Information Processing System (Information Processing Theory)

Information processing theorists did challenge the notion inherent in behaviourism, that learning doth involve the forming of associations 'twixt stimuli and responses. These theorists do not reject associations outright, for they postulate that the formation of associations 'twixt sundry bits of knowledge doth facilitate their acquisition and storage in memory. Rather, they are less concerned with external conditions, focusing instead upon internal (mental) processes that intercede 'twixt stimuli and responses. Learners, then, are active seekers and processors of information. Unlike behaviourists, who maintained that persons respond merely when stimuli do impinge upon them, information processing theorists contend that persons select and attend to features of the environment, transform and rehearse information, and relate new information to that which hath come before.

Information processing theories do diverge in their views upon which cognitive processes hold greatest import and how they operate; yet they share certain common assumptions. One such assumption is that information processing doth occur in stages, which intercede 'twixt the reception of a stimulus and the production of a response. A corollary to this is that the form of information, or how it is represented mentally, doth differ depending upon the stage. These stages are qualitatively distinct from one another.

Another assumption doth hold that information processing is analogous to computer processing, at least metaphorically speaking. The human system doth function in a manner similar to a computer: it receiveth information, storeth it in memory, and retrieveth it as necessity dictates. Cognitive processing is remarkably efficient, with little waste or overlap. Researchers do differ in the extent to which they extend this analogy. For some, the computer analogy is naught but a mere metaphor, while others employ computers to simulate the activities of humans. The field of artificial intelligence is concerned with programming computers to engage in human activities, such as thinking, employing language, and solving problems.

Researchers also assume that information processing is involved in all cognitive activities: perceiving, rehearsing, thinking, problem-solving, remembering, forgetting, and imaging (Farnham-Diggory, 1992; Matlin, 2009; Mayer, 1996; Shuell, 1986; Terry, 2009). Information processing extends beyond human learning as traditionally delineated. This lesson is concerned primarily with those information functions most germane to learning.

The 'Information processing model of learning and memory' schema doth present an information processing model, incorporating sundry processing stages. Albeit this model be of a generic nature, it doth closely correspond to the classic model proposed by Atkinson and Shiffrin (1968, 1971).

Information processing commences when a stimulus input (e.g., visual, auditory) doth impinge upon one or more senses (e.g., hearing, sight, touch). The appropriate sensory register doth receive the input and hold it briefly in sensory form. It is here that perception (pattern recognition) occurs, which is the process of assigning meaning to a stimulus input. This typically doth not involve naming, for naming taketh time, and information stayeth in the sensory register for but a fraction of a second. Rather, perception involveth matching an input to known information.

The sensory register transferreth information to short-term memory (STM). STM is a working memory (WM) and correspondeth roughly to awareness, or that of which one is conscious at a given moment. WM is limited in capacity. Miller (1956) proposed that it holdeth seven plus or minus two units of information. A unit is a meaningful item: a letter, word, number, or common expression (e.g., “bread and butter”). WM also is limited in duration; for units to be retained in WM, they must be rehearsed (repeated). Without rehearsal, information is lost after a few seconds.

Whilst information is in WM, related knowledge in long-term memory (LTM), or permanent memory, is activated and placed in WM to be integrated with the new information. To name all the state capitals beginning with the letter A, students recall the names of states—perhaps by region of the country—and scan the names of their capital cities. When students who do not know the capital of Maryland learn “Annapolis,” they can store it with “Maryland” in LTM.

It is a matter of debate whether information is lost from LTM (i.e., forgotten). Some researchers contend that it can be, whereas others aver that failure to recall reflecteth a lack of good retrieval cues rather than forgetting. If Sarah cannot recall her third-grade teacher’s name (Mapleton), she might be able to if given the hint, “Think of trees.” Regardless of theoretical perspective, researchers agree that information remaineth in LTM for a long time.

Control (executive) processes regulate the flow of information throughout the information processing system. Rehearsal is an important control process that occurreth in WM. For verbal material, rehearsal taketh the form of repeating information aloud or subvocally. Other control processes include coding (putting information into a meaningful context—an issue being discussed in the opening scenario), imaging (visually representing information), implementing decision rules, organising information, monitoring level of understanding, and using retrieval, self-regulation, and motivational strategies.

