Telescoping Effect: Why Old Events Feel Recent and New Ones Feel Old

Telescoping is what you get when uncertain dates are squeezed against a boundary. Rubin and Baddeley's 1989 model needs no bias in memory at all. It assumes three things: that recent events are better retained than old ones; that dating errors are unbiased but grow with the age of the event, so a two-year-old memory can be off by months while a two-week-old memory is off by days; and that when a question names a period, events from outside the period can intrude into it, but nothing can intrude from the future. Put those together and the errors pile up in one direction. An old event can be misdated forward into a survey's reference period but has nowhere to go backward, so old events crowd toward the present.

The same arithmetic predicts the less familiar half of the effect. A very recent event cannot be dated into the future, so its errors can only run backward, and it is remembered as slightly older than it is. Rubin and Baddeley found exactly this when they asked psychologists to date departmental talks given over two years: the errors grew by about four tenths of a day for every day since the talk, and they pointed toward the middle of the interval from both ends. Janssen, Chessa and Murre reported the same shape for public news events in 2006, a small backward telescoping for recent news and a large forward telescoping for remote news. That is the pattern the demo tests, and it is why the slider for a 21-year-old hurricane tends to land short while the slider for a two-year-old Olympics tends to land long.

Two things sharpen the effect. The first is the form of the question. A date given in absolute terms, July 2004, is more accurate than one given as an elapsed time, three weeks ago, and people prefer the elapsed form for distant events, which is where the errors are largest. The second is memory for the event itself. A vivid, well-rehearsed memory feels close, because the ease of recalling it is read as recency, which is one reason big public events feel like yesterday. The effect is a property of dating, not of forgetting: people are not losing the event, they are losing its place on the calendar.

John Neter and Joseph Waksberg published "A study of response errors in expenditures data from household interviews" in the Journal of the American Statistical Association in 1964. The work was done for the United States Census Bureau, which was trying to measure how much households spent on alterations and repairs to their homes. The design compared two ways of asking. In an unbounded interview, respondents were asked about the previous month with nothing to anchor the start of the period. In a bounded interview, the interviewer first read back the jobs the household had reported in the previous visit and then asked only about anything since.

The unbounded interviews produced far more. Compared with bounded recall of the same one-month period, respondents without the anchor reported roughly 40 to 55 percent more jobs and spending, and the excess was concentrated in large jobs that had in fact been done before the period began and were being pulled forward into it. Neter and Waksberg introduced the word telescoping for this and proposed bounded recall as the remedy. The procedure became standard in panel surveys, including the National Crime Victimization Survey, where an earlier interview with a household bounds the next.

Loftus and Marburger showed in 1983 that a public landmark does the same work as a bounding interview. Across five experiments with 1,694 participants, asking "since the eruption of Mount St. Helens, has anyone beaten you up?" reduced the forward telescoping of crime reports compared with asking about a plain reference period, and a personal landmark or New Year's Day worked as well. Rubin and Baddeley's 1989 paper supplied the explanation summarized above, and its title, "Telescoping is not time compression," states its claim: the dates are not being squeezed by a distorted sense of time, they are being scattered by ordinary uncertainty and trapped by a boundary.

Any question that asks when, or how many in the last so-many months, invites the effect. The rows below are illustrations, not findings from the studies cited on this page.

The counters follow from the model: give the memory a boundary it cannot cross, or replace the memory with a record.

• Bound the period with a landmark. Before asking yourself how many times something happened in the past year, fix the start of the year to an event you can date, then ask what happened after it. This is the technique Loftus and Marburger tested.
• Date in absolute terms. Work out the month and year rather than how long ago, since the elapsed-time form carries larger errors for distant events.
• Reason from other dates. Where were you living, what job did you have, what season was it? A date reconstructed from context is usually closer than one read off the feeling of recency.
• Distrust vividness. A memory that is easy to bring to mind feels recent; the two are not the same.
• Use records where they exist: receipts, photos, messages, calendars. For anything that will be counted, such as spending or medical visits, a record beats a bounded memory and a bounded memory beats an open question.

For the people who write questions, the lesson of the 1964 study is to bound every recall period, either with a previous interview or with a landmark the respondent supplies.

Telescoping is a dating error. Several neighbors change what is remembered rather than when.

• It is not rosy retrospection, which changes the evaluation of a past event, making a vacation better in memory than it was at the time. Telescoping leaves the content alone and moves the date.
• It is not the hindsight bias, the sense after the fact that an outcome was predictable. That is an error about what you knew when; telescoping is an error about when.
• It is not the availability heuristic, judging frequency by how easily examples come to mind, although the two meet in surveys: an event that comes to mind easily is both over-counted and dated too close.
• It is not the recency effect, the better recall of the last items in a sequence, which is about what is remembered rather than how it is dated.