Randomise, but force 1st item per talker to be a filler

JanV

Hello,

OpenSesame is a such a cool program! I'm currently trying to set up an experiment in which the participants are confronted with three different talkers. Each talker utters a variety of sentences, among which some fillers. I'd like for the order of presentation to be essentially random but with the constraint that the first sentence heard per talker is a filler.

Here's a slimmed down stimulus list.

SentenceID	Talker	SentenceType
S1	A	filler
S2	A	filler
S3	B	filler
S4	B	filler
S5	C	filler
S6	C	filler
S7	A	target
S8	A	target
S9	B	target
S10	B	target
S11	C	target
S12	C	target

Permissible orders include:

• S1, S7, S6, S8, S3 etc.
• S4, S3, S9, S1, S11 etc.
• S1, S3, S6, etc.

There's no need for the first three items to be fillers by different talkers, just for each talker's first item to be a filler.

Non-permissible orders include:

• S7 etc. (Talker A is introduced with a target rather than with a filler.)
• S1, S8, S9 (Talker B is introduced with a target rather than with a filler.)

Any suggestions? Thanks for your time and effort!
Jan

JanV

Hello,

I haven't fully implemented the following solution in OpenSesame yet, but this Python code generates lists that satisfy the constraint (see OP):

# This script generates a pseudorandom order of stimuli.

# We need the shuffle function from the random library.
from random import shuffle

item_list =     [("S1", "A", "filler"),
                 ("S2", "A", "filler"),
                 ("S3", "B", "filler"),
                 ("S4", "B", "filler"),
                 ("S5", "C", "filler"),
                 ("S6", "C", "filler"),
                 ("S7", "A", "target"),
                 ("S8", "A", "target"),
                 ("S9", "B", "target"),
                 ("S10", "B", "target"),
                 ("S11", "C", "target"),
                 ("S12", "C", "target")]

# Restrictions on randomisation:
# - The first stimulus for each talker (second column) needs to be a filler.

# Pick the first stimulus. This needs to be a filler.
items_notpicked = item_list

# Only retain the stimuli that may be picked, viz., filler items
under_consideration = [row for row in items_notpicked if 'filler' in row[2]]
# Shuffle these and take the first item
shuffle(under_consideration)
items_picked = [under_consideration[0]]

# Remove the items picked from the available items, i.e., only
# retain a row if it doesn't contain the sentenceID.
sentences_selected = [i[0] for i in items_picked]
talkers_selected = [i[1] for i in items_picked]
items_notpicked = [i for i in items_notpicked if not i[0] in sentences_selected]

for items in range(len(items_notpicked)):
# only consider filler items or sentences from talkers already selected
    under_consideration = [row for row in items_notpicked if ('filler' in row[2]) or (row[1] in talkers_selected)]
# shuffle, take the first one, append to item list
    shuffle(under_consideration)
    items_picked += [under_consideration[0]]
# update sentence and talker lists
    sentences_selected = [i[0] for i in items_picked]
    talkers_selected = [i[1] for i in items_picked]
    items_notpicked = [i for i in items_notpicked if not i[0] in sentences_selected]

A couple of runs produced lists such as these:

[('S5', 'C', 'filler'),
 ('S3', 'B', 'filler'),
 ('S9', 'B', 'target'),
 ('S11', 'C', 'target'),
 ('S4', 'B', 'filler'),
 ('S6', 'C', 'filler'),
 ('S12', 'C', 'target'),
 ('S10', 'B', 'target'),
 ('S1', 'A', 'filler'),
 ('S8', 'A', 'target'),
 ('S7', 'A', 'target'),
 ('S2', 'A', 'filler')]

[('S5', 'C', 'filler'),
 ('S2', 'A', 'filler'),
 ('S4', 'B', 'filler'),
 ('S1', 'A', 'filler'),
 ('S12', 'C', 'target'),
 ('S9', 'B', 'target'),
 ('S8', 'A', 'target'),
 ('S11', 'C', 'target'),
 ('S7', 'A', 'target'),
 ('S10', 'B', 'target'),
 ('S3', 'B', 'filler'),
 ('S6', 'C', 'filler')]

[('S6', 'C', 'filler'),
 ('S12', 'C', 'target'),
 ('S4', 'B', 'filler'),
 ('S5', 'C', 'filler'),
 ('S9', 'B', 'target'),
 ('S11', 'C', 'target'),
 ('S2', 'A', 'filler'),
 ('S3', 'B', 'filler'),
 ('S1', 'A', 'filler'),
 ('S10', 'B', 'target'),
 ('S8', 'A', 'target'),
 ('S7', 'A', 'target')]

JanV

Quick follow-up:At the end of the inline script, I've now added

global sentences
items_picked.reverse() # the "pop" method used later works from the back to the front,so just reverse the list here
sentences = [i[0] for i in items_picked]
global talkers
talkers = [i[1] for i in items_picked]

Then I used an inline script before the trial presentation with the following content:

global sentences
current_sentence = sentences.pop()
global talkers
current_talker = talkers.pop()

exp.set("my_sentence", current_sentence)
exp.set("my_talker", current_talker)

my_sentence and my_talker can then be used in the trial's sketchpad.

This seems to do the trick. Hopefully this is useful to other users. Thanks again for developing OpenSesame!

eduard

Hi Jan,

nice that you sorted things out yourself. I haven't looked at your code in detail, but just a few small remarks.

1) It shouldn't be necessary to declare your variables as globals. All the inline_scripts share a common workspace

2) In a comment you say: # the "pop" method used later works from the back to the front,so just reverse the list here. Actually, you can specify which item should be popped from a list. The default value is -1 I believe, but if you pass 0 it will pop from the beginning.

3) exp.set("my_sentence", current_sentence) can be replaces with var.my_sentence = current_sentence

Hope this is a little useful to you.

Good luck,

Eduard

JanV

Very much so, thank you!