Cognitive Psychology
Two streams arrive at once. One of them is your job; the other is not. The argument that ran through cognitive psychology for two decades was not about whether you can follow one — everybody can — but about how much of the other one is processed before it is thrown away.
A visual analogue of a listening task — read the limits panel before drawing conclusions
By the end you should be able to separate detecting a physical change in an ignored channel from recognising its meaning, and say why that distinction divides the selection theories.
About 30 minutes. Nothing you do here is saved or sent anywhere.
This is a visual analogue, not dichotic listening. Two written streams presented side by side are not two spoken messages arriving at two ears. Reading is voluntary in a way listening is not, and you can move your eyes. The optional stereo tones give you genuine left–right separation, but they carry no words. Treat every result here as an illustration of the logic of the paradigm.
Your first trial only works once. Run it before reading on. From then on you know what to listen for, and everything afterwards measures divided attention rather than a failure of selection. The results panel marks the first trial separately.
You will follow one stream of words and press a button whenever it shows an animal. A second stream of ordinary words runs beside it, and you are asked to ignore it completely. Two short questions follow each trial.
Watch your side. Press Target — the button or the space bar — whenever your side shows an animal. Ignore the other side entirely. Two short questions follow each trial.
One practice trial, then a run of four trials. Results are withheld until the run ends, so that seeing them cannot change what you report next.
Your rule Choose a rule and start.
Status Not started.
Answer the question above to unlock the experiment.
| Position | Left stream | Right stream | Tone |
|---|
| Planted word | Rule | Trials | Looked different | Word recognised |
|---|
| Rule | Trials | Targets hit | Targets missed | Presses that were not targets |
|---|
Three accounts of selection, each proposed to handle findings its predecessor could not. The exercise is not to pick a winner. It is to say, for each finding, whether it sits comfortably inside an account or is the sort of thing that made people abandon it.
"Did you notice anything?" and "what was it?" are not the same question. Reporting that the ignored stream changed in appearance is a physical judgement; reporting which word it was requires that the word's meaning survived long enough to be identified. Almost every argument in this literature turns on keeping those two apart, and almost every summary of it runs them together.
Early selection put a filter before meaning was extracted: unattended input is selected on physical grounds — which ear, which voice, which side — and its content never gets analysed. That explains the striking ignorance of the unattended message very cleanly, and it is embarrassed by anything meaningful getting through.
Attenuation replaced the filter with a volume control. Unattended input is weakened rather than blocked, and words with low thresholds — your own name, something relevant to what you are doing — can still get over the line. That is a smaller change than it sounds, and a much harder theory to falsify.
Late selection moved the bottleneck after identification: everything is recognised, and selection decides what reaches awareness, memory and response. It handles breakthrough easily and has to work harder to explain why so little of the unattended stream can be reported.
A later and productive reading is that the question was badly posed: selection is early or late depending on how much the attended task demands. When the relevant task is easy, spare capacity spills into the irrelevant stream and meaning gets through; when it is hard, it does not. That is why the rule setting in this tool is not decoration. It is the variable that made "early or late?" the wrong question.
It can show you the difference between noticing a change and identifying a word, and it can let you feel what a load manipulation does. It cannot adjudicate, because the classic evidence involves shadowing continuous speech at two ears, and this is written text you can look at. Take the logic from here and the evidence from the reading.