How to make Anki image occlusion test the right visual knowledge

Check permission, source quality, and the task. A diagram copied from a course handout may be allowed for personal use without being allowed for public redistribution. Record the source and intended audience before sharing. Inspect resolution at the size used on your actual device. Tiny text, compressed lines, and low contrast can make a correct identification impossible even when the underlying source was readable on a large lecture screen.

Next, write the expected response. Are you naming a labeled part, locating it, describing a relation, or explaining a function? One mask usually does not test all those objectives. In the fictional diagram exercise, the learner wants to identify components of a simple apparatus. They keep the image title and orientation visible, then inspect whether the masked label asks for a single component. This separates the retrieval task from a general desire to memorize the picture. The current editing manual documents the native occlusion tools; use it for the mechanical controls.

Look at the full front with the target hidden. A legend, caption, abbreviation, or duplicate label elsewhere can give away the answer. Neighboring labels can also make the task easier in a way that may or may not fit your objective. The correct response is not to hide every surrounding feature automatically. Remove accidental answer leakage while keeping the orientation and structural context needed for a fair identification.

Imagine that the apparatus diagram names a component in both a leader-line label and the caption. Covering only the leader label tests reading the caption. If the objective is identifying the component visually, that is a defect. The learner revises the mask arrangement and tests the front again. Then they ask a classmate to explain how the answer was inferred, if sharing is permitted. That explanation can distinguish recognizing the component from using a leftover textual cue. An image card should be evaluated through the visible task, not just by the area covered in the editor.

The Anki editing manual describes native modes that differ in which covered regions remain hidden while one target is tested. Inspect the mode rather than assuming every masked image behaves the same way. Choose based on the intended information available during the task. If neighboring labels are present in the real assessment, their presence may be appropriate. If the learner must identify a component without those labels, a different mask arrangement may be needed.

Test the same target under two deliberately different contexts. In the apparatus example, the learner first identifies a component with neighboring names visible, then attempts the task with those names unavailable. They record what changed in their reasoning. This is an original comparison exercise, not proof that one mode is universally superior. If the answer depends entirely on a neighboring label, add independent visual practice. If hiding the neighbors removes essential orientation, restore enough context to make the question fair. More concealment is not automatically better question design.

Open an actual review on the phone or tablet you intend to use. Check whether the image fits, the relevant line remains visible, and the mask does not cover the feature needed to identify the target. Test zooming or device orientation as part of the planned routine. If reading the diagram requires constant panning, use a clearer crop or a more suitable device. A technical display problem should not be recorded as a retrieval failure.

In the fictional example, a leader line ends near two adjacent components. On the laptop, the endpoint is visible; on the phone, compression makes it unclear. The learner replaces the source with a permitted higher-quality version or redraws a personal schematic. They preserve the relationship and source rather than simply enlarging a mask. After editing, they inspect every generated prompt because one layout correction can change what is covered. Review usability is part of question correctness when the answer depends on a small visual distinction.

After reviewing the masked diagram, attempt a second view or draw a simplified version from memory. Change orientation only when the subject permits it, and retain labels needed to understand the view. Ask for a relationship or function separately from identification. The purpose is to find out whether the learner recognizes the concept or only the position of a patch on a repeated picture. Neither a successful card nor a successful redraw guarantees performance in every context, but the comparison reveals a specific gap.

For the apparatus example, the learner can name a component in the original diagram but cannot explain why it connects to the next one. They add a mechanism question and revisit the source explanation. They do not generate more masks over the same label as a substitute for that reasoning. A useful audit sheet separates visual identification, spatial relationship, and explanation. Mark which task each prompt serves, then compare the sheet with the actual course requirements. This prevents a visually impressive deck from omitting the skill the assessment will ask for.

Add one orientation check to the audit: name the view and point to a stable landmark before identifying the target. If those preliminaries are uncertain, the problem may be orientation rather than the hidden label itself. In the apparatus example, tracing the connection between two components can clarify the visual relationship. Keep that exercise separate from memorizing where a rectangular mask appears, so the diagnosis identifies the missing visual skill instead of merely adding another label.

Keep the original image, source reference, and native collection when the occlusion behavior matters. A text export may show some written fields, but it is not a complete representation of masks and generated visual prompts. Before any transfer, inspect a representative sample and state exactly what you want to preserve: labels, explanations, or the full review behavior. Those are different requirements and may call for different workflows.

Elly’s CSV, TSV, and TXT import workflow does not preserve Anki image occlusion, media, native templates, scheduling, or review history. Evaluate a separate source-based explanation task only with material you may use. For example, you could draft a written question about a component’s function while keeping visual identification in Anki. Check the result against the source and do not describe it as a native occlusion migration. A useful division of tasks can be better than forcing a single tool to imitate behaviors that the text transfer does not carry.