1. Create a multiple choice question with pairs: control - experimental - control - experimental, etc. (the experimental can be in any order)


2. Use advanced randomization to randomize only the experimental choices


3. Hide the question with JS (and click the Next button if it is on its own page)


4. Base your Loop & Merge on the displayed choices in the hidden question (do NOT randomize the loop). The loops will show in control - experimental - control - experimental order where the experimental are randomized.



---

Rather than using loop and merge, I'm wondering if the thing to do would be to put in a randomizer step, at the very beginning of the flow, that is only to SET all the the randomized selections you'll need (as embedded data), then use what it randomly selected at that step as Pair1 and Pair2 (etc as needed) to inform the rest of the flow.
That would mean that (a) you have all of these set, so it's now in your data which Pairs were to have been assigned at each variable step and (b) you can use those selections for logic in the actual presentation of content later in the flow.
So, what I did for my example was assume we have three "Options" that you want to assign randomly to different "Pairings", and then you would then use Pair1, Pair2, Pair3 later in the survey without worrying about randomizing them later because they were pre-randomized. (I used three just because I wasn't sure if you actually only had 2 instances, or were just simplifying for purpose of posting.)
I just LABELED them as Option1, Option2, Option3, but you could use more meaningful content you actually intend to pipe in, etc, as needed.
Then I created Groups for each option (to keep everything together and then used nested randomizers and logic to assign the Pairings to options. The top level randomizer is going to process all THREE of the Option level randomizers, which will each assign just ONE Pair to that option. Here's "Group1" which assigns Option 1 to one of the Pairs at random:
Basically, what it is doing is taking Option 1 and then I used a randomizer to assign that to one of the Pairings IF THAT PAIRING IS EMPTY.
Then I Duplicated this entire group, editing all the Option1 to Option2, and again for Option3. When I ran test responses through, none of the Pairings were blank, and all of them had a randomly assigned "Option" number assigned (no duplicates).
So for your experiment, you would then set it up as:
1. Control sentence-picture pair - item 1]
2. Experimental sentence-picture pair – Pair1 (randomly assigned at the start of the survey)]
3. Control sentence-picture pair – item 2]
4. Experimental sentence-picture pair –Pair2 (randomly assigned at the start of the survey)]
And you can use the Option or other indicator for analysis, for display logic within the survey, etc.

TomG Thank you so much for your answer! Just to make sure I understand correctly: the multiple choice question (step 1) is only 1 question with the answer options being the complete structure of the type of items in the experiment?
The only thing I don't understand yet, is how I can base my Loop & merge on the options in the hidden multiple choice question. Does this mean I have one block which starts with the hidden multiple choice question and then separate questions for the control and experimental items? So something like:
[Block 1

1. Multiple choice question (hidden)


• Control item 1


• Experimental item 1 (random)


• Control item 2


• Experiment item 2 (random)


• Control item 3


• ....

https://community.qualtrics.com/XMcommunity/discussion/comment/47206#Comment_47206Yes, the MC question is one question with a choice for each of the 'pairs' (both control and experimental). All you need is an identifier for each pair - you can add additional fields for each in the loop & merge setup.
The hidden (w/ JS) MC question must be in a block before the loop & merge block. Then base the loop & merge on that question.
Your loop & merge block only needs three questions:

1. sentence recording


2. picture (yes/no question)


3. picuture (typed responsed)

TomG That works great, thanks! The only issue left now, is that the pseudorandomization does not work when I add an additional experimental condition. So I have experimental condition 1 and experimental condition 2. I have fixed locations in my items list for both type of conditions. For example:
1. Control item 1
2. Experimental item - condition 1
3. Control item 2
4. Experimental item - condition 2
5. Control item 3
6. Experimental item - condition 1
7. ...
I would like to randomize the items belonging to condition 1 and the items from condition 2, but I don't want to mix up the positions of the two types of experimental items. When I use the choice ranomizer I cannot distinguish between the conditions of the experimental items. Is there an (easy) way to do this?

CarolK Thanks for your answer as well! I'm not familiar with using embedded data and branching in Qualtrics, but I'm going to see whether I can make it work in my questionnaire.
One thing is not entirely clear to me. When you talk about Option 1, 2 and 3, does that translate to experimental condition 1, 2 and 3? I indeed have different types (4) of experimental sentence-picture pairs and I want to control in which position of the experiment each type of experimental sentence-picture pair appears. So if I understand correctly, I'd need to create 4 'Options'. However, I do not want to randomly assign one of these experimental options to my sentence-picture pairs. Instead, each sentence-picture pair is already of a particular type (1, 2, 3 or 4). If I understand the flow of your example structure correctly, Qualtrics would randomly assign the type to the different sentence-picture pairs. Instead, I just want to tell Qualtrics that:
(1) particular sentence-picture pairs are of type 1, others are of type 2, etc.;
(2) the order of appearance of the experimental types in the questionnaire is fixed (e.g., type 1 appears in position 2, 8 and 20; type 2 in position 4; 6 and 16; etc.); and
(3) the sentence-picture pairs belonging to the different experimental types can be chosen by Qualtrics at random (so if I have 3 different sentence-picture pairs of type 1, these sentence-picture pairs should appear in position 2, 8 and 20 in the questionnaire, but I don't mind in which order).
I hope I made the structure of my experiment a bit more clear. Do you think this kind of set-up could also work with your solution? Or did I misunderstand your explanation?

https://community.qualtrics.com/XMcommunity/discussion/comment/47236#Comment_47236No, there is not an easy way to do it. Also, the difficult ways (included the embedded data approach suggested) will result in data that will be difficult to report and analyze because it will have to be decoded to match answers to conditions.

https://community.qualtrics.com/XMcommunity/discussion/comment/47241#Comment_47241Ah I see, that's too bad... I'll think about different solutions then, either pseudorandomizing the lists beforehand and assigning participant to one list at random or randomizing sets of questions. Thanks for you feedback!