Updated on
October 9, 2026
Top-Down vs Bottom-Up Processing: Definitions and Examples
Top-down processing uses prior knowledge and context; bottom-up processing builds meaning from sensory detail. Compare both with clear classroom examples.
What is the difference between top-down and bottom-up processing?
Top-down vs bottom-up processing describes two influences on perception: existing knowledge and expectations, and incoming sensory information. Top-down processing helps interpret what you encounter using context. Bottom-up processing builds from details such as letters, sounds or shapes. Both influences usually work together, with sensory evidence checking expectations.
Write a shape between A and C and it may look like the letter B. Put the same shape between 12 and 14 and it may look like the number 13. The marks stay the same, but the surrounding sequence changes what you expect. In class, the same tension appears when a learner reads the word they expected instead of the word on the page.
For psychology learners and trainee teachers, the useful question is how evidence and expectations shape the same perception. This guide compares the theories, their history, everyday examples and the research that tests them. Reading is one application, explored in our wider guide to reading comprehension strategies. Classroom examples give year groups in UK and US terms. If you mean a top-down or bottom-up approach to management or curriculum decisions, see the teaching and leadership approaches answer below.
Key takeaways
- Top-down processing is knowledge-driven. Bottom-up processing is driven by incoming information.
- Neither route is better. Accurate understanding usually depends on both.
- Prior knowledge can speed interpretation, but it can also produce confident mistakes.
- Detailed decoding matters when material is unfamiliar, ambiguous or technically precise.
- These are descriptions of processing, not fixed learner types or diagnoses of special educational needs and disabilities (SEND).
On this page
- What is the difference between top-down and bottom-up processing?
- What is top-down processing?
- What is bottom-up processing?
- History: Gregory, Gibson and interactive models
- Examples of top-down and bottom-up processing in everyday life
- How top-down and bottom-up processing work together
- Limitations of the top-down and bottom-up distinction
- Classroom application: reading and learning
- Check your understanding
- Frequently asked questions
- References
What is the difference between top-down and bottom-up processing?
The main difference between top-down and bottom-up processing is the source of influence. Top-down processing draws on knowledge and expectations, while bottom-up processing draws on details reaching the senses. Both can shape the same experience. The table compares what each contributes and where each can mislead.
| Question | Top-down | Bottom-up |
|---|---|---|
| Where does it start? | What you already know, expect or are trying to do | The letters, sounds, features or measurements reaching you |
| What does it help with? | Prediction, inference, gist and connecting details to a wider idea | Accurate decoding, discrimination, observation and checking detail |
| What can go wrong? | An expectation overrides contradictory evidence | Details remain separate from a useful wider interpretation |
| How can you check it? | Identify the knowledge or context behind the expectation | Locate the exact feature that supports or changes the interpretation |
Neither direction is better. Gregory and Gibson are often taught as rival theories. The interactive models in the next section treat perception as both at once: the detail constrains the guess, and the guess helps organise the detail.
What is top-down processing?
| Term | Definition |
|---|---|
| Top-down processing | Top-down processing uses prior knowledge, expectations and goals to shape how you interpret sensory information, especially when the signal is incomplete, ambiguous or mixed with noise. |
| Sensation | Sensation concerns the incoming signal. |
| Perception | Perception concerns how that signal is organised and understood. |
| Schema | A schema is an organised body of knowledge that helps supply an expectation. |
The psychologist Richard Gregory described perception as a set of hypotheses in The intelligent eye (1970). He developed the account in Gregory (1980). A hypothesis here means a possible explanation of the signal. Visual illusions show how an interpretation can diverge from the stimulus.
Two classic studies show the effect. Warren (1970) replaced a speech sound in a recorded sentence with a cough. Listeners believed they heard the missing sound, and the cough seemed to occur at a different point in the sentence. When silence replaced the sound instead, they noticed the gap.
Image description
Two aligned schematic strips mark the same missing speech-sound location. A cough replaces it in one condition; silence replaces it in the other. Sound can seem present with a cough and a gap can be noticed with silence. These are qualified reported perceptions, not universal responses. There is no time/amplitude scale, invented sentence, simulated audio or participant data.
