Contributions:Human Association Task (HAT)

Synopsis

HAT studies how the brain forms mental associations between things, and adapts those associations when the uncertain environment changes. It combines uncertain sensory evidence, a brief memory delay and learning from feedback, allowing these processes to be studied during iEEG recording.

Participants judge whether a moving dot cloud contains more red or blue dots. After a delay, they choose a car, dog or tree image using the current color-category association; feedback helps them discover when that association has changed.

The association, not the individual picture

The initial association is red-car / blue-dog: more red dots means choose the car; more blue dots means choose the dog. Images and their positions vary across trials. The three possible associations are:

Association	Red	Blue
Initial	Car	Dog
Alternative	Dog	Tree
Alternative	Tree	Car

Session Flow

Figure 1. Guided session flow. Basic practice can be repeated. In the main task, association changes are not announced.

An on-screen association hint supports the start of unannounced practice and the main task, then fades. Participants continue using feedback to guide their choices.

Visual Representation

Participants view a gray screen with stimuli near the center. At choice onset, one image from each category appears; the Up, Left and Right arrow keys select the corresponding screen position.

Figure 2. Full task display, uniformly reduced. Figures 2 and 3 were captured from the standalone C++ task using the supplied 256 x 256-pixel images. The white edge patch is a timing marker.

Figure 3. Enlarged views from one trial, using identical central crop bounds and magnification. Dots disappear before choices appear. Holding a response key enlarges that image; no elements have been repositioned.

Display settings in the documented configuration: 452 dots; red RGB [199, 72, 50], blue RGB [77, 123, 161]. ImageSizePixels = 70 and SelectedImageScale = 1.5 give a 70-pixel base image and a 105-pixel selected image. Source-image resolution is separate from on-screen size.

Single-Trial Timeline

Figure 4. Single-trial sequence and requested timings for the listed configuration; boxes are not to time scale. The 100 ms key hold occurs within the choice window.

The fixation marker remains during the delay. With the listed settings, a confirmed choice stays visible for 150 ms, followed by the intertrial interval (ITI). Correct and incorrect responses request distinct feedback sounds at ITI entry; a timeout skips confirmation and receives no feedback sound.

Practice and Association Switching

Both switch exercises start with red-car / blue-dog.

Practice stage	Progression
After Video 1	Red-car / blue-dog; finish after 7 correct responses in total.
After Video 2: announced change	5 correct in total, then red-tree / blue-car with a New association notice; finish after 6 consecutive correct.
After Video 2: unannounced change	5 correct in total and initial hint fade complete, then red-tree / blue-car without a change notice; finish after 6 consecutive correct.

The initial hint begins fading during ITI after 2 consecutive correct responses in unannounced practice and the main task. Errors and timeouts reset streaks, but preserve total correct counts. Stage changes follow ITI. Practice does not count toward the main-task limit.

Main-task criterion

With the listed settings, the accuracy threshold is 70% over a rolling window of 15 trials, requiring at least 11 correct responses; they need not be consecutive. Performance is evaluated after each completed trial once 15 trials are available within the current association. Timeouts count as incorrect. If the criterion is met and the session continues, the next trial uses a new association and the window starts again.

Color coherence is the percentage of dots in the more common color. Values near 50% are harder to judge, so an error can reflect color uncertainty as well as an association change. Association selection balances exposure rather than following a fixed cycle.

Breaks, completion and controls

The listed settings specify two scheduled breaks and a 450-main-trial cap. The round target counts completed rounds after the initial association, not simply the number of switches. The task ends at the configured target or cap; staff may stop earlier.

Space continues instruction and break pages; C repeats basic practice from its repeat page. Shift pauses the task. Release it, then press and release Shift again to resume. During dots, pausing waits until they disappear. In videos, Left / Right seeks by 5 seconds. Space or Enter acknowledges task completion.

Parameters

The values below describe a recorded task configuration, not universal defaults. In Operator, open Config to inspect or edit parameters, then use SetConfig to apply them. Accepted values depend on the installed build.

GuidedMode

Value: 1. Enables guided videos and practice before the main task.

SwitchCriterionWindow

Value: 15 trials. Number of trials in the rolling accuracy window, which restarts after an association change.

SwitchCriterionPercent

Value: 70%. Accuracy threshold within SwitchCriterionWindow. At 70% over 15 trials, at least 11 responses must be correct; they need not be consecutive.

CoherenceMin

Value: 51%. Lower bound of main-task dominant-color percentage.

CoherenceMax

Value: 100%. Upper bound of main-task dominant-color percentage.

CoherenceSigma

Value: 20 percentage points. Gaussian scale around 50% before folding; not the observed standard deviation of presented coherences.

