Fluency Lab helps you collect, review, and respond to oral reading. Students record themselves reading an assigned passage; you review the recording with the help of AI-assisted error detection and a prosody rubric, adjust anything that needs your judgment, and release feedback. Class and student reports let you watch growth over time.

Use Fluency Lab as an instructional data point — not as a standalone high-stakes assessment.

Fluency Lab is:

Fluency Lab is not:

Fluency Lab supports two read types — a cold read (with a short comprehension check) and a warm read (fluency-only). For details on when to use each and how they fit into a chapter, see the Cold Reads and Warm Reads and Where Fluency Lab Fits in Instruction sections below.

After a student records and submits, Fluency Lab produces feedback for you to review. It includes:

The three scores measure different things — and why you need all three

Each 0–100 score answers a different question about the same reading:

Why all three are needed. A student can be strong on one and weak on another, and each gap points to a different instructional next step:

If the three scores were redundant, these profiles couldn't happen. The fact that a student can look very different across them is exactly why we report them separately.

A note on overlap. These are distinct measures, but they are not perfectly independent — reading is one connected act, and the research they're based on treats accuracy, automaticity, and prosody as related components of fluency (Rasinski, 2004). Where a behavior reasonably bears on more than one score, it does — but each score has a single primary focus. The clearest example is pace: Fluency uses pace as a broad check on whether the overall rate sits in a reasonable range, while Prosody's Pace dimension judges whether the rate and pausing sound natural and support meaning. Related questions, deliberately kept distinct.

Core rule: Scores update as you confirm/dismiss errors and adjust rubric inputs, so your review directly affects the final result the student sees.

How Accuracy is scored — and how we handle insertions and substitutions

Accuracy uses the standard Oral Reading Fluency (ORF) formula:

Accuracy = (total words − total errors) ÷ total words × 100

This matches what teachers are familiar with from running records, DIBELS, and AIMSweb. Every error type counts equally as one error — there is no weighting between error types.

How insertions and substitutions are handled:

Open a submission to review it in one of two modes. The mode is set when the product loads the review (AI-assisted by default; teacher-scored when configured for a no-AI experience).

AI-Assisted mode

The app pre-populates likely reading errors and prosody suggestions. To complete your review:

For full submission in AI-Assisted mode, every filtered error must receive a verdict (confirm or dismiss) and all four prosody dimensions must be filled.

Note: Changing sensitivity after you have made edits can reset confirmations, dismissals, and annotations, so finalize sensitivity before deep review.

Teacher-Scored mode (no AI surface)

This mode is useful when teachers or districts want a no-AI review experience while still using Fluency Lab recording, playback, and reporting. See the Disabling AI section below.

In AI-Assisted mode, your job is to rate and correct what the AI proposes before it reaches the student:

As you confirm/dismiss and adjust rubric scores, WCPM, Fluency, Accuracy, and Prosody recalculate live so the final numbers reflect your decisions.

Fluency Lab's prosody rubric is built on a well-established, research-based instrument: the Multidimensional Fluency Scale (MDFS), developed by Zutell & Rasinski (1991) and presented by Rasinski (2004, Assessing Reading Fluency; also in The Fluent Reader). The working copy we built our rubric from is the reproduction published by Flow Reading Fluency (2019, www.FlowReadingFluency.com) under the name Multidimensional Prosody Scale. It rates oral reading on four dimensions — Expression & Volume, Phrasing, Fluidity (Smoothness), and Pace — each 1–4 points, for a total of 4–16.

This is also the scale Florida specifically calls for: the state's Secondary Foundations: A Focus on Fluency materials present the Multidimensional Fluency Scale as the prosody rubric for the B.E.S.T. ELA fluency benchmark (ELA.612.F.2.4). Our rubric uses the same instrument.

Fluency vs. prosody, and why we use Rasinski's scale

Fluency is not one thing — and prosody is one part of it. Reading researchers, including Rasinski, define reading fluency as multidimensional: it is made up of three related components (National Reading Panel, 2000; Rasinski, 2004, Assessing Reading Fluency):

Rasinski describes fluency as a bridge between word decoding and comprehension: at one end it connects to accuracy and automaticity in decoding; at the other end it connects to comprehension through prosody — the "expressive interpretation" of the text (Rasinski, 2004). In other words, prosody is the expressive component of fluency, and it is distinct from the rate/automaticity component. That distinction is exactly why Fluency Lab reports them as separate scores rather than one number:

This three-part framing is not unique to Rasinski — it is codified in state reading standards. Florida's B.E.S.T. ELA fluency benchmark (ELA.612.F.2.4), for example, defines fluent reading as reading "with accuracy, automaticity, and appropriate prosody or expression" — the same three components Fluency Lab reports as separate scores.

