Label choice task

Let participants assign on-screen text labels to images by drag and drop, requiring a minimum amount of labelling before they can continue.

When to use it

Use this when the task is to pair names with pictures rather than to pick a single answer: naming several objects on one screen, categorising a set of faces, or matching terms to diagrams. A separate trial per picture with typed answers is slower and loses the simultaneous comparison between items, while dragging labels keeps every option and every picture in view at once. If each trial has exactly one correct picture to choose, a plain click-response trial is simpler.

Trial file

type	stimFormat	stim1	stim2	stim3	responseType	responseOptions	required
test	.jpg	apple	banana	cherry	label	apple;banana;cherry	labels:3;stim:3
test	.jpg	dog	cat	horse	label	dog;cat;horse	labels:3;stim:3
test	.jpg	car	bus	train	label	car;bus;train	labels:3;stim:3
test	.jpg	shirt	shoe	hat	label	shirt;shoe;hat	labels:3;stim:3

How it works

responseType set to label turns the trial into a drag-and-drop labelling display: the images from the stim columns are laid out on screen, and the responseOptions cell supplies the text labels, semicolon-separated, that the participant drags onto them. This response type only works when the stimuli are images; there is nothing to drop a label onto in a word trial. The number of labels does not have to match the number of pictures, so distractor labels or spare pictures are both allowed.

required uses a grammar specific to label trials: labels:3;stim:3 means at least 3 labels must be used and at least 3 stimuli must be labelled before the submit button appears. With three labels and three pictures per row, that forces a complete labelling on every trial. Unknown keys inside the cell are silently ignored, so a typo like label:3 weakens the requirement without any warning.

One inheritance gotcha: label minimums leak forwards. A later label trial whose required cell is blank silently inherits the previous label trial’s minimums, so fill the cell in on every label row (as this file does) rather than relying on blank meaning off.

Label trials have no documented key scoring grammar, so this recipe leaves the trials unscored (the correct column stays empty) and the placements are scored offline from each trial’s recorded response.

Variations

• Allow partial labelling by lowering the thresholds, for example labels:2;stim:2 in required, so the participant may leave one picture unlabelled.
• Add distractor labels by listing more options than pictures in responseOptions, keeping stim:3 in required so every picture still gets labelled.
• Drop the required cell on the first label trial only if labelling should be optional there, and remember that later blank cells inherit earlier minimums rather than clearing them.