Lesson 5 of 11 · 15 min
The levels of processing model
Explain Craik and Lockhart's levels of processing model, classify processing as structural, phonemic or semantic, and explain why deep, elaborative encoding makes memories last.
What you will be able to do
- Describe the levels of processing model and the role it gives encoding in long-term memory.
- Classify a task as structural, phonemic or semantic processing.
- Distinguish maintenance rehearsal from elaborative rehearsal in the model.
- Describe investigations of encoding and deep (semantic) processing.
Think of a coin you have handled thousands of times. Can you say which way the Queen or King faces on it, or what is written around the edge? Most people cannot, although they have seen the coin far more often than the face of a friend they met once. Repetition alone did not make the memory. What mattered was what you did with the information when you met it: you only ever needed to know what the coin was worth.
Craik and Lockhart (1972) built a model on that idea. Instead of asking which store information is in, the levels of processing (LOP) model asks how deeply it was processed when it was encoded.
Definition
The levels of processing model
The levels of processing model (Craik & Lockhart, 1972) says that how well information is remembered depends on the depth to which it is processed at encoding. Shallow processing of its physical features leaves a weak, short-lived memory trace; deep processing of its meaning leaves a strong, lasting one. Memory is a by-product of processing, not a set of stores.
| Level | What is processed | Example question about the word BREAD | Memory |
|---|---|---|---|
| Structural (shallow; also called physical or visual) | What the word looks like: its letters, case, shape | Is the word in capital letters? | Poorest |
| Phonemic (intermediate; also called acoustic) | What the word sounds like | Does the word rhyme with head? | Better |
| Semantic (deep) | What the word means and how it links to other knowledge | Is it something you can eat? | Best |
Choose a level to see what is processed, then sort the orienting questions into their levels.
Classify each question asked about a word by the level of processing it requires.
Rehearsal in the model. Craik and Lockhart distinguished two kinds of rehearsal.
- Maintenance rehearsal repeats information at the same level, usually by its sound. It keeps the information available while you use it, but adds little to long-term memory.
- Elaborative rehearsal processes the information more deeply: linking it to its meaning, to what you already know, to examples and images. This is what makes lasting memories.
This is a sharp break from the multi-store model, where any rehearsal transfers information to LTM and more rehearsal means better memory. In the LOP model, ten repetitions of a phone number can leave less in memory than one moment spent noticing that it ends in your birth year.
Multi-store model
- Memory is a set of stores.
- Information reaches LTM by rehearsal: the more the better.
- Explains why recent items are recalled (recency) and why STM is limited.
Levels of processing model
- Memory is a by-product of processing; no separate stores.
- Information lasts according to the depth of encoding; repetition at a shallow level adds little.
- Explains why we remember meaningful things we never rehearsed.
Investigating encoding. If the stores encode information differently, errors in recall should show it. When people recall a list straight away, the words they confuse tend to sound alike; when they recall it later from LTM, the confusions tend to share meaning. Craik and Levy (1970) investigated encoding with lists of words for free recall, analysing recall by serial position. Semantic similarity between the words (words with related meanings) affected recall of the earlier list positions, which come from long-term memory, but not the recency portion, which comes from short-term memory. This fits the idea that STM relies mainly on acoustic encoding and LTM on semantic encoding.
Investigating deep processing. Elias and Perfetti (1973) gave participants an orienting task to do with each word in a list, without telling them a memory test would follow (incidental learning). For some participants the task required semantic encoding (thinking about the word's meaning); for others it required acoustic encoding (thinking about its sound). In a surprise recognition test, words encoded semantically were recognised better: deep processing produced better memory than phonemic processing, even though nobody had tried to learn.
Tip
Why incidental learning?
If participants know a test is coming, they may rehearse or use their own strategies whatever task they are given, and the level of processing is no longer controlled. Not telling them about the test (incidental learning) means the orienting task is the only processing the words get. It involves a mild deception, so participants are debriefed afterwards.
Answer a structural, phonemic or semantic question about each word, then take an unannounced recognition test. Results stay in this browser and are for learning only.
Worked example
Classify orienting tasks and predict recall
Three groups of participants saw the same 30 words. Group A counted the letters in each word. Group B thought of a word that rhymed with each. Group C rated how useful each object would be on a deserted island. None was told there would be a memory test. Then all were asked to recall the words.
Identify the level of processing in each group, and predict the order of their recall from best to worst, with a reason.
Step 1: Group A
Counting letters uses only the word's appearance: structural (shallow) processing.
Step 2: Group B
Finding a rhyme uses the word's sound: phonemic (intermediate) processing.
Step 3: Group C
Judging usefulness requires thinking about what each object is and does: semantic (deep) processing.
Step 4: Predict the order
Group C > Group B > Group A. Deeper processing at encoding produces a stronger, more durable memory trace (Craik & Lockhart, 1972).
Because no group expected a test, the orienting task is the only processing the words received, so differences in recall can be put down to the level of processing.
Try it yourself first, then check the working one step at a time.
Answer
A structural, B phonemic, C semantic. Recall: C best, then B, then A, because deeper processing gives more durable memories.
Method
Appearance → structural; sound → phonemic; meaning → semantic.
Predict recall in order of depth: semantic > phonemic > structural.
Watch out for
Calling rhyming deep processing: it uses sound, not meaning.
Which study method uses elaborative rehearsal?
Summary
The levels of processing model (Craik & Lockhart, 1972): memory depends on the depth of processing at encoding, not on stores.
Structural (appearance) is shallow, phonemic (sound) intermediate, semantic (meaning) deep; deeper processing gives longer-lasting memory.
Maintenance rehearsal repeats information at the same level; elaborative rehearsal processes its meaning and makes lasting memories.
Encoding studies such as Craik and Levy (1970) link STM with sound and LTM with meaning.
Elias and Perfetti (1973): words encoded by meaning were recognised better in a surprise test than words encoded by sound.
Finished the lesson?
Try the 2 quick checks above and the lesson completes itself, or mark it complete now.