Episodic vs semantic memory: what Tulving meant, and where the line blurs

Episodic memory is your memory of events you personally experienced, tied to when and where they happened. Semantic memory is your general knowledge of facts, words and concepts, with no record of when or where you learned them. Remembering your first psychology lecture is episodic; knowing what "cognitive dissonance" means is semantic. Both are forms of explicit, or declarative, long-term memory. Endel Tulving named the distinction in 1972.

The short definitions are easy. The hard part is the exam question that sits on the line between the two, and the fact that Tulving changed his own definition a great deal after 1972. Here is the distinction as he drew it, a test that sorts most examples, and the places where it still blurs.

• Tulving's 1972 chapter defined episodic memory as memory for "temporally dated episodes or events" and semantic memory as a "mental thesaurus" of words, concepts and rules.
• Both are explicit (declarative) memory. Procedural memory, such as how to ride a bike, is implicit and belongs to neither.
• By 2002, Tulving defined episodic memory partly by how recall feels: you re-live the event in your mind. A semantic fact only feels known.
• Facts about your own life, such as your date of birth, are usually semantic, not episodic. Researchers call them personal semantics.
• Whether the two are separate brain systems is still argued. Tulving said yes; several critics said one system is enough.

The difference in one table

A quick memory aid: an episode is one scene from your life, like an episode of a series. Semantic comes from semantics, the study of what words mean, and semantic memory is where word meanings live.

Where the distinction came from

The distinction started at a conference. In March 1971, Tulving and Wayne Donaldson ran a meeting in Pittsburgh on organisation in memory. Three of the papers there talked about "semantic memory", a term from M. Ross Quillian's 1966 doctoral dissertation, which was never published. Tulving thought these speakers described something different from the memory that psychologists had studied since Ebbinghaus. He went home and wrote a chapter about the difference for the conference book, Organization of Memory, published in 1972.

He also needed a name for the memory that was not semantic. He did not like "event memory", because another field already used it. In a small book by S. Munsat he found the phrase "nonepisodic memory", and he took the positive form: episodic.

Tulving did not claim the idea was new. In a 1987 commentary on the chapter, he wrote that the distinction "was not at all original": philosophers and researchers of memory disorders had made similar points before. He compared his readers to the character in Molière's play who finds out he has spoken prose all his life. Memory researchers were pleased to learn what they had been studying.

How Tulving changed the definition

The 1972 version was modest. Tulving called it a "pre-theoretical" position. It sorted memory by the kind of information it holds: personal events on one side, general knowledge on the other. He did not yet claim that two different systems existed in the brain. He also described the two as "partially overlapping", and in 1983 as "closely interdependent".

In his 1983 book, Elements of Episodic Memory, he went further. He proposed that episodic and semantic memory were two functionally separate systems. He also made a confession. In 1972 he had assumed that the classic study-and-test experiments, the kind that go back to Ebbinghaus's forgetting-curve experiments, measured episodic memory. He later said they did not. Those experiments asked people what they remembered, but almost never where or when. They also ignored what remembering felt like.

That second point became the centre of the theory. In 1985, Tulving described three kinds of awareness:

• Autonoetic (self-knowing) awareness goes with episodic memory. You mentally travel back in time and re-experience the event.
• Noetic (knowing) awareness goes with semantic memory. You know the fact, but you do not re-live anything.
• Anoetic (non-knowing) awareness goes with procedural memory. You are aware only of the present.

By 2002, the definition rested on three ideas: a self, autonoetic awareness and a sense of subjective time. Tulving wrote that episodic memory "makes possible mental time travel through subjective time". He also wrote that its operations "require, but go beyond, the semantic memory system". In his account, episodic memory grew out of semantic memory, not the reverse.

Where they sit among the other types of memory

Episodic and semantic memory are the two branches of explicit memory. The open textbook Psychology 2e from OpenStax says explicit memory is also called declarative memory "because it can be put into words". Implicit memories are "not part of our consciousness". They include procedural memory, which "stores information about the way to do something", priming and emotional conditioning.

So if an exam asks for "episodic vs semantic vs procedural memory", the first two are explicit and the third is implicit. Episodic and semantic memory are both declarative.

Classifying exam scenarios

Two questions sort most examples:

1. 1.Is the memory tied to one time and one place in your own life?
2. 2.When you recall it, do you re-live the event, or do you only know the fact?

If both answers point to the event, the memory is episodic. If you only know the fact, it is semantic. If you can do something but cannot put it into words, it is procedural.

