Graduate test format

Verbal reasoning test with your cannot-say errors counted separately

Read a short passage, judge three statements about it as true, false or cannot say, and do that six times inside nine minutes — free, no account, and the result appears the moment the clock stops. Where every other version of this test hands back a mark out of eighteen, this one hands back the table underneath it: how many statements you called false when the passage was merely silent, and how many you called cannot say when the passage had settled them. Those are opposite mistakes with opposite fixes, and a single total hides which one you make.

100% free
No signup
18 statements, 9 minutes
3 x 3 error table
Every answer explained

Six passages, three statements each, 9 m 00 s for all 18. Every statement gets one of three verdicts, and the whole difficulty of the format lives in the third one.

The rule you are judging under

True — the passage, taken as given, makes the statement so.
False — the passage contradicts the statement.
Cannot say — the passage neither. Silence is not denial: a statement the passage never touches is Cannot say, however implausible it sounds, and a statement the passage makes obvious is True, however surprising.

Treat every passage as true even where it conflicts with what you know. Bringing in an outside fact is the same error as bringing in an outside guess.

The worked statements run on a seventh passage that is not in the set, with the reasoning shown after each one and no countdown anywhere. Press 1, 2 or 3 for the three verdicts, or T, F and C, or use the buttons.

How to take the true / false / cannot-say test

Six passages, eighteen judgments, one clock across all of them.

Fix the meaning of the third verdict before you start
True means the passage as written makes the statement so. False means the passage contradicts it. Cannot say means neither — and it is the verdict that decides the score, because a passage being silent about a claim is not the passage denying it. Two worked statements on a seventh passage are offered first, with the reasoning and the deciding words shown after each, and neither of them touches the mark.
Judge each statement from the passage alone, keys 1, 2 and 3
Take every passage as true even where it clashes with something you know: an outside fact is as much a foreign import as an outside guess. Press 1 for true, 2 for false, 3 for cannot say, or T, F and C, or tap the three buttons. The passage stays on screen for all three of its statements, so the second and third are answered by someone who has already read it, which is what makes the reading overhead measurable at the end.
Read the table rather than the mark
The result opens with a grid of the verdict the passage supports against the verdict you gave. Everything off the diagonal is grouped into three families — silence filled in, judgment withheld, verdict reversed — because coaching for one of them is the opposite of coaching for another. Below it sits your median time on a passage's first statement against its later two, and a full review that quotes the words in each passage that decide the item.

Technical specifications

Statements	18 scored — three under each of six passages — plus 2 worked statements on a seventh passage that enters no figure
Verdicts	Three, so blind guessing averages 6.0 of 18. That floor matters when reading a percentage: 12 of 18 is 67% and exactly twice chance
Answer key balance	Six true, six false, six cannot say. The balance is fixed so the third verdict cannot be inferred from how often it has already come up, and the six are drawn in a fresh order every run
Time	9 minutes across the whole set, which is 30 seconds a statement including the reading. Spend it where you like; there is no per-passage allowance, and an answered statement does not come back
Passages	55 to 78 words each, 394 in total, written for this page. The registers are a bridge closure, lake monitoring, a payroll change, apprentice training, archive access and a road tender — four of the six turn on arithmetic the passage never performs
Headline output	A 3 x 3 table of asked against given, with the three off-diagonal families counted apart: True or False where the passage was silent, Cannot say where it was not, and True given for False or the reverse
Reading overhead	Median response time on the first statement under a passage against the median on its second and third. Same reader, same passage, once cold and twice warm — the difference is the read, and over six passages it is paid six times rather than eighteen
Reference figure	None, deliberately. A verbal reasoning test names a question format rather than a published instrument, so there is no population behind it to rank you in; the score is reported raw and the errors are reported by type

Frequently asked questions

Does cannot say mean the statement is probably false?

No — cannot say and false are opposite findings, and confusing them is the single largest source of lost marks in this format. False requires the passage to contradict the statement: something in the text has to be incompatible with it. Cannot say requires only that the text does not settle the matter either way, which is the position you are in for anything the author simply did not mention. A passage that lists three reasons a bridge closed has not denied a fourth, and a rule stated without its penalty has not told you what the penalty is.

One passage says something I know to be untrue. What do I do?

Judge it as though the passage were true, because that is the instruction the format runs on. These items are tests of whether you can hold a text at arm's length and read only what it commits to; the moment your own knowledge is admitted as evidence, two readers with different backgrounds get different keys and the item stops measuring anything. In practice this bites hardest on causal statements, where what you know about how the world usually works makes an unstated cause feel established.

Why is the clock across all six passages instead of one per passage?

Because that is where the real difficulty of the format sits, and a per-passage timer would remove it. A single budget forces you to decide how much a stubborn statement is worth relative to the four you have not read yet, and that allocation is most of what separates people on a timed verbal test. It also lets the arithmetic in the report work: minutes you spend reading are recoverable from the statement-level times only because you controlled where they went.

Is 12 out of 18 a good score?

