Administration Phase (3)

Rating Methods

Once the elements and the constructs are obtained, they are written down on the protocol's two-way table. The resulting intersection or grid gives the technique its name. This protocol can be administered to the subject with the explanation that the elements are placed vertically on the top half of the table and that they constitute the columns while the constructs are placed horizontally on the left side of the grid and constitute the rows. The reason for this arrangement is so that each construct can be related to every element via a rating system. It is therefore necessary for all the elements to fall within the range of convenience of the constructs. The subject's task is to choose a score (according to the rating method chosen) that accurately defines how each element relates to each construct. The subject is generally advised to fill in the rows from left to right, continuing on to a new row when the previous one has been completed. Various rating methods have been suggested, each with its advantages and disadvantages:

A) The dichotomous method involves placing all the elements on either one pole or the other of the construct. Kelly suggested that when most elements were located on one pole of the construct, leaving only one or two elements on the opposite pole, then the construct should be discarded due to its poor discriminative capacity and the resulting lack of balance in the numerical calculations. On the other hand, Bannister's (1960) "split half" solution (placing half of the elements on one pole and half on the other) considerably limits the interviewee's range. The instructions for the use of the dichotomous method are the following:

"Place these elements on one of the two construct poles. If the element is best described by the left pole then place a zero in the box. If the right pole best describes the element then place a cross in the box."

This is the simplest rating system to administer and is recommended for use with children or with people suffering from cognitive deficits, since it does not involve the use of numbers (e.g., 0 or 1). Flash cards can also be used with the elements but require that the interviewer write down the scores:

"Place the cards that include the people best described by the right pole of the construct in the right pile and place the people who are best defined by the left pole of the construct in the left pile."

This method greatly simplifies the respondent's judgments and the limited power of the applicable mathematical analyses. It should only be used if more complex methods cannot be applied. A greater number of elements and constructs than normal should be used (e.g., 20 x 20) to partially compensate for these disadvantages.

B) The ordinal method, as described by Bannister (1963), emerged as an attempt to avoid the aforementioned difficulties with the dichotomous approach. In applying this method, the person is instructed to order the elements from one pole of the construct to the other, for example, from the most "honest" to the most "dishonest":

"Arrange the elements from the left pole to the right pole so that you put a 1 in the box indicating the most 'honest' element, followed by a 2 until you get to n (total number of elements) which would indicate the most 'dishonest' element."

This procedure can be very simple and has been succesfully used in the administration of the repertory grid to children, especially if the procedure includes the use of flash cards: "Out of all these cards, which one describes the most `honest' and therefore least `insincere' person?" Once the subject has given an answer, the chosen card is removed and a '1' written in the appropriate box. The above question is then repeated with one element less and a '2' is written in that box, and so on until no elements are left.

The ordinal system offers more discriminative power than the dichotomous one but may force the subject to score differences between elements without there necessarily being any. Furthermore, it is more difficult to administer as the number of elements increase. There are also limitations in terms of the mathematical analysis. For example, the GRIDCOR programme can only compute grids using nine elements or less when this system is used.

C) The rating scale method is the most widely used. Each element is assigned a value in a Likert-type scale delimited by both poles of the constructs. For example, the applications of a construct within a seven-point scale would be:

  1. Very honest (left pole)
  2. Quite honest (left pole)
  3. Slightly honest (left pole)
  4. Middle point
  5. Slighty dishonest (right pole)
  6. Quite dishonest (right pole)
  7. Very dishonest (right pole)

These scales can range from 3 to 11 or more intervals. Although three-point intervals allow for very little movement and force the subjects to simplify their answers, increasing the number of intervals progressively induces the subject to increase the complexity of the judgments made. Therefore, the three-point interval scale is only recommended for those people who may find difficulties with larger scales. Our experience tells us that 7-point interval scales are preferable. If a different number of intervals should be used, then the user must change the protocol shown in this manual because it is designed for a seven interval scale (this was also the procedure used with Daniel). Furthermore, the GRIDCOR programme can only admit a maximum of nine intervals. This procedure allows for more freedom when applying the constructs to the elements and it does not force distinctions where they do not exist. However, interval scales are not without problems. There is nothing to indicate that the distances between scale points are metrically equivalent. For example, it has been found that scores located near the extremes of constructs are clearly more meaningful than intermediate scores which are, at times, imprecise (Yorke, 1985).

Regardless of the system used, there are times when the subject cannot score an element along a construct because the construct is simply not applicable to that particular element. In our example, Daniel used a construct "adjusted-unmanly" which is only applicable to males. Although this construct is very idiosyncratic ("adjusted" refers to a socially adapted man and "unmanly" to a man with an unclear sexual identity) it does help us to see why the female elements of Daniel's grid cannot be applied to either pole of this construct. In these cases, the space can be left blank (or "N/A" written) indicating that it does not apply. Leaving a blank space poses a series of problems for the mathematical analysis of the grid. To avoid this, two alternatives exist, neither of which has proven to be completely satisfactory:

Another problem that can be encountered while administering the grid is that the subject may not know enough about the element to score it on a particular construct. In this case, it is necessary to insist on the following:

"Not only are we interested in judgments that you feel certain of but also in any inferences, intuitions or fantasies that you may have. If you have any idea, no matter how subjective it may be, try to answer with a score."

At times, people refuse to write down a score until they are absolutely certain. However, inferences that have not been repeatedly put to the test may also be of interest as they are predictions based on a particular construct system. They may not provide information on the scored element but do provide an insight into the way the interviewee infers things. We have already stated that the repertory grid does not provide "objective" data on the elements as much as on the person's construing processes.

If the subject is still unable to imagine whether element A is "honest" or " dishonest," he/she is allowed to write a "?" resulting in a blank box. A "?" response poses the same problems as a "N/A" response and therefore requires the same solutions as those explained above.