5. Sequencers in box element notation

Topics in this Section :

5.1. The operator’s main task.

5.2. Converting the 'identify control state' diamond into box element notation.

5.2.1. An imaginary example.

5.2.2. The operator’s actual behaviour.

5.3. A complex sequencer example.

5.4. A complex 'routine'.

5.5. The general properties of a 'sequencer'.
5.5.1. Conditionals.

5.5.2. Context independence.

5.5.3. Elements or 'routines' ?

5.6. Generality of these diagrams and mechanisms.

Section 5 : 'Sequencers’ in box element notation

Building up behavioural complexity from a cognitive processing element

Lisanne Bainbridge

Section 3.c showed how processing elements (introduced in Section 2) can be organised into larger wholes called 'routines', which meet a cognitive need.

Section 4 identified the determinants of the sequence in which the 'routine' topics were thought about.

And suggested the main routines are called on by sequencers which build up an overview of the task situation, which is referred to in deciding what to do next.

There's an early discussion of some of the issues here in :

(1975) The representation of working storage and its use in the organisation of behaviour.

This Section expands on what has been said earlier by suggesting a mechanism for this sequencing - for choosing which cognitive need to think about next. A grouping of elements which determine the sequence of topics thought about is called a 'sequencer'. (I previously called them 'sequences', but that is rather confusing.) 'Sequencers' contain elements and refer to ’routines' which find the main cognitive needs. 'Sequencers' may just contain ordinary elements, but the most important of them also contain conditional choice elements, which arrange the switching between alternative cognitive needs in different contexts. The 'boxes’ in the conditional elements in a 'sequencer' contain the items which determine the behaviour sequence, so are the 'location' of the working storage which provides the context for choosing appropriate behaviour. (These are just place holders, no implication intended that this is done by a particular type of mechanism.)

The furnace operator's behaviour could be most simply accounted for, not by one big 'sequencer', but by a main 'sequencer' which identifies the main contexts of the task, e.g. whether the state of the control task was acceptable or not, and on the basis of this leads to several other 'sequencers' which chose the most appropriate behaviour within the given sub-context. In the furnace operator's behaviour, the 'sequencers' lead to each other, and they refer to each other for data which is needed by one 'sequencer' but has been found by another. It is suggested these cross-referenced data items are the operator’s 'overview' of the task.

This may seem rather a strong claim at this point. There is more evidence on it in Section 6, which has more discussion of the overview.

Section 5.3 below presents in detail one of the 'sequencers' devised to account for part of the furnace operator's behaviour. And Section 5.5 describes the main properties of sequencers.

Section 6, on Working Storage, summarises all the 'sequencers' used in describing the furnace operator's behaviour, how the sequencers build up and cross refer to an overview of the situation which determines the sequence of behaviour, and how this leads to a contextual model of behaviour. Section 6 expands on Section 4, which first made those points.

5.1. This operator’s main task

The furnace power supply controller’s main concern was to ensure the total amount of power used by the steel making furnaces within a given half hour did not exceed a limit set by the electricity board. He controlled the amount of power available to each furnace, whether it gets 100% of the ideal power usage in this steel making stage, or a smaller amount. The limit is set because in the melting stage, furnaces use a huge amount of power, and it is very expensive for the electricity company unexpectedly to have to bring power generating processes on-line which can work immediately, so they pass that cost onto the steel company. This operator does not himself control the furnaces.

(Electricity power generation controllers have a task somewhat similar to gas grid controllers, who were studied by Umbers (1976). At that time there was only one TV channel, and the gas grid controllers kept an eye on the BBC TV programme schedule, so when a popular programme ended they were prepared for the whole nation putting the kettle on to make a cup of tea. But that is another story ! Though it is a good example of anticipatory control based on directed attention to specific information sources.)

Figure 4.2 from Section 4 is a simplified representation of the furnace controller’s main sequences of activity, in conventional (at that time) flow diagram form.

Repeat of Figure 4.2 from Section 4 : main activity sequences in the furnace power control task, Bainbridge (1974) Figure 3.

The main changes in direction of thinking are all determined by the current control state, coded 0, 1/2, 1 in the figure.

The main control state check is top left, and the operator cycles back to that between segments of other activity :

Left column - when control state acceptable (coded 0 in diagram) - the operator reviewed the state of the furnaces.

Middle column - when control state unacceptable but action not needed immediately (coded 1/2 in diagram) - the operator reviewed the available action choices and chose one.

