Filter Theory

What it is

Broadbent set the theory out in Perception and Communication (1958), the book credited as a significant factor in the development of cognitive psychology. His attention research grew out of a wartime concern with improving communication between aircraft and control centres, and he combined his accounts of selective attention and short-term memory into a single channel hypothesis: one stream of information is processed at a time, and the filter decides which.

His own experiments used dichotic listening. Participants heard three pairs of digits at once, one set in each ear, and usually recalled them ear by ear rather than pair by pair. Recalling by ear, a physical feature, fitted a filter that selects on physical features.

The model fitted what Colin Cherry reported in 1953. Listeners shadowing one of two simultaneous messages could say that the voice in the other had changed from a man's to a woman's, a physical change, and could not say what its words had been.

In effect

The theory made a sharp prediction: if meaning is analysed only after the filter, nothing meaningful can get through from the message being ignored. Neville Moray's 1959 study tested it by placing the listener's own name in the ignored ear. His abstract reports that almost nothing from the rejected message was retained, and that the only stimulus found that would break through the barrier was the subject's own name. About a third of listeners noticed it, and replications put the figure at 34.6 per cent in 1995 and 29 per cent in 2021.

Recognising a name requires analysing its meaning, so a filter that blocks meaning early has no account of that. Anne Treisman's attenuation theory kept the filter and changed what it does: it turns the ignored message down instead of switching it off. For a plain-language account of why this matters to anyone who has heard their name across a room, see the article Why You Hear Your Name in a Crowd.

What it does not say

It does not say that listeners notice nothing about the ignored message. The theory expects physical features, such as a change of speaker, to be registered, and that is what Cherry's listeners did.

It does not stand as a complete account today. It is best read as the baseline that the own-name finding and Treisman's attenuation account were measured against, not as the last word on when selection happens.

It does not explain who hears their name. Later work found that low-span listeners hear it more often than high-span listeners (65 per cent against 20 per cent in a 2001 study of 40 undergraduates), which is a question about listeners rather than about the filter's position.

---