How We React Like Computers When Stressed



The Effect Of Increased Demand During Divided Attention




Abstract
Introduction
Method
Results
Discussion
References


ABSTRACT


   What is the limiting factor on our ability to "pay attention". Attention is the criteria which determines which stimuli achieve conciousness.

   The two main theories are Filter and Capacity. Neither theory is all-encompassing. Some of the differences in opinion arise from liberal interpretations of "at the same time".

   A computer can appear to do many things concurrently (even though it can NEVER process 2 inputs simultaneously) due to it's fast speed. Conceivably,the brain can operate at a similar speed and use similar selection criteria.

   119 Psychology students performed a Response Task whilst concurrently reading words on a computer screen. The hypothesis that increasing the difficulty of the reading task would increase Reaction Time for the other task was supported. Treisman's (1971) experiment which concluded parallel processing was possible can be re-interpreted without parallel processing and should be re-conducted with additional manipulations of the variables.


INTRODUCTION

   Why is it that if we are doing something difficult (e.g. fixing something) we don't notice someone talking to us - even if they are right next to us? Or we stop talking in a car when turning into heavy traffic? How many aircraft at once can an Air Traffic Controller handle? Situations like this suggest that humans have a limited amount of "attention" to go around amongst tasks,and if one becomes quite difficult,then the other(s) suffer a reduction in attention. What are the biological bases of this phenomenon?

   Before proceeding,we need to define just what is meant by "attention". Kahneman (1973) defined "attention" as "A label for some of the internal mechanisms that determine the significance of stimuli,and thereby make it impossible to predict behaviour by stimulus considerations alone" (p2). Johnston and Heinz (1978) called it "The systematic admission of perceptual data into conciousness".

   Most previous theories have been built on Broadbent's (1958) Filter Theory. The three stages involved are;
1. Sensory Registration (e.g. the ear "registers" that there is a sound)
2. Perceptual Analysis (e.g. is the sound my name? A noise?)
3. Response Selection (take appropriate action).

   At one or more stages there is a "bottle-neck" - if there is more than one stimulus - causing a filtering of the stimuli. Different theories place the bottleneck(s) at different stages. Leading proponents of this model are; Treisman (1960,1964),Erdelyi (1974),Deutsch and Deutsch (1963),and Norman (1968). They have made considerable "improvement" on this model,as the original theory "per se" didn't hold up too well under experimentation.

   These theories talk of "Modes Of Selection". An Early Mode involves an easily discriminated stimuli (e.g. a voice vs. noise) which makes little demand on the system - it is filtered at or near Stage 1. A Late Mode involves harder to discriminate stimuli (e.g. 2 sounds - both the same voice,similar sounding words,but different meanings) which makes a high demand on the system - it isn't filtered until at or near Stage 3.

   An alternative theory (e.g. Kahneman,1973) states that there is a limited capacity of attention,and that increased capacity requirements by a task reduces the capacity available for other tasks.

   The key problem in the author's view is the definition of "at the same time". Numerous experiments show that subjects exposed to 2 words simultaneously (literally at the same time) almost invariably only recall hearing one word. Often they are aware of another sound,but don't comprehend it. This is taken as evidence that stimuli CANNOT be processed concurrently (they are processed serially). Kahneman (p5) states that there is a CONTRADICTION since people CAN do things concurrently,such as driving and talking (the stimuli are processed in parallel).

   To reconcile this apparent contradiction,consider a mainframe computer (as used by banks,etc.). At any ONE instant of time it can only do ONE thing. This is known as Executing an Instruction. However,it can Execute 30,000,000 Instructions PER SECOND (30MHz). Thus HUNDREDS of people can use the computer AT THE SAME TIME,since each person only requires a small number of Instructions,each one only taking 0.000000033 of a second! Note that this is STRICTLY serial processing.

   This leads One to the question "How fast can the brain operate?" Studies in the 50s (which subsequently led to the advent of Sub-liminal Advertising - which was subsequently banned) showed that the brain can visually perceive 100,000 bits of information per second (100kHz),but only some of that information makes it to conciousness (as per Divided Attention theory). The rest remains SUBLIME (from SUB-LIMEN meaning "below the Limen or Threshold"- it is unconcious).

   In fact - it could be argued that visual stimuli are only admitted into conciousness at a rate of less than 30Hz (every 0.033 seconds). Consider a movie - which consists of still pictures being shown at a rate of 30Hz. If visual stimuli were making conciousness at a faster rate,then at least 2 consecutive stimuli would be of the same picture,and "motion" would not appear fluid. We would really see lots of still pictures (which clearly isn't the case - except for those old movies which are very "jerky").

   Therefore (I propose),when driving you are constantly changing your "attention" from sight (watch the road) to sound (listen to your passenger). Moray (1969) first proposed this "changing channel" process,but at a much slower rate. Treisman (1971) challenged this based on her Parallel Processing findings,but her findings can be re-interpreted using one of her own ideas(!) - that of a "Dictionary Unit" (in fact,many researchers have had the same idea,but under different names).

   The "Dictionary Unit" is a pre-conciousness structure which recognizes specific stimuli. This could over-ride the current "channel setting" for something important (e.g. hearing one's own name at a party while listening to someone else). This is in fact what happens in computers - a newly-arrived task can cause a LESS IMPORTANT task to be interrupted (this will be discussed further).

