Memory Models and Theories

Memory Models and Theories

This note provides an overview of memory models and theories, drawing from the works of Eysenck, Keane, Baddeley, and Anderson. It explores the complex nature of memory and poses questions about the implications of remembering everything versus not remembering at all, using examples such as Luis Borges' "Funes, the Memorious" and the real-life case of Clive Wearing to illustrate these contrasting conditions. The study of memory involves understanding the mechanisms and processes that allow us to encode, store, and retrieve information.

Memory Metaphors

Early conceptualizations of memory often use metaphors to explain its function. These include the wax table (Plato, Aristotle), gramophone (Pear), aviary (Plato), house with objects (James, Freud), library, and web or computer. These metaphors are further detailed in Roediger (1980) and provide a way to simplify and relate memory to more tangible concepts. For instance, the wax tablet suggests memory is like writing on wax, where impressions can be made and retained over time.

Historical Milestones in Memory Research

Key milestones in memory research include Ebbinghaus's (1880) study of memory using meaningless syllables, which led to the behaviorist verbal learning approach. Ebbinghaus's work was pivotal in applying scientific methodology to the study of memory. Bartlett (1932) introduced the concept of schemas, emphasizing the role of prior knowledge in memory organization. Schemas are mental frameworks that help us organize and interpret information. Tolman (1948) proposed cognitive maps, contributing to the cognitive tradition focused on mental representations and information processing, akin to the computer metaphor. These maps represent mental layouts of environments.

Measuring Memory

Memory is measured through several methods. Neuropsychological cases demonstrate how specific brain injuries can lead to particular memory impairments, offering insights into which brain areas are critical for different types of memory. Neuroimaging techniques, such as fMRI and EEG, help identify brain areas associated with different memory tasks and their collaborative functions. These tools allow researchers to observe brain activity in real-time. Experimental psychology employs various experimental methods to reveal causal relationships in memory functions through manipulations and their observed effects. Experimental manipulations help isolate and test specific variables affecting memory.

Three Stages of Memory

The traditional information processing model describes memory in three stages: encoding (putting information into memory), storage (maintaining information in memory), and retrieval (recovering information from memory). Encoding involves converting information into a usable form. Storage refers to retaining the encoded information over time. Retrieval is the process of accessing stored information. Forgetting can occur at any of these stages, as noted by Atkinson & Hilgard (2005).

Multi-Store Model (Atkinson and Shiffrin, 1968)

The Multi-Store Model posits that environmental stimuli are initially received by sensory stores that are modality-specific. Sensory stores are temporary and modality-specific, meaning they handle information from different senses separately. Information is held briefly in these stores, and attention directs some of it to the short-term store. Attention acts as a filter, selecting which sensory information to process further. Rehearsal in the short-term store can transfer information to the long-term store. Rehearsal involves repeating information to keep it active in short-term memory.

Sensory Stores

Sensory stores include iconic memory (visual store) and echoic store (transient auditory store). Iconic memory lasts about $0.5$ seconds, possibly longer (Landman, Spekreijse, & Lamme, 2003). It holds visual information briefly. Mechanisms of visual perception operate on the icon, and attention can shift within iconic memory in approximately $55$ ms (Lachter, Forster, & Ruthruff, 2004). This allows for quick processing of visual input.

Partial Report Experiments

Sperling's experiment on iconic memory showed that participants could report more information when prompted with a partial report cue compared to a whole report, indicating more information is available than can be verbally reported (Atkinson & Hilgard, 2009). The partial report involves cuing participants to report only a subset of the information. Temporal integration experiments (Di Lollo, 2001) show that performance decreases with increased delay between frames, suggesting rapid decay of iconic memory (Atkinson & Hilgard, 2005). This decay highlights the transient nature of iconic memory.

Short-Term Memory (STM)

Classical Modal Model (Atkinson and Shiffrin, 1968)

The Classical Modal Model assumes intact STM is necessary for learning. According to this model, STM acts as a gateway to LTM. However, it has been observed that patients with STM deficits can still show learning ability. This challenges the idea that STM is essential for all learning.

Working Memory (Baddeley, 1986)

Working Memory describes the short-term store as a workspace for task completion, featuring different subsystems for different modalities (visual and auditory). Unlike STM, working memory emphasizes active processing and manipulation of information.

Working Memory Model (Baddeley and Hitch, 1986)

The Working Memory Model includes the phonological loop (short-term storage and manipulation of auditory information), the visuo-spatial sketchpad (short-term storage and manipulation of visual information), and the central executive (an attentional control system that integrates information from the slave systems and long-term memory). These components work together to manage and process information.

