Retrieve Rename
Archive Display Compact
Memo Table
Directory Record-card Chart
Message Letter
Message Area:
Figure 7.1
Example of an 'option' screen
There were a num ber of requirements that the simulated interface had to fulfil. It was necessary that a num ber of different versions of the interface could be generated that differed only in specified ways. There had to be three versions of the interface which differed in their syntax to yield the two groups object => action syntax and action => object syntax and a further version for the second test session which allowed either syntax sequence to be used, the 'free' syntax interface.
Also, different versions had to be generated to provide balance of presentation between the groups. The factors to be balanced were:
i) arrangem ent of objects and actions on the screen (objects in top section; actions in top section) ii) arrangem ent of particular command words (at least two different arrangem ents were needed)
All versions of the interface represented the functionality in the same way, i.e. they were identical apart from the initial 'option' screen. Also, all versions had a data capture facility which recorded the commands selected and the time taken to select the first command.
v - i /
5. M ethod 5.1 Design
The experiment was conducted in two sessions. In the first session, subjects used an interface which had a fixed syntax and in the second, they used an interface with free syntax (i.e. the subject could choose commands in any order which was convenient). This scheme was chosen because in previous studies (Barnard et al 1982), it has been shown that free syntax systems require more learning effort. Here, subjects would already be accustomed to the system before they encountered the free syntax condition.
In the first test session there were two independent variables: Required syntax (object => action; action => object)
Task structure (object-type; action-type)
For each subject, the required syntax was constant for the entire test session. Half the subjects had an object => action syntax and half had an action => object one. Subjects were informed of their particular syntax in the instructions for the test session. The task structure was varied so that each subject received object-type tasks on half of the trials and action-type tasks for the other half. This variation meant that on half the trials the task structure and the
syntax were consistent. If a subject using an object => action syntax interface received an object type task, such as:
MEMO
You w ant to re-use this memo, so make a copy of the original, called "Memo/Storage". the task type and the required syntax would be consistent - the subject would specify the command sequence, 'memo-copy1.
If the same subject had received an action type task, such as :
You need two draft quality print-outs of the memo to stick up on noticeboards around the department.
the two variables (task structure type and required syntax) would be inconsistent - the subject would have to specify the command sequence, 'memo-print’.
phases differed in their order of presentation of the two task types. In phase one, subjects 1-16 received a block of six trials of one type followed by a block of six trials of the other type. In phase two, subjects 17-32 received a mixed set of twelve trials following a pattern of alternation between the two types (e.g. one object-type; two action-types; two object-types; one action-type etc.). It was believed that subjects in the first phase had an opportunity to adopt a proceduralised response strategy and after a few trials, errors due to an inconsistency between task structure and syntax would be unlikely. There would be only one opportunity to observe what happens when the procedure became redundant i.e. when the task structure was changed (Task 7).
There were three dependent variables: Errors
Difficulties or sub-optimal responses
Time taken to select the first element of a command
There were a number of balancing factors between groups. These were: i) the position of the action set and the object set on the screen -
for half the subjects the objects appeared in the top section and for the other half the actions appeared in the top section.
ii) the position of each particular command word within the set - there were two
arrangements of the command word sets so that words that appeared in the centre position on one set appeared at the periphery in the other set.
iii) the sequence of tasks used - for all phases there were two sequences of the tasks. Each task was composed of four object-action pairings from the set of sixty-four possible pairings (8 objects x 8 actions); with no pairing being repeated within the set of tasks. Each subject was given twelve tasks, so a total of forty-eight object-action pairings were
performed in the first test session. Details of the tasks are given below.
In this first session, it was of interest whether consistency between the task structure and syntax facilitated performance. If this was the case, one would expect greater syntax error and difficulty scores on inconsistent trials, as well as greater response latencies in selecting the first command because subjects would have to apply additional processes to translate their representation into an appropriate form.
In the second test session, there was one independent variable: Task structure (object-type; action-type)
Here, the subjects experienced a single structure for the entire session; this was object-type for half the subjects and action-type for the other half. Each subject had eight tasks, composed of four sub-tasks (32 pairings).
The subjects were told that they were to use a slightly different computer system. The
content of the actual items differed in the second test session - e.g. the letter was concerning a different subject and it was addressed to someone different. Subjects were instructed to choose the commands in any order they liked and their preference was scored as a dependent
variable. The extent to which the syntax chosen was consistent with the task structure was measured.
Expressed generally, the experimental hypothesis asserts that:
a particular syntax sequence should be easier to use when it is compatible with the task structure, i.e. when the first element of the sequence expresses the repeated item of the task structure.
52 Subjects
Subjects' participation in this study was voluntary and each subject was paid £2.
There were 32 subjects (17 female and 15 male). The mean age of the subjects was 23.6 years (sd = 7) with a range of 17-47 years. The majority of the subjects were students and research staff from University College London. All but 6 of the subjects spoke English as their first language. Only one subject had no previous computer experience. The other 31 subjects had used computers for wordprocessing, data analysis, running experiments and playing games. Half the subjects had used a mouse before.
53 Domain
The domain of common office tasks was chosen, as for previous experimentation. It includes the set of common tasks undertaken by professional workers regardless of their particular area of expertise e.g. filing letters, writing memos, producing graphs, keeping diaries etc. The domain remained constant for all groups.
5.4 Tasks
As described above, the set of tasks was devised from combinations of the 8 objects and the 8 actions. There were two types of task: object-type tasks and action-type tasks. The
object-type tasks were focussed around one object which was to have four operations performed on it whereas the action-type tasks focussed on a particular operation which was to be performed on four objects.
The manner of presentation of the task instructions was the same for all tasks. The subject was shown a card by the experimenter. The card contained 8 options (objects or actions) with one option highlighted. The subject was asked to keep the highlighted option in mind (as this was what the task was about). Then the subject could look at the details of the task in a folder - the instructions for each task were typed on a separate page. By combining the highlighted item and the instructions the subject could work out w hat was required. Twelve tasks were used in the first test session and eight in the second session. More tasks were included in the first test session to compensate for the subjects' unfamiliarity with the device, in general.
See Appendix H for a complete description of the task set. Sample tasks appear below: Object-type task (the highlighted option, 'Memo1, is the 'topic' of the task)