“As such, it is not surprising that the way in which information is organized, i.e., the sequence of events, can have a deep impact on what we learn, as well as how well we learn it.”
Carvalho and Goldstone (2015)
I was drawn by the title of this paper and the fact that one of the co-autors Robert Goldstone is interested in complex adaptive system models as applied to understanding how we learn and perceive.
Inductive learning
Inductive learning refers to learning categories and concepts from exemplars of those categories or concepts (with students then making generalisation, leading to an understanding of the rule.). Category learning takes place across time, by studying a series of examples rather than getting all category-relevant information at once.
| Inductive | Deductive |
| examples – rules | rules – examples |
If I recall correctly, it was when exploring inductive learning design that Kornell and Bjork (2008) discovered, in contrast to their hypothesis, that spaced practice of examples from a category, resulted in better posttest performance than massed practice.
Understandably, researchers interested in the field of Concept Learning, are drawn to interleaved practice as a sub-component sequencing (or spacing). Hence I read and re-read, and re-read the paper, surfacing the key points from the complexities of the results presented.
An over of sequencing
“What you learn is more than what you see” is essentially a paper investigating different learning sequences and how this impacts learning. Notably, “temporal spacing” and “temporal juxtaposition” then “category structure,” and “retention interval.” Disentangle spacing and interleaving effects is by no means easy.
Temporal Spacing
Larger spacing is better considered better than smaller spacing intervals. With larger spacing increasing the difficulty of repeated tests and this increased retrieval difficulty proposed to increase long-term retention.
Carvalho and Goldstone (2015) suggest increased encoding variability that comes with spacing and a broadening of learning context are two reasons spacing might benefit long-term retention. Conversely, massed presentation increases item familiarity, decreasing attention, resulting in less efficient encoding and poorer memory. Finally, that spacing is of itself, a desirable difficulty.
Temporal Juxtaposition
Interleaved and blocked study sequences differ in how closely in time items from the same versus different categories are experienced. Interleaved are temporally closer together, blocked study sequences are further apart.
Here Carvalho and Goldstone (2015) consider both habituation and discriminative contrast hypotheses. That repetitive presentation of a stimulus leads to decreased responsiveness (habituation hypothesis). That during blocked study participants become habituated to the common relation between successive stimuli of the same category, taking longer to achieve criterion than in an interleaved study sequence.
In the case of high discriminability stimuli (interleave practice), less habituation takes place between repetitions, greater re-encoding, leading to increased memory consolidation.
Category Structure
Category structure has been proposed as a key factor modulating sequencing effects in learning. Carvalho and Goldstone (2014b) demonstrated that by changing only the type of categories presented, participants could show improved learning following interleaved or blocked study.
More precisely, interleaved study resulted in better performance for low discriminability categories (categories in which all the stimuli were highly similar, both within and between categories), whereas blocked study resulted in better performance for high discriminability categories (in which all the stimuli were dissimilar, both within and between categories).
Category structures have an important modulating effect over which study sequences might result in improved learning.
Retention Interval
Is the advantage of interleaved over blocked sequences of study – time lag between study and test?
Carvalho and Goldstone (2015) ask why would interleaved study potentiate long-term retention of information to a greater degree than blocked study? Is the answer contextual interference and greater discrimination when tested?
What do Carvalho and Goldstone (2015) advise? In situations that require learning differences between categories, interleaved study will accelerate learning by promoting encoding of exactly these properties of the objects. On the contrary, for situations that require learning similarities within categories, blocked study will accelerate learning by promoting encoding of these similarities. It is important to recognise what the sequence of study is doing and that changing the relative frequency of different temporally proximate similarities and differences. This, in turn, affects the normal learning process resulting in differential encoding of stimuli properties.
In it’s most accessible form: interleaved study benefits learning low discriminability categories while blocked study is beneficial when learning high discriminability categories.
If this is proving interesting, then Yan and Sana (2020) is worth reading. Investigating the interleaving effect from two unrelated domains: physics and statistics. This led Prof Veronica Yan to conclude an email conversation with the following recommendation.
“Interleaving is clearly effective for inductive learning of material with high similarity between categories but not within categories.”
Prof Veronica Yan
References
Birnbaum, M.S., Kornell, N., Bjork, E.L. et al. Why interleaving enhances inductive learning: The roles of discrimination and retrieval. Mem Cogn 41, 392–402 (2013). https://doi.org/10.3758/s13421-012-0272-7
Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is spacing the “enemy of induction”? Psychological Science, 19(6), 585-592.
Sana, F., Yan, V. X., & Kim, J. A. (2017). Study sequence matters for the inductive learning of cognitive concepts. Journal of Educational Psychology, 109, 84-98, 1.