Palmer (1975) showed people a scene, then briefly showed a drawing of a real-world object. The scene they had just seen influenced how well they identified the object.
What is bottom-up processing?
Bottom-up processing builds an interpretation from incoming features such as lines, edges, sounds or letters, which constrain the words, objects and patterns you can recognise. Think of letters combining into words, or sounds into syllables. Familiarity and task demands affect speed. The term identifies a source of information rather than a fixed pace.
The psychologist James Gibson went furthest in this direction. He argued that perception is direct: the mind picks up what the environment offers without having to construct a picture from guesses (Gibson, 1979). He also introduced the idea of affordances, meaning what the environment offers the person perceiving it. A chair affords sitting; a handle affords pulling.
Bottom-up detail becomes especially useful when context offers little help. Reading a new surname, checking a telephone number or learning an unfamiliar alphabet requires close attention to what is present. Existing knowledge may still help you recognise a letter or a familiar spelling pattern.
Image description
An illustrative E/F comparison enlarges the bottom-stroke difference. It demonstrates feature inspection, not an anatomical pathway. The caption preserves the contribution of letter knowledge, so the image does not imply pure processing without learning.
What is feature detection in bottom-up processing?
Feature detection means responding to a particular property of a signal, such as an edge or line, so that information about parts can contribute to recognising an object. Consider the strokes that distinguish one printed letter from another. The question is which features are present before you choose a letter.
McClelland (2013) reviews models with feature, letter and word levels. Signals pass between levels in both directions. Such models help explain letter recognition, while recordings of activity in the nervous system provide a separate kind of evidence. The Walsh et al. (2020) review tests how strongly neural findings support predictive accounts. A model diagram and a neural recording answer different questions.
History: Gregory, Gibson and interactive models
The history of top-down and bottom-up processing brings together Gregory’s work on how we use knowledge and Gibson’s account of direct perception. Later models link context with detail. The theories ask different questions. They explain perception in different ways. The dates below place each view in its own setting.
| Date | Account | Contribution |
|---|---|---|
| 1970 and 1980 | Gregory: The intelligent eye, then Perceptions as hypotheses | Perception uses possible explanations informed by stored knowledge. |
| 1979 | Gibson: The ecological approach to visual perception | Perception picks up information available in the environment, including affordances. |
| 1981 | McClelland and Rumelhart’s model | Feature, letter and word levels interact rather than forming a purely one-way chain. |
| 2013 | Clark’s predictive-processing account | Perception involves testing predictions against incoming signals. |
Gregory vs Gibson: hypotheses or direct perception?
Gregory argues that we use what we know to form a best guess, while Gibson stresses what the world offers directly to the person who sees it. Gregory asks why the same signal can mean different things. Gibson asks what we can pick up directly from the world. Later models let signals and knowledge shape the same result.
Examples of top-down and bottom-up processing in everyday life
Everyday examples of top-down and bottom-up processing include spotting a familiar face, hearing a tune in noise and reading an unclear mark. Each draws on both the signal and what you know. To explain a case, separate what you saw or heard from what you expected. Then ask what extra detail could change your answer.
| Example | Bottom-up detail | Top-down influence |
|---|---|---|
| Recognising a face | Visible features such as the outline of the eyes and mouth | Knowing the person or expecting to meet them supplies a possible identity. A closer view can change it. |
| Following music in noise | The pitch, timing and loudness of the sounds that reach you | Knowing a melody helps you anticipate a phrase. Listening closely checks whether the expected notes are present. |
| Interpreting an ambiguous mark | The same visible strokes between different surrounding symbols | The letter sequence A and C suggests B; the number sequence 12 and 14 suggests 13. |
A face or a tune you know can give you a useful first guess. The shape you see or the notes you hear can help check it. On a new task, the balance may change. Follow what each source adds to the answer.
How top-down and bottom-up processing work together
Top-down and bottom-up influences work together when a possible interpretation is checked against incoming information, and the result helps refine what the observer expects to see or hear. The distinction points to two sources of information. Interactive models explain how they can shape the same experience.
In predictive processing, expectations help interpret a signal. A prediction error is the mismatch between an expected signal and the signal that arrives. Clark (2013) describes how such mismatches could help update an interpretation. Rauss and Pourtois (2013) discuss how the top-down and bottom-up terms fit predictive coding. This account links predictions with the incoming signals used to check them.