GuidedPracticeCoherenceMin

Value: 70%. Lower bound of practice dominant-color percentage.

GuidedPracticeCoherenceMax

Value: 85%. Upper bound of practice dominant-color percentage.

StopRule

Value: 0. 0 selects a completed-round target; 1 selects trial count. The trial cap still applies.

PostInitialMappingRounds

Value: 21. Completed-round target excluding the initial round; not a guaranteed switch count.

TrialLimit

Value: 450. Maximum completed main-task trials; excludes practice.

MidpointBreak

Value: 2. Number of scheduled breaks in the documented configuration.

PreDotMinMs

Value: 500 ms. Minimum requested pre-dot fixation duration.

PreDotMaxMs

Value: 500 ms. Maximum requested pre-dot fixation duration. Equal bounds give a fixed duration.

DotDurationMinMs

Value: 500 ms. Minimum requested moving-dot display duration.

DotDurationMaxMs

Value: 500 ms. Maximum requested moving-dot display duration.

DelayMinMs

Value: 1600 ms. Minimum requested post-dot delay.

DelayMaxMs

Value: 1600 ms. Maximum requested post-dot delay.

ResponseDuration

Value: 5000 ms. Maximum choice-response window, including the required key hold.

ChoiceHoldDuration

Value: 100 ms. Required response-key hold within the choice window.

ChoiceConfirmedDuration

Value: 150 ms. Confirmed-choice display duration after the required hold.

ITIMinMs

Value: 900 ms. Minimum requested intertrial interval.

ITIMaxMs

Value: 1100 ms. Maximum requested intertrial interval; the interval varies within the two bounds.

DisplayRefreshRateHz

Value: 60 Hz. Configured display refresh rate, not a physical timing measurement.

AutoDetectDisplayRefreshRate

Value: 0. Automatic refresh-rate detection is disabled. Check requested durations against physical timing with the site photodiode.

BCI2000 Integration

HumanAssociationTask.exe is the native C++ Application. It runs with the site's Signal Source, Signal Processing and Operator modules. Guided playback requires the task's AV adapter and packaged media.

Running the task and resources

Use the site launcher; load acquisition and task parameters, select SetConfig, then Start. The documented setup launches the Signal Source on the response-keyboard computer with --LogKeyboard=1 --FileFormat=BCI2000, supplying KeyDown / KeyUp and native recording. Do not also enable KeyLogFilter.

The supplied package places resources under tasks/HumanAssociationTask_BCI2000/resources: cifar contains BMP images, audio contains feedback WAVs, and guided contains videos/instructions. Match paths to the loaded parameters. Categories are 2 = car, 6 = dog and 58 = tree, with 500 images each.

States

The native BCI2000 .dat file stores acquired signals, loaded parameters and task states. Selected states for identifying the main events and outcomes are listed below.

State or group	Meaning
HATPhase / HATGuidedStage HATTrialKind	Current phase and guided stage; trial kind: 0 none, 1 practice, 2 main task.
HATTrial / HATPracticeTrial	Main-task counter (0 during practice) and separate practice counter.
HATMappingID HATSwitchCount / HATPostSwitchCount	Current association; switch counts before and after trial evaluation.
HATDominantColor HATRealizedCoherence	Color: 1 red, 2 blue; rounded realized dominant-color percentage.
HATResponseCategory HATResponseOutcome	Category: 2 car, 6 dog, 58 tree; 0 no accepted response. Outcome: 0 pending, 1 incorrect, 2 correct, 3 timeout.
HATResponseOnset HATReactionTimeMs	Confirmation/timeout pulse; RT in ms from choice-window start to that event, excluding internal pauses. Confirmed-response RT includes the key hold; it is not first-keypress latency.
HATTrialStart / HATTrialCompleted	Trial-entry and post-ITI trial-evaluation pulses.
HATFinished	Endpoint reached; final acknowledgement may still be pending.

Task states, including HATPhotodiode, are software markers. Check build-specific block alignment and use verified photodiode/audio recordings for physical timing. HATFinished may precede final acknowledgement and file closure. Keep all recording parts together.

Versioning

Author: Jiancao Huang. Behavior, controls and state meanings follow the inspected C++ source (protocol 6.0); Figures 2 and 3 use its standalone renderer. Parameter values come from a recorded configuration declaring ProtocolVersion = 6.2-bci-hospital1500-flex; TaskReleaseID = 0.4.0-preintegration-hospital1500-flex; DataSchemaVersion = 2. These sources do not verify the currently installed hospital build.

Location

Source (authorized access): HumanAssociationTask in BCI2000 SVN

See Also

See also: BCI2000 Operator Module