Right - when action needed (coded 1 in diagram) - the operator made the action chosen previously.

There is more discussion of this diagram in Section 4.3 above. More detail of these processes is in Figure 4.3. But both Figures 4.2 and 4.3 are simplifications for clarity.

5.2. Converting the 'identify control state' diamond into box element notation

5.2.1. An imaginary example

To illustrate how the box element notation works for alternatives, this section gives a simplified example, rather than an example derived from data. This imaginary example was derived for the purposes of writing this review, not before the data analysis.

Basically, any decision about what to do next depends on various factors, and this decision and its effect on what happens next can be described as a 'conditional' (see figures 3c.2 and 3c.5 in Section 3c).

The sequencers contain box elements, ’routines’, and conditionals which call on 'routines' to find the data needed in deciding what best to do next.

So the overall behaviour could be described as 'routines' below a level of ’sequencers' which include conditionals.

The main determiner of the next best behaviour is the control state, see top left of Figures 4.2 and 4.3.

This leads to 3 main types of behaviour, according to whether the control state is acceptable, trending to needing action, or action needed.

A simplified box element representation of this main decision could be as in Figure 5.1.

Figure 5.1: Example of an imaginary 'sequencer' which describes how the 'identify control state' processing done by 'routine' 1 leads to other behaviours

In this example there are two main factors which determine the sequence of events : the size and direction of the control error, and whether an action has been made before. The values of these sequencing determinants are found :

- by 'routines' (labelled as 1, 2, 3),

- by cross reference to data in the working storage in another box, represented by an arrow in the box. In this Figure, the lower left box containing an arrow is referring back to the top control error box, i.e. it also uses the result found by 'routine' 1.

The decisions in the 'sequencer' lead to other 'sequencers', called A, B and C. This 'sequencer' determines the order in which the cognitive needs are thought about, by the other sequencers.

Labels in the Figure : the follow-on processing is in dashed boxes :

Numbers in these boxes represent 'routines' :

1 - the main control state checking routine (see Figure 5.4).

2, 3 - choosing the best action - routines for centre column of Figure 4.3

Letters indicate ’sequencers’ :

A - right column of Figure 4.3

B - left column of Figure 4.3

C - cycles back to checking the control error.

So there are two ways a sequencer leads to further behaviour, by :

- either using a 'routine' with a 'head box' (diagrammed as in a numbered box) to find the information needed. After using a routine, processing returns to the sequencer which called on it.

- or going to another subsidiary sequencer which may contain more processing (labelled in the diagram with a letter or full name).

- - -

5.2.2. The operator’s actual behaviour

Reminder - in all box-element diagrams derived from data :

- the processing 'routines' were identified from the contents of the verbal report, for method see Sections 2 and 3c.

- the conditionals/ sequence determinants were identified by finding which aspect(s) of the context influenced the choice of behaviour at this point, for method see Section 4.1.

Figure 5.2 : Reaction to control state assessment (made by 'routine' 1) by sequencer OVERALL. Bainbridge (1972) Figure 7.3.a.

Apologies for the poor quality of most of the ’sequencer’ Figures.

These figures are from my thesis (1972), and so were typed and drawn by hand long before there were even word processors, let alone graphics software.

Dotted boxes are the labels of the descriptors of more detailed activity.

(The sequencer names were chosen by the analyst for easy reference.)

This OVERALL sequencer finds the control state, using the thinking represented by 1 in a dotted box.

1 is a 'routine' - it returns the needed control state information, see Figure 5.4 below.

Then the operator chooses his next activity depending on whether there has been a step change in furnace power usage, and what the control state is. These are the data used to decide between possible best next activities.

In Summary, the further sequencers are :

OVERALL (main determinants of what happens in reaction to control state) is in Figure 5.2.

Figure 5.4 shows the 'routine' identifying the control state.

ONGOING (activity when control state is acceptable) is in Figure 5.5.

UNACCEPT (decision making when control state is unacceptable) is in Figure 5.3.

UPDATE and ASSESS diagrams are in Section 6b.

Diagrams for STEP CHANGE, ACTION, END TIME, with associated routines, are not currently available, see notes in final section.

How does Figure 5.2 OVERALL relate to the previous figures (4.2 and 4.3) describing the operator’s behaviour ? This doesn’t map easily onto the flow diagrams in Section 4, which were simplified from the actual behaviour to give a clear version of the operator’s key activities for purposes of presentation. For example :

UNACCEPT describes what the operator does in the mid and right columns of Figure 4.2 repeated above.

