Never Leave a Blank: The Arithmetic
There is no penalty for a wrong answer in this section. An empty drop-down is a mark handed back, and the arithmetic of guessing well is worth knowing before test day.
The rule, and what follows from it
CELPIP Reading does not deduct marks for wrong answers. Your score is the number of questions you answered correctly out of 38, and a wrong answer and a blank are worth exactly the same: nothing.
That asymmetry is the whole lesson. A wrong answer costs you nothing you had. A blank costs you the chance of a mark you could have had for the price of one click. There is no situation anywhere in this section in which leaving a drop-down untouched is the better choice.
Candidates still leave blanks, and almost always for one of two reasons. The first is habit from tests that do penalise guessing, where the arithmetic is genuinely different. The second is running out of time with the intention of coming back, and then not coming back. The first is fixed by knowing the rule; the second is fixed by a routine, which is most of this lesson.
The arithmetic of a guess
Every question in Reading offers four options, except Part 3, where each statement offers five. A completely uninformed guess on a four-option question therefore returns one mark in four on average, and on a Part 3 statement one in five.
Across a handful of guesses that averages out to real marks. Four uninformed guesses at four options is one expected mark. Eight is two. Given that the bands near the top of the scale are two questions wide, that is not a rounding error, it is potentially a band.
Informed guessing does considerably better, and it improves faster than people expect. Eliminating one option of four takes the return from one in four to one in three. Eliminating two takes it to one in two. You do not need to solve a question to double its value, you only need to kill two options, and killing two options is often possible in ten seconds when solving would take ninety.
This is why the instruction is never simply to guess. It is to guess last, having spent whatever seconds you had on elimination rather than on solution.
What a guess should be based on
Eliminate the overstated option. If one option contains all, never, always, proves or decided and the text was hedged, it is very likely wrong. This is the cheapest elimination available and it costs about two seconds.
Eliminate the option about the wrong subject. Name the person, the paragraph or the column the question is about, then discard any option that is about a different one. In Parts 2 and 4 this frequently removes two options on its own.
Eliminate the two-clause option with a bad clause. If you can check half an option and the half fails, the whole option fails. Half an option is often checkable in the time available when the whole question is not.
Prefer the cautious option. When nothing else separates two survivors, the one making the smaller claim is the better bet, because overstatement is a large distractor family and its mirror is small.
Do not prefer the longest option, or the one with unfamiliar vocabulary. These are folk rules from other tests and they have no basis here. Length in these options is a function of the fact being described, not of correctness.
Never leave a drop-down empty, even on a question you intend to return to. Put in your best current option first and improve it later, because the return often does not happen.
The thirty second routine
Build a fixed end-of-part routine and run it in every part, including the ones that go well. It takes about thirty seconds and it is the most reliable mark-saving habit in the section.
When roughly thirty seconds remain, stop answering and start filling. Go through every drop-down in the part and make sure none is empty. Do not think about them yet, just fill them with the best option you can assess in two seconds each. Four empty boxes take about ten seconds to fill and convert a guaranteed zero into roughly one expected mark.
Then, with whatever remains, go back to the question you were closest to solving, not the first one you skipped. Order is not a virtue. A question you had narrowed to two is worth several times one you have not read.
Run this routine in practice as well as on test day. It has to be automatic, because the moment it is needed is the moment you are least able to plan.
With thirty seconds left in a part, stop solving and fill every empty box. Ten seconds of filling beats ninety seconds on one unresolved question, and it is the same ten seconds in every part.
Marking and returning, safely
Skipping a hard question is good practice. Skipping it without a system is how blanks happen.
The safe version is to never leave a question empty even when you intend to return. Put an answer in, any answer, then move. If you come back you improve it, and if you do not you have lost nothing. The habit costs one click and removes the entire failure mode.
Be realistic about how often returning actually happens. Inside a part with a clock this tight, the questions you skip are usually the ones there is no time to revisit, and the parts that overrun are exactly the parts with skipped questions in them. Plan as though you will not return, and treat any return as a bonus.
One place where returning is genuinely likely is Part 3, where each statement you resolve narrows what is left in the four paragraphs. A statement parked at 40 seconds really is cheaper to solve after the others. Even there, fill it before you park it.
Why two marks matter more than they look
It is tempting to treat one or two guessed questions as noise. The conversion from raw score to level says otherwise, and it is worth seeing the numbers.
The bands are not evenly spaced. Near the top they are narrow: 31 or 32 correct is CLB 9, and 28 to 30 is CLB 8. So at the level most immigration applications care about, the gap between two bands can be a single question. Lower down the bands are wider, and 24 to 27 all sit at CLB 7, so three questions there move nothing.
The practical consequence is that guessing discipline is worth most to the candidates closest to their target, which is to say most candidates sitting the test. If you are two questions from the band you need, filling four blanks is, on average, the band.
It also argues for aiming a couple of questions above your target rather than exactly at it. A passage that goes badly on the day is a normal event, and a margin absorbs it where a precise aim does not.
Spend your last seconds on the question you had narrowed to two options, not the first one you skipped. Eliminating two of four doubles a question's expected value, and that is usually achievable when solving is not.
Practise it
Five questions on the arithmetic of guessing and the routine that stops blanks happening.