Someone recently asked me about the spacing effect. More precisely, about the time best left between study sessions (or the optimal inter-study interval if we want to be technical about it). My only plausible answer was one that few people want:
It depends.
The spacing effect posits that for any material, information is better remembered if there is a larger rather than a small interval between the first time it is studied and the second (Smith and Firth, 2018, p17). It’s most often associated with Ebbinghaus, who published his results in 1885, but something remarkably close to it was described by Robert Hooke in 1682 - this is the oldest reference I can find, but I suspect there might be earlier ones.
Adolf Jost confirmed Ebbinghaus’s findings in 1897, encapsulating them in Jost’s Law of Forgetting. Then, in a series of studies carried out between 1911 and 1912 at the University of Michigan, Nellie Perkins added even more support to the phenomenon, publishing her finding in 1914 as The Value of Distributed Repetition in Rote Learning.
Of course, the spacing effect isn’t confined to rote learning. In the 1970s, Alan Baddeley assisted the Post Office in the design of a program to train postal workers to type. He found that spacing out training sessions was more effective than blocking them.
More recent research has affirmed the robustness of the phenomenon.
The spacing effect certainly has a long history, but it’s only in the last few years that people have looked at its practical applications. Interestingly, there really isn’t much research looking at its impact on real learners in actual classrooms.
Indeed, the vast majority of studies have involved graduate and undergraduate volunteers and tightly controlled environments, allowing the researchers to isolate variables and draw conclusions. Most look at the spacing effect within a relatively short time-scale. That’s not to dismiss the research outright, but we should always remain mindful that we might not see the same results in real classrooms and common real-world learning scenarios.
The spacing effect happens when we apply a distributed style of leaning (or distributed practice). This is true regardless of the type of learning in which we’re engaged. Most studies ask participants to learn random words or titbits of trivia, but a few have looked at foreign language vocabulary, medical facts, and even proficiency of medical procedures.
All have reached the same general conclusions - spreading out our leaning is much more efficient than trying to cram it all in at once.
Studies on the effect often concern themselves with identifying an optimal inter-study interval (or ISI). Testing participants can then assess the effectiveness of the interval on the learned content some time after the final presentation of the material, known as the retention interval (RI) - how long you need to remember the information. We, therefore, consider the time span between the final learning session and, say, an end of stage test (see figure below). How, for example, will teaching and learning look if the to-be-learned information is tested tomorrow, next week, or in several months’ time?
Let’s look at a study by Nicholas Cepeda and colleagues, published in 2008. The researchers had 1354 participants learn 32 obscure trivia facts, such as ‘snow golf was invented by Rudyard Kipling,’ or ‘What European nation consumes the most spicy Mexican food? Answer: Norway.’ The facts were obscure to minimise the possibility that volunteers could access prior knowledge to help them recall the facts later.
The study used 26 ISI and RI combinations - the researchers manipulated the gap between the first and second presentation of the fact, along with the gap between the final presentation and the test.
While distributed learning fares much better than blocked learning, there are important factors to consider.
The inter-study interval matters: excessively spaced learning, for example, is no different to blocked learning - both prove detrimental.
Increasing the ISI leads to better retention, but only up to a point, after which further increases either have no effect or decreases retention.
Optimal ISI increases as desired RI increases. This means that if you want to remember something for a few minutes, opt for a short ISI (less than one minute). If you want retention to last much longer (weeks, months and even years) a longer ISI is going to be more beneficial. For long-term learning, you ideally need to be spacing out the content over multiple days.
This is fine, but how can we gauge the optimal inter-study interval? According to Cepeda, a 1 day ISI is optimal for a 7 day RI, but a 21 day ISI is optimal for a 350 day RI. A 5 to 10 per cent delay is probably optimal.
Rohrer and Pashler (2007) suggest that the gap between studying and restudying should be between 10 and 30 per cent of the time between the first presentation of the material and the time in which the material is to be needed, such as a test of exam.
In a 2014 study, Carolina Küpper-Tetzel, Melody Wiseheart, and Irina Kapler had participants (210 undergraduate and graduate students) learn 28 word pairs using 56 concrete and familiar nouns, ensuring there was no semantic relationship between the words in the pair.
They designed learning schedules over 3 levels, representing expanding, contracting, and equal inter-study intervals (Küpper-Tetzel, Wiseheart and Kapler, 2014).
In the expanding schedule, intervals increased (1 day then 5 days); in the contracting schedule, intervals decreased (5 days then 1 day); in the equal schedule, intervals were kept constant at 3 days and 3 days. They divided retentions intervals into 15 minutes, 1 day, 7days, and 35 days.
Results confirmed that the optimal schedule depended on the retention interval. For shorter RIs of 1 to 7 days, a contracting learning schedule (that is, decreasing intervals between learning sessions) resulted in the best free recall performance on the final test.
For longer RIs of up to 35 days, an equal or expanding learning schedule (constant or increasing intervals) resulted in better performance compared to a contracting schedule.
Surprisingly, there was no difference between equal and expanding schedules across any RI. This might suggest that the interval between the last 2 learning sessions is more critical for the final memory performance than the interval between the first and the second session.
How, then, can we explain these findings?
One explanation has to do with context dependent retrieval. For shorter retention intervals, learners experience a greater overlap between the last two learning sessions and the final test. Contextual variability theory posits that memories are encoded along with contextual cues, so if you study the material in the same room and same time of day, that information will be encoded along with the learned information.
However, this means that as the time between study sessions increases, the context also changes, providing fewer cues and making it harder to retrieve the memory. A shorter retention interval favours contracting schedules because they take advantage of the contextual similarity between the last study session and the test. Longer retention intervals favour equal or expanding schedules because they promote the encoding of a wider variety of contextual cues.
Like I said at the beginning:
It depends.