Nothing on this page can tell you, and any page that answers that question with a number is inventing one. Three verdicts put blind guessing at 6.0, so 12 is twice chance, and that comparison is real. What is not available is a comparison with other people: the score depends entirely on how hard these particular eighteen statements are, so an average quoted for it would describe this page's difficulty rather than a population. Publishers of the commercial tests in this format do hold norms, and they sell them to licensed customers and cut them by role.

Why does the report break my errors into three groups?

Because the fix for one group makes the other worse. A reader who fills silences needs to become more literal and check that every element of a statement has a corresponding element in the text. A reader who withholds judgment needs the opposite — to accept that a fact assembled from two sentences is still a fact the passage gives you, and that cannot say is not a safe default. Told only that they scored 11 of 18, both readers do the same thing next time: read more slowly, which helps neither.

Some statements need arithmetic. Is that still verbal reasoning?

Yes, and the commercial tests in this format do it constantly. The arithmetic here is never harder than one division — 240 out of 330, six from a group of 90 against seven from a group of 240 — and its purpose is to make the statement follow from the passage without appearing in it. That is what separates comprehension from lookup: a statement you can confirm by matching a phrase tests reading speed, while a statement you have to derive tests whether you noticed that two figures in different sentences belong to each other.

Can I write notes, and can I go back to a statement?

Write whatever you like on paper; the page has no view into it and no rule about it. You cannot revisit a statement once answered, because that is how the computer-delivered versions of this test are built and because a set with free navigation measures a different thing — reviewing is a strategy in its own right, and mixing it in would make the time figures unreadable. If the tab goes to the background mid-statement, that statement is struck out rather than marked wrong, the countdown stops, and a replacement joins the end so the passage still receives three verdicts.

Where the three-verdict format came from, and what versions of it get wrong

The true / false / cannot-say item is not a comprehension question with an extra option bolted on. It descends from the critical-thinking appraisals built for graduate and professional selection, whose whole subject is the boundary of what a text supports, and the third verdict is the item. Take it away and you have a reading test: every statement is either matched in the passage or not, and a fast reader wins. Put it back and you are measuring something a fast reader can be very bad at — noticing that the passage which reported two facts in the same sentence never claimed one caused the other. That is why employers keep buying the format for roles where the cost of an error is a confident inference nobody checked, and why the named appraisal built around it stretches the same judgment across five points rather than three.

Two conventions vary between publishers and both change the answers, so a page using this format owes you its own. The first is what you may bring: this test asks you to treat the passage as true and to use nothing else, which makes several statements that are true in the world into cannot-say items. The second is how far a permitted inference may reach. Some instruments accept only what the text states, others accept what follows beyond reasonable doubt, and the gap between those two readings is exactly the set of statements that need one arithmetic step or one combination of two sentences. This page takes the second reading: a payroll of 240 plus 90 shrinking to 318 is a fall the passage gives you, even though no sentence contains 330. Where a version of this test never asks you to combine anything, it has quietly collapsed into the first reading and made cannot say the answer to almost everything.

The failure common to free versions of this test is not the items, it is the report. A mark out of twenty tells you nothing you can act on, because in a three-way format the wrong answers are informative and the total throws that information away. Eighteen statements produce six distinct ways to be wrong, and they fall into three families that want three different responses. So the table comes first here, and the mark second. If you want the same discipline applied to arguments rather than passages, the five sub-scale version drops the clock and reports a profile instead of a total; if you want to see how much of your score here was language at all, the figural contrast poses matched logical structures with no words in them. Two adjacent formats are worth a run for a different reason: the logical reasoning items replace the passage with bare premises, and the numerical set replaces it with a table, which is the same sift at three different levels of prose.

A reaction time here is the interval between the frame that painted the stimulus and the timestamp the browser attached to your key, both read from the same monotonic clock. What neither can see is the display pipeline behind it, so on a 60 Hz screen roughly 16 ms of every figure below is the machine rather than you. That is the timing floor: two numbers closer together than that are the same number, and this page reports no precision it cannot support.

Nothing on this page is reported at that resolution — the two medians it prints are in seconds — but the floor is the reason they are printed in seconds. The reading-overhead figure is a difference between two medians taken from the same clock on the same machine, so the display latency is in both of them and cancels; what does not cancel is your own decision about when you had finished reading, which is worth several seconds rather than several milliseconds.

This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish a reading difficulty, or a pass at the verbal stage of any real application. Only a qualified professional, working with more than a browser, can make that judgment.

Where the eighteen verdicts are kept

Every number on this page is worked out by JavaScript running in the tab you are reading it in. Your answers, your reaction times and your score are never uploaded, logged or kept — which is also why the test carries on working after you disconnect from the network, and why nothing here can be held back behind an email address.

The eighteen verdicts, their timestamps and the error table built from them exist as one JavaScript array in this tab and go no further. That array is why reloading loses a finished set, and why the copy button carries counts — how many silences were filled, how many judgments withheld — rather than which statements you were shown or how you answered them.