The two-store model can account for many research results. One of the most consistent research findings is that when people have a list of items to learn, they tend to recall best the initial items (primacy effect) and the last items (recency effect), as portrayed in 'Serial position curve showing errors in recall as a function of item position.' According to the two-store model, initial items receive the most rehearsal and are transferred to LTM, whereas the last items are still in WM at the time of recall. Middle items are recalled the poorest because they are no longer in WM at the time of recall (having been pushed out by subsequent items), they receive fewer rehearsals than initial items, and they are not properly stored in LTM.

Research suggesteth, however, that learning may be more complex than the basic two-store model stipulateth (Baddeley, 1998). One problem is that this model doth not fully specify how information moveth from one store to the other. The control processes notion is plausible but vague. We might ask: Why do some inputs proceed from the sensory registers into WM, and others do not? Which mechanisms decide that information hath been rehearsed long enough and transfer it into LTM? How is information in LTM selected to be activated?

Another concern is that this model seemeth best suited to handle verbal material. How nonverbal representation occurreth with material that may not be readily verbalised, such as modern art and well-established skills, is not clear.

The model also is vague about what really is learned. Consider people learning word lists. With nonsense syllables, they have to learn the words themselves and the positions in which they appear. When they already know the words, they must only learn the positions; for example, “cat” appeareth in the fourth position, followed by “tree.” People must take into account their purpose in learning and modify learning strategies accordingly. What mechanism controllth these processes?

Whether all components of the system are used at all times is also an issue. WM is useful when people are acquiring knowledge and need to relate incoming information to knowledge in LTM. But we do many things automatically: get dressed, walk, ride a bicycle, respond to simple requests (e.g., “Do you have the time?”). For many adults, reading (decoding) and simple arithmetic computations are automatic processes that place little demand on cognitive processes. Such automatic processing may not require the operation of WM. How doth automatic processing develop, and what mechanisms govern it?

These and other issues not addressed well by the two-store model (e.g., the role of motivation in learning and the development of self-regulation) do not disprove the model; rather, they are issues to be addressed. Albeit the two-store model is the best-known example of information processing theory, many researchers do not fully accept it (Matlin, 2009; Nairne, 2002). Alternative theories covered in this lesson are levels (or depth) of processing and activation level, and the newer connectionism and parallel distributed processing (PDP) theories. Before components of the two-store model are described in greater detail, levels of processing and activation level theories are discussed (connectionism and PDP are covered later in this lesson).

Levels (Depth) of Processing

The theory of levels (or depth) of processing posits that memory is contingent upon the manner in which information is processed, rather than its specific locus within the cognitive architecture (Craik, 1979; Craik & Lockhart, 1972; Craik & Tulving, 1975; Lockhart, Craik, & Jacoby, 1976). This perspective eschews the postulation of discrete stages or structural entities such as Working Memory (WM) or Long-Term Memory (LTM) (Terry, 2009). Instead, it proposes a continuum of processing modalities, ranging from the superficial (physical or surface processing) to the profound (semantic processing or that pertaining to meaning), with acoustic (phonological, sound-based) processing occupying an intermediate position. These levels exist along a dimensional scale, wherein physical processing constitutes the most rudimentary level (akin to the symbol “x” devoid of inherent signification, as adduced by the pedagogues in the introductory vignette), and semantic processing represents the apex of cognitive engagement. To illustrate, consider the act of reading, wherein the word “wren” is encountered. This lexical item may be subjected to processing at a superficial level (e.g., noting its lack of capitalization), a phonological level (e.g., recognizing its rhyme with ”den”), or a semantic level (e.g., identifying it as a small avian creature). Each successive level embodies a more sophisticated (or deeper) form of processing than its predecessor; the extraction of semantic content from “wren” engenders a greater augmentation of the item's informational density than does acoustic processing, which, in turn, surpasses the contribution of surface-level processing.