Reading comprehension provides a clear example. Visual features help us identify letters, and letters help us identify words. At the same time, a familiar word can make one of its letters easier to recognise. The interactive activation model developed by McClelland and Rumelhart (1981) explains this two-way flow between features, letters and words.
This interaction helps explain why a reader may read a familiar sentence fluently but slow down at an unfamiliar name. Context narrows the possibilities. Attention to the letter sequence helps check whether the prediction fits. Skilled performance depends on a useful exchange between knowledge and evidence.
Diagram · Build It · Perception
Two directions, one interpretation
Knowledge flows down, evidence flows up, and what you perceive is the interpretation that fits both. When they disagree, the evidence should win.
- A 13 C
Knowledge decides
The same written shape reads as the letter B between A and C, and as the number 13 between 12 and 14. The marks do not change; the expectation does.
- A cough in a word
Knowledge fills a gap
When a cough replaces a speech sound in a recorded sentence, listeners believe they hear the missing sound. When silence replaces it, they notice the gap (Warren, 1970).
- An unfamiliar name
Evidence leads
With little context to go on, a reader has to work through the letters of a new name. Knowledge checks the result afterwards: does it sound like a name I have heard?
Drawn from interactive models of perception such as McClelland and Rumelhart (1981) and the predictive-coding account in Rauss and Pourtois (2013).
The diagram shows an interpretation being checked and revised. Expectations help make sense of a signal, while the signal can challenge them. In a psychology example, identify both influences and explain what happens when they disagree. This gives a clearer account than naming one route alone.
Limitations of the top-down and bottom-up distinction
The terms are used in different ways across perception, attention, reading and education. A process called top-down in one model may not fit the same label in another. Reviews of predictive coding and attention warn that a simple split can hide other effects.
Rauss and Pourtois (2013) point out that both terms are used everywhere in psychology and neuroscience but rarely defined, and propose recasting them in terms of predictive coding. Awh, Belopolsky and Theeuwes (2012) call the split in attention research a failed dichotomy, meaning an overly simple division into two categories. Stimuli associated with reward can capture attention even when that works against a person's current goals, which neither goals nor the features that stand out can fully explain. They add past selection history, meaning what previously drew someone's attention, as a third source of bias.
Image description
Three unranked contributors point towards an attention-selection label. The illustration refers to the attention framework in Awh, Belopolsky and Theeuwes, not a universal three-part theory of perception or learner type.
What are the critiques of predictive processing?
Critics of predictive processing ask whether studies of the brain show the specific processes the theory claims, and whether all predictions have to come from higher levels. The challenge is to test the process. An effect of expectation alone leaves that question open.
| Question | Critique |
|---|---|
| Neural evidence | Walsh and colleagues (2020) examine studies of brain activity. They weigh how far the findings support its key claims. |
| Sources of prediction | Teufel and Fletcher (2020) show that predictions can also be built into incoming processing. A prediction may thus reflect more than one kind of influence. |
| Testing the theory | To test the theory, ask which finding would support it over a rival account. |
Is the top-down vs bottom-up split an oversimplification?
The top-down and bottom-up split is too simple when it leaves out past experience, reward or current goals, because these can jointly shape what attracts our attention. It is useful to name two influences. It is less useful to treat them as separate kinds of person. State what the labels mean in the model you are using.
Saghravanian (2025) argues for one system of attention shaped by needs over three timescales: evolution, what we learn and the task we face now. The paper reviews ideas and evidence. It presents an account to test, rather than reporting a new experiment.
For a psychology answer, say whether you are discussing what we see, what we attend to or how we read. Name the signal and the knowledge behind a guess. Showing that context changes a guess supports an effect of context. To explain the brain process that caused it, we need more evidence.
Classroom application: reading and learning
How does top-down processing affect reading?
Reading is where the two directions have caused the fiercest argument in education. In the 1960s, Goodman (1967) described reading as a psycholinguistic guessing game. This language-based account proposed that a good reader samples the print and predicts the rest. The same idea, that readers predict words from meaning, sits behind three-cueing. This prompts beginners to identify a word from the picture or the sentence.
