What is episodic memory if it is a natural kind?

Abstract

Colloquially, episodic memory is described as “the memory of personally experienced events”. Even though episodic memory has been studied in psychology and neuroscience for about six decades, there is still great uncertainty as to what episodic memory is. Here we ask how episodic memory should be characterized in order to be validated as a natural kind. We propose to conceive of episodic memory as a knowledge-like state that is identified with an experientially based mnemonic representation of an episode that allows for a mnemonic simulation thereof. We call our analysis the Sequence Analysis of Episodic Memory since episodes will be analyzed in terms of sequences of events. Our philosophical analysis of episodic memory is driven and supported by experimental results from psychology and neuroscience. We discuss selected experimental results that provide exemplary evidence for uniform causal mechanisms underlying the properties of episodic memory and argue that episodic memory is a natural kind. The argumentation proceeds along three cornerstones: First, psychological evidence suggests that a violation of any of the proposed conditions for episodic memory amounts to a deficiency of episodic memory and no form of memory or cognitive process but episodic memory fulfills them. Second, empirical results support a claim that the principal anatomical substrate of episodic memory is the hippocampus. Finally, we can pin down causal mechanisms onto neural activities in the hippocampus to explain the psychological states and processes constituting episodic memory.

Appendix

In Sect. 3.4 we distinguish between an actual and temporally enduring mnemonic representation and a possible and only instantaneous temporally explicit mnemonic simulation. The mnemonic representation is a compositional representation in which the temporal succession of the events in the episode is encoded by some structure among the representational constituents. This can be formally accounted for in the following way:

• (R1) Let ${\mathbb {R}}=\left( {\hbox {R}, \{\hbox {s}\}} \right)$ be a representational structure, where $\hbox {R}$ is the set of representational states and $\hbox {s}:\hbox {R}\times \hbox {R}\rightarrow \hbox {R}$ a thereon defined binary and recursive structure building operation; if $\hbox {r}^{{\prime }{\prime }}=\hbox {s}\left( {\hbox {r}, \hbox {r}^{\prime }}\right)$, $\hbox {r}$ and $\hbox {r}^{\prime }$ are called the representational constituents of $\hbox {r}^{{\prime }{\prime }}$.

• (R2) Let ${\mathbb {V}}=\left( {\hbox {V}, \{\langle \cdot ,\cdot \rangle \}} \right)$ be an event structure where $\hbox {V}$ is the set of events (including episodes) and $\langle \cdot ,\cdot \rangle$ the previously defined binary and recursive episode building operation $\langle \cdot ,\cdot \rangle :\hbox {V}\times \hbox {V}\rightarrow \hbox {V}$.

• (R3) Let $\upmu :\hbox {R}\rightarrow \hbox {V}$ be a mapping from the set of representational states into the set of events. $\upmu (\hbox {r})=\hbox {e}$ means that $\hbox {e}$ is the event or episode represented by the primitive or complex representational state $\hbox {r}$.

• (R4) Then the mapping $\upmu$ is a homomorphism from the representational structure ${\mathbb {R}}$ into the event structure ${\mathbb {V}}$. That is, for every $\hbox {r}, \hbox {r}^{\prime }\in \hbox {R},\,\upmu \left( {\hbox {s}\left( {\hbox {r}, \hbox {r}^{\prime }}\right) }\right) =\langle \upmu ( \hbox {r} ),~\upmu (\hbox {r}^{\prime })\rangle$.

Statement (R4) is equivalent to the claim that the representational structure is compositional: The content $\upmu$ of a complex representational state $\hbox {s}\left( {\hbox {r}, \hbox {r}^{\prime }}\right)$ is a structure-dependent function, namely $\langle \cdot ,\cdot \rangle$, of the contents of its representational constituents $\hbox {r}$ and $\hbox {r}^{\prime }$, i.e., $\upmu \left( \hbox {r}\right)$ and $\upmu \left( {\hbox {r}^{\prime }}\right)$.

In a temporally explicit mnemonic simulation the temporal succession of events in the object domain is represented itself by a temporal succession of events in the representational domain. We can formally account for this in the following way:

• (Q1) Let ${\mathbb {V}}=\left( {\hbox {V}, \{\langle \cdot ,\cdot \rangle \}} \right)$ again be our event structure, where $\hbox {V}$ is the set of events and $\langle \cdot ,\cdot \rangle$ the episode building operation.

• (Q2) Let ${\mathbb {Q}}=\left( {\hbox {Q}, \{\langle \cdot ,\cdot \rangle \}} \right)$ be the simulational structure, which itself contains a set of events $\hbox {Q}$, i.e., $\hbox {Q}\subseteq \hbox {V}$, together with the operation $\langle \cdot ,\cdot \rangle$, which builds temporally succeeding sequences of events. ${\mathbb {Q}}$ hence is a substructure of ${\mathbb {V}}$.

• (Q3) Let the mapping $\upnu : \hbox {Q} \rightarrow \hbox {V}$ assign each primitive or complex simulational state the primitive or complex event it, by simulation, represents.

• (Q4) Then the mapping $\upnu$ satisfies the following condition: For every $\hbox {q},\hbox {q}^{\prime }\in \hbox {Q}, \upnu \left( \langle {\hbox {q}, \hbox {q}^{\prime }}\rangle \right) = \langle \upnu \left( \hbox {q} \right) , \upnu (\hbox {q}^{\prime })\rangle$.

In this account, the representational content of a temporal succession of simulational states is just the episode comprising the temporally succeeding representational contents of each simulational state. Mathematically speaking, $\upnu$, unlike $\upmu$, is not only a homomorphism but an endomorphism.