The prior knowledge paradox: what to do with what they know

Making learner's prior knowledge work for them

Photo by Gabriella Clare Marino on Unsplash

Remember Ausubel’s famous quote?

“If I had to reduce all of educational psychology to just one principle, I would say this: The most important single factor influencing learning is what the learner already knows. Ascertain this and teach him accordingly’’ (Ausubel 1968, p. vi).

This felt for a long time like a maxim we could live by. But two relatively recent pieces of research complicate this orthodoxy – not overturning it entirely, but showing the story is much less straightforward than some assumed.

The two studies

First, a meta-analysis of studies about the role of prior knowledge in learning (Simonsmeier et al., 2022) showed that –

a) Most studies focused on final test scores not learning gains. They measured whether those with higher test scores to begin with (more prior knowledge) had higher results at the end. Most did not measure whether participants with greater prior knowledge actually learned more, that is, whether they made larger learning gains from pre- to post-test.

b) Of the studies that did measure learning gains, the correlation between higher prior knowledge and learning gains was near 0 (i.e. greater prior knowledge did not appear to predict greater learning gains). Even though most effect sizes hovered around zero, the variability was huge: sometimes prior knowledge predicted strong positive gains, and sometimes it predicted negative ones.

And, given that most of us aren’t that interested in whether an intervention keeps knowledge stable (those with higher scores at the start have higher scores at the end), it seems strange that learning gains haven’t been the standard measure across most studies. (The reason seems to be that learning gain isn’t that easy to measure).

If that wasn’t enough, a well-designed RCT this year provided the first causal test of whether domain-specific prior knowledge improves new learning (Buchin & Mulligan, 2025). In their experiment, participants were randomly assigned to learn three out of four topics within a domain (either Sensation and Perception or Historical Geology) over three days. This created a high prior knowledge (HPK) domain for each participant and a low prior knowledge (LPK) domain where they had no training. Crucially, within the HPK domain, one topic had been deliberately left untrained. That untrained topic became the test of new learning in a domain where participants already had a fair amount of background knowledge.

During the next phase (learning phase), everyone studied new text passages from all eight topics including –

• the untrained topic from their trained domain (the all important “new learning within domain” test), and

• four completely untrained topics from the untrained domain (the comparison condition).

If prior knowledge genuinely helps you learn new material in the same domain, the prediction is clear: learners should perform better on the untrained topic in the trained domain than on the topics from the untrained domain.

But that’s not what happened.

Across every measure of new learning, participants did no better on the new topic in the domain they’d been trained in than on topics from the completely untrained domain. In other words, having more background knowledge in a domain did not help them learn new material in that domain. Learners felt the within-domain material was easier (lower mental effort ratings), but they didn’t learn it any better.

At first glance, this looks like yet another gut punch to Ausubel.

Taken together, these two studies suggest that it isn’t nearly good enough – or accurate­ – to say “prior knowledge helps” (the so-called ‘knowledge is power’ hypothesis). Clearly, the range of effects found in Simonsmeier et al. (202) speak to at least two issues:

1. Prior knowledge is defined very differently in different studies. If we aren’t measuring a similar construct, it’s going to be hard to compare the results.

2. A nice clean line between prior knowledge and learning doesn’t exist. There are mediators and moderators that run interference, and these ultimately explain whether and how prior knowledge helps, hinders or doesn’t do very much at all. Like shining a light through a prism, the path alters.

Let’s take these two things one at a time.

Defining prior knowledge

PK is defined in different ways across studies. Across studies, ‘prior knowledge’ can mean anything from procedural and conceptual knowledge to beliefs, tacit understandings, episodic memories, or semantic networks. Sometimes it also includes misconceptions too. For Ausubel of course, prior knowledge, or ‘existing knowledge’ as he called it, refers to concepts and propositions that ideally exist in organised hierarchies where superordinate concepts are capable of subsuming more detailed ideas.

In fact, Hattan et al. (2023) looked at many definitions of PK used in various studies and decided on the following:

“prior knowledge can be understood as the sum of individuals’ existing knowledge, including personal, domain, topic, strategic, social, cultural, and linguistic knowledge (Alexander et al., 1991; Hattan & Lupo, 2020). Further, individuals’ existing knowledge extends beyond academic knowledge to knowledge of self and the world outside the classroom. Moreover, individuals’ prior knowledge affects their learning and development even when it is incomplete or inaccurate.”

Crikey.

A definition this broad makes Ausubel’s seem almost surgical by comparison and highlights just how differently ‘prior knowledge’ is operationalised across studies. And ultimately, if they’re measuring different constructs, it’s no wonder we end up with a range of results.

Mediators and moderators

So, we have somewhat of a prior knowledge paradox on our hands: decades of theory suggests prior knowledge affects learning from schema-based comprehension studies to expert–novice research and beyond. But the two studies described here don’t appear to support this.

Clearly the link between prior knowledge and learning gains is complex. Schneider and Simonsmeier (2025) propose a set of moderators and mediators:

At least 16 mediators, including attention, encoding, retrieval, interference, knowledge restructuring, cognitive load, motivation, etc.

And…

4 sets of moderators:

• learner characteristics (e.g., working memory, age, misconceptions),

• knowledge characteristics (e.g., accuracy, coherence, specificity),

• task characteristics (e.g., complexity, representational format), and

• instructional characteristics (e.g., scaffolding, feedback, generative vs receptive demands).

