final memory tasks for NSD experiment
placeholder description from slack: Currently, we are planning an episodic memory test with several components and Ian is planning for a similarity judgement task on an overlapping set of items (with the episodic component neccesarily happening before the similarity component to prevent memory contamination). In terms of the episodic memory task, it would be something like: 250 (or more if possible) old items and 50 new items for a total of 300 test items. For each item, respond on 1 to 6 scale how confident you are that you have seen the item before (e.g. 1 = High Conf Old, 2 = Med Conf Old, 3 = Low Conf Old, 4 = LC New, 5 = MC New, 6 = HC New) If you say new, the trials ends and on to the next item. If you say old, then you perform two additional steps: first, you say how many times you think you saw the item over the course of the experiment (e.g, 1, 2, 3, 4, 5+ times) second, you make a temporal position judgement along a timeline at the bottom of the screen (with landmarks, per @Sarah DuBrow's comment above), with the ability to indicate uncertainty via a 'cone' which you can control the size of. Critically, this test should be in at least two blocks. The blocks are identical except the final temporal position judgement. For one half of the blocks, you are asked to indicate when the first time you saw the item was and for the other half you indicate the last time you saw it. In terms of the similarity task (Ian can chime in here), it would be done using Meadows, I believe, and would use ~80/100 of the items. In terms of item selection, we're thinking that the old items should be selected from the shared 1000. Ideally, they would span the whole timeline of the experiment and those used in the similarity task would ideally be sampled somewhat evenly from the overaraching similarity space. Ian suggested using the similarity scores calculated by W2V or WMD similar to how @Yufei did as a way to guide sampling. Perhaps the relevant group of us can touch base some point before the final memory test. If others have thought on the best way of sampling items, please chime in!