Episodic Memory
Twin or New Game
Also called: same, similar or new memory game, drawing memory game for older adults, pattern separation game, mnemonic discrimination game
Measure memory for detail in adults aged 50 and over: the player sees drawings of objects one at a time and says whether each is the same as one seen before, a look-alike of one seen before, or new, scored as how well look-alikes are told from new drawings.
5-7 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 195
- Evidence grade
- C
- Registered clinical studies
- 62
Figures for the Mnemonic Similarity Task, the published test this game is compared with.
How these are countedWhat it measures
A drawing memory game built for repeated visits in research with adults aged 50 and over. The player sees drawings of everyday objects one at a time and says for each whether it is the same as one seen before, a look-alike of one seen before, or new. Scores are raw research measures with no percentile.
Constructs
- Mnemonic discrimination
- Pattern separation
- Recognition memory
Research uses
- Research on memory for detail in older adults
How the paradigm works
The Twin or New Game is a drawing memory game built for repeated visits in research with adults aged 50 and over. The player sees 128 drawings of everyday objects in one stream, each for 2 seconds and then gone: 64 are shown for the first time, 20 are shown again exactly, and 44 are shown as look-alikes of a drawing seen earlier, a little different in one way. For each drawing the player taps Same, Look-alike or New, with no time limit and no feedback. The paper counts and grades on this page are those of the Mnemonic Similarity Task, the published test the game is compared with, not the game's own.
More about the paradigm
Each session takes one of 6 alternate forms. The order of the forms is drawn once for each participant, and each session takes the next form after the sessions that participant began before it, so a reload keeps its form, a missed visit skips none, and no form comes back until all 6 are used. A first session opens with a practice of 8 drawings that names the right answer each time, once more if fewer than 6 are right; later sessions open with a reminder.
Keep each participant on one device class and one input at every visit; every session records its device, so a change shows in the export. Add one practice session first, before the baseline: in the source task only the first of four weekly sessions was lower (Stark and colleagues 2023). It is kept and reported, never dropped. For the retest, choose two later sessions 7 to 28 days apart in the study view, because its default pair, visits 1 and 2, starts with the practice session.
The drawings are made in code from stated numbers, with no picture file and none of the Mnemonic Similarity Task's pictures: 12 kinds of object in 8 colours, each with 3 stripes and 3 spots, so no count can be read to answer. A look-alike keeps the kind and colour and moves one dimension by 1 level (close) or 2 levels (wider): the width of the stripes, the size of the spots or the width of the object, each level 15% of the middle value. The steps are our own choice and have not been matched to how people rate similarity, so the record reports the look-alike answers by step.
The game follows the short continuous form of the Mnemonic Similarity Task (64 first showings, 20 repeats and 44 look-alikes, 2-second pictures, and the answers same, similar or new), which kept r = .69 with the full task in 50 older adults in 5 to 6 minutes (Stark and colleagues 2023). The main score is how well look-alikes are told from new drawings: the share of look-alikes answered Look-alike minus the share of new drawings answered Look-alike. The game has no norm of its own: its reliability and norms come from use. The 6 forms are used in turn; their scores are pooled only after first sessions show the forms are equally hard.
Scores
- Look-alikes Told From New Drawings%
- The share of look-alike drawings answered Look-alike, minus the share of new drawings answered Look-alike.
- Higher is better; compared within a study
- Same Drawings Told From New Drawings%
- The share of repeated drawings answered Same, minus the share of new drawings answered Same.
- Higher is better; a low value can mean the drawings were not looked at
- Look-alike Answers, Close Look-alikes%
- The share of look-alikes one step from the drawing seen before that were answered Look-alike.
- Informational
- Look-alike Answers, Wider Look-alikes%
- The share of look-alikes two steps from the drawing seen before that were answered Look-alike.
- Informational; shows whether two steps are easier than one
- Usable Trialstrials
- Trials counted in the scores. Left out: a trial during which the screen was left, the window changed size, or a second finger or a palm touched the screen.
- Quality check
Use and evidence
195 papers from 2021 to 2025 name this task in their title or abstract, up 40% compared with 2016 to 2020. That ranks it 155 of 582 test families in our census. These counts belong to the Mnemonic Similarity Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 54 |
| 2024 | 41 |
| 2023 | 36 |
| 2022 | 32 |
| 2021 | 32 |
| 2020 | 36 |
| 2019 | 25 |
| 2018 | 34 |
| 2017 | 31 |
| 2016 | 13 |
| 2015 | 18 |
| 2014 | 19 |
| 2013 | 8 |
| 2012 | 5 |
| 2011 | 0 |
| 2010 | 1 |
| 2009 | 0 |
| 2008 | 2 |
| 2007 | 1 |
| 2006 | 1 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade C
100 to 299 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 4
- Papers on reliability or validity
- 21
- Registered clinical studies using it as an outcome
- 62
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on the Mnemonic Similarity Task, the published test this game is compared with, not the game's own.
Source: ConductCognition test census 0.39.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Drawings
- Drawn in code, with no picture file: 12 kinds of object (a vase, a mug, a lamp, a bottle, a teapot, a bag, a hat, a boot, a clock, a kite, a jar and a watering can) in 8 colors, each with 3 stripes and 3 spots, so no count can be read to answer. A practice uses a thirteenth kind, a flowerpot. No drawing is a picture of the Mnemonic Similarity Task.
- Look-alikes
- A look-alike keeps its object's kind and color and moves one of three things by 1 step (close) or 2 steps (wider): the width of the stripes, the size of the spots or the width of the object. A step is 15% of the middle value. The steps are our own and are not matched to ratings of similarity, so the look-alike answers are reported by step.
