4.2.6 Alternatives to Rating

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 
119 120 121 122 123 124 125 126 127 128 129 130 131 

However they’re arrived at, the constructs in a grid tell you something about

how the interviewee thinks about a topic. The ratings tell you what s/he thinks

about individual elements.

So far, all our examples have used ratings to indicate what the person thinks.

However – as long as the interviewee indicates where, with respect to a

construct, a given element is positioned – anything will do for this purpose!

Grouping the Elements

Kelly’s original grids were very simple, the elements being divided into two

groups: those described by the emergent (left-hand) pole of the construct, and

those described by the implicit (right-hand) pole of the construct. If you think

about it a moment, you’ll see that this is simply a 2-point rating scale, with

ratings of 1 and 2 defining the scale from emergent to implicit pole.

Hitherto, we have been using a 5-point scale to position elements on constructs

and thereby identify the ‘what’s of the situation. You may wonder about the

alternatives.

A scale with an odd number of points allows for the indication of a ‘middle’

position, which may convey ‘neutrality’ depending on the construct in

question and the topic of the grid. A scale with an even number of points

enforces a preference, no matter how slight, for one pole of the construct rather

than the other.

This may be useful with topics in which the constructs have poles which stand for

preferred and non-preferred characteristics. Assessments and evaluations of

various kinds spring to mind (employee and managerial performance appraisals,

teaching practice appraisals, and training course evaluations); indeed, any situation

inwhichitmay be tempting or politic tomake a‘neutral-point’ rating, or inwhich a‘drift

towards themiddle’occurs evenwheremore extreme ratings arewarranted.

However, as we saw in Section 3.2.4, the use of more than 8 or 9 points in the

scale is excessive. To group elements into so many different categories along a

construct is to assert that people can make very fine judgements, consistently

for all the elements, regardless of the construct in question, and this is not the

case. It is spurious precision.

Ranking the Elements

And that is the rationale used by people who avoid ratings and use ranks. In

fact, they’re a bit suspicious of the apparent precision claimed for any rating

system: when a person rates any three elements on a construct as follows,

1 John Mary Alan 5

Predictable 3 1 5 Unpredictable

do they really mean that Mary is more predictable than John, exactly to the

same degree as Alan is less predictable than John? Probably not, yet that’s

what the use of ratings implies. Three minus 1 equals 5 minus 3. This, they

feel, is a misleading use of numbers. (Anyone who has done a course in

statistics will recognise the issue. We’re talking about the properties of

interval, as opposed to ordinal, data here.)

Why not simply rank the elements along the construct to reflect their relative

positions, without making claims about the distances between the elements?

Exactly the same meaning would be conveyed about the elements, but without

the spurious precision, by:

smaller/higher John Mary Alan larger/lower

ranks ranks

Predictable 2 1 3 Unpredictable

Ranking systems are sometimes used with repertory grids; it’s worth doing if

you are concerned to obtain, and eventually analyse, a single grid, or several

grids with the same number of elements.

There is a problem, though. It becomes very awkward to compare grids in

which there are different numbers of elements (ranks would run between 1

and 3 for 3 elements, but 1 and 10 for 10 elements: so a given rank, ‘3’, say,

cannot be easily compared between the two grids). You would use the

conventional ratings, rather than rankings, in such circumstances.