ONGOING and STEP CHANGE describe what the operator does in the left column of Figure 4.3.

Figure 5.2 and later Figures are a closer representation of what the operator actually did.

Figures 4.3 and 5.1 are simplifications for expository purposes.

5.3. A complex sequencer example

Figure 5.3 below represents the most complex of the 'sequencers' : UNACCEPT describing the order in which the furnace power controller thought about topics when the control error was unacceptable (control state coded 1 in Figure 4.2). There were many possible behaviours depending on details of the context. So this 'sequencer' consists mainly of conditional elements which lead to other sequencers.

(In other behaviour contexts there may be fewer alternatives, so the sequencers for them include fewer conditionals : e.g. Figure 5.5 below has none, Figure 6b.4 has a couple.)

Figure 5.3 : Sequences when control error unacceptable (Bainbridge, 1972, Figure 7.3.1).

Numbers top left and right indicate the phrases at the beginning of the protocol sections with evidence for behaviour in this context. So this diagram is based on analysis of 18 sections of the verbal report.

Reminder of symbols introduced in Section 3.c on 'routines' :

- double lined box - head box of routine, which returns value needed.

- arrow in box indicates value found elsewhere :

- single line arrow - item needed available locally,

- double line arrow (Figure 3c.5) - item needed can be found in another routine or sequencer. In this sequencer, double line arrows happen to all be in conditionals.

- double arrow box (Figure 3c.4), e.g. 'F cut' in above diagram - value found either by carrying out routine, or by reference to an item stored elsewhere (i.e. in working storage). As discussed in Section 3 and below, this provides a mechanism for the observed flexible mixture of serial and parallel processing.

This box element representation describes the actual details of the operator’s behaviour, not the very generalised description in the classic flow diagrams such as Figure 4.3.

It may help to understand to understand this diagram if you know what this 'sequencer' does.

This sequencer determines what best to think about next, on the basis of a combination of three aspects of the situation :

i. the size of the control error, and whether it is so large that an action is required,

ii. whether or not the operator has anticipated that an event (change of furnace stage) will happen later during this half-hour, which will compensate for the current mis-usage,

iii. whether or not any of the furnaces have, or previously have had, a cut to their power supply (as if so they will be the best to make an action on).

Those 3 factors are known, they’re the result of previous operator thinking (indicated by a double-line arrow in a box, which refers to where this information was found/ stored elsewhere).

The decisions made by the conditional elements lead on to several other 'sequencers' :

i. OVERALL (Figure 5.2 above), in which the operator returns to assessing the size of the control error, and chooses an appropriate activity on the basis of it,

ii. ACTION when action is required (diagram not available, see final note), which chooses the size of, and assesses the future effect of making, an action,

ii. ONGOING (Figure 5.5 below), in which the operator reviews the state of all the furnaces considered together, which he does when the control state is acceptable,

Each of those 'sequencers' may also call on 'routines' to find data needed, and lead on to other 'sequencers' depending on the evidence obtained.

For ASSESS see Section 6b.

2. The UNACCEPT sequencer in Figure 5.3 above contains one main cognitive need/ link to a 'routine', to find which furnace would be the best one to make an action on, i.e. to cut the power to. This is found by 'routine' number 2, when the control state is such that action is needed.

The diagrams for 'routines' 2 and 5.c are not available, see note at end.

The double-arrow items in the boxes in some conditionals and some elements in a sequencer indicate cross referenced items, key factors which need to be available in 'working storage' for reference, as the main determinants of the sequence of behaviour. They are the main descriptors of the task context, the operator’s 'overview'. See Table 4.4 column B. This is discussed more in Section 6.

The main feature of this diagram is that it makes clear how the choice of next activity is determined by several descriptors of the current state of the environment and task thinking, the 'overview'.

In most cases this cross reference to what is already known is represented by a double line arrow in the box in the conditional (see e.g. Figure 6a.3). In other cases there is a simple activity to find what’s needed.

In fact everything is a result of the operator’s previous thinking, as the sequence is affected by the operator’s assessment of the control state, not by the actual value of current power usage.

There is evidence on the ways in which sequencers follow on from each other in the next Section, 6a.

There are more examples of sequences and routines in Section 6.b, on updating the overview.