Why a rubric, and why this one. Because expression is subjective and hard to quantify, the field measures prosody with listening-based rubrics rather than lab software. Rasinski notes these rubrics "provide valid measurements of the third component of reading fluency – prosodic reading" and, "in the hands of knowledgeable teachers, provide valid and reliable information" on a student's expressive reading (Rasinski, 2004). He presents the Multidimensional Fluency Scale (developed with Jerry Zutell; Zutell & Rasinski, 1991) as the more precise, dimension-by-dimension option, rating oral reading on four dimensions — Expression & Volume, Phrasing, Smoothness, and Pace — each 1–4.

In a later interview he reiterates why a rubric is the practical choice:

"You can put kids on a spectrograph and do some analysis there, but … teachers [can] just listen to students read and rate their reading along some dimensions on a rubric. Actually, I have one. I developed it with my colleague, Jerry Zutell … It's a multidimensional fluency scale, which looks at four different dimensions of fluency … It's actually been found to be quite reliable and valid."

— Timothy Rasinski, An Interview with Timothy Rasinski, The Language Teacher (JALT), 2021

We chose the Multidimensional Fluency Scale as the baseline for Fluency Lab's prosody rubric because it is research-grounded, already familiar to teachers, diagnostic dimension-by-dimension, and practical to approximate from acoustic analysis. The sections below show the source scale, the adapted rubric we actually use, and exactly what we changed and why.The source scale (Multidimensional Fluency Scale)

Interpreting the total (source scale): Scores range 4–16. Totals below 8 may indicate the need for direct prosody instruction; totals of 8 and above indicate the student is reading with adequate prosody (per the Flow Reading Fluency, 2019 reproduction we built from).

Fluency Lab's adapted rubric (our chart)

This is the rubric as it actually appears in the tool. It keeps the four dimensions and the 1–4 / 16-point total, but the descriptors — especially Pace and Fluidity — are rewritten to match how Fluency Lab derives them from acoustic analysis.

You can accept the AI-suggested level for each dimension or override it. The four dimensions add to a total out of 16 — the same scale teachers already recognize from the MDFS.What we changed, and why

We deliberately kept the structure (four dimensions, 1–4 each, /16) so the rubric stays familiar and the total can be interpreted the traditional way. The changes below adapt a human, holistic scale to acoustic, AI-assisted scoring.

Note — two different "Pace" ideas: The Fluency score also has an internal pace component based on gross words-per-minute (an "ideal" band of roughly 80–160 WPM). That is separate from the Prosody rubric's Pace dimension, which reflects how the reading sounds (rate naturalness + pausing). They are complementary, not duplicates.

How the prosody score is produced

After you finish your review and submit, the student can view read-only feedback:

Students never see raw AI error details, dismissed errors, or AI prosody suggestions — only your final, approved result.

When students see their results

Whether and when a student can open their feedback is controlled by the "Show results to students" setting on the assignment (in Create/Edit Assignment, just under the start/due dates). The options are:

Note: With the default ("after the assignment is graded for all students"), if any assigned student hasn't completed the assignment or hasn't been graded yet, results stay hidden for the whole class. If you want each student to see feedback as soon as you finish reviewing their submission, choose "as students' assignments are graded."

How the assignment is recorded in the gradebook

When you create or edit a fluency assignment, you also choose its category — the dropdown just under the Start Date / Due Date fields and the Enable Grace Period option. The choices are:

This category controls how the fluency assignment is recorded in the gradebook (how it's grouped and weighted). Pick the category that matches how you want the reading to count — for example, Practice for low-stakes progress reads, or Quiz / Assignment when the reading should carry more weight. The reading is scored and reviewed the same way regardless of category; the category only affects gradebook treatment.

The score submitted to the student reflects your judgment, not raw AI output. Before releasing results:

Teacher judgment remains the final authority for all submitted feedback.

Fluency Lab can be used with no AI surface. In Teacher-Scored mode:

AI feedback can be turned off at the admin/district level. When it's disabled, Fluency Lab still runs fully — recording, playback, rubric scoring, and reporting all work — just without AI suggestions. The scoring mode is applied when the product loads the review.

Fluency Lab is used with Connections Reading passage sets spanning grades 6–12, with cold reads near the start of a chapter and warm reads near the end. It also includes the Additional Passage Library of Lexile-banded practice passages (see Where Fluency Lab Fits in Instruction below).

Student recordings are analyzed by two components:

Districts remain responsible for any required parental consent (e.g., COPPA for students under 13).

Data retention. Student recordings and results are retained for the school year so students and teachers can reference and replay them, and are deleted at the end of the school year.

[Confirm before publishing] Confirm where student recordings and results are stored within Perfection Learning's systems, and that end-of-school-year deletion is wired up to the actual audio files (not just database records). The retention policy itself is confirmed: retained for the school year for student/teacher reference, deleted at the end of the school year.

