Figure 1.
The Matrix game (a) used in [4] and its Marble Drop equivalent (b) described in [3].
Figure 2.
An example array of 12 set cards.
The cards with the solid highlight form a level 4 set (all attributes are different), and cards with the dashed highlight form a level 1 set (Shape is different, and all other attributes are the same).
Figure 3.
An example of a picture trial (a) used in the experiment and its equivalent word version (b).
Cards highlighted with a border are the cards that form set (not visible for subjects). The card with a dashed border is a highlighted card.
Figure 4.
Mean reaction times averaged for different groups.
(a) Reaction times grouped by trial difficulty and type. (b) Each subject's mean reaction times averaged in two trial types.
Figure 5.
Proportions of the counts of consecutive fixations on the same card.
Proportions have been calculated separately for picture and word trials.
Figure 6.
Subjects' overall usage of dimension reduction in two trial types.
(a) The usage of attribute types in similarity-based scanning as a proportion of the trial's fixations sequence. (b) The probability of using dimension-reduction based on the fixation's position within a trial. The probability at fixation x is calculated as: N(dr(x))/N(x). N(dr(x)) is a number of trials that have dimension-reduction blocks that include fixation x; and N(x) is a total number of trials that have at least x number of fixations. Fixation sequences in word trials are significantly longer than in picture trials. For comparison purposes, fixation sequences and dimension-reduction blocks from word trials were transformed into shorter lengths to match the lengths of corresponding picture trials.
Figure 7.
The mean overall similarity of all cards in a particular subsequence to the highlighted card.
The values are calculated separately for picture and word trials.
Figure 8.
A radar chart for the proportions of saccades in each saccade category.
Figure 9.
An example of an annotated raw fixation sequence produced by wst03 during the trial shown in Figure 3a.
Each lane with solid boundaries represents a card, whereas each sublane with dashed boundaries represents an attribute within a card. The lane labeled as C7 is the highlighted card. The other two cards that belong to the set are C3 and C12. Each rectangular block represents a fixation on a card's attribute value. Red blocks represent fixations where a subject paired a highlighted card with another card, while blue blocks represent consecutive search for a third card. Green blocks are fixations where the subject found a set and made final verifications.
Figure 10.
A simplified depiction of ACT-R architecture.
Internal workings and external connections between vision, declarative, goal and procedural modules. These four modules provide the most of the functionalities necessary for modeling SET tasks.
Figure 11.
A visualization of the content of the model's iconic memory.
Contents were visualized after the first fixations on the highlighted cards (cards with dashed boundaries) in (a) picture and (b) word trials. Those are the same trials as shown in Figure 3.
Figure 12.
Comparison of (a) subjects' and (b) models' reaction times.
Models' reaction times were calculated from 100 runs of picture and word set experiments.
Figure 13.
Models' overall usage of dimension reduction in two trial types.
(a) The usage of attribute types in similarity-based scanning as a proportion of the trial's fixations sequence. (b) The changing proportion of trials in which dimension-reduction was used. The proportions are calculated as a function of the fixation position within a trial. The proportion on fixation x is calculated by counting the trials that have a dimension-reduction block that include fixation x. The lengths of blocks from word trials are also normalized to match the length scale of picture trials.
Figure 14.
The mean overall similarity of all cards in a particular subsequence to the highlighted card.
The values are calculated separately for picture and word trials.
Figure 15.
A radar chart for proportions of saccades in each saccade category.