5.4. A complex 'routine'

The main difference between 'routines' and 'sequencers' is that 'routines' meet a cognitive need - they are focused on finding the value of a 'head' box. And they were identified during the first phase of analysis. While 'sequencers' determine the order in which these needs are thought about, and contain the overview. They describe the sequences of thinking found during the second phase of analysis. The difference is not that 'sequencers' consist of conditional elements and 'routines' do not. 'Routines' may contain conditional elements. And sequencers may not.

For example, the furnace operator assesses the present control state by a 'routine' which includes and chooses between three different methods, depending on circumstances. Figure 5.4 is the full 'routine'. This is the routine labelled '1' in Figures 5.1and 5.2.

The main electricity usage control task is to limit the amount of power used in each half hour section of time. The method the operator uses to assess the control state depends on where time is within that 1/2hour : within roughly the first 5 minutes, 10-15 minutes, or later.

None of these decisions are mentioned in the verbal report, as mentioned before. They were identified by checking the task context at the points where what was in the verbal report changed in content/ topic, see Section 4.1.

Figure 5.4 : 'Routine' describing the assessment of control error. Different methods of assessment are used at different times in the half-hour (Bainbridge, 1972, Figure 7.2.1).

Conditionals or meta-knowledge ?

The operator used three different methods for assessing the control state at different times in the half hour. So this 'routine' actually describes the choice of 'how' to achieve an aim. In Section 7 the choice of how to do something is accounted for by a meta-knowledge mechanism.

The choice between these two types of description of the process of deciding how to do something (decision by the analyst making the description, not the operator !) is based on some assumptions by the analyst and some simple evidence. Many of the choices between alternative methods appear to be unconscious (no evidence in the verbal protocol), and could be described by a multi-dimensional look-up table, as in Section 7. But when the decision between alternatives requires some conscious processing to find the information on which the decision is based (i.e. there is evidence for thinking as it is mentioned in the verbal protocol), it is described by a 'routine', as here.

This diagram might be taken to imply the control state is identified by some serial processing, in which the operator asks some questions. But the conditional symbol represents decisions which are never mentioned in the protocol, but have been identified by a separate analysis, as described in Section 4. As these decisions are not mentioned, it is assumed they are done by unconscious and perhaps parallel processing.

The current task context, the awareness the operator has about the time in this half hour, determines the path the cognitive processing takes through this diagram. But choosing the path is parallel not serial.

5.5. The general properties of a 'sequencer'

Sequencers may include many or no conditionals, depending on how many different types of activity may be relevant in sub-contexts of this situation, see 5.5.1 below.

I used to think there was a clear difference between routines and sequencers in the mechanisms involved, that routines have a focus on finding a data item, while sequencers include conditionals and decide between alternative sequences. But looking at them now, it is obvious there are many items which don’t fit this clear distinction.

The first stage of the analysis made sense of repeated sections of the verbal protocol, which were identified as 'routines’. These had two characteristics :

- they did mainly have 'one cognitive goal', as the protocol was divided into sections according to when the focus of attention changed.

- the division into segments of behaviour showed this processing could occur in different local contexts.

Once identifying the local sequences of items had got as far as was possible, and was described by the 'routines', the analysis moved onto a second phase when the 'sequencers' were identified. This also had three characteristics :

- it identified the sequencing determinants/ choice points, which needed a different type of analysis, see Section 4.1.

- the sequencers contain the 'overview', the person's 'mental picture' of the state of the task, which is used at the choice points.

- the sequencers integrate longer sections of behaviour, describing what was done in different control contexts.

All sequencers need information. How this is acquired may be described in the diagrams in two ways (again, analyst decision about how to make the description) depending on whether the processing needed to find the data is :

- small enough for there to room on the diagram to include the elements used.

- large enough to need to be described elsewhere as a separate 'routine', e.g. '1' above, indicated in many sequencers by small dashed boxes with a number in them. They may also be labelled as 'routines' when the same behaviour may occur in different contexts, though this happens with ’sequencers' as well.

See more on this in Section 5.5.3 below.

5.5.1. Conditionals

'Routines' always have a 'head box' and are oriented to delivering some needed information.

Sequencers describe the sequence of activity in a particular context.

In some situations there are many alternative behaviours, so many conditionals, as when the control error was UNACCEPTable, see Figure 5.3.

In other contexts the operator’s alternative actions may be much simpler, so the diagram describing the sequence of behaviour only has no or few conditionals, as in the next example.

What the operator does when condition are stable, ONGOING - Figure 5.5 below, actually has no conditionals, all the reviewing behaviours are needed but they can be done in any order.