Meaning, structure and visual cues (MSV) are the three sources used in three-cueing. Meaning comes from context, structure from grammar and visual cues from what the reader sees. The quoted posts below use this shorthand when discussing word recognition.
Stanovich (1980) reviewed the research on good and poor readers and reached a different conclusion. His interactive-compensatory model proposes that when one process is weak, a reader leans harder on the others. Beyond the first stages of learning to read, rapid, context-free word recognition and general comprehension strategies distinguished stronger readers. Stanovich (1984) set out how less-skilled readers can use context as much as, or more than, skilled readers when the task makes it available.
The bar counts only the eight studies checked for this guide, not the whole literature. Seven of the eight were published between 1979 and 1986, so the picture is an old one. Newer work on reading has built on it rather than replaced it. The one clear dissent, from Willows and Ryan (1981), found that skilled readers made more use of grammar and meaning when filling gaps in text. That is a different task from recognising a word.
The finding needs one refinement. Nation and Snowling (1998) ran two studies with children aged 7 to 10 (roughly UK Years 3 to 6, US Grades 2 to 5).
| Study | Finding |
|---|---|
| First study: 92 children | better readers gained relatively more from sentence context, and comprehension predicted that gain better than decoding did. |
| Second study | In the second, children with dyslexia used context to compensate for poor decoding, while children with poor reading comprehension benefited from context less than typical readers. |
So context is not the enemy. Leaning on it to identify words is a sign of weak decoding; using it well depends on comprehension.
That distinction is the teaching point. Teach the code so learners read words from their letters.
A meta-analysis combines results from several studies. The effect size, d, measures the size of the difference between groups. A review of systematic phonics found a moderate overall effect on reading (d = 0.41). Effects were larger when teaching began early (d = 0.55) (Ehri et al., 2001).
Then use context, pictures and knowledge to work out what the words mean. Our guides to the science of reading, theories of reading and the Simple View of Reading set out the wider debate.
What teachers are saying · September 2026
Context: clue or crutch?
Nobody in these posts says top-down or bottom-up. They argue about the same thing in other words: should context help a learner identify a word, or explain what it means? Three posts from September 2026, quoted exactly.
Many of my students could read the words aloud, even those they were not familiar with, but they could not use context clues to determine the meaning of the words. I taught during the "whole language" era.
The post you are amplifying describe MSV/three-cueing. This is not supported as a model of skilled word recognition.
- objection
Guessing from context clues is a valuable skill that adults do without thinking (like when you're reading street signs from far away) and also kids gotta be able to sound out words to help them identify them.
The retired teacher's point is the one Nation and Snowling's results predict: pronouncing a word is not the same as knowing what it means. The objection is also fair. Skilled adults do use context when the print is degraded, which is exactly what the compensatory model says happens when the letters cannot be read.
When prior knowledge helps understanding
Top-down processing is only as good as the knowledge behind it. The evidence that relevant prior knowledge and schemas help comprehension is strong, and some of it is striking.
Recht and Leslie (1988) divided 64 US junior high school learners (roughly UK Years 8 to 10) by reading ability and by how much they knew about baseball. The learners then read an account of part of a game. Prior knowledge had a significant effect on every measure of recall and summary. It did not interact with reading ability, so it helped weaker and stronger readers alike.
Knowledge also helps at the level of the word. Priebe, Keenan and Miller (2012) matched children on decoding skill and varied only whether they knew the topic of a passage. In poor readers, knowing the topic increased fluency and reduced reading errors, especially errors based on the letters.
Reviews add the conditions. Smith et al. (2021) reviewed 23 studies of primary-aged readers and found that the effect of background knowledge depends on the text and on any misconceptions the reader holds. Hattan, Alexander and Lupo (2024) reviewed 54 articles and identified 30 techniques for activating prior knowledge. How well a technique worked depended partly on the amount, accuracy and specificity of what learners already knew. Activating vague knowledge does little; activating the right knowledge does a lot.
How does unequal prior knowledge affect understanding?
Unequal prior knowledge can give readers different starting points. A text’s topic, vocabulary and assumptions may be familiar to one reader and unfamiliar to another, even when their decoding skill is similar. The baseball study above shows why topic knowledge matters. Knowledge of the subject and skill in reading the words answer different questions.
