Eleven question types, one bank

Ask for the evidence you actually need

StudyDrome Exam Manager holds eleven question types in one shared bank. Six mark themselves the moment a candidate submits. Essay, Short Answer, File Upload and OSCE Station reach the grading queue. Likert Scale collects opinion and stays out of the scoring maths by design. Every type carries points, tags and an optional difficulty band.

Book a pilot

Which question types can an exam use?

Eleven types, defined by one enum and mirrored value-for-value between the API and the manager app. The names below are the ones on screen.

Type

What the candidate does

Marking

When it earns its place

Multiple Choice

Picks one option from several

Automatic, all or nothing

Broad coverage at scale. One best answer, and a near miss earns nothing

True/False

Accepts or rejects one claim

Automatic, all or nothing

Fast checks on a rule or a fact. Cheap to write, cheap to sit

Checkboxes

Picks every option that applies

Automatic, partial credit optional

When the answer is a set, such as every drug that interacts

Sequence

Puts steps in the right order

Automatic, partial credit by position

Procedures and protocols, where the order is the competence

Matching

Joins pairs across two lists

Automatic, partial credit by pair

Terms to definitions, findings to diagnoses, drugs to classes

Fill in the Blank

Types the missing word into the stem

Automatic

Recall you want produced, not recognised from a list

Short Answer

Writes a sentence or two

Grading queue

Reasoning too short for an essay and too open for a key

Essay

Writes an extended answer

Grading queue

Judgement, structure and argument, marked by a person

File Upload

Submits a file they produced

Grading queue

Work made outside the browser: a drawing, a scan, a spreadsheet

OSCE Station

Performs at a station, live

Examiner app, scored on a scale or a checklist

Clinical skill watched in person by an examiner

Likert Scale

Rates a statement on a fixed scale

Collected, and excluded from the scoring maths by design

Course feedback or self-rating beside an exam, with no effect on the mark

Six types mark themselves: Multiple Choice, True/False, Checkboxes, Sequence, Matching and Fill in the Blank. Four route to the grading queue. Likert collects opinion. One bank feeds both exam types, so written exams and OSCE exams draw from the same items, tags and skills.

For the craft of writing the stem and the options, read how to write flaw-free MCQs and how many distractors an MCQ should have.

How are the automatic types scored?

Every question carries a decimal point value. The value is guarded non-negative at the domain boundary, so a saved question always holds a real weight.

An account can also set a minimum point value that every question must meet. It is checked when a question is created, when it is edited, and again when a reviewer re-weights a published question. Set the minimum to 0 and no floor applies. The authoring screen calls the field "Mark point"; the bulk bar calls the action "Update Questions".

Partial credit is a toggle on the question itself. Turn it on and the auto-grader awards proportional credit on three types:

Type

How the credit is worked out

Checkboxes

Wrong selections are subtracted from right ones, then divided by the size of the correct set. The result is clamped at zero

Sequence

The share of items placed in the correct position

Matching

The share of pairs joined correctly

Multiple Choice and True/False stay all or nothing: one selection, one key. The formulas are published and fixed, so two markers of the same paper always agree.

How do you record how hard a question is?

Each question can carry a difficulty from 0 to 10. The editor calls the field "Difficulty Level". It groups the scale into four named bands, so an author picks a meaning rather than a number:

  • Easy, 0 to 2: from "0 - Trivial recall" to "2 - Simple comprehension".
  • Moderate, 3 to 5: from "3 - Applied knowledge" to "5 - Analysis required".
  • Hard, 6 to 8: from "6 - Multi-step reasoning" to "8 - Complex integration".
  • Very Hard, 9 to 10: "9 - Rare/atypical scenarios" and "10 - Expert-level synthesis".
  • Optional: difficulty is a label, not a requirement. A question can sit in the bank without one.

Who decides the difficulty, the author or the data?

Both, and they answer different questions. The division of labour is clean.

The author sets the intended difficulty while writing. That label is what a blueprint reads, and it is what difficulty-weighted scoring uses. It says how hard the item was meant to be.

The measured difficulty arrives later, from real candidate responses. Item analysis reports a p-value for each item after delivery, beside its discrimination and its distractor behaviour. That says how hard the item turned out to be.

So the band records intent and the p-value records the result. Reading the two side by side is how a bank gets better each year.

