Attention Network Simulator: Selection, Filtering & Neural Competition

simulator intermediate ~10 min
Loading simulation...
acc = 87% — reliable target selection with moderate load

With default parameters (signal 1.0, 5 distractors, moderate top-down bias), the attention network selects the correct target 87% of the time with a reaction time of ~450ms.

Formula

acc = Φ((S × (1 + β) - μ_d) / (σ × √N))
RT = 200 + 50 × log₂(N + 1) / S ms
SNR = S × (1 + β) / (σ × √N)

Selecting What Matters

Every second, your brain is bombarded by millions of sensory signals — yet you experience a coherent, focused stream of consciousness. The attention network accomplishes this remarkable feat by selectively amplifying task-relevant information while suppressing irrelevant distractors. This simulator models the neural competition that underlies attentional selection, letting you explore how signal strength, distractor load, and top-down expectations shape perception.

The Biased Competition Framework

Desimone and Duncan's biased competition model proposes that multiple stimuli compete for limited neural representation. When two objects fall within the same receptive field, their neural responses are mutually suppressive. Attention resolves this competition by injecting a bias signal — stronger signals or top-down expectations tip the competition in favor of the target, producing winner-take-all selection.

Top-Down vs Bottom-Up

Bottom-up attention is driven by stimulus salience — a bright flash or sudden motion captures attention automatically. Top-down attention is volitional, guided by goals and expectations stored in prefrontal cortex. The bias parameter β models this top-down influence: higher values simulate strong attentional templates that pre-activate target features, enabling efficient search even in cluttered visual fields.

Neural Noise and Errors

Neural processing is inherently noisy — stochastic firing rates mean that distractor activity sometimes exceeds target activity by chance. The noise parameter σ controls this variability. When signal-to-noise ratio drops below a critical threshold, selection errors become frequent, modeling phenomena like inattentional blindness and change blindness that reveal the limits of human attention.

FAQ

What is the attention network?

The attention network encompasses brain regions — including the frontal eye fields, intraparietal sulcus, and temporoparietal junction — that control the selection of relevant information from cluttered sensory input. Posner identified three sub-networks: alerting, orienting, and executive control.

How does top-down attention work?

Top-down (endogenous) attention uses prefrontal cortex signals to bias competition among sensory representations in favor of task-relevant stimuli. This 'attentional template' amplifies matching features and suppresses distractors via feedback connections to visual cortex.

What is biased competition in attention?

Desimone and Duncan's (1995) biased competition model proposes that stimuli compete for neural representation, and attention resolves this competition by biasing activity toward relevant items. Both bottom-up salience and top-down goals influence the competition outcome.

Why does performance decline with more distractors?

Each additional distractor increases the probability that noise-driven neural activity will exceed the target signal. In serial search, reaction time increases linearly with set size; in parallel search (pop-out), the effect is minimal because the target's salience dominates.

Sources

Embed

<iframe src="https://homo-deus.com/lab/cognitive-neuroscience/attention-network/embed" width="100%" height="400" frameborder="0"></iframe>
View source on GitHub