attentional allocation

  • attention capacity is limited (1-4 items)

  • necessary for conscious recognition

attentional allocation determined by interaction of:

  • bottom up saliency based system: how distinctive items features are relative to other items in scene

    • guidance of attention/eye movements by low level visual saliency (Itti and Koch)

    • independent brain areas code for diff features → local, competitive interactions within these areas results in feature specific saliency maps → these are combined into master saliency map → within this map there is “winner take all” competition for attention → attention drawn to location of highest activation → attention results in binding of individual features at that location

  • top down volitional system: our goals influence attentional allocation

    • eye tracking

      • Yarbus- taped peoples eyes open and glued mirror on cornea, shot laser at mirror and saw where it bounced to, did math to figure out where the eye was looking

        • observers received diff instructions before viewing

        • conclusions: eye movements aren’t random, people look at information parts of scene for their task, appears to be top-down influence on eye movements

      • magnetic search coils

      • now: video-based eye trackers- either head mounted or remote (integrated, standalone)

    • volition attentions controlled by attentional control settings (top down bias of features) via working memory, visual routines, long term memory/knowledge of environment

    • good control of eye movements based on cues

    • attentional control settings and detection (Steve Most)

      • driving simulator task- drive thru unfamiliar city, actual stimuli in color, tasked w following route indicated by colored arrows, then motorcycle cuts driver off (could match or not the color of arrow you were looking for) → when arrow and motorcycle color matched, more likely to see motorcycle because you are already looking for specific color arrow

    • working memory and attention (Soto, Humphreys, Blanco) dual task- memorized flashed object, then find one line that was not vertical and indication whether it was to right or left of vertical

      • conclusion: early, reflexive process guides attention toward objects held in working memory, some limits to effectiveness otherwise you wouldn’t need to search for something in your environment

    • visual routines (Hayhoe and colleagues): making tea/sandwich, locating/directing/guiding/checking behaviors, only maintain info that is necessary for current stage of behavior (conveyer belt), change blindness for stop sign

      • examples: changed no parking sign to stop sign during eye movements based, then changed back during next eye movement

      • conclusions: people use visual routines to direct attention to task relevant locations to guide behavior, only represent information that is important for current behavior

    • constraining attention based on likely locations: we extract “gist” of scene and can use that info to constrain search to likely locations for a target - Torralba et al., - look for diff objects in scene and measure first 2 eye movements ppl make

  • history system

    • contextual cuing: during search, some layouts predict location of repeated target, in others the distractors are random, become faster w repeated

      • conclusion: consistent layout drives attention to target

      • implicit: people can’t pick out repeated layouts

    • value-driven attention capture:

      • during training- report orientation of line in red or green object, after response earn either small or big reward

      • during test- report orientation of line in unique shape (color irrelevant), slow when distractor appears in high reward color

      • conclusion: automatic capture of attention by rewarded item

      • implicit learning: works even if unaware that one color was highly rewarded

    • spatial likelihood

      • target happens to appear more often in one quadrant, people become faster at detecting it in that quadrant → implicit learning works even if they don’t explicitly realize there is a rich quadrant

    • can you learn multiple statistical regularities?

      • task- find C with break on left or right and indicate direction (in each quadrant there are 4 Cs one of each color)

      • statistical contingency: target appears in “high” quadrant 50% of trials, target appears in “high” color in 50% of trials

      • results: when target appeared in dominant quadrant and color it is faster to find, suggests history/statistical regularity biases attention

conclusions: many mechanisms ensure attention is allocated to relevant objects

  • Attentional Set (perhaps controlled by WM)

  • Prior knowledge of likely locations

  • Visual Routines

  • Memory for specific environs (the change detection lab)

  • Emotional Sets (inattentional blindness experiment from before MT).

  • Historical effects (contextual cuing, value-based capture, and spatial probability).

divided attention and behavior: system must perform 2+ discrete tasks at the same time. Can we do this?

  • people cannot do two simple tasks without any interference

  • you can break down simple task into 3 cognitive processes - perception, response selection/decision, and response production

  • reason we can’t multitask is because the decision phase of the first task cannot overlap with the decision phase of the second task

automacity?

  • with practice tasks that required lots of attention/decision making become more automatic (ex. driving)

  • why and do they ever become completely automatic?

    • why? able to generate larger repertoires of behavior per attention guidance, able to schedule shifts of attention better between competing tasks, central process shortening (each decision is made more quickly)

    • controversy if they ever become completely automatic

Ruthruff, Johnston, Van Selst, 2001

  • Subjects had over 14,000 trials of practice pairing two tasks (high/low tone; ABCD)

    • showed minimal PRP effect

  • paired a new task 1 w old task 2

    • really long RT2 - PRP effect

  • paired old task 1 with new task 2

    • RT2 not effected, little PRP

Psychological refractory period:

  • Only one decider – decision must be done serially, so if you are making one decision the other must wait.