Fluency Lab does not use a generative AI / large language model to write feedback. It uses two automated analysis components:

Both produce suggestions that you review, confirm, override, or dismiss. Because they rely on automated speech analysis, they can be wrong in predictable ways — see the next section.

The sections below go beyond the core workflow above. They cover the AI's limitations, how Fluency Lab compares to traditional scoring, and the instructional use of results.

AI speech recognition flags potential errors and suggests prosody scores. Always listen to the recording and apply your judgment before releasing results. Below are the known failure modes (also surfaced in the in-app AI grading disclaimer).

Errors the AI may flag incorrectly (false positives)

Errors the AI may miss (false negatives)

Scoring notes to keep in mind

Bottom line: The score submitted to the student reflects your judgment, not raw AI output.

The takeaway: Fluency Lab reproduces the familiar four-dimension rubric and /16 total, but the numbers come from acoustic analysis plus your review. Treat it as an accelerator for teacher judgment, not a substitute for it.

Cold read (near the beginning of a chapter)

Use the cold read to understand how students approach the text before instruction has done most of the support work.

Warm read (near the end of a chapter)

In the current product flow, the warm read is fluency-only:

Because warm reads do not include comprehension questions, interpret them primarily as fluency evidence and combine them with classroom comprehension evidence.

A typical chapter sequence can be:

You can use Fluency Lab with the whole class, small groups, or selected students. For students receiving Tier 2 support, use it as one part of a larger support cycle alongside classroom evidence and existing MTSS processes.

Tip: Avoid over-testing. Keep the routine short and connected to instruction.

Passage sources: Connections Reading and the Additional Passage Library

Fluency Lab passages (available for grades 6–12) come from two sources:

Scoring is identical for both sources — Accuracy, Fluency, Prosody, and WCPM are computed the same way. Because Additional Library passages are chosen by reading level (not grade), they carry no grade label; combine their results with your own grade-level expectations (see the Scoring notes above).

Fluency Lab handles incomplete reads explicitly, while leaving the final call with you.

Use patterns across multiple passages and time points, not a single score.

Always combine Fluency Lab results with classroom evidence, existing screening data, and MTSS procedures.

Suggested classroom supports

These instructional moves can be used outside the scoring tool when the data indicate a need (they mirror evidence-based fluency practices: teacher/peer modeling, repeated reading, phrase-cued reading, echo/choral/partner reading, and text dissection).

From the main app, open Reports to review performance over time. Reports support class-level snapshots, student-level trend views, assignment/gradebook views, and band-based interpretation.

Performance bands (the color legend shown in Reports):

Note: The Reports legend labels these as numeric ranges with colors — it does not show named proficiency tiers in the product. (If you want tier names for your own communication you might call them Below / Approaching / Proficient / Advanced, but those names are not surfaced in the app.)

Reports display raw WCPM and class-average comparisons; the bands above apply to the 0–100 scores. (Percentile-vs-norm rankings are computed internally but are not surfaced in the current Reports build.)

Best practice: Use reports to identify trends and plan instruction — not to make placement decisions from a single attempt.

Use this quick sequence after each reviewed submission:

You may choose to share simple, teacher-approved progress patterns with students — for example, whether WCPM, accuracy, or prosody improved over time. Keep this reflective and instructional, not punitive.

Possible student reflection prompts:

Is Fluency Lab a benchmark or screener?
No. It's a progress-monitoring and instructional tool — teachers grade individual readings, but you shouldn't base a high-stakes decision such as placement or eligibility on a single attempt.

Why is the AI's WCPM lower than my hand count?
Fluency Lab's WCPM includes hesitation errors, not just miscues, so it can read lower than a traditional ORF count.

A correctly read word is flagged as an error — what do I do?
Use Play Error to listen, then dismiss it. Proper nouns, self-corrections, and natural pauses at punctuation are common false positives.

Can I turn the AI off?
Yes. Teacher-Scored mode hides AI suggestions, the sensitivity selector, and AI badges while keeping recording, playback, rubric scoring, and reporting.

Does Fluency Lab use ChatGPT / a generative AI model?
No. It uses Azure speech recognition for word accuracy and DisVoice acoustic analysis for prosody — not a generative language model.

Why are there two "Pace" ideas?
The Fluency score has an internal pace component based on gross WPM; the Prosody rubric's Pace dimension reflects acoustic rate naturalness and pausing. They're complementary.

A prosody dimension shows as missing — is that a bug?
No. When a measure can't be reliably derived (e.g., unreliable pitch tracking from noise), Fluency Lab shows it as missing rather than guessing. You can enter that dimension yourself.

Does the student see the errors I dismissed?
No. Students see only confirmed errors and approved scores.

Fluency Lab works best as an instructional feedback loop:

It should strengthen teacher decision-making, not replace it.