Table

The components and actions of the restaurant script proposed by Schank and Abelson (1977)

Script name Component Specific action

Eating at a restaurant Entering Walk into restaurant

Look for table

Decide where to sit

Go to table

Sit down

Ordering Get menu

Look at menu

Choose food

Waiter arrives

Give orders to waiter

Waiter takes order to cook

Wait talk

Cook prepares food

Eating Cook gives food to waiter

Waiter delivers food to customer Customer eats Talk

Leaving Waiter writes bill

Waiter delivers bill to customer

Customer examines bill

Calculate tip

Leave tip

Gather belongings

Pay bill

Leave restaurant there was considerable agreement in the lists produced. At least 73% of subjects mentioned sitting down, looking at the menu, ordering, eating, paying the bill, and leaving. In addition, at least 48% included entering the restaurant, giving the reservation name, ordering drinks, discussing the menu, talking, eating a salad or soup, ordering dessert, eating dessert, and leaving a tip. So there appear to be at least 15 key events involved in people's restaurant-visiting knowledge. Other evidence from Galambos and Rips (1982) has shown that when subjects have to make a rapid decision about whether or not an action is part of a script (e.g., determining that "getting to a restaurant" is part of a restaurant script), they answer rapidly when the action is part of the script but take longer when it is not a script action. Evidence for script theory has also been found in more applied contexts concerning eyewitness testimony for robberies (see Holst & Pezdek, 1992).

However, in later extensions of this theory (see Schank's, 1982, 1986, dynamic memory theory, in Eysenck & Keane, 1995) the specific organisational structure of scripts was somewhat modified. Psychological evidence had shown that the script idea was wrong in some respects. Bower et al. (1979) found that subjects confused events that, according to script theory, were stored separately and should not have interfered with one another. For example, recognition confusions were found between stories that called on distinct but related scripts; visits to the dentist and visits to the doctor. As scripts had been defined as structures that were specific experiences in specific situations, one clearly could not have a "visit to a health professional" script. In response to these problems, Schank revised script theory, in his dynamic memory theory. Abbot, Black, and Smith (1984) have found support for this new type of organisation proposed in dynamic memory theory by showing that various parts of what were formerly called scripts are hierarchically organised. At the top level is the general goal (e.g., eating at a restaurant), at the intermediate level are scenes that denote sets of actions (e.g., entering, leaving, ordering), and at the lowest level there are the actions themselves.

General evidence for schemata

There is considerable evidence in several different areas for the operation of schema-like knowledge structures (see e.g., Alba & Hasher, 1983; Graesser, Woll, Kowalski, & Smith, 1980). After Bartlett, many studies have shown that when people have different expectations about a target event they interpret and recall it in different ways (see e.g., Anderson & Pichert, 1978; Bransford & Johnson, 1972; see Chapter 12).

Furthermore, schemata have also been implicated in perception, where they reduce the need to analyse all aspects of a visual scene. When we view everyday scenes, like our bedroom or a lecture theatre, we have clear expectations about what objects are likely to be present. Schemata reduce the amount of processing the perceptual system needs to carry out to identify expected objects (see Chapter 4), thus freeing up resources for processing more novel and unexpected aspects of the scene (like the lecturer's dress-sense). Friedman (1979) has shown this by presenting subjects with detailed line drawings of six different scenes (from a city, a kitchen, a living room, an office, a kindergarten, and a farm). Each picture contained objects you would expect in the setting and a few unexpected objects. Friedman found that the duration of the first look was almost twice as long for unexpected as for expected objects, indicating the role of schemata in processing the latter. The differences between expected and unexpected objects were even more marked on a subsequent recognition memory test. Subjects rarely noticed missing, or partially changed, expected objects even when only those expected objects that had been looked at were considered. In contrast, deletions or replacements of unexpected objects were nearly always detected. As Friedman concluded, "The episodic information that is remembered about an event is the difference between that event and its prototypical, frame representation in memory" (p. 343).

These effects regarding the recollection of unexpected items have been found repeatedly in a number of different experiments, although they can be modified by conditions that interfere with subjects' attention to the processing of the unexpected objects (see Henderson, 1992; Mantyla & Backman, 1992).

Fundamental problems with schema theories

Schema theories are not without their problems. While they remain the most overarching set of proposals on the structure and organisation of knowledge in long-term memory, they have a number of faults.

The unprincipled nature of schema theories

There is a broad consensus among many researchers that schema theories are unprincipled. This stems from the fact that it is often possible to create any particular content for the knowledge structures used, to account for the pattern of evidence found. Schank deals, in part, with this problem by attempting to delimit all the possible structures in long-term memory, but the theory is still underspecified. Problems still remain; for example, what are the specific contents of all of these structures? In general, then, schema theories tend to be good at accounting for results in an ad hoc fashion, but are not as predictive as one would like them to be.

There are two remedies to this situation. First, the theorist could specify the content of structures that are used; at least for a definable set of situations. That is, if you were using dynamic memory theory, you could specify all the possible scripts that might be used by a person. Unfortunately, this is probably impossible given the breadth of human knowledge and the possible variability in knowledge structures from one individual to the next. The other option is to be clearer about how these structures are acquired (see Chapter 14). If we knew more about this issue then we could begin to test how different selected experiences might be combined to form hypothetical structures in a more controlled fashion.

