Every TOEFL score, each section and the overall, sits between 1.0 and 6.0 in half-band steps. This guide explains how each section earns its band, how the four bands become an overall score, what the bands mean, and how they line up with the 0–120 scale they replaced. Everything here comes from ETS’s own pages, linked at the end.
The test has 120 items across the four sections and takes about 90 minutes of testing time; ETS says to allow about two hours in all. Reading and Listening are answered on screen and scored by machine. Speaking and Writing produce recordings and texts that ETS rates against its scoring guides. All four end up on the same 1–6 scale, which is what makes the overall average possible.
ETS built the 2026 Reading and Listening sections on a multistage adaptive design: the section adjusts to your performance as you go, so the questions you see later depend on how you handled the earlier ones. Your band is derived from how many questions you answered correctly and how demanding those questions were. That has two consequences worth knowing before test day.
What you can control is the same as on any test: answer every question, since a blank cannot score, and move on when a question is taking too long. The section guides for Reading and Listening cover each task type.
Spoken and written responses are not marked right or wrong. ETS scores them centrally, in its scoring network rather than at the test center, against the scoring guides for each task: how well the response does the job set by the prompt, how clear and well organised it is, and how accurate and varied the language is. Each section’s ratings are combined into one band from 1.0 to 6.0. Our sample answers for the email, the academic discussion and Listen and Repeat show what the levels look like in practice.
The overall score is the arithmetic mean of the four section bands, rounded to the nearest half band. ETS’s own example: an average of 5.25 reports as 5.5. Here is another. Reading 4.5, Listening 4.0, Speaking 3.5 and Writing 4.0 average 4, which rounds to an overall 4.0. Because the rounding step is half a band, small differences matter: averages of 4.74 and 4.76 report as different overall scores. The score calculator shows where any combination lands.
Keeping the overall on the same scale as the sections is the point of the design. Under the old system a total of 100 and a section score of 25 were numbers of different kinds; now a 5.0 overall and a 5.0 in Reading say the same thing about proficiency, and both read directly against the CEFR.
ETS aligns the scale to the Common European Framework of Reference and prints the level on the score report. The alignment is the same for every section and for the overall score.
For what each level means in day-to-day English, see TOEFL CEFR levels; for the requirements universities actually set, see what counts as a good TOEFL score. There is no failing score: every band is reportable, and institutions decide their own minimums.
ETS published a concordance from each band to the previous section scores (0–30) and total (0–120). The sections map differently: the same old score can sit on different bands depending on the section, because the old sections had different score distributions. For 2 years from January 21, 2026, reports also show a comparable overall score on the old scale, the midpoint of the total range for your band.
ETS publishes the scale, the concordance and the design, but not the arithmetic that turns answers into a band. What follows is our reconstruction from what ETS has released. It is reasoning, not a leak, and it is labelled as such wherever we use it.
The concordance table gives, for every band, the previous 0–30 section score it corresponds to. The previous test’s relationship between points earned and that 0–30 score was never published by ETS either, but it was stable from form to form and widely documented from ETS practice material: Reading earned a 24 (the floor of band 5) at roughly 33 of 45 points, an 18 (band 4) at about 27 of 45, and reached 0 below about 9 of 45; Listening earned a 22 (band 5) at about 28 of 34 points and a 17 (band 4) at about 24 of 34. Chaining the two gives the share of scored points that each band most plausibly represents on a set of mixed difficulty. Listening needs a higher share than Reading for the same band, which is the concordance talking: a 22 in Listening and a 24 in Reading were the same band.
Three corrections apply on the real test. First, adaptivity: a multistage section routes you to a harder or easier later stage from your early answers, and the harder stage carries more credit per correct answer, so a strong start followed by a good showing on hard material scores above the table while a weak start probably caps the band reachable from the easier stage. Second, the bottom of the scale is compressed: on four-option questions, guessing alone earns about a quarter of the points, which is why the lowest bands sit so close together. Third, the 2026 task types are new (Complete the Words, Build a Sentence and the rest), so the item mix is not the old one. Expect the table to be right to within about half a band in the middle of the scale and less certain at the ends.
Each spoken or written task is rated against its own scoring guide on a short scale (the email and academic discussion guides run from 0 to 5), and a section band comes from the combined task ratings through the cut scores ETS set in its standard-setting study. Reading those guides against the CEFR descriptors, a response that fully does the task with clear organisation and accurate, varied language sits at the top rating and the top bands; a response that does the task with noticeable errors that rarely obscure meaning sits around the middle ratings and bands 4 to 4.5; a response that is on topic but short, repetitive or hard to follow in places sits around bands 3 to 3.5. Because the mapping runs through cut scores on combined ratings, one weak task is diluted by the others, and one strong task does not carry a section.
Suppose a test taker earns about three quarters of the Reading points and about four fifths of the Listening points, both on a mixed-difficulty route, and their written and spoken responses are consistently “does the task with some errors”. The reconstruction puts Reading at band 5 (73–81%), Listening at band 5 (82–90%), and Speaking and Writing around 4.5. The overall is the average, 4.75, which rounds to 5.0, a C1 on the report, and about 95 on the retired scale.
Because ETS does not publish the calculation, no practice test can produce a TOEFL score. Our score reports and the TOEFL dashboard give an estimated band range for Reading and Listening only, using the reconstruction above: your accuracy on scored questions is read through the section’s thresholds, the range covers the statistical uncertainty of your sample and narrows as you answer more, and nothing is shown under ten scored answers. Speaking and Writing responses are kept for your own review against the scoring guide and a model answer; they get no number, because a number there would be a guess. Treat the estimate as a direction of travel, not a prediction of your report.
Applying to US graduate programs?
Most also require the GRE, which ETS administers alongside the TOEFL. Our GRE course covers Verbal and Quant in full, with adaptive practice tests and instant scoring.
Official pages checked for this guide. Last reviewed October 2026.