Noticing when something is present much more easily than when it's absent — nonoccurrences are nearly invisible.
Imagine you have a box of crayons. It's really easy to notice when there's a bright red crayon in the box—you see it right away. But if someone quietly takes the red crayon out, you probably won't notice it's missing for a long time. Our brains are really good at noticing things that ARE there, but really bad at noticing when something ISN'T there.
The Feature-Positive Effect describes the asymmetry in how organisms—including humans—process information about what is present versus what is absent. People find it significantly easier to learn associations involving the presence of a cue than associations involving its absence, leading them to underweight or overlook the informational value of things that do not happen. This bias extends beyond mere attention; it distorts recall, judgment, and decision-making, because nonoccurring events are encoded more weakly in memory and perceived as less important even when they carry equivalent or greater diagnostic value. The effect has been demonstrated across species, from pigeons and monkeys to children and adults, suggesting a deeply rooted processing asymmetry rather than a learned cultural habit.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Investors tend to react strongly to the presence of positive earnings signals or visible risk indicators but fail to notice the absence of expected disclosures, dividend payments, or regulatory filings—missing information that can be equally diagnostic of a company's health.
Clinicians more readily identify and act on symptoms that are present (a visible rash, an abnormal lab value) while systematically underweighting the diagnostic value of absent symptoms, even when the absence of certain findings is a strong differential diagnostic clue.
Teachers and students more easily learn rules defined by the presence of a feature ('words with a double letter are X') than rules defined by absence ('words without a vowel cluster are Y'), leading to systematic difficulty with concepts framed around what is missing or excluded.
People readily notice and remember hurtful actions or comments from partners (presence) but chronically fail to appreciate the absence of conflict, criticism, or neglect—leading them to undervalue stable, low-drama relationships.
Product teams track and respond to bug reports and feature requests (present signals) but struggle to identify usability problems from what users are NOT doing—such as features that are never used or flows that are silently abandoned—because absence generates no salient data.
Managers easily recognize and reward visible achievements (closed deals, shipped features) but fail to notice or credit preventive work—the security breach that didn't happen, the compliance violation that was avoided—because success through prevention produces no observable event.
News coverage gravitates toward events that occurred (attacks, disasters, scandals) rather than non-events (prevented attacks, avoided crises, successful deterrence), distorting public perception of risk and causing preventive policies to appear wasteful.
First described by Sainsbury and Jenkins (1967) in pigeon discrimination learning experiments. Extended to adult human subjects by Newman, Wolff, and Hearst (1980) at Indiana University.
In ancestral environments, present stimuli—a predator in view, ripe fruit on a branch, a moving shadow—demanded immediate response and carried urgent survival information. Absent stimuli (the lack of a predator, the absence of food) were less actionable in the moment. Organisms that quickly attended to present features and acted on them had a survival advantage, so neural systems evolved a processing bias toward what is here and now rather than what is not.
Machine learning models are typically trained on positive examples (features that are present in the data) and struggle to learn from absent features or missing data patterns. Anomaly detection systems have difficulty identifying threats defined by what is missing rather than what is present. Additionally, training datasets encode the feature-positive effect from their human annotators, who are more likely to label and flag present features than absent ones, creating systematic gaps in model knowledge about nonoccurrences.
50 cards are free to preview. Buyers unlock the rest of the deck plus the interactive training — Spot-the-Bias Quiz unlimited, Swipe Deck with spaced repetition, My Blindspots, Decision Pre-Flight, the Printable Deck + Cheat Sheets, and the Field Guide e-book. $29.50$59.