A familiar sport, story or technical term can supply an expectation that helps organise a passage. When that experience is missing, the text asks the reader to build more from its details. A useful classroom check separates what learners know about the topic from what they can decode on the page.
When top-down processing causes errors
Prior knowledge is useful only when it is relevant and sufficiently accurate. Expectations can make an ambiguous signal easier to interpret, but they can also make contradictory details easier to miss. In the classroom, a learner may read the word they expected rather than the word on the page. They may assume a graph confirms a familiar claim, or treat a new problem as identical to a worked example. Use three checks:
- Locate:What information is actually present?
- Interpret:What does that information mean here?
- Reconcile:Does it support or change the first expectation?
These checks make top-down influence accountable to evidence. Learners compare what they expected with the information actually present.
Research shows how stubborn the effect can be. In a study by Lipson (1982), young readers who held wrong prior knowledge used it rather than the text, even when the text contradicted them. They were better at learning new information than at correcting old information. That is why a misconception needs to be named and tested directly, not simply overwritten by the correct version.
Posts on X · September and October 2026
Guessing from context
Four short posts from X users on the same question: when does using context to read a word help, and when does it replace the print?
- example
We see a lot of kids guess words from the pictures.
- angle
context clues help you understand what the word means, not how to pronounce it.
- angle
context reading is a real skill. It is used best to fill in blanks in an otherwise relatively robust reading skill.
- objection
Most kids do not need intensive phonics to learn to read. They need useful phonics, word recognition, and context.
Examples of top-down and bottom-up processing in class
Top-down processing is visible when learners use a known story pattern, concept or problem type to predict meaning. Bottom-up processing is visible when they inspect the exact word, sound, measurement or step. The strongest tasks require an initial interpretation and a clear check against the available evidence.
Reading
| Influence | Reading example |
|---|---|
| Top-down influence | A learner sees the sentence, “The gardener planted the bulbs in autumn.” Knowledge of gardens and seasons helps them anticipate the meaning. That is top-down influence. |
| Bottom-up processes | A grapheme is a letter or letter group that represents a speech sound, such as b or sh. Recognising the letter sequence in bulbs, distinguishing it from buds and blending the sounds are bottom-up processes. Context can compensate for weak decoding (Stanovich, 1980). A meta-analysis found that planned, explicit teaching of letter-sound relationships supports learning to read (Ehri et al., 2001). |
Listening
Before a weather forecast, you expect words about rain, wind and temperature. That gives you a frame for what you hear. To note the time of a warning, you must listen to the exact words and numbers. For a learner working in a second language, a short phrase may need several listens. Pair a first guess about the message with a check of the words.
Science
A learner may know that metals conduct electricity. That prior knowledge helps them predict which material will complete a circuit. The result of the test is bottom-up evidence.
If the bulb stays dark, they should inspect the circuit, connections and material before accepting the original prediction. The teaching value lies in the comparison between “what I expected” and “what I observed”.
Mathematics
Recognising that a problem concerns proportion is top-down because it draws on a known structure. Reading the units, locating the given values and checking the operation are bottom-up checks on that interpretation. A learner who rushes to a familiar method may solve the wrong problem. A learner who records every number without recognising the structure may struggle to choose an operation.
Worked examples at two phases
Year 3 (US Grade 2): reading a science text
Illustrative worked example
Year 3 science: reading a seed-growth text
- Task and first responseThe class read a short text about how seeds grow. Before reading, the teacher asked one specific question: what do plants need to grow? Learners suggested water and light. That knowledge helped them follow the text, which is the top-down half.
- Check the printed wordOne learner read seed as bean, because the picture showed a bean. The teacher covered the picture and asked the learner to read the word again from its letters: s, ee, d, seed. The learner then read seed from its letters. Only then did the class go back to the picture to check what the sentence meant.
- What the completed response showedThe revised reading matched the printed word. The routine kept the two jobs in order: letters to identify the word, knowledge and pictures to understand it. When a picture overrides the print, cover it and return to the letters before checking meaning.