How do several questions share one case vignette?

A context item is a titled stem shared by many questions: a clinical vignette, a passage, a data set. It lives in a question bank, and each question links to it. Candidates read it in a side panel while they answer, so the case stays on screen.

A test can keep context-linked questions together when it shuffles, and that grouping is on by default. Randomisation then moves the case as one unit. A candidate never meets question three of a vignette before question one.

The structure is one level deep on purpose: one vignette, many questions. You will find the screens under "Context Items", either account-wide or inside the bank that owns the case.

How is an OSCE Station question scored?

A station is marked by an examiner against a scoring scale: a named, ordered set of levels with a label and a description each, such as Not done, Partially done, Done. You build them under "Scoring Scales".

  • One scale per station: the whole station is judged on a single scale. This is the default.
  • Or one scale per item: each checklist item gets the scale that fits it, chosen with a mode flag on the question.
  • Reusable: scales are account-level. Build a scale once, then use it across stations and cohorts.
  • Retired without loss: a scale you no longer use can be deactivated, so past papers keep the scale they were marked on.
  • Clinical skills: stations also carry skills from a per-account library, managed under "Clinical Skills". It is the classification axis medical schools use alongside tags.

What holds the eleven types together?

  • One enum defines every type, mirrored value-for-value between the API and the app. A type means the same thing on both sides.
  • All eleven live in the same bank, under the same tags, the same points and the same version history. Nothing is a separate product.
  • Likert items collect opinion and are excluded from the scoring maths by design, so a survey block never moves a candidate's mark.
  • The partial-credit formulas are published and fixed, so two markers of the same paper always agree.
  • Difficulty is the author's label. The empirical p-value is measured separately, in item analysis.
  • A context item is one vignette shared by many questions, one level deep, and it travels with them into the published paper.
  • Clinical skills sit in a flat library per account, so a station's skill means the same thing across every exam.

Who needs a bank this wide?

  • Medical and health-sciences schools, who blueprint written papers and OSCE stations from one bank.
  • Exam offices running mixed papers, where an essay and an MCQ sit on the same paper.
  • Course teams collecting feedback beside the exam, with the survey kept clear of the mark.

Frequently asked questions

How many question types does StudyDrome Exam Manager support?

Eleven, defined by one enum and mirrored value-for-value between the API and the manager app. They are Multiple Choice, True/False, Checkboxes, Sequence, Matching, Fill in the Blank, Short Answer, Essay, File Upload, OSCE Station and Likert Scale. Six mark themselves, four reach the grading queue, and Likert collects opinion. All eleven live in the same question bank.

Which question types mark themselves?

Six: Multiple Choice, True/False, Checkboxes, Sequence, Matching and Fill in the Blank. Multiple Choice and True/False are all or nothing, with one selection and one key. Checkboxes, Sequence and Matching can award partial credit when the toggle on the question is on. The rest of the paper reaches a grading queue for a person to mark.

How does partial credit work on a Checkboxes question?

Wrong selections are subtracted from right ones, then divided by the size of the correct set. The result is clamped at zero, so a guess-everything answer earns nothing rather than a penalty. Sequence scores the share of items in the correct position, and Matching the share of pairs joined correctly. The formulas are published and fixed.

What is a Likert Scale question for?

Opinion, gathered in the same tool as the exam. A Likert item asks a candidate to rate a statement on a fixed scale, and it is excluded from the scoring maths by design. That means a course-feedback block or a confidence self-rating can sit beside real questions, and it will never move a candidate's mark.

What does the difficulty level on a question do?

It records the author's intent. The scale runs from 0 to 10, shown on screen as four named bands. Every level carries a written descriptor, from "0 - Trivial recall" to "10 - Expert-level synthesis". The label feeds difficulty-weighted scoring and blueprint reporting. The measured p-value arrives separately, from real candidate responses, in item analysis.

Can several questions share one case vignette?

Yes. A context item is a titled stem, held in a bank and linked from many questions. Candidates read it in a side panel while they answer. A test can keep context-linked questions together when it shuffles, and that grouping is on by default. A case therefore reaches the candidate as one block, in the right order.

Book a pilot

Bring one mixed paper. We will build it from the eleven types, publish it, and show you what marks itself and what reaches the queue.

Book a pilot

Related