Dear Maria Josefina,

I suggest you to see links and attached files in subject.

-Primacy and Recency Effects in Learning - RapidBI

https://rapidbi.com › Human Resources & Talent

- Relevance of the serial position effect in the differential diagnosis of ...

https://www.sciencedirect.com/science/article/.../S21735808130006...

- Proceedings of the 2007 Academy of Marketing Science (AMS) Annual ...

https://books.google.dz/books?isbn=3319118064

- The Psychology of Learning and Motivation: Advances in Research and ...

https://books.google.dz/books?isbn=0080463630

- Serial-position effect - Wikipedia

https://en.wikipedia.org/wiki/Serial-position_effect

Best regards

As you know, the primacy effect predicts that the earliest words will be learned relatively quickly, while the recency effect predicts the same for the later ones. You have a 15-word list and 5 learning trials. So you need to decide (a) how many of the words count as "early" or "late" [as opposed to "middle"] and (b) how many of the learning trials to include.

As for (a), you could select the first 3 & last 3, with either all 9 others as "middle" or just the middle 5. Or you could just go 5/5/5. It's up to you to decide whether (say) words #4 and #5 really count as "early." I think you can make out a case for any of these approaches.

As for (b), you could look only at the first learning trial or at all 5, or really anything in between. Here I would recommend including more than the first, but again, it's up to you - you just need to be consistent.

Let's say, for purposes of argument, you opt for a 5/5/5 division across all 5 learning trials. (I'm not saying to do this, just using it as an example.) The primacy effect should be revealed by subtracting the number of times the test-taker recalled each of the first 5 words minus the number of times they recalled each of the middle 5 words. The recency effect likewise, only use the last 5 instead of the first 5. So in your data set, you would enter the "count" for each set of words across the 5 trials. (Alternatively, just make each word on each trial be its own variable. Then you can use transform-compute to create the variables of interest.) So if a test-taker did this on the first 5 (3-4-5-5-5) and this on the middle 5 (1-1-2-3-4), then 22-11 = 11 for that person.

As I think on it, what would be really interesting would be to know which breakdown leads to the clearest primacy and recency effects. In other words, you could look at the results using just the first & last 3 words, or 4, or 5; the middle 5, or 7, or 9; one trial, two, or more. If you entered the results for each word on each trial for each person (a lot of data points, to be sure), you could compute these results quite easily. I'm not sure this has been done. If it hasn't, it would be a publishable piece of work if you had a decent sample. With a title like "How should we measure the primacy and recency effects on the RAVLT?" Not the kind of project that makes you feel brilliant, but solid, useful work that might become widely read and cited by people doing clinical work with the test. If you do something like that, please let me know your results!