When the control state was acceptable, the operator reviewed what all the furnaces were doing, in groups according to how much power they were using :

melting stages - high power needed, if the power is reduced the melt quality won’t be affected, the stage will just take longer. 'Baskets left' means baskets of raw materials still to be loaded and melted, so gives a rough idea of how much longer the melt stage will take.

quality stages - don’t use as much power, but if the power is reduced the quality may be affected.

non-power using stages - such as filling or emptying the furnace.

This diagram is based on analysis of 17 protocol fragments.

Figure 5.5 : ONGOING review of furnace stages when control state acceptable. Bainbridge (1972) Figure 7.3.3.

Figure 5.5 ONGOING describes the operator’s thinking when the control state is acceptable and no immediate step changes are anticipated.

It makes much use of a sequencer called UPDATE which checks on the time within this half hour. UPDATE and its associated routines are discussed in Section 6.b on working storage.

For descriptions of the other behaviours this ONGOING 'sequencer’ leads on to, see:

OVERALL is the general control state checker, Figure 5.2 above,

5.b and 5.c are 'routines' which I have not currently got access to the box element diagrams for.

5.5.2. Context independence

Like 'routines', some sequencers do not happen only in one context.

Another reason why they are described separately, rather than linked into one big diagram.

For example :

OVERALL (Figure 5.2) i.e. assessing and reacting to the control state, happens after nearly every other activity. The operator constantly returns to doing this.

In some contexts it is appropriate for ONGOING (Figure 5.5), a general review of what is happening, to follow on from either OVERALL (Figure 5.2) or UNACCEPT (Figure 5.3).

This situation-action independence is more important with 'routines' for different reasons, see more below on ASSESS in Section 6.b and more on goal-action independence in Section 7.

5.5.3. Elements or 'routines' ?

Whether cognitive processes are described in a sequencer by explicitly showing the elements, or by referring to a 'routine' which contains the elements, may be for two reasons :

- a sequence of activity must be described as a separate sequence when it happens in more than one context, as that means there must be situation- action independence,

- otherwise separating the elements into a separate diagram may be just the analyst’s decision, based on the possible complexity of the diagram :

- - When a cognitive need in a sequencer can be met by a single or a few elements, they may be diagrammed direct in the sequencer, see the first 2 steps in the UPDATE figure in Section 6b.

- - When the processing is more complex, it may be separated into a routine, e.g. routines 5.b and c in Figure 5.5.

5.6. Generality of these diagrams and mechanisms

Some general comments on the strength of the points made so far.

The main focus of this review is on the general nature of the cognitive processes underlying the observed verbal report from a process operator doing a complex dynamic industrial task.

A 'sequencer' describes how decisions are made about what to do next, on the basis of various aspects of the situation. In making these decisions there are :

- cognitive needs, needs for the information on which the decision on what next to think about are based,

- calls on elements or 'routines' to meet those information needs.

- or reference to the result of previous processing (arrow in box) which meets this need (the overview).

These decisions lead to other 'sequencers', which also may call on 'routines', and so on.

The decisions between alternatives about what to do next are represented in the diagrams as conditional elements.

Carrying out the 'routines' is explicitly mentioned in the protocol. However the sequencing decisions are not explicitly mentioned in the verbal report : the presence of a decision, and the variables and criteria used in the decision, were inferred by the method described in Section 4.1.

So there is no direct evidence about the decisions in a 'sequencer'. The structure given in the diagram just represents the simplest form for describing the choices identified by this type of analysis. The diagram shows the influencing factors, and the combinations of them which lead to particular outcomes. It is convenient to use conditional elements to describe the decisions, not only because this improves the internal consistency of this approach to cognitive modelling, but also because the choices do have to be expressed in a way which represents how 'routines' are called on if the data needed is not already available, and to show how other 'routines' fit into the sequencing. If there was not this requirement, the decisions between behaviours could just as well be described by a multidimensional table, or by comparing lists, or by states in working storage which lead to transitions, or in several other formats. The importance of these provisos will appear in Section 7.

There appear to be two types of 'parallel processing'.

As the data needs which are met by reference to working storage are never mentioned in the verbal protocol, this suggests that the decisions represented by the 'sequencer' might be done by unconscious parallel processing, rather than by a sequence of distinct decisions each of which considers only one variable.

The two-arrow box (see Section 3c.3.4, arrows to both working storage and a routine) implies that the conditional processing is done in parallel if the data required is available in working storage, and as serial processing if a 'routine' has to be carried out to find the data.