The aforementioned tripartite stratification bears a conceptual resemblance to the sensory register, WM, and LTM components of the two-store model. Both perspectives acknowledge the incremental elaboration of processing through successive stages or levels. However, the levels of processing model diverges in its rejection of the notion that these processing modalities constitute discrete stages. Within this framework, one is not compelled to progress sequentially to the subsequent process in order to achieve more intricate processing; the depth of processing may vary within a given level. For instance, “wren” may undergo either low-level semantic processing (identification as a small bird) or more expansive semantic processing (examination of its similarities and dissimilarities to other avian species).

A further point of divergence between these two information processing models concerns the sequence of processing operations. The two-store model posits a linear progression wherein information is initially processed by the sensory register, subsequently by WM, and ultimately by LTM. Conversely, the levels of processing model does not presuppose such a sequential arrangement. Processing at the semantic level does not necessitate prior processing at the surface and sound levels, save for the minimal processing requisite for initial reception of the information (Lockhart et al., 1976).

The two models also proffer disparate explanations for the influence of processing modality on memory. According to the levels of processing model, the depth at which an item is processed correlates positively with the fidelity of memory recall, owing to the greater consolidation of the memory trace. The aforementioned pedagogues, in the opening scenario, express their concern regarding the facilitation of deeper processing of algebraic information among their students. Once an item has been processed to a specific point within a given level, additional processing at that same point is not expected to enhance memory performance. In contrast, the two-store model posits that memory may be enhanced through repeated processing of the same type. This model predicts a positive correlation between the extent of rehearsal of a list of items and the likelihood of subsequent recall.

Empirical evidence lends support to the levels of processing model. Craik and Tulving (1975) presented participants with a series of words, accompanied by questions designed to elicit processing at a specific level. For surface processing, participants were asked, “Is the word in capital letters?” For phonological processing, the question was, “Does the word rhyme with train?” And for semantic processing, the question was, “Would the word fit in the sentence, ‘He met a _____ in the street’?” The duration of processing at each level was meticulously controlled. The results indicated that recall was most accurate for items processed at the semantic level, followed by the phonological level, and least accurate for the surface level. These findings suggest that forgetting is more likely to occur as a consequence of shallow processing, rather than as a result of information attrition from WM or LTM.

The levels of processing model implies that student comprehension is enhanced when material is processed at deeper levels. Glover, Plake, Roberts, Zimmer, and Palmere (1981) demonstrated that instructing students to paraphrase concepts while reading essays significantly improved recall, relative to activities that did not draw upon prior knowledge (e.g., identifying key words within the essays). Instructions to read slowly and attentively, however, did not yield any discernible improvement in recall performance.

Notwithstanding these corroborative findings, the levels of processing theory is not without its lacunae. One concern pertains to the universality of semantic processing as the deepest level. The acoustic properties of certain words (e.g., “kaput”) may be as salient as their semantic content (e.g., “ruined”). Indeed, recall is contingent not only upon the level of processing, but also upon the nature of the recall task itself. Morris, Bransford, and Franks (1977) observed that, in a standard recall task, semantic coding yielded superior results to rhyming coding; however, in a recall task that emphasized rhyming, questions eliciting rhyming during coding produced better recall than semantic questions. Moscovitch and Craik (1976) proposed that deeper processing during learning engenders a higher potential for memory performance, but that this potential is realized only when conditions at retrieval mirror those extant during learning.

Another reservation concerning the levels of processing theory pertains to the effect of additional processing at the same level on recall. Nelson (1977) presented participants with one or two repetitions of each stimulus (word), processed at the same level. Two repetitions resulted in improved recall, contrary to the predictions of the levels of processing hypothesis. Further research corroborates that additional rehearsal of material facilitates retention and recall, as well as promoting automaticity of processing (Anderson, 1990; Jacoby, Bartz, & Evans, 1978).