Year 9 (US Grade 8): reading a graph in science
Illustrative worked example
Year 9 science: checking a reaction-rate graph
- Task and predictionLearners are given a graph titled Temperature and rate of reaction. Most predict that the rate keeps rising with temperature, because that is what they learnt last term. The teacher asks them to write the prediction down, then to read the axes, the units and the last three points before saying anything.
- Read the evidence and reviseThe line rises, then falls above about 40 °C, because the reaction is controlled by an enzyme that stops working at high temperatures. Learners who read the data first spot the fall; learners who trusted the prediction describe a steady rise. The discussion is about the gap between the two, which is the top-down error made visible and corrected by bottom-up evidence.
Five teaching strategies that connect knowledge with evidence
Activate specific prior knowledge
Ask for the knowledge that the new task needs. “What do you remember about evaporation?” is more useful than “What do you know about the water cycle?” The first question narrows attention to a relevant idea.
Prior knowledge often supports later work, although its measured effect depends on the task and assessment (Dochy et al., 1999). Follow activation with an example, text or result. Ask learners which part confirms, extends or contradicts what they recalled.
Model the code before asking for meaning
Make hidden conventions visible through explicit instruction. In reading, this can mean modelling a spelling pattern. In graphs, it can mean checking the title, axes, units and scale before drawing a conclusion. In algebra, it can mean identifying the variable and operation before attempting a solution.
The purpose is to make the details reliable enough to support a sound interpretation. Learners can then connect the parts to the meaning of the task.
Compare close examples
Place two examples side by side that differ in one important feature. Compare a correlation with a causal claim, a square with a non-square rhombus, or two sentences separated by one punctuation mark. Ask learners to identify the difference and explain why it changes the meaning.
This sequence joins bottom-up discrimination to top-down understanding. The learner first notices the feature, then links it to a rule or concept.
Separate prediction from observation
Before a demonstration, ask learners to record a prediction and its reason. Afterwards, ask them to record only what they observed. Finally, ask what the evidence does to the prediction. Keeping these statements separate supports metacognition in the classroom and makes it harder for expectations to rewrite the result.
Fade prompts when the routine is secure
Use a short checklist while learners acquire a way of inspecting a task. Remove one prompt at a time when their work shows that the routine is becoming reliable. The checklist is temporary support. Fading it as knowledge grows can release working memory for the meaning of the task.
The slide deck below turns these strategies and the psychology behind them into a set of slides for a staff briefing or a psychology lesson. It includes visual illusions, the reading research and four rules for lesson design.
Identifying the processing demand
Look at the task and the error before describing the learner. Use brief formative assessment to test whether the missing part is prior knowledge, vocabulary, decoding, sensory access or attention to evidence. Change one feature at a time, then check whether the learner’s response changes.
| Learner signal | Next teaching move |
|---|---|
| An answer the source does not support | Ask which exact word, value or feature supports the answer. Return to the source and mark the evidence before keeping or changing the answer. |
| Details listed, main idea missed | Give the main idea first. Use a frame, worked example or concept map to show how the details fit it. |
| Understands when told, cannot read the terms | Teach the relevant code or key terms. Rehearse them in context, then return to the same text. |
| A familiar method on the wrong problem | Put the problem beside a near-miss example. Ask which feature differs before any working starts. |
| A word guessed from the picture | Cover the picture and read the word from its letters. Then use the picture and sentence to check its meaning. |
| Numbers recorded, operation unclear | Ask which known problem structure connects the values. Check the units and given information before choosing an operation. |
Mini app · Build It · Processing demand
What is the task asking for?
Choose what you saw and the phase you teach. You get the likely demand, a question to test it, a next move and an example. It describes the task, not the learner.
A planning aid built from the task analysis on this page. It does not identify a learning difficulty. If a pattern persists after good teaching, talk to your special educational needs co-ordinator (SENCO) or special education team.
The planner analyses the task and the observed response. A learner may depend on context in one subject and careful detail in another. Prior knowledge, vocabulary, familiarity, sensory access and the quality of the explanation all affect the route they can use.
Top-down and bottom-up processing in SEND
Do not label a learner as a “top-down processor” or “bottom-up processor”. The terms describe what shapes a task, not fixed learner types. They do not diagnose dyslexia, autism, attention difficulties or a sensory impairment.