This flexibility is represented in some diagrams by dashed lines.

The sequencers do often refer to data which has been obtained in some previous processing. These data together make the overview, discussed more fully in Section 6.

The results may not seem very tidy or principled. Indeed, now I look at the box element diagrams more than 50 years after they were first drawn, I can find many small inconsistencies.

That is because the representation is a result of the way the analysis was done. And the analyses involved many small judgements.

I first identified the 'routines', by combining together the evidence in different protocol fragments which I had judged were meeting the same cognitive need, as described in Sections 2 and 3c.

The second phase of the analysis was to identify what affected the order in which these items were thought about, using the technique described in Section 4.1. The sequences observed and choice points identified were then described by the 'sequencers', as in Sections 5 and 6b.

The 'overview' was then identified as the main items which determine the choice of what to do next and are located in the sequencers, as in Table 4.4 Column B and Section 6 next.

So these diagrams emerged from a great deal of detailed analysis, and many little judgements, to make some sort of sense of what underlies real complex flexible and adaptive behaviour. The diagrams were not constructed by starting from any rules, principles, or pre-conceptions about what an overview, sequencer or routine should be like.

Rather the reverse. I did not know when I started the protocol analysis that components like box elements, routines, sequencers, or overviews would emerge as the best way of accounting for the data. These are all mechanisms that emerged from trying to represent the cognitive behaviour. And once they have emerged, I am focusing in this review on :

- making clear the analytic techniques used, how the mechanisms suggested came out of the analysis of the protocol data.

- the evidence for the possible general properties of these mechanisms, and how they account for flexible context-sensitive behaviour.

There also does not appear to be any over-riding principle of different mechanisms for routines and for sequencers, instead they represent different levels of integration of the behaviour.

Several routines and sequencers mentioned in the diagrams are missing here. I have not got usable computer files of all of them. This may be for several reasons :

- the diagrams may be in my Ph.D. thesis but I did not think it sufficiently important to translate them to my computer when I first wrote this review (in 1993 that was not straightforward process). My thesis is now in storage and would take a lot of work to find.

- I may have a computer image but it has not been opened since the 1990s, and is in such an aged and obscure format (an undocumented version of .PP format) that image software specialists have not found a way of reading it on a modern 64-bit Mac.

- There are of course also some aspects of behaviour which there was not enough verbal report evidence about to be able to say with any certainty what the cognitive processing was in this context.

The advantage of studying a control task using verbal reports is that the same situations do keep recurring, so there is substantial evidence for many of the underlying processing activities which have been identified.

Also these diagrams represent the cognitive activities of a specific person doing a specific task.

And these are also the results found by one analyst. This analysis has not been checked by another person using the same techniques, to see if different judges come to the same conclusions about what underlies the protocol. As the results involve many small judgements by the analyst : once these techniques of analysis had been developed, the analysis should strictly have been done by several people to test the validity of the results.

- - -

Summary of main points in Section 5

* Doing a complex dynamic task involves meeting several main classes of cognitive need.

* These main cognitive needs may be thought about in a very varied order.

* 'Sequencers', made up of conditional and ordinary processing elements, make the decisions which determine the sequence of thinking about the main cognitive needs

* The 'sequencers' contain the main cognitive needs, i.e. the elements which call on the main 'routines'.

* The data found which meets the main cognitive needs in the 'sequencers' are also the items which are referred to by the sequencing decisions, so they act as the context for those decisions, called here the 'overview'.

* There is a cycle in processing : the overview determines which cognitive needs are thought about. These needs call on other 'routines', which update the overview, and so on (summarised in Section 4, Figures 5-6).

* The factors involved in making a decision about what to do next are considered in parallel, unless the necessary data is not available, when a 'routine' is used to find it. This flexibility between parallel and serial operation is represented by the two arrow box.

* So this cyclic contextual processing can act as a serial processing device when there is no context or when the contextual overview has not yet been built up, even though this standard serial sequence of activity is not explicitly represented as serial in the working methods (see more in Sections 6, 7, 8).

Access to other papers via Home page

In this review, there are 3 main groups of topics.

1. Introduction.

Basic element, sources of data which meet cognitive needs.

2. The cognitive processing element.

Meeting an information need :

3a. by finding it in the environment.

3b. from a stored knowledge base.

3c. by working through a 'routine',

or referring to the result of using a routine elsewhere.

3d. more on knowledge bases.

Choosing what to do.