zinked

From the shelf at zinked.ai

Rereading After a Delay

A Small Study of Recall for Technical Text. Mira Castellan and Jonas Ostrander, 2025.

An invented paper, written to test listening. Its study, numbers and references are made up. Press Listen and it is read the way zinked reads a paper: the orientation first, then the paper, with an easy listening edition of the method.

Before you read

The orientation. In the app it is heard and not seen: two or three minutes on what the paper asks, how it goes about it, what it finds and how it is laid out.

This paper asks a narrow question. If you are going to read a difficult technical explanation twice, when should the second reading happen: straight away, or a day later?

The background is that rereading is the most common study habit and one of the least effective. Spacing two readings apart is known to help with short facts, but there is little evidence for long technical passages.

To find out, the authors ran one small experiment. Forty-eight graduate students, none of them computer scientists, read an explanation of a sorting algorithm. Half reread it at once. Half reread it after one day. A week later everyone wrote down what they remembered, and the accounts were scored against a list of thirty key ideas.

The main finding is that the delayed group remembered more: sixty-one percent of the key ideas, against forty-seven percent for the group that reread immediately. The gap was widest for readers who had found the text hard. For them, waiting a day was worth twenty-one points. For readers who found it easy, the two schedules were about the same.

The authors are open about the limits. It is one study, with one text and one delay, and recall was measured only once.

The paper is short and laid out in the usual order. After the abstract, the introduction sets out the question. The method section describes the students, the text and the scoring, with one simple formula and a table of the two schedules. The results section has the numbers, including a second table that splits them by difficulty; that table is worth a look on the page. The discussion draws the practical conclusion for reading software: when a reader marks a passage as hard, offer a second reading tomorrow, not now.

Abstract

Readers of technical material often reread a passage straight away when it does not make sense. We ask whether waiting before rereading helps more. Forty-eight graduate students read a 2,400-word explanation of a sorting algorithm, then reread it either at once or after a delay of one day. One week later, the delayed group recalled 61% of the key ideas and the immediate group 47%. The gain was largest for readers who rated the text as difficult. We conclude that a delayed second reading is a cheap way to improve recall, and we describe how a reading app could schedule it.

1 Introduction

Rereading is the most common study habit, and one of the least effective [1, 2]. Its appeal is easy to understand: the second pass feels fluent, and fluency is mistaken for learning (Arkwright and Bell, 2011). Spacing the two readings apart is known to help with word lists and short facts [3], but the evidence for long technical passages is thin.

This paper reports one small experiment. Our question is narrow: if a reader is going to read a technical explanation twice, when should the second reading happen? We compare two schedules, immediate and delayed by one day, and measure what is remembered a week later.

We make two contributions. First, we show a clear advantage for the delayed schedule on free recall. Second, we show that the advantage depends on how difficult the reader found the text, which suggests when a reminder to reread is worth sending.

2 Method

This section has an easy listening edition. Listen to it in Easy listening and then As written to hear the difference.

2.1 Participants and material

Forty-eight graduate students took part, none of whom had studied computer science. The material was a 2,400-word explanation of merge sort, written for this study and checked by two instructors. It contained 30 key ideas, e.g. that the list is split in half until each part holds one item.

2.2 Procedure

Participants were assigned at random to one of two groups of 24. Both groups read the text once at their own pace. The immediate group then read it again straight away; the delayed group read it again 24 hours later. After the first reading, each participant rated the difficulty of the text from 1 (easy) to 7 (hard).

One week after the second reading, participants wrote down everything they could remember. Two raters, who did not know the groups, counted how many of the 30 key ideas each account contained. They agreed on 93% of ideas; disagreements were settled by discussion.

2.3 Measure

The recall score for a participant is the share of key ideas recalled:

R=k30×100

where k is the number of key ideas in the written account. We compare mean scores between groups with a two-sided t-test, and report Cohen’s d as the effect size.

GroupParticipantsSecond readingMean reading time
Immediate24Straight away11.2 min
Delayed24After 24 hours12.9 min
Table 1: The two reading schedules.

3 Results

The delayed group recalled more. Mean recall was 61% (SD 14) for the delayed group and 47% (SD 15) for the immediate group, a difference of 14 points, t(46) = 3.3, p < 0.01, d = 0.96.

Rated difficultyImmediateDelayedDifference
Easy (1–3)55%60%5
Hard (4–7)41%62%21
Table 2: Mean recall by rated difficulty and group.

Table 2 splits the result by how hard readers found the text. For those who rated it easy, the schedules differed little. For those who rated it hard, the delayed schedule was 21 points ahead. The delayed group also spent longer on the second reading, about 1.7 minutes more, which fits the view that a delayed rereading is slower because it is doing more work [4].

0%20%40%60%80%Easy (1–3)Hard (4–7)ImmediateDelayed
Figure 1: Recall one week later, by group and rated difficulty. Bars show means; lines show one standard error.

4 Discussion

Waiting a day before rereading a technical text improved what readers remembered a week later, and it helped most when the text had felt difficult. The result agrees with earlier work on spacing (Lindqvist et al., 2014) and extends it to a longer, more demanding passage.

The study is small and used a single text. We did not test delays other than one day, so we cannot say whether a longer gap would do better. Recall was measured once; we do not know how long the advantage lasts.

For software that people read in, the practical point is simple. When a reader marks a passage as hard, or slows down sharply, that is the moment to offer a second reading tomorrow, not now. More details are at https://example.org/reread-study.

References

  1. Okafor, T. and Lindqvist, P. (2013). What students do when a page does not make sense. Journal of Invented Studies, 14(1), 4–58.
  2. Arkwright, S. and Bell, R. (2011). The comfort of the second pass. Annals of Imaginary Psychology, 34, 30–41.
  3. Lindqvist, P. et al. (2014). Spacing short facts. Bulletin of Made-Up Research, 132, 354–380.
  4. Marchetti, D. (2016). Slow rereading. Proceedings of Nowhere, 31, 187–194.

That is the whole paper

It was written to test how zinked reads a paper aloud. Nothing in it was found out by anyone.