Inference in Reading: Supported, Not Guessed
Several questions in this section ask what a text implies rather than what it says. There is a line between a supported inference and a reasonable guess, and the exam sits exactly on it.
What an inference question is asking
Stems like the writer suggests, it can be inferred that, the writer implies and the writer's attitude is best described as are all asking the same thing: what must be true given what the text says, even though the text does not say it.
The operative word is must. An inference question is not asking what is plausible, what is likely, or what you would conclude if you were the person in the letter. It is asking what the text forces.
This is a narrower target than most candidates aim at, and that narrowness is the entire difficulty. Reading is a generative activity: you build a picture as you go, filling gaps with reasonable assumptions, and by the end of a 400 word letter your picture contains a great deal the letter never stated. An inference question asks you to separate the part of that picture the text put there from the part you supplied yourself.
The one-step rule
A supported inference in this section is almost always one step from the text. Not two. The text gives you a fact, and the inference is the immediate consequence of that fact, with nothing added.
She has started the letter four times since March. One step: writing this has been difficult for her. That is forced, because starting and abandoning something four times is a stated behaviour with an immediate reading.
Two steps: she is unhappy in her marriage. That may be true in the world of the letter, and the next paragraph may even reveal it, but it is not one step from four abandoned drafts. It requires an extra assumption you supplied, and if the letter does support it, it supports it through some other sentence you should be pointing at instead.
The practical test on test day: for the option you are about to choose, can you name the specific line that forces it, and does it follow without an extra assumption? If you need two lines and a premise, the option is very likely a distractor, because distractors in inference questions are built exactly that way.
What the text will not carry
Your own knowledge of how things work. A passage about a municipal budget does not entitle you to conclude what councils usually do. What is generally true is not what the text says, and options built from general plausibility are common precisely because they are so easy to accept.
Emotional filling. A letter mentions that the writer has not spoken to her sister since June. It is natural to conclude they have argued, and the letter has not said so. People stop speaking for many reasons and the exam knows this.
Causation from sequence. The text says the programme was cut and attendance fell. An option says attendance fell because the programme was cut. That is a second claim, and unless the text makes the link, it is not there.
Intention from action. Someone did something; an option says they did it in order to achieve something. Purpose is a separate claim from action, and a text that reports one has not necessarily reported the other.
For any inference answer, name the specific sentence that forces it before you commit. If you need two sentences and an assumption of your own, you are one step too far and looking at a distractor.
Where inference lives in each part
Part 1. The heaviest concentration in the section. Personal messages carry their meaning in what people avoid saying, in which topic gets one sentence and which gets a paragraph, and in tone shifts. This is why reading the letter properly once, rather than scanning, is the whole opening move of the part.
Part 2. Very little pure inference, but questions 6 to 8 ask why the email reached its conclusions, which is one step from the argument rather than from the chart. The evidence is the reasoning the writer showed, not a bullet in a column.
Part 3. Essentially none, and this is worth knowing because candidates import inference into it and lose marks. Part 3 asks where a fact is, not what follows from it. If you are reasoning rather than locating, you have misread the task, and the paragraph you are reasoning toward is usually the wrong one.
Part 4. Constant, and of a specific kind. The inferences are about people: whether a speaker would agree with a statement they never addressed, what a concession implies about their position, whether the commenter is defending or attacking. One step still applies, and the step is almost always through the ledger.
When two options both look supportable, choose the more cautious one. A weaker claim has fewer ways to exceed what the text actually established.
The would-they-agree question
Part 4 has a recurring stem type worth handling separately: X would most likely agree that.... It looks like pure speculation and it is not.
The correct approach is to treat the option as a statement and check it against the person's stated position and their stated reason. If the option follows from the reason they gave, it is supported even though they never said it. If it merely aligns with their side, it is not.
An example of the distinction. The critic opposes a lighting bylaw because the costing is unrealistic. Would they agree that the council should get firm supplier quotes before voting? Yes: that follows directly from their stated reason. Would they agree that the town has too many streetlights? No: that is their side's general view, and the ledger says their objection was about arithmetic.
This is also where the expert trap lives. Options of the form the expert would agree that the council should approve the plan are almost always wrong, because the expert's stated position is that they are not taking a side, and nothing follows from a refusal.
Training the distinction
Inference is not improved by reading more. It is improved by making your reasoning visible so you can see where you added something.
The drill is small and it works. For every inference question in a practice set, write the line number or the first four words of the sentence that forces your answer, before you check whether you were right. Then, when you check, look at two things: whether you were correct, and whether the line you named is the line the explanation names.
Three patterns show up in that record. If you were right but named no line, you are guessing well, which will not survive harder passages. If you named a line but needed an extra assumption to get from it to the answer, you are inferring two steps and will keep choosing the plausible distractor. If you named the wrong line, the problem is comprehension rather than inference, and it is a different fix.
One last habit for the day itself. When two options both seem supportable, the correct one is nearly always the weaker, more cautious claim. The exam does not reward boldness, and an option that says less is an option with fewer ways to be unsupported.
Do not import inference into Part 3. That part asks where a fact is, not what follows from it, and reasoning your way to a paragraph is how the forced-match trap catches people.
Practise it
Five questions on the line between what a text forces and what it merely allows.