The third and fourth rows show the most useful point for students. One term can leave two memories. The lecture is an episode. The definition you take into the exam is semantic knowledge, and it no longer needs the lecture.

There is one caveat. Tulving himself did not think that a single memory always belongs to one system. In 2002 he wrote that episodic memory "is not a particular kind of memory task", and that most memory tasks draw on more than one system. Exams still ask you to classify examples, and the two questions above give the expected answer. But the theory describes systems, not labels for single memories.

Personal semantics: the category in between

Facts about your own past are the hardest cases. You know the name of your primary school, the street you grew up on and your date of birth. These facts are personal, like episodic memory. But you know them without re-living anything, like semantic memory.

Renoult and colleagues called this category personal semantics in 2012. It is usually counted as semantic memory. However, their review found that the brain activity behind personal semantics can look more like semantic memory, more like episodic memory, or like neither. The result depends on how researchers test it. For an exam, "semantic" is the usual answer. For the research, personal semantics is an open question.

The evidence that they can come apart

Tulving's strongest evidence came from people with brain injuries. Such cases are rare, because most brain damage affects both kinds of memory together.

The best-known case is K.C., a man Tulving's group studied in Toronto. At the age of 30, he had a serious head injury in a motorcycle accident. His intelligence and language stayed normal. He knew many facts about his own life, such as the address of his childhood home and the location of his family's cottage. He also kept most of the school knowledge he had learned before the accident. But he could not recollect a single event from his own life. He also could not imagine what he would do later that day or in the future. After the injury, he had great difficulty learning both new events and new facts.

A second line of evidence comes from Faraneh Vargha-Khadem and colleagues in 1997. They described three patients with damage to the hippocampus on both sides of the brain, from birth, by age 4 and at age 9. All three had pronounced amnesia for everyday events. Yet all three attended mainstream schools. Their language, literacy and factual knowledge were in the low average to average range. The authors concluded that episodic and semantic memory are "partly dissociable".

The opposite pattern also exists. In a 2010 review, Greenberg and Verfaellie describe semantic dementia as a severe loss of semantic memory with relatively unimpaired episodic memory. If you need the clinical terms that go with these cases, our article on how interference differs from anterograde and retrograde amnesia keeps them apart.

Are they really two separate systems?

Not everyone agrees that they are. Tulving's 1983 claim of two separate systems "quickly became controversial", in his own words. McKoon, Ratcliff and Dell published a critical evaluation of the distinction in 1986. Later critics, among them Glenberg in 1997 and Craik in 2001, argued that one memory system can explain the results. In their view, facts are easier to recall than events because people use them more often.

The K.C. case is open to the same objection. He may recall old facts better because he rehearsed them more often than single events. Tulving answered that K.C. also kept facts he had rarely rehearsed, such as the difference between stalactites and stalagmites. But Tulving agreed that the rehearsal explanation is logically possible.

Current researchers mostly stress how much the two depend on each other. Greenberg and Verfaellie concluded that the two forms of memory "can affect each other both at encoding and at retrieval". Renoult and Rugg, in a 2020 review of Tulving's work, report the trace transformation hypothesis: semantic knowledge emerges from episodic memories with time and experience. This is a hypothesis, not a settled result. An older form of the idea is Alan Baddeley's suggestion that semantic memory is the "accumulated residue" of many episodes.

The developmental cases also have limits. Renoult and Rugg note that people with developmental amnesia learn new facts slowly in laboratory tests. They also note that a damaged but working hippocampus may still support some of that learning.

What this means when you learn psychology terms

For a student, the practical point is the one from the exam table. Every term starts as an episode: a lecture, a page or a conversation. The goal is to turn it into semantic knowledge that you can use without the original scene. If Baddeley's "accumulated residue" idea is right, that knowledge comes from many encounters with the term, not from one.

That fits the evidence on studying. In a 2013 review of ten study techniques, Dunlosky and colleagues gave a high utility rating to only two: practice testing and distributed practice. In other words, test yourself instead of rereading, and spread the sessions out. Spacing reviews out explains why separate sessions beat one long session, and our guide to memorising psychology terminology shows how to use both methods for an exam.

Psychology World is built for this kind of repeated contact. You look up a term, save it to one of your own lists, and then drill the list as flashcards, as multiple-choice questions once the list has four or more terms, or by typing the answer. A daily word also brings a term back without a search. The four steps of the learning loop and what each study mode does are on the home page.