Execute on_start only at start of parent timeline

[dschreij]

I am creating an experiment in analogy of https://groups.google.com/forum/#!topic/jspsych/fCKiH-HRvGM. I want the trials to keep going for a limited amount of time, after which the block is ended. Because I am using loop_function to determine if a trial should be repeated. I cannot use timeline_variables (as far as I know?) so I have to implement the sampling of the trial variables myself. I have this so far:

function init_pool() {
    var conditions = [
        {cue: 'A',          probe: 'X',             trial_type: 'AX'},
        {cue: 'A',          probe: 'X',             trial_type: 'AX'},
        {cue: randomLetter, probe: 'X',             trial_type: 'BX'},
        {cue: 'A',          probe: randomLetter,    trial_type: 'AY'},
        {cue: randomLetter, probe: randomLetter,    trial_type: 'BY'},
    ];
    return jsPsych.randomization.repeat(conditions, 3);
}

var distractors = ['B','C','D','E','F','G','H','I','J','L','M','N','O','P','Q','R',
    'S','T','U','V','W','Y','Z'];

function randomLetter() {
    return distractors[Math.floor(Math.random()*distractors.length)];
}

var trials = {
        timeline: [fixation, first_letter, blank, second_letter, feedback],
        loop_function: function(){
            return block_duration - Date.now() + block_start_time >= 0;
        },
        on_start: function(trial) {
            if (condition_pool.length === 0){
                condition_pool = init_pool();
            }
            var currentTrial = condition_pool.pop();
            // The below obviously doesn't work, but it would be nice ;)
            // I just created global variables for the properties I want to be 'active' for the current trial
            trial.timeline_variables = currentTrial;
        }
}

The main problem is that the on_start function is propagated to all elements in the timeline, so fixation, first_letter, etc. all execute this on_start function.

Is it also possible to specify a function that only fires at the start of the current element, and not at the start of each of its child components? I tried on_load too, but it has the same behavior.

[dschreij]

Oh wait. You can use timeline_variables... 🤦 . I thought it would only repeat the current trial (and its values) and not sample new ones. But I guess in a different context my question still stands. Is it possible to execute a function only at the start of its timeline, and not propagate it to all of its child components?

[jodeleeuw]

No, not yet. This is an open issue that we are planning to address with the 7.0 release. #585

[dschreij]

Alright, good to know. Thanks!

[dschreij]

Alright, I have to revisit this again. I found out that the loop_function is only called after all timeline_variables have been processed, while I hoped it would evaluate after every trial. It seems you cannot use timeline_variables after all in a situation which you want to evaluate after each trial if the block should be stopped or continue depending on the time passed.

[jodeleeuw]

This is a helpful use case for us to consider. Maybe adding something like break_function to timelines would help.

I think you can use the on_finish event of the last trial in the timeline and call jsPsych.endCurrentTimeline() (docs here) if you want to break out of that timeline. Caveat: this kind of timeline control hasn't been as thoroughly tested as most features of the library, so it's possible that there are edge cases we haven't caught yet.

[dschreij]

Thanks for your suggestion. That's the first approach I took too, but somehow that gave unexpected behavior. It also depends a lot on how deeply you nest your timeline. If you are in timeline X which contains Y, and you want to stop X, then I have found if you call jsPsych.endCurrentTimeline() at the moment the experiment is in timeline Y, it only breaks out of that segment, and not out of X. This is well described in the docs, but this makes the usage of endCurrentTimeline limited in this sense. I solved it now by just using functions in the 'timed' timeline for all dynamic parts (e.g. which stimulus to display and which response to accept). It's a bit more work, but works perfectly!