When learning same answer shows for all option

The last few days I have encountered a bug when using decks.memrise, when learning a new word it gives me the same answers for all the options.


I have already looked at other solutions that have been given in the past.
Such as this

I have also searched my current levels for any duplicate and only found one (which i added today, so the bug was occuring before then). Which I have deleted from the level.

I also searched my database for the last 5 days worth of words i’ve added for any duplicats and not found any. There is no simple way to search every word i’ve ever entered into my database as it is currently 17 pages long.

Are there any other fixes? Or am I going to have to be forced to either delete my entire database or move to a different service other than memrise?



No answer, I’m afraid but the same problem has started appearing now and again. On different courses. Also the ‘could not load’ message is showing up.

I have been experiencing the same problem. Very annoying.

I have fixed mine, but still annoying.

I went through my database (which have been building for a very long time). My format for it has always been. Kanji, English, Hiragana. So 3 columns.

However, for some reason memrise had decided to completely remove all the data in the Hiragana field for 15 pages worth of words.

I deleted all words from those 15 pages, and things are now working properly.

I have no idea why memrise felt the need to delete the ‘Hiragana’ words from my database, or why they feel the need to keep changing things. If things work fine, don’t change them. Change isn’t always ‘good’ or ‘innovative’. Anytime memrise has changed something in the last ~5 years that I’ve used it, it’s caused more problems and headaches than anything else.

1 Like

I thought for a while that this bug is related to the size of the course, ex. 500+ words, but it seems to happen with smaller sized courses as well. Usually you see five of the same (right one) and one additional one in the 3x3 grid. Same with web UI and Memrise app. Started about three weeks’ ago.

I’ve been experiencing the same problem with most of my courses, and some other people’s courses.

Has anyone been able to fix this?

It seems to come and go, the later entries I’m adding now seems to have a better spread. Anyway, it’s really up to the back-end software engineers to fix this regression bug.

Regressions are considered P1 bugs in most professional software companies.

1 Like

Same issue happening in my Swahili course. Just posting here in case the devs are interested.

When learning new words recently, the Swahili multiple choices almost always have the correct answer duplicated one or more times. When the multiple choices are on the English side this problem does not occur.

I have the same problem.
My database works the same as @Drar, so I checked the database but I don’t have any missing data.
My course is VERY long and I don’t have the time to re-build it before my exam.
Any support from Decks or Memrise or whoever is behind this???

What I wonder is, is there any professional reading at this? Or is it just users’ mutual support?

The problem really is probably linked to so called phantom entries (see i. e. this thread). That is, duplicate answers show up due to the fact that someone edited the course’s content, in this case this is probably mostly due to corrections/fixes.

The problem really is that Memrise seems to store whatever entries have been made in their database and presents these as answers to questions/tasks. However, once somebody updates/fixes stuff, those entries are not being removed and, consequently, they show up as possible answers. This again is probably connected to the way their matching system works, i. e. they try to present you with close matches to the right answer, probably using fuzzy word/sentence search (i. e. SoundEx). This would result in (nearly) exact matches being presented instead of other ones.

But that’s all just my personal assumption …

1 Like

That’s another bug. This is a new one we are discussing. You add a new word and it’s repeated as a possible response six times in the grid. It’s a regression. If this was Google or some other company project managers would sit in my office until it was fixed. But this is Memrise.

I’d agree with that - the ones in my own courses were definitely not edited, so are not phantom entries (it’s entirely possible that the mechanism is similar, though).
The fact that these show up on the web and the (unchanged) app version suggests it’s something to do with the underlying database or its indices.
Oh, and I’m sure Memrise is busy fixing things, like creating even better colour schemes and increasing gamification for beginner users :stuck_out_tongue_winking_eye:

you’re probably right
:sob: :sob: :sob:

Hi, sorry you’re all having issues with this, I’m digging into why this might be the case. We have changed some of the backend around how we generate distractors, as they should be generated much faster, but it is possible we introduced a bug while doing so.
Some questions that might help us reach the bottom of this:

  • Once this happens for an item, does it always continue happening for that Item? (for example, @Drar, I had a look at your course and am currently seeing different distractors for ‘examine’, see attached images)
  • Does this happen more frequently for recently added items?
  • If you have more examples of this happening, could you please link the course and the specific Item it happens on?
    Screenshot 2019-12-20 at 14.15.59

As I had wanted to continue studying but hadn’t received a response from any dev I instead purged my entire database other than the most recent words. Which fixed the issue - this is why it doesn’t give multiple correct answers anymore (and is also when I found our that for some reason an entire column had been removed by the system on ~15 pages worth of words in my database, as I mentioned further up)

Regarding your questions I can answer based on the week I experienced this.

1: It happened for every word that I added and tried to learn, regardless of if I tested on Kanji or English. However, when just ‘reviewing’ the word, it wouldn’t do it all the time, just sometimes. During the ‘learning’ phase, it was always happening.
2: It happened 100% of the time for any word I added. I was also the only one who added words to the course as I had my course closed to the public.
3: I only use my own created courses, so unsure on this.

For information based on what column was deleted… My database and courses are always setup as
Kanji - English - Hiragana

and I then test on kanji > English and english > kanji. Hiragana is just there for the learning/if I make a mistake with the kanji, so I can check the reading faster (rather than having to input it into a dictionary).

I have used this format for ~6? 7? years. However once the repeating correct answers started showing and I decided to delete the database to try and fix it, that is when I found the entire Hiragana column had been removed (the category was still there, just ever word was removed from that column).

what used to be;
勉強する to study べんきょうする

had randomly become;
勉強する to study

Once I had removed all of those broken words, the problems dissapeared, and have not arisen again.

Hope this information helps~

Thanks, it does help. I’ll definitely investigate this more, it’s particularly worrying that you lost data, I’m very sorry to hear that.

What I suspect might be happening, which I can’t confirm without more investigation, is that when adding new elements, we didn’t find enough related content to generate the possible answers, potentially due to the missing data.

Has anyone else lost data from columns in their courses?


It started about six weeks ago or so. Any new entries added after that are repeated. I saw this last time yesterday. It does not change. Old entries added before that possible change six weeks’ ago are properly randomized.

My guess is that this new algorithm does not pick up entries from this previous date, maybe because they were generated from the Memrise original database or something similar. Or otherwise it does not use the whole set of entries from the full course database based on a criteria which is not applicable, or makes a very narrow selection. Sometimes so narrow the same word is used. It would be good to have code such as let’s say in a 3x3 grid, if four entries or more are the same, use random words from the whole database. Actually if there are two similar words then some random shuffling should be done.

In my case, my database was entirely created many months ago and was not edited after that.
I have no missing elements in any of the columns, as far as I could check, but the bug is only happening with the last items I’m learning -I stopped for a while and only started again around November 2019 with levels 15-16.
It happens both in the learning session and in the review ones, only a few words work properly. However, it does not happen with words I learnt before.

Not sure if you can check directly, however just to give you some items to test on, my course is 2181178 - shokyu nihongo II vocabolario e kanji
On level 15: it happens for 眠い - こわい
it does not happen for: まるで(氷のようだ) - いやな

Thanks for the detailed descriptions of the bugs, it gives me a place to start digging and investigating. If anyone else has reports like the above they really do help us get to the bottom of issues. So please share if you can.