Attention Allocation in Self-Directed Learning

Abstract. This working paper reports preliminary results from a survey of self-directed learners (n = 84) on how attention is distributed across concurrent subjects. We find that reported focus is heavily concentrated: a median of 68% of weekly study time went to a single subject, with the remainder fragmented across three or more. We discuss implications for interleaving and offer a cautious interpretation, given the self-report design.¹

1. Introduction

Self-directed learners rarely study one thing. They carry several subjects at once and decide, week to week, where attention goes. Classroom research has studied the scheduling of practice — in particular the contrast between blocking and interleaving — but usually under conditions the learner does not control.² This study asks a narrower question: left to their own devices, how do learners actually allocate attention?

2. Method

We recruited 84 respondents from online self-study communities. Each reported, for the prior week, the subjects studied and an estimated share of total study time per subject. Estimates were normalized to sum to 100%.³ This is a convenience sample, and every figure below inherits the limits of self-report.

3. Results

Three patterns stood out:

Pattern	Share of respondents
One dominant subject (>60% of time)	61%
Two balanced subjects	24%
Three or more, evenly split	15%

The concentration is the headline: most learners are, in practice, near-single-tasking across a week, whatever their intentions.⁴

4. Discussion

If interleaving benefits retention, and learners tend to concentrate, then the gap between optimal and actual practice may be an allocation problem, not a knowledge problem — learners may know interleaving helps and still not do it.⁵ The obvious next step is a diary study with objective time logs rather than recalled estimates.

Footnotes

1. All results rest on recalled, self-reported time. Treat effect sizes as indicative, not precise. ↩

2. Doug Rohrer and Kelli Taylor, “The Shuffling of Mathematics Problems Improves Learning,” Instructional Science 35 (2007): 481–498. ↩

3. Respondents whose estimates summed to within ±10% of 100 were rescaled; 12 responses outside that band were excluded. ↩

4. Compare stated study plans with realized time; the divergence is the subject of a planned follow-up. ↩

5. See also my Short Literature Review on Flow States on what predicts a productive session. ↩