Start with access. Check whether the learner can see or hear the material clearly, recognise the code, understand the vocabulary and retrieve the knowledge the task assumes. Then alter one part of the task and observe what changes. For example, provide the main idea before a dense source, read a question aloud without changing its wording, or use a graphic organiser while keeping the intellectual demand.
If a decoding difficulty is persistent, use the school’s assessment and support process rather than inferring a cause from this model. If a learner understands the topic when it is explained but cannot read the text accurately, that discrepancy is useful evidence for the next professional conversation. It is not a diagnosis on its own.
A short planning routine
Before teaching, write the lesson’s main idea in one sentence. List the knowledge learners need to understand that sentence. Then list the details, codes or conventions they must read accurately. This prevents a broad discussion of prior knowledge from replacing the specific preparation the task needs.
- Frame the whole.State what learners are trying to understand or produce.
- Expose the parts.Model the features, words or data that carry the meaning.
- Make a prediction.Ask learners what they expect and why.
- Check the source.Require a word, feature, result or step that supports the answer.
- Revise the model.Ask what the new evidence adds or changes.
Use the routine selectively. A familiar task may need only a quick evidence check. New or ambiguous material may need a worked model and several close examples.
The aim is to make the route from knowledge to evidence visible enough to teach. Choose the step that addresses the task in front of the learner.
Try it in your next lesson that involves reading. Before learners use any context, ask them to read the key word from its letters. Then ask what the word means in the sentence. For the comprehension strategies that build on that first step, go back to our guide to reading comprehension in the classroom.
Check your understanding
Five questions check the definitions, the examples and the classroom routine. Each answer explains why it is right, so the quiz works as revision for a psychology course as well as a check for teachers.
Print the revision sheet. Use its three cases to practise separating the incoming detail from the expectation.
Frequently asked questions
How does unequal prior knowledge affect top-down processing?
Relevant prior knowledge can help readers interpret a text, giving them different starting points for understanding. In Recht and Leslie’s (1988) baseball study, topic knowledge helped both weaker and stronger readers. Knowing the subject and reading its words accurately are separate contributions to comprehension.
What does feature detection show about bottom-up processing?
Feature detection describes a response to a particular property of a signal, such as an edge or line. McClelland (2013) reviews models that connect features, letters and words in both directions. These models explain interactions in recognition, while neural recordings provide a different kind of evidence about brain activity.
What is an example of top-down processing?
Reading a shape as the letter B between A and C, but as 13 between 12 and 14, is a classic example. So is hearing a sound that a cough has replaced in a sentence (Warren, 1970). In class, a learner who predicts a story's ending from its genre is using top-down processing.
What is an example of bottom-up processing?
Sounding out an unfamiliar word letter by letter and reading a graph’s scale before interpreting it are bottom-up processing. So is noticing that a gap in speech is silence rather than a sound. Each builds an interpretation from the details rather than from an expectation.
Which is better, top-down or bottom-up processing?
Neither. Perception and learning depend on both, working together. Top-down processing makes interpretation fast; bottom-up processing keeps it accurate. Problems arise when one runs without the other, such as a strong expectation that is never checked against the evidence.
Is top-down processing the same as three-cueing?
No. Three-cueing is a teaching method that prompts beginners to identify words from pictures, sentence meaning and grammar. It rests on the older idea that skilled readers rely on prediction. The research reviewed by Stanovich (1980) points the other way: skilled readers recognise words from their letters, and context helps most with meaning.
Is top-down vs bottom-up processing the same as top-down vs bottom-up learning?
Top-down and bottom-up processing describe influences on perception. Top-down and bottom-up learning can describe how teaching moves between whole ideas and their parts or specific examples. A lesson might start with a model, then examine its details. Another might start with examples, then build a general rule. Either sequence can involve both sources of influence on perception.
What are top-down and bottom-up approaches in teaching or school leadership?
In teaching and school leadership, a top-down approach can mean starting from a whole plan or central decision. A bottom-up approach can mean building from examples or local contributions. These describe organisation or teaching sequence. The processing comparison on this page concerns perception. A centrally planned lesson still involves both detail and expectations when someone reads its material.