   The hypothesis then is that increasing the difficulty of a primary task (the most important task) will cause the performance of other tasks to suffer (an increase in reaction time) due to a reduction in space capacity.


METHOD


Subjects: 119 Male and Female first year Psychology Students at the University of Sydney.
Apparatus: Apple Macintosh computers with mouse (the program is described below)
Procedure: In the first condition subjects were required to click on the mouse whenever a black-box appeared on the screen. This occured at random intervals. The computer recorded base-line reaction times (R.T.) for subsequent comparison. Subjects were then split into two random groups (A/B). In the second condition,in addition to the black-box,words appeared in the corners of the screen,in a random fashion. Group A subjects then had to read these words aloud (this being the primary task) while still responding to the box (this being being the secondary task). Group B subjects monitored Group A with a list of the words being presented,making note of any errors. In the third condition,Group B subjects had to say aloud the Super-Ordinate Category the words belonged to,while Group A monitored for errors. This therefore increased the difficulty of the primary task. The categories used were Animal,Seascape,Cookery. One problem with the program was that if a word and the box appeared simultaneously,then clicking in response to the box also caused the word to dis-appear.


RESULTS


   Mean Reaction Times for the three conditions were calculated on the computer,as well as Standard Deviation,and Standard Error of Measurement. These are summarised in Table 1. The error-rate was analysed and not found to be significant.

A One-way Analysis of Variance was performed on the results yielding: F=106.65 p<0.0001
                                                                              
                                    Table 1                                  
                                                                              
                    Mean Reaction Times In The 3 Conditions                   
              --------------------------------------------------              
              |               |     | mean RT |  Std  |        |               
              |   condition   |  N  | in secs |  Dev  |  SEM   |
              |---------------|-----|---------|-------|--------|
              | 1- simple     | 119 |  0.316  | 0.065 | 0.006  | 
              | 2- words      | 119 |  0.486  | 0.148 | 0.014  |             
              | 3- categories | 119 |  0.722  | 0.336 | 0.031  |             
              --------------------------------------------------              


DISCUSSION


   The results of this experiment support the hypothesis that as the difficulty of a primary task increases (from no vocalization,to saying words,to saying their super-ordinate category),spare capacity available for other tasks decreases (increasing R.T. to the appearance of the black-box - due to serial processing of stimuli).

   There is no indication at all of parallel processing in this experiment,so let's revisit Treisman's (1971) claim that this is possible in more detail.

   In the first condition subjects listened to a single list consisting of numbers and nonsensical syllables,and had to respond to hearing a number. In the second condition subjects listened to two simultaneous lists,and had to press the appropriate key for no numbers or one number. The two conditions were then repeated,but the subjects were first told what number to listen for - this is called pre-cuing. Pre-cuing is known to reduce R.T. The logic was that the following steps would occur if there was only serial processing;
1. Determine if the item on one of the ears is a number.
2. If it's not - switch to the other ear.
3. Determine if that's a number.

   The first condition is finished after stage one (only one list). Treisman therefore expected that the reduction in R.T. for the second condition would be twice as much as the first. In fact the R.T. in both conditions reduced by the same amount (0.115s),however overall R.T. in the second condition was 0.080s longer than the first. Treisman therefore concluded that there was parallel processing occuring in the second condition,but that it was less efficient than in the first.

   The problem here is that an active strategy is assumed (i.e. listen to one ear first). In light of our computer theory consider this:-
The subject knows that he only has to listen for A number (NOT 2 numbers),therefore the importance of the number Dictionary Unit is raised (in the same way as our "Name" Dictionary Unit always has a high importance),THEREFORE a PASSIVE strategy can be adopted. i.e. which stimulus has a high importance? "listen" to that one! The data supports this hypothesis also (increase in R.T. when going from actively listening to passively listening).

   The author would suggest re-running the experiment in the following way;
1. Present a single list of numbers and syllables for a base-line R.T.,AND ask the subjects to recall what numbers they heard (active listening).
2. Present 2 lists simultaneously and again ask for recall (passive listening).
3. Again present 2 lists BUT this time allow 2 numbers to co-incide. Don't tell the subjects to expect this (remember in the original experiment there was never more than one number at a time). Thus the subject will still be passively listening (only expecting no more than one number). NOW when the subject is asked to recall I would predict that he will only be able to remember one number of a pair OR he will notice that there were 2 numbers and complain ("which button do I press?"). Either way there will be the same R.T.
4. NOW tell the subject to expect 2 numbers. He will now revert back to an active strategy - listen to one ear,then listen to the other. There will now be a noticeable increase in R.T. (double the processing) and an increase on recall from condition 3 (if the subject didn't notice the duality at that point).

   Done in this manner there should then be very strong evidence that parallel processing is not possible.


   In conclusion,we have seen that when required to perform multiple tasks,if one becomes harder our performance on the others decreases. This suggests a limited capacity for attention. Further research can quantify exactly what this is - and then we can make sure that we're employing sufficient Air Traffic Controllers to safely accomodate the number of aircraft in the sky.


REFERENCES

               Johnston, W. A. & Heinz, S. P. (1978)
               Flexibility and Capacity Demands on Attention
               Journal of Experimental Psychology:General,107,420-435

               Kahneman, D. (1973)
               Attention and Effort
               Engelwood Cliffs, NJ: Prentice-Hall


E-mail me!

Back to BIG DON'S Home-page