You are stranded on a desert island or in the grasslands of a foreign country, without food or water or transportation. Your only friend is a volleyball. Somewhat annoyingly, a prescient memory researcher has pinned a list of words to a tree with instructions for you to rate their relevance to your situation. How relevant is “sofa” to your desperate situation? How about “knife”, might that help you out?

In a surprise retention test after your rescue, which of the words will you remember?

When this type of experiment is conducted in the laboratory—where participants merely imagine being stranded in a remote place rather than being flown to Marble Bar—an intriguing pattern of results emerges: A few seconds of “survival processing” consistently leads to better recall than a number of other techniques known to enhance memory, such as forming a visual image of an item, relating the item to the self, or engaging in so-called “deep” processing that focuses on embellishing the meaning of the material.

This survival effect appears to be robust and replicable, and—remarkably—it does not appear to matter much whether the words being presented for the rating task are actually relevant to survival (“knife”) or relate to a trip to the hairdresser (“shampoo”) or any other random topic (“sofa”).

What remains controversial, however, is its explanation. Why do we remember things better after we consider their survival value?

A recent article in the Psychonomic Society’s journal Memory & Cognition seeks to shed further light on this issue. In three experiments, researchers James Nairne and colleagues focused on memory for prior processing of items, in addition to memory for the items themselves.

Participants were initially given the standard survival task for some items—“how relevant is ‘sofa’ to your survival”?—whereas other items were presented for a pleasantness rating task—“how pleasant is ‘chocolate’?” In a second phase, people were presented with the same words again together with an equal number of never-before-seen items, and were asked to indicate whether each word had been processed previously, and if so, how (i.e. survival vs. pleasantness). In a final surprise recall phase, people were asked to recall as many items as possible from the second phase—that is, irrespective of whether or not they had been rated during the first phase.

Although the results differed slightly between the three studies, several common themes emerged. First, when asked about prior processing during the second phase (“did you rate this item for survival initially”?) people failed to show an enhanced ability to identify survival items. In other words, there was no “survival effect” in memory for the source of earlier processing.

Second, in striking contrast to the source memory, a “survival effect” was found on the final surprise memory test. That is, when people had to recall the items themselves, they recalled them better when they had responded to a survival query during the second phase, as opposed to a control query. Notably, this advantage extended to items that were not seen initially—that is, items that were never rated for their survival value but were only queried whether or not such a rating had taken place.

Specific examples help to clarify this complex pattern of results:

Suppose you are given two words, ‘sofa’ and ‘shampoo’, and you rate the former for its relevance to surviving in the desert and the latter for its pleasantness. You are then shown ‘sofa’ and ‘shampoo’ again, together with ‘elixir’ and ‘kumquat’, and for each item you are asked to decide whether you rated it either for survival or pleasantness. Note that you must say ‘no’ to when asked whether you rated ‘elixir’ for survival, and no also to the question of whether you rated ‘kumquat’ for pleasantness—after all, you never rated either of the items. For this source memory task, there is no survival advantage—your ability to differentiate ‘sofa’ (“yes, I rated it for survival”) from ‘elixir’ (“no”) is identical to the ability to differentiate ‘shampoo’ (“yes, I rated its pleasantness”) from ‘kumquat’ (“no”).

Now you are asked to recall all of the words seen previously in the experiment, at both occasions. It is here that the survival effect again emerges: ‘sofa’ is more likely to be recalled than ‘shampoo’. Intriguingly, ‘elixir’ is also recalled better than ‘kumquat’, even though neither item was rated at the outset—however, you were queried about whether or not you rated ‘elixir’ for survival. Even though the answer was no, the mere query boosted subsequent memory for this item compared to ‘kumquat’, which was queried by asking whether you had rated its pleasantness.

The latter outcome is particularly remarkable, because it reveals a memorial advantage that results from a retrieval attempt—testing your memory for an earlier source of processing has benefits for items not seen before simply because of the way the question is phrased: By focusing on survival-relevant attributes of a word during questioning, a stronger memory trace is formed than if the question directs attention to other features.

By extension, if you want to remember a famous actor’s name, ask yourself if Humphrey Bogart appeared in Cast Away, rather than in Animal House. Survival-related processing helps memory even when it’s part of a question about items not previously considered.