Inference: Certainty, Likelihood and Missing Evidence
Separate certain conclusions from model-based likelihood and unresolved information, using a five-option response rule and checkable original examples.
This complete public guide is identical for everyone and available before answering. Your separate private page record shows correct, wrong and unanswered out of six. No percentile, pass band or weakest ability is inferred from these unequated parts.
Understand this original rule
- This part uses five choices: True, Probably true, Insufficient data, Probably false and False. In this original edition, certain means the supplied facts settle the claim, while probable choices use an explicit sampling model or probability supplied in the prompt. The bank does not ask you to import an unstated social stereotype or personal impression to estimate likelihood. Read the rule before applying a familiar label from another practice test.
- The distinction is a deliberate boundary, not a reconstruction of a publisher’s scoring system. Historical Watson–Glaser practice instructions illustrate five response categories, but the scenarios, probability conventions and short six-item allocation here are independently authored. An employer invitation may specify different instructions or forms. Use this page to practise matching evidence to a claim, not to infer what score you would receive on a named commercial assessment.
Identify a certain conclusion
- A conclusion is certain within the exercise when every situation allowed by the supplied facts supports it. If a complete list says every tagged package was weighed and package K is tagged, K was weighed. You need not know whether the statement describes a real warehouse; the exercise tells you to use it as given. Keep the conclusion within that list instead of extending it to packages never mentioned.
- A claim can also be certainly false when it directly contradicts a supplied fact or a necessary consequence. If an actual tagged, weighed package exists, “no tagged package was weighed” is false. Notice the existence statement: a universal rule without any instance does not automatically establish that an object exists. Write the small chain supporting your decision rather than selecting True because the statement sounds ordinary or reassuring.
Use the declared sampling model
- For a fresh model, imagine a closed bag with six circles and four squares, one token selected uniformly. “The next token is a circle” has probability six tenths, so it is more likely true than false, but not certain. A square remains possible. Uniform selection is an essential stated assumption; without it, merely knowing the inventory would not necessarily determine how a person chooses a token.
- Likewise, if a stated model gives a proposed event probability one quarter, Probably false is the bounded choice: the event is less likely than its complement, yet possible. Do not replace a known probability with certainty because the event happened several times in your imagination. The fictional probabilities here are task inputs, not evidence about actual device reliability, employee behaviour or the future performance of a real organisation.
Leave missing information unresolved
- Insufficient data means the passage supplies no justified direction for this particular inference. A note that several teams held meetings does not establish whether a named person attended when no person or attendance information is given. Attendance and non-attendance both fit. Missing information should remain missing; a plausible story about what usually happens is not a fact supplied by the exercise.
- Ask exactly what extra information would change the classification: an attendance list, a complete inventory, an operating duration or a defined selection process. Naming that missing input is a useful reasoning skill. It prevents a common shortcut in which unfamiliar claims are labelled probably false merely because they are unfamiliar. Your own confidence and the evidence’s ability to resolve a claim are related questions, but they are not interchangeable.
Check the scope before the label
- Underline words that change strength: every, some, none, next, always and exactly. A statement about one observed event may support “at least one” without supporting “every event”. A known chance for one defined draw is not a guarantee about every future draw. Copy the smallest evidence phrase that supports the scope of your proposed answer, then check whether you silently added a wider population or time period.
- Try changing just one word in a fresh claim. Replace “some recorded items” with “all items ever produced”, or “more likely” with “certain”. The required evidence often changes sharply. This exercise is not about choosing cautious language automatically: an explicit contradiction deserves False, and an explicit fact can deserve True. Good restraint means matching the strength of the evidence, including when that evidence is decisive.
Use the shared clock thoughtfully
- Timed mode uses one site-selected thirty-minute clock for all thirty questions across five parts. Moving between parts, reading their rules or returning to an earlier item does not pause it. There is no separate inference timer. Six minutes per part is a rough planning division, not a measured target, equal-difficulty claim or instruction to spend the same amount on every question. Untimed study uses the same bank without a deadline.
- You can explicitly leave an item unanswered and revisit it before finishing. At the deadline, the whole attempt closes and blanks remain separate from selected wrong answers. A detected device-clock inconsistency clears the attempt rather than producing a timed result. These controls support local practice, not invigilation. Follow the exact instructions in an actual invitation for navigation, permitted help, equipment and accessibility arrangements.
Review the reason behind each choice
- After finishing, inspect the explanation even for a correct choice. Was your answer supported by an explicit fact, a probability comparison or the absence of a justified direction? Build a new small example for each label and explain why the neighbouring labels do not fit. This is more informative than memorising option letters. A correct selection alone cannot reveal whether you reasoned, guessed or recognised a previous question.
- The record shows correct, wrong and unanswered out of six, with no calibrated inference level or percentile. Five-option items differ from the two-option parts, and difficulty is not equated. Do not diagnose a weakest ability from the smallest count or call an individual count proof of guessing. This full public guide is the same for everyone; sharing its link shares learning content rather than private responses or results.