Phonological Loop

Verbal information is stored based on phonological code. Phonological code involves storing information based on how it sounds. It is harder to memorize acoustically similar letters (e.g., TBCGVE vs. RLTKSJ). This effect demonstrates the reliance on auditory coding. Blocking rehearsal leads to rapid decay of information (Brown+Peterson task). This task involves preventing participants from rehearsing information. The digit span test measures the capacity of the phonological loop. It assesses how many digits a person can remember in order. Miller (1956) proposed $7 +/- 2$ chunks, instead of individual elements, can be retained. Chunking involves grouping information to increase memory capacity.

Primacy and Recency Effects

Primacy effect refers to better recall of initial list items, and the recency effect involves better recall of last list items, potentially due to item discrimination. The primacy effect is attributed to more rehearsal for early items. The recency effect is due to these items still being in STM.

Visual-Spatial Sketchpad

The visual-spatial sketchpad provides short term storage and manipulation of visual information, separated into the visual (what) and spatial (where) subcomponents. The visual component deals with object recognition. The spatial component deals with location information. Measurements include mental rotation tasks and the Corsi block-tapping task. These tasks assess visual and spatial memory abilities.

Central Executive

The central executive is an attentional control system located in the frontal lobe. It oversees and coordinates the other components of working memory. Its functions include controlling subsystems, directing attention, and integrating information. It acts as a supervisor. It is measured using the Wisconsin card-sorting task and verbal fluency task to assess executive functions. These tasks evaluate cognitive flexibility and verbal skills. The episodic buffer (Baddeley, 2000) integrates and briefly stores information from various sources, including the phonological loop, visuo-spatial sketchpad, and long-term memory. This buffer provides a temporary storage space for integrated information.

STM vs. LTM: Neuropsychological Evidence

Neuropsychological evidence, such as the cases of H.M. (Henry Molaison), who had an LTM deficit but intact STM, and K.F., who had an STM deficit but intact LTM, supports the independence of STM and LTM. H.M.'s case showed that STM and LTM could be separated. K.F.'s case provided further evidence for this separation. This double dissociation contrasts with simple dissociation, where damage to one area impairs one function while leaving others intact. Double dissociation provides stronger evidence for functional independence.

Case of H.M.

Experimental operation led to difficulty in forming new memories, but intact short-term memory and ability to learn new motor skills. H.M.'s case was crucial for understanding memory systems. His inability to form new declarative memories highlighted the role of the medial temporal lobe.

Long-Term Memory

Components (Squire, 1992)

Long-term memory components include explicit (declarative) memory (episodic and semantic) and implicit (nondeclarative) memory (conditioning, skills, priming, etc.). Declarative memory is conscious and verbally accessible, involving rapid learning and fallible recall, while nondeclarative memory is unconscious, involves skills, gradual learning, and is often infallible. Explicit/declarative memory is measured with direct tests, while implicit/non-declarative memory is measured with indirect tests. Direct tests involve conscious recall. Indirect tests assess unconscious influences on performance.

Non-Declarative / Implicit Memory Forms

Forms of non-declarative/implicit memory include classical conditioning (e.g., blink conditioning), priming (conceptual and perceptual e.g., word-stem completion), skills (motor skills, motor sequence learning), and statistical learning. Classical conditioning involves learning associations between stimuli. Priming enhances the processing of a stimulus due to prior exposure. Skill learning involves acquiring new motor abilities. Statistical learning involves extracting patterns from the environment. Priming specificity includes stimulus specificity, associative specificity, and response specificity. These specificities highlight the context-dependent nature of priming. Skill learning involves sequence learning, such as the Alternating Serial Response Time (ASRT) Task. This task assesses the learning of motor sequences. Visual Statistical Learning (Fiser+Aslin, 2001) uses unsupervised statistical learning using arbitrary complex shapes. This demonstrates how we can learn patterns without explicit instruction.

Long Term Memory: Declarative vs Nondeclarative

Declarative memory includes episodic and semantic memory, while nondeclarative memory encompasses procedural memory, priming, conditioning, habituation, and sensitization. Episodic memory involves personal experiences. Semantic memory involves general knowledge. The medial temporal lobe is key for declarative memory, while the basal ganglia, neocortex, amygdala, and cerebellum are involved in nondeclarative memory. Different brain areas support different types of memory.

Processing-Based Taxonomy

The processing-based taxonomy identifies three basic processing modes of memory: rapid vs. gradual encoding, associative vs. single item encoding, and flexible and compositional vs. rigid and unitized representations. These modes highlight different ways we process and store information.