A final point of contention concerns the very nature of a “level.” Critics have argued that the notion of “depth” is nebulous, both in its definition and in its operationalization (Terry, 2009). Consequently, our understanding of how processing at different levels influences learning and memory remains incomplete (Baddeley, 1978; Nelson, 1977). Time, as a metric, is an unreliable indicator of level, as certain surface processing tasks (e.g., “Does the word conform to the following letter pattern: consonant-vowel-consonant-consonant-vowel-consonant?”) may consume more time than semantic processing (e.g., “Is it a type of bird?”). Moreover, processing time within a given level does not necessarily signify deeper processing (Baddeley, 1978, 1998). This ambiguity surrounding the concept of levels (or depth) diminishes the overall utility of this perspective.

Resolving these issues may necessitate an integration of the levels of processing framework with the two-store model, culminating in a more refined model of memory. For example, information within WM may be related to knowledge in LTM in either a superficial or a more elaborate manner. Furthermore, the two memory stores themselves may incorporate levels of processing within each respective store. Semantic coding in LTM may engender a more extensive informational network and a more meaningful means of encoding information than either surface or phonological coding.

Activation Level

An alternative conception of memory, yet one which shares commonalities with both the two-store and levels of processing models, posits that memory structures are characterized by varying degrees of activation (Anderson, 1990). This viewpoint eschews the notion of discrete memory structures, instead proposing a unitary memory system with differing activation states. Information may exist in either an active or an inactive state. When active, information is readily accessible. The active state is sustained so long as attention is directed towards the information. In the absence of attention, the activation level will diminish, though the information may be reactivated upon reactivation of the corresponding memory structure (Collins & Loftus, 1975).

Active information encompasses both newly acquired information entering the information processing system and information that has been previously stored in memory (Baddeley, 1998). Irrespective of its provenance, active information is either currently undergoing processing or is capable of being processed expeditiously. Active material is broadly synonymous with WM, though the former category is more encompassing than the latter. WM pertains to information within immediate consciousness, whereas active memory includes both this information and material that can be accessed with relative ease. As an illustration, if I were to visit Aunt Frieda and we were to admire her flower garden, that information would reside in WM, while other information associated with Aunt Frieda’s yard (e.g., trees, shrubs, canine companions) might exist in an active state.

Rehearsal serves to maintain information in an active state (Anderson, 1990). Analogous to working memory, only a finite number of memory structures may be simultaneously active. As one’s attention shifts, the level of activation fluctuates accordingly.

The concept of activation level resurfaces later in this discourse (i.e., in the context of Anderson’s ACT theory), as it is integral to both the storage of information and its subsequent retrieval from memory. The underlying principle involves spreading activation, whereby one memory structure may trigger the activation of another structure that is adjacent (or related) to it (Anderson, 1990). Activation propagates from active to inactive regions of memory. The degree of activation is contingent upon the strength of the pathway along which activation spreads, as well as upon the number of competing (or interfering) pathways. The likelihood of activation spread increases with heightened practice, which reinforces structures, and diminishes with the passage of time (or length of the retention interval), as strength attenuates.

One advantage of activation level theory lies in its explanatory power regarding the retrieval of information from memory. By dispensing with the construct of discrete memory stores, the model obviates the potential problem of transferring information from one store to another. Short-Term Memory (STM) or WM is simply that segment of memory that is currently active. Activation decays with the effluxion of time, unless rehearsal sustains the activation of the information (Nairne, 2002).

Concurrently, the activation level model has not entirely escaped the pitfalls inherent in dual-store models, as it too dichotomizes the information system (into active and inactive components). Moreover, we encounter the problem of determining the threshold of strength necessary for information to transition from one state to another. Intuitively, we recognize that information may exist in a state of partial activation (e.g., the phenomenon of having a word “on the tip of one’s tongue”—where one possesses knowledge of the word but is unable to recall it). Thus, we might inquire as to the requisite degree of activation for material to be deemed active. These considerations notwithstanding, the activation level model offers valuable insights into the processing of information.

We shall now proceed to examine, in greater detail, the constituent elements of the two-store model: attention, perception, encoding, storage, and retrieval (Shuell, 1986). The ensuing section will address the topic of attention; while perception, encoding, storage, and retrieval will be explored in subsequent sections.