Since each mediator is shaped by moderators (and many mediators operate simultaneously) it becomes almost impossible to predict, in advance, exactly how a learner’s prior knowledge will affect new learning.

This is why, I think, Ausubel had the right idea all along…

Ausubel’s advance organisers

Yes, if we just take his quote on face value we can get hung up. But, if we read his theory as a whole, we see that Ausubel too realised not any old knowledge will do.

To be useful, knowledge must be relevant, available/activated and organised. In Buchin and Mulligan (2025), having background knowledge in the domain was not enough to lead to greater learning gains. I can hear Ausubel (if still alive) saying, “Well, obviously.”

Ausubel had the right idea about how to mobilise this knowledge. As we’ve seen, Schneider and Simonsmeier (2025) propose a long list of mediators and moderators. One way to cope with them is to shape and constrain, as far as is possible and desirable, the prior knowledge a learner uses to understand the new material:

“a scaffolded ‘‘boot strapping’’ may be necessary and useful, especially for naive learners with a low level of domain-specific prior knowledge.” (Gurlitt, 2012, p.354).

This is exactly what Ausubel attempted when he invented his tool called the advance organiser.

I think about it like this:

When you come to something new, be it a journal article, a talk at a conference, a book, there’s the very real possibility you’ll get completely confused or make take away the least important ideas. I think of my university self, showing up to lectures on ‘contract law’ or ‘equity and trusts’ having done none of the reading and hoping, by osmosis, I’d take something in.

Instead, I experienced a sort of cognitive shock – the feeling of having nothing in my head for the new ideas to latch onto. In short, the lecture was wasted on me. I’d have done better to stay in bed and read a paper on the topic.

To avoid this shock, what learners benefit from is a structuring and activation of relevant prior knowledge before encountering the new information.

Enter advance organisers.

An advance organiser is a tool Ausubel invented that provides the learner with a temporary, high-level schema (a conceptual ‘hook’) through which the new material can make sense (Gurlitt et al., 2012).[1]

Advance organisers can come in many forms: graphical, narrative, comparative, etc. And, which form to use when will depend on the learner and the materials. For example, graphical organisers may work better for learning materials that require the learner to understand relationships.

Personally, when I deliver training, I like to use a concept map advance organiser to show learners the hierarchy of ideas:

I then add to it in advance of the next stages as the ideas get more detailed. What’s nice is that this becomes a generative activity later in the training where they try to reconstruct the map, using their own words and adding additional links.

It’s not rocket science, but it provides some organisation to what otherwise can become a mish mash of concepts. It helps because, when we’re new to something, we need this organisational structure.

To make advance organisers even more effective, we can emphasise the key concepts in them too. Gurlitt et al., (2012) found that when they emphasised the key concepts in their advance organisers (which were written paragraphs), by having students fill in the missing letters of the key concepts words, the success rate doubled. Again, this shows us that advance organisers work best for learners new to a topic when they are designed to shape, constrain and direct their prior knowledge as much as possible.

So, where does all this leave us?

Simonsmeier et al. (2022) and Buchin & Mulligan (2025) have thrown a cold bucket of water on lazy proclamations that ‘prior knowledge supports learning. They’ve woken us up to the stark reality that prior knowledge affects learning through multiple channels. And, like anything, (dual coding, retrieval practice, spacing, etc.) it must be thought of as part of a holistic learning process rather than examined in isolation.

Saying that, given the complexity of the relationship between prior knowledge and learning and the bloated definitions we are left with when we try to account for every facet of the construct, I think we are left with only one option: shape, constrain and direct prior knowledge as much as possible before introducing new ideas.

Assuming background knowledge from familiarity with content in the same domain won’t be good enough (Buchin and Mulligan, 2025). Neither is simply allowing students to ‘form an impression’ of what we say or hope they will construct the same organised schemas we have.

We have to engineer what we want: students using relevant, organised and activated prior knowledge to understand new ideas.

References

Ausubel, D. P. (1968). Educational Psychology: A Cognitive View. Holt, Rinehart and Winston.

Buchin, Z. L., & Mulligan, N. W. (2024). Prior knowledge and new learning: An experimental study of domain-specific knowledge. Journal of Experimental Psychology: Applied.

Gurlitt, J., Dummel, S., Schuster, S., & Nückles, M. (2012). Differently structured advance organizers lead to different initial schemata and learning outcomes. Instructional Science, 40, 351-369.

Hattan, C., Alexander, P. A., & Lupo, S. M. (2024). Leveraging what students know to make sense of texts: What the research says about prior knowledge activation. Review of Educational Research, 94(1), 73-111.

Schneider, M., & Simonsmeier, B. A. (2025). How does prior knowledge affect learning? A review of 16 mechanisms and a framework for future research. Learning and Individual Differences, 122, 102744.

Simonsmeier, B. A., Flaig, M., Deiglmayr, A., Schalk, L., & Schneider, M. (2022). Domain-specific prior knowledge and learning: A meta-analysis. Educational psychologist, 57(1), 31-54.

---

[1] Note, this is an updated explanation of the role advance organisers play proposed by Gurlitt et al. (2012).