How do I know if I’m a bottom-up processor?
You can describe the information you used on a particular task. A label such as bottom-up processor overlooks how your knowledge, the context and the available signal change between tasks. Ask which details shaped an interpretation and which expectations helped. The quiz checks understanding of the terms. The mini app analyses classroom tasks, rather than identifying a fixed type of person.
Are top-down and bottom-up processing learning styles?
No. They describe how a task is processed, not a fixed type of learner. The same learner uses top-down processing on a familiar topic and bottom-up processing on an unfamiliar one, so matching teaching to a supposed processing style has no basis.
References
- Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences.
- Gregory, R. L. (1980). Perceptions as hypotheses. Philosophical Transactions of the Royal Society of London. B, Biological Sciences.
- McClelland, J. L. (2013). Integrating probabilistic models of perception and interactive neural networks: a historical and tutorial review. Frontiers in Psychology.
- Saghravanian, S. J. (2025). Rethinking attention: A unified perspective on top-down and bottom-up processes. Acta Psychologica.
- Stanovich, K. E. (1984). The interactive-compensatory model of reading. Remedial and Special Education.
- Teufel, C. and Fletcher, P. C. (2020). Forms of prediction in the nervous system. Nature Reviews Neuroscience.
- Walsh et al. (2020). Evaluating the neurophysiological evidence for predictive processing as a model of perception. Annals of the New York Academy of Sciences.
- Awh, E. S., Belopolsky, A. V. and Theeuwes, J. (2012). Top-down versus bottom-up attentional control: a failed theoretical dichotomy. Trends in Cognitive Sciences, 16(8), 437 to 443.
- Dochy, F., Segers, M. and Buehl, M. M. (1999). The relation between assessment practices and outcomes of studies: the case of research on prior knowledge. Review of Educational Research, 69(2), 145 to 186.
- Ehri, L. C., Nunes, S. R., Stahl, S. A. and Willows, D. M. (2001). Systematic phonics instruction helps students learn to read: evidence from the National Reading Panel’s meta-analysis. Review of Educational Research, 71(3), 393 to 447.
- Gibson, J. J. (1979). The ecological approach to visual perception. Boston: Houghton Mifflin.
- Goodman, K. S. (1967). Reading: a psycholinguistic guessing game. Journal of the Reading Specialist, 6(4), 126 to 135.
- Gregory, R. L. (1970). The intelligent eye. London: Weidenfeld and Nicolson.
- Hattan, C., Alexander, P. A. and 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 to 111.
- Lipson, M. Y. (1982). Learning new information from text: the role of prior knowledge and reading ability. Journal of Reading Behavior, 14(3), 243 to 261.
- McClelland, J. L. and Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: I. An account of basic findings. Psychological Review, 88(5), 375 to 407.
- Nation, K. and Snowling, M. J. (1998). Individual differences in contextual facilitation: evidence from dyslexia and poor reading comprehension. Child Development, 69(4), 996 to 1011.
- Palmer, S. E. (1975). The effects of contextual scenes on the identification of objects. Memory and Cognition, 3(5), 519 to 526.
- Priebe, S. J., Keenan, J. M. and Miller, A. C. (2012). How prior knowledge affects word identification and comprehension. Reading and Writing, 25(1), 131 to 149.
- Rauss, K. and Pourtois, G. (2013). What is bottom-up and what is top-down in predictive coding? Frontiers in Psychology, 4, 276.
- Recht, D. R. and Leslie, L. (1988). Effect of prior knowledge on good and poor readers' memory of text. Journal of Educational Psychology, 80(1), 16 to 20.
- Smith, R., Snow, P., Serry, T. and Hammond, L. (2021). The role of background knowledge in reading comprehension: a critical review. Reading Psychology, 42(3), 214 to 240.
- Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading fluency. Reading Research Quarterly, 16(1), 32 to 71.
- Warren, R. M. (1970). Perceptual restoration of missing speech sounds. Science, 167(3917), 392 to 393.
- Willows, D. M. and Ryan, E. B. (1981). Differential utilization of syntactic and semantic information by skilled and less skilled readers in the intermediate grades. Journal of Educational Psychology, 73(5), 607 to 615.