- Forms
- 6 alternate forms of 128 drawings: 64 shown for the first time, 20 shown again exactly and 44 shown as look-alikes of a drawing shown earlier, 22 close and 22 wider. In every form the 64 first showings hold each kind 5 or 6 times and each color 8 times, and no drawing is in two forms. Each person gets the forms in an order drawn once for them, and each session takes the next form.
- Stream
- One continuous stream: each first showing is followed once, by its repeat or its look-alike, half at a short gap (4 to 11 drawings later) and half at a long one (20 to 40).
- Trial
- Each drawing shows for 2,000 ms and is then gone. Three buttons, Same, Look-alike and New, open when the drawing appears and stay until one is tapped. There is no time limit and no feedback, and the next drawing comes 500 ms after the answer or after the drawing's end, whichever is later.
- Practice
- In the first session only: 8 drawings (4 new, 2 repeats and 2 look-alikes) with a tick or a cross and the right answer named, once more if fewer than 6 are right. Later sessions open with a short reminder.
- Drawings and buttons
- Every drawing is at least 160 px and every button at least 64 px tall, laid out from the window before each trial. The first tap on a button answers, and a later tap changes nothing. Older-adult mode makes the words 1.2 times larger and the three buttons up to 1.2 times taller, and changes no drawing, timing or score. The game look adds a soft bar along each edge of the screen and the plain look has none. The drawing, its place and the white around it are the same in both looks.
- Schedule
- Built for repeated visits: add it to a study's visits and each session takes the next of the 6 forms. Add one practice session first and keep it in the results: in the source task only the first of four weekly sessions was lower. For a retest, choose two later sessions 7 to 28 days apart in the study view, because its default pair starts with the practice session. Keep each participant on one device class and one input.
- Scores
- How well look-alikes are told from new drawings (the main score: the share of look-alikes answered Look-alike minus the share of new drawings answered Look-alike), how well repeats are told from new drawings, the look-alike answers for close and for wider look-alikes, and the number of usable drawings. A drawing is left out when the screen was left, the window changed size, or a second finger or a palm touched the screen.
The design follows the short continuous form of the Mnemonic Similarity Task (Stark and colleagues 2023); the drawings, steps, forms, buttons and looks are our own. This game is built for repeated visits and is a new measure for research with adults aged 50 and over: its reliability and norms come from use, the figures published in those studies are not its own, and it computes no percentiles.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: twin_or_new_game test answers, one of three for each drawing (Same, Look-alike or New), leaving out a trial with the tab hidden, the window resized, or a second finger or a palm on the screen, or with no answer. Fewer than 96 scored drawings of 128, or answers under 200 ms on at least half the trials, invalidates. One answer on 90% or more of the answers is flagged, as is a recognition score under 0.5. A hidden tab withholds the percentile.
Why these rules
The rates need answers: the look-alike rate rests on 44 look-alikes and 64 first showings, so a session with fewer than 96 of its 128 drawings scored is too thin to give them (a ConductScience choice: three quarters of the stream). An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the drawing; following response-time effort checks for untimed tests (Wise and Kong 2005), the session is invalid when half or more of the answers are that fast. One answer on 90% or more of the drawings means the drawings were not compared. Accuracy is not judged against chance, because a low look-alike rate can be a real result in older adults. The recognition score, the share of repeats answered Same minus the share of new drawings answered Same, is flagged for review under 0.5, the value the source used to screen engagement: a person who cannot tell a repeat from a new drawing may not have attended. The 96-trial floor and the 90% limit are ConductScience choices, and the 0.5 flag is the source's own engagement cut. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Look | Game | Game, Plain |
Every plan can change the look, which never changes how a result is scored.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Twin or New Game in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Stark and colleagues, 2013), with the standard settings (look: Game). The main score was look-alikes told from new drawings (the share of look-alike drawings answered Look-alike, minus the share of new drawings answered Look-alike).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Stark, S. M., Yassa, M. A., Lacy, J. W., & Stark, C. E. L. (2013). A task to assess behavioral pattern separation (BPS) in humans: Data from healthy aging and mild cognitive impairment. Neuropsychologia, 51(12), 2442-2449. https://doi.org/10.1016/j.neuropsychologia.2012.12.014The task this game follows: old, similar or new answers to pictures, scored as the share of look-alikes called similar minus the share of new pictures called similar; scores fell with age from 20 to 89 (abstract read).
- Stark, C. E. L., Noche, J. A., Ebersberger, J. R., Mayer, L., & Stark, S. M. (2023). Optimizing the mnemonic similarity task for efficient, widespread use. Frontiers in Behavioral Neuroscience, 17, 1080366. https://doi.org/10.3389/fnbeh.2023.1080366The short continuous form: 64 first showings, 20 repeats and 44 look-alikes, 2-second pictures and the answers same, similar or new, in 5 to 6 minutes. In 50 older adults it correlated r = .69 with the full task, and an old or new answer in place of the three answers weakened that agreement (methods read in full).
Source papers
Questions
How long does the Twin or New Game take?
About 5 to 7 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Look-alikes Told From New Drawings, Same Drawings Told From New Drawings, Look-alike Answers, Close Look-alikes, Look-alike Answers, Wider Look-alikes and Usable Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Mnemonic Similarity Task used in research?
These counts belong to the Mnemonic Similarity Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. 195 papers from 2021 to 2025 name the task in their title or abstract, up 40% compared with 2016 to 2020. That ranks it 155 of 582 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.39.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Mnemonic Similarity Task have?
These counts belong to the Mnemonic Similarity Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. In the published research, 4 papers report the task's test-retest reliability and 62 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Look (Game, Plain). Every plan can change the look, which never changes how a result is scored.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.