Common Spaced Repetition Mistakes (And How to Fix Them)

The short answer

Most spaced repetition failures come from the same handful of mistakes: treating SRS as an onboarding tool instead of a maintenance tool, feeding the scheduler dishonest grades, piling on new cards faster than you can encode them, and drilling recognition instead of recall. Fix those and FSRS (or any SRS algorithm) can do what it is good at: bringing items back right before you forget them, with minimal daily load.

Mistake 1: Putting brand-new words straight into your review queue

SRS algorithms assume a memory trace already exists. Drop a word you saw five seconds ago into today's reviews and you either pass using working-memory echo or fail in a loop and poison the card's initial difficulty rating. Spaced repetition maintains memory; it does not create it from zero.

Fix: Run an acquisition pass before scheduling: meaningful exposure, active recall (not multiple choice), a short delay before the first blind test, then graduation into SRS. See why you should not put new words straight into spaced repetition for the full breakdown.

Mistake 2: Skipping a structured acquisition phase

Even learners who know SRS is for review often rush first contact: see the word, flip the card, rate yourself. Without priming, a working-memory buffer, and blind retrieval, the first "review" is not a review at all. The scheduler receives garbage telemetry on day one.

Fix: Treat the first seconds and minutes as a separate pipeline from long-interval SRS. In Anki, use learning/intraday steps. In research terms, close the First-Turn Bottleneck with priming, a 5 to 10 second interleaved distractor, then blind active recall before FSRS takes over. Details in Fixing the First-Turn Bottleneck.

Mistake 3: Training recognition instead of recall

Multiple-choice cards, heavy cloze context, and "tap the right answer" flows feel productive. They mostly measure passive recognition. Recognition is easier than generation, builds weaker traces, and wrong options can encode false associations.¹ You "know" the word on the card and cannot produce it in conversation.

Fix: Default to production: type or say the target from a minimal cue. If you use cloze, pair it with full-sentence production elsewhere. See cloze card drawbacks and passive vs active vocabulary.

Mistake 4: Adding too many new cards per day

New cards are exciting; reviews are chores. Cranking daily new-card limits to 30 or 50 feels like progress until review debt compounds. Every new item adds future reviews on a growing exponential schedule. Within weeks your due count explodes and you either quit or start skipping days, which defeats spacing entirely.

Fix: Set new cards based on sustainable review time, not ambition. A common heuristic: cap new cards so daily reviews (new + due) stay within the time box you will actually honor. When debt spikes, stop adding new cards until dues stabilize. FSRS cannot save you from a deck you outgrew your calendar.

Mistake 5: Doing new cards before clearing reviews

Many apps default to "new first." That inverts the spacing effect. Due cards are the ones closest to forgetting; delaying them while you harvest dopamine from fresh words lets retrievability drop and failures cluster. You also mix acquisition (new) with maintenance (due) in one cognitive pile.

Fix: Reviews before new, every session. Protect due cards while you still have attention. Add new words only when today's maintenance queue is under control.

Mistake 6: Lying to the scheduler

Pressing "Good" or "Easy" when you guessed, peeked, or only recognized the answer trains FSRS (or SM-2) on fiction. Intervals stretch, you fail later, then you hammer "Again" and wonder why the card is "hard." The algorithm is not broken; the input is.

Fix: Grade honest retrieval only. If you needed a hint, that is "Again" or "Hard." Consistent honesty keeps Difficulty, Stability, and Retrievability meaningful (see how spaced repetition works). Over time you do fewer reviews, not more, because intervals match real memory.

Mistake 7: Letting leech cards clog your queue

Leech cards (items you fail repeatedly) are usually bad prompts, wrong granularity, or words that never got a proper acquisition pass. Leaving them in rotation burns minutes per session and drags morale. Suspended leeches often mean "this card design is wrong," not "my brain is wrong."

Fix: At the first leech threshold, stop repeating blindly. Rewrite the card (split meanings, add context, swap recognition for production), re-acquire the item outside SRS, then reintroduce it. Delete cards that cannot be fixed.

Mistake 8: Cramming reviews to zero

Clearing a 400-card backlog in one Sunday feels virtuous. It is also massed practice: dozens of items tested back-to-back with little spacing between them. Short-term performance rises; long-term retention does not benefit the way distributed reviews do.²

Fix: Cap daily reviews, spread the backlog over days, and prevent the next pile-up by fixing mistakes 4 and 5. Spacing only works if you show up consistently at moderate load, not in heroic bursts.

Building a sustainable SRS workflow

A workflow that avoids most of these traps looks like this:

1. Acquire new items with production and a short delay before blind recall.
2. Graduate into SRS only after honest first retrieval.
3. Review due cards first, at a daily cap you can sustain.
4. Add new cards slowly, tied to real available time.
5. Grade honestly and fix or suspend leeches instead of muscling through.

Spaced repetition is not a willpower sport. It is a scheduling system that needs clean data and realistic volume. Get those right and SRS stays boring in the best way: a few focused minutes a day, compounding for months.

LinGoat is designed around this workflow. New vocabulary goes through a pre-SRS laddering system (priming, a 5 to 10 second interleaved distractor, blind active recall) before FSRS scheduling, so you avoid the First-Turn Bottleneck and dishonest day-one ratings. You practice in full sentences with word- and grammar-level feedback, which keeps review production-based rather than recognition-only. Read Fixing the First-Turn Bottleneck for the acquisition pipeline, why new words should not enter SRS immediately for the science, or see how LinGoat works and open the app.

References

1. Roediger, H. L., & Marsh, E. J. (2005). The positive and negative consequences of multiple-choice testing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(5), 1155-1159. https://doi.org/10.1037/0278-7393.31.5.1155
2. Karpicke, J. D., & Roediger, H. L. (2007). Expanding retrieval practice promotes short-term retention, but equally spaced retrieval enhances long-term retention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(4), 704-719. https://doi.org/10.1037/0278-7393.33.4.704