of the science.
The premise
Comprehension is a system, not a score.
When a person reads, several things happen at once. They hold earlier words in mind while new ones arrive. They connect ideas across sentences. They build meaning from vocabulary and structure, assembling a whole from parts. When comprehension fails, it fails somewhere specific in that system, and a single score can never tell you where.
ARIA is grounded in Cognitive Load Theory, the well-established science of how working memory manages information under demand. Rather than asking only whether a reader got the answer right, ARIA is designed to reveal how the answer was reached, and where the process broke down.
A traditional test tells you a reader is struggling. A cognitive profile tells you which part of the system, and what to do about it.
How it works
01
02
03
From a paper booklet to a cognitive profile.
One comparable scale
Each response resolves into a structured profile of how the reader understood the text, then onto a single 100 to 800 scale that means the same thing across grades, schools, and years.
Credibility safeguards
A built-in validity layer flags profiles where the response pattern itself is unreliable, so that every reported score can be trusted as a measurement.
The specific cognitive model and scoring method are proprietary. What is shared here is the scientific approach, not the algorithm.
The finding
The Comprehension Cliff.
Across every institution measured so far, the same shape appears. The share of students reading at grade level holds through the early grades, then falls sharply across the transition years, toward almost none by Grade 8. We call it the Comprehension Cliff.
Based on the schools we have measured. Gujarat sample, 2026. Stated as a specific finding, not a national claim.
A Tier-3 school, share of students at grade level, by grade
A metro school, share of students at grade level, by grade
57 to 0%
At one Tier-3 school, the share of students reading at grade level fell from 57% in Grade 3 to none at all by Grade 8.
74 to 0%
A well-resourced metro school recorded 74% at grade level in Grade 5, and zero by Grade 8. Resources delayed the fall. They did not prevent it.
Every school
The same shape recurs in every institution measured, regardless of size, fees, or location. That consistency is the finding.
Why it stays hidden
Conventional testing never sees it coming.
In nearly every case, the students whose comprehension was quietly collapsing kept posting unremarkable scores on conventional reading measures. The cliff is invisible to surface assessment, because surface assessment measures speed and accuracy, not understanding.
This is why the cliff matters. It is not that schools are failing to notice a visible problem. It is that the problem is structurally invisible to the tools they have. ARIA was built to make it visible while there is still time to act.
Why the data matters →
What it produces
A record built to a standard.
Because every profile sits on one universal scale and is captured under standardised conditions, ARIA data is comparable across individuals, cohorts, grades, and years. The result is a structured, longitudinal record of cognitive development that is ordinarily impossible to assemble at scale.
We are interested in research collaboration, validation studies, and partnerships with teams who take measurement seriously. If that is you, the conversation starts below.
Talk to us →Measurement first. Everything else follows.
Once you can see the process of understanding, intervention stops being guesswork. That is the order Agnira is built around. Measure precisely, then act.