Suffix Effect: Why One Extra Spoken Word Erases the Last Digit

When a list is spoken, the last item or two are recalled far better than the ones before them, better than the same items would be if the list had been read from a screen. Crowder and Morton's explanation, published in 1969, is that a spoken item leaves a brief raw trace of its sound, an echo, that outlives the word by a second or two. At the moment recall begins, the last spoken items are still available in that store, and the listener can read them off without having to retrieve them. Printed items leave no such echo, so a printed list ends without the extra boost. They called the store precategorical acoustic storage, precategorical because it holds sound before it has been sorted into words.

A spoken suffix enters the same store and overwrites what is in it. The word go arrives in the slot where the last digit's echo was, and the listener is left with whatever they had managed to encode by ordinary means, which for the final item is often very little. That is why the damage lands where it does: the earlier items were never relying on the echo, so the suffix costs them nothing, while the last item loses the one advantage it had. The suffix does not need to be attended to or remembered; it only needs to be heard.

The effect is selective about what counts as a suffix, which is where most of the evidence for the account comes from. A tone or buzzer of the same length and loudness does little or nothing. A suffix in a different voice from the list, or coming from a different place in the room, does less than one in the same voice from the same place. A suffix that follows the list after a pause of a couple of seconds does less than one that follows at once, because by then the echo has faded on its own. And none of this happens with printed lists, where there is no acoustic trace to overwrite. Later work found suffix-like effects with lip-read lists and mouthed suffixes, which pushed the explanation away from a purely acoustic store toward something more like a record of speech, and James Nairne's 1990 feature model treats the suffix as overwriting the modality-specific features of the final item rather than an echo of its sound. The pattern of results itself has held.

The effect belongs to a family. The advantage of spoken over printed lists at the end of the list is the modality effect, and the suffix effect is the most direct way of removing it. The two together are the reason this page's demo asks about printed lists: a suffix can only take away an advantage the list has.

The observation came first from Kent Dallett in 1965 and from Robert Crowder's 1967 work on redundant prefixes and suffixes in immediate memory: a spoken digit list followed by a spoken item the listener was told to ignore was recalled worse at the end than a list with nothing after it. Robert Crowder and John Morton gave the finding its theory in "Precategorical acoustic storage (PAS)," published in Perception and Psychophysics in 1969.

The 1969 paper brought three observations together. Spoken lists are recalled better than printed lists at the final positions. Within spoken lists, the last item shows a pronounced advantage of its own. And a spoken suffix that need not be recalled removes that advantage, while leaving the earlier positions where they were. Crowder and Morton proposed a short-lived acoustic store that holds the last second or two of speech before it has been categorized into words, feeds recall of the final items, and is overwritten by the next speech sound to arrive. A printed list never enters it, which is why the visual curve has no recency bump for the suffix to remove.

John Morton, Robert Crowder and Harvey Prussin then tested the store's properties in "Experiments with the stimulus suffix effect," a 1971 monograph in the Journal of Experimental Psychology reporting 17 experiments with 383 adult participants. The suffix was varied in voice, in apparent location, in timing and in whether it was speech at all. Speech suffixes in the same voice from the same place did the most damage; changes of voice or location reduced it; nonspeech sounds did little. The pattern fit a store that was specific to speech and sensitive to its physical features, and the stimulus suffix became a standard tool for probing auditory sensory memory.

The acoustic account did not survive unchanged. In 1980 Ruth Campbell and Barbara Dodd showed a recency advantage and a suffix effect for lists that were lip-read with no sound at all, which a purely acoustic store cannot produce, and Catherine Penney's 1989 review set the suffix effect within a broader account of separate streams for heard and seen verbal material. Nairne's feature model of 1990 accounts for the suffix effect as the overwriting of modality-specific features of the final item by a following item of the same kind. What remains undisputed is the finding the demo asks about: a spoken suffix erases the end-of-list advantage, a tone does not, a printed suffix has nothing to erase, and a change of voice weakens the effect.

The effect appears wherever a short spoken sequence is followed by one more sound of the same kind. The rows below are illustrations, not findings from the studies cited on this page.

The effect depends on timing and on the kind of sound, and both can be managed by whoever is speaking.

• End on the content. When you read out a number, a code or a list, let the last item be the last thing the listener hears for a second or two.
• Put a pause between the list and anything that follows. A suffix that arrives after a couple of seconds does much less damage, because the echo has already been read off or has faded.
• Ask for a readback before you add anything. If the listener repeats the list first, the final item is secured before any sign-off can land on it.
• Change the voice or the channel. A closing remark in writing, or from a different speaker, interferes less than the same words in the same voice.
• When you are the listener, say the last item back immediately, out loud or under your breath. Rehearsing it moves it out of the store the suffix overwrites.
• For anything that matters, write it down or send it. A printed list has no suffix effect because it never relied on the echo.

The suffix effect is one specific finding about the end of a spoken list. Its neighbors describe the advantage it removes and the larger curve it sits on.

• It is not the modality effect, the better recall of the last items of a heard list compared with a seen one. The suffix effect is the removal of that advantage by one more spoken item, and the two are usually studied together.
• It is not the recency effect in general, the better recall of the last items of any list, which appears with printed lists too and is removed by a delay or a distraction task rather than by a single sound.
• It is not the serial-position effect, the whole U-shaped curve of recall by position. The suffix effect flattens only the rising end of that curve, and only for heard lists.