The problem of inflexibility and connectionist schemata

Although dynamic memory theory was developed to overcome many of the inflexibilities of script theory, some prominent theorists still consider that the intuitive flexibility of the schematic approach has not been realised in any of the present schemes (see Rumelhart, Smolensky, McClelland, & Hinton, 1986a). For example, Rumelhart and Ortony (1977) had proposed that the slots/variables in schemata should have two distinct characteristics. First, as stated earlier, they should test to see whether a certain object is an appropriate filler for the slot or provide a default value. Second, there should be interdependencies among the possible slot fillers. That is, if one slot is filled with a particular value then it should initiate changes in the default values of other slots in the schema. For example, assume that you have a schema for rooms that includes slots for the furniture, the small objects found in it and the usual size of the room. So, a kitchen schema would have the following structure and defaults:

Furniture: kitchen table, chairs.

Small objects: coffee pot, bread bin.

Size: small

Other rooms would have different defaults; for example a bathroom would also be small but would have a toilet, bath, and sink as furniture and toothbrushes as small objects. Rumelhart and Ortony's proposal was that, when the small-objects slot is filled with coffee pot, there should be an automatic change in the default value for the furniture slot to kitchen table and chairs. However, this second characteristic of schemata was never realised in the schema theories of the 1970s and 1980s.

Rumelhart et al. (1986a) proposed to remedy this state of affairs with a connectionist treatment of schemata. In this view, schemata emerge at the moment they are needed from the interaction of large numbers of parallel processing elements all working in concert with one another (for a treatment of connectionist ideas, see Chapter 1). In this scheme, there is no explicitly represented schema, but only patterns of activation that produce the sorts of effects attributed to schemata in previous research. When inputs are received by a parallel network, certain coalitions of units in the network are activated and others are inhibited. In some cases where coalitions of units tend to work closely together, the more conventional notion of a schema is realised; but where the units are more loosely interconnected the structures are more fluid and less schema-like.

Rumelhart et al. have illustrated the utility of such a scheme by encoding schema-type knowledge in a connectionist network. First, they chose 40 descriptors (e.g., door, small, sink, walls, medium) for five types of rooms (e.g., kitchen, bathroom, and bedroom). To get the basic data to construct the network they asked subjects to judge whether each descriptor characterised an example of a room type they were asked to imagine (e.g., a kitchen). When they built a network that reflected this information, they found that when activation was kept high in the sink unit and then some other unit (e.g., oven), the network settled into a state with high activation in units that corresponded to the typical features of a kitchen (e.g., coffee-pot, cupboard, refrigerator). Similarly, runs starting with other objects resulted in the emergence of descriptors for other prototypical rooms.

This connectionist work could solve the problem of the unprincipled nature of schema theories in that it promises to specify a means by which schemata acquire their contents. Ironically, it does this without having to specify these schematic contents.

WHAT IS AN IMAGE? SOME EVIDENCE

The first half of this chapter has dealt with propositional representations and the ways they have been used to represent relations and events. In the latter half of the chapter we turn to analogical representations, specifically visual images, to consider how they have been studied in the literature.

Historically, visual imagery has been studied for a long time. Over 2000 years ago, Aristotle regarded imagery as the main medium of thought. Furthermore, orators in ancient Greece used imagery-based, mnemonic techniques to memorise speeches (see Yates, 1966); a technique that is still used today as an aid to improving one's memory. This interest in imagery can be traced in a continuous line through philosophers, like Bishop Berkeley at Trinity College Dublin, to the 19th-century research of Galton (see Mandler & Mandler, 1964). Galton (1883) distributed a questionnaire among his eminent scientific colleagues, asking them to, for example, imagine their breakfast table that morning. Surprisingly enough, several reported no conscious mental imagery at all.

As in Galton's studies, much of this early research relied on the use of introspective evidence. During the behaviourist era, when introspection fell into disrepute and mental representations were in a sense "banned", research on imagery lay fallow for a number of years. However, with the emergence of cognitive science, the study of mental representations once again became respectable. The main motivation behind this push was the perceived necessity to be representationally precise about the possible cognitive mechanisms.

Nowadays, many researchers are working on the structure of imagery. In this section, we report on three sets of studies which illustrate several important properties of mental images. First, studies on mental

The different degrees of rotations performed on the materials in Cooper and Shepard (1973) for mirror-imaged letters (on the right) and normal letters (on the left).

rotation show how people can rotate visual images. Second, studies on image scanning give us some idea of how people can "mentally scan" a visual image. Third are studies on re-interpreting the images of ambiguous figures.

Stop Anxiety Attacks

Stop Anxiety Attacks

Here's How You Could End Anxiety and Panic Attacks For Good Prevent Anxiety in Your Golden Years Without Harmful Prescription Drugs. If You Give Me 15 minutes, I Will Show You a Breakthrough That Will Change The Way You Think About Anxiety and Panic Attacks Forever! If you are still suffering because your doctor can't help you, here's some great news...!

Get My Free Ebook


Post a comment