My survey will serve as an item sorting task for scale development.

Respondents will rate the applicability of survey items on five dimensions. Because the item list is long (91 items), I want to randomly assign 30 items to a respondent.

Doing so works with the "statement randomization": I get the survey to randomly present 30 items, but it does this for each block independently. However, I would like that the 30 items that are initially selected remain the same when evaluating the other dimensions, such that every respondent has rated an item 5 times, once on each dimension.

Because I am randomizing the order in which the blocks of the five dimensions are shown, the "carry forward statements" option doesn't seem like the right option, as I cannot place the same block first every time (prohibits full randomization). Therefore, I don’t know how to use that initial set of items for the next blocks.

An option would be to create 3 groups of 30 items and randomly assign respondents to a group, which then has the 5 dimensions. Unfortunately, that would take away part of the randomization, as every respondent should have an equal opportunity to receive every item.

Is there a way that I can randomly assign 30 out of the 91 items to a respondent, that are then carried on throughout the next 4 blocks/dimensions, which are shown in random order?