RESEARCH METHOD COHEN ok
RESEARCH METHOD COHEN ok RESEARCH METHOD COHEN ok
ALLOTTING ELEMENTS TO CONSTRUCTS 437 would be strengthened by the additional statement that ‘persons resemble each other in their construction of events’. Can the practice of providing constructs to subjects be reconciled with the individuality corollary assumptions A review of a substantial body of research suggests a qualified ‘yes’: [While] it seems clear in the light of research that individuals prefer to use their own elicited constructs rather than provided dimensions to describe themselves and others ... the results of several studies suggest that normal subjects, at least, exhibit approximately the same degree of differentiation in using carefully selected supplied lists of adjectives as when they employ their own elicited personal constructs. (Adams-Webber l970) However, see Fransella and Bannister (1977) on elicited versus supplied constructs as a ‘gridgenerated’ problem. Bannister and Mair (1968) support the use of supplied constructs in experiments where hypotheses have been formulated and in those involving group comparisons. The use of elicited constructs alongside supplied ones can serve as a useful check on the meaningfulness of those that are provided, substantially lower inter-correlations between elicited and supplied constructs suggesting, perhaps, the lack of relevance of those provided by the researcher. The danger with supplied constructs, Bannister and Mair (1968) argue, is that researchers may assume that the polar adjectives or phrases they provide are the verbal equivalents of the psychological dimensions in which they are interested. Allotting elements to constructs When a subject is allowed to classify as many or as few elements at the similarity or the contrast pole, the result is often a very lopsided construct with consequent dangers of distortion in the estimation of construct relationships. Bannister and Mair (1968) suggest two methods for dealing with this problem which we illustrate in Box 20.2. The first, the ‘split-half form’, requires the subject to place half the elements at the similarity pole of each construct, by instructing the subject to decide which element most markedly shows the characteristics specified by each of the constructs. Those elements that are left are allocated to the contrast pole. As Bannister (1970) observes, this technique may result in the discarding of constructs (for example, male–female) which cannot be summarily allocated. A second method, the ‘rank order form’, as its name suggests, requires the subject to rank the elements from the one which most markedly exhibits the particular characteristic (shown by the similarity pole description) to the one which least exhibits it. As the second example in Box 20.2 shows, a rank order correlation coefficient can be used to estimate the extent to which there is similarity in the allotment of elements on any two constructs. Following Bannister (1970), a ‘construct relationship’ score can be calculated by squaring the correlation coefficient and multiplying by 100. (Because correlations are not linearly related they cannot be used as scores.) The construct relationship score gives an estimate of the percentage variance that the two constructs share in common in terms of the rankings on the two grids. Athirdmethodofallottingelementsisthe ‘rating form’. Here, the subject is required to judge each element on a 7-point or a 5-point scale, for example, absolutely beautiful (7) to absolutely ugly (1). Commenting on the advantages of the rating form, Bannister and Mair (1968) note that it offers the subject greater latitude in distinguishing between elements than that provided for in the original form proposed by Kelly. At the same time the degree of differentiation asked of the subject may not be as great as that demanded in the ranking method. As with the rank order method, the rating form approach also allows the use of most correlation techniques. The rating form is the third example illustrated in Box 20.2. Alban-Metcalf (1997: 317) suggests that there are two principles that govern the selection of elements in the repertory grid technique. The first is that the elements must be relevant to that part of the construct system that is being Chapter 20
438 PERSONAL CONSTRUCTS Box 20.2 Allotting elements to constructs: three methods Source:adaptedfromBannisterandMair1968
- Page 406 and 407: PROCEDURES IN ELICITING, ANALYSING
- Page 408 and 409: DISCOURSE ANALYSIS 389 completeness
- Page 410 and 411: ACCOUNT GATHERING IN EDUCATIONAL RE
- Page 412 and 413: STRENGTHS OF THE ETHOGENIC APPROACH
- Page 414 and 415: A NOTE ON STORIES 395 instruments t
- Page 416 and 417: INTRODUCTION 397 the physical s
- Page 418 and 419: STRUCTURED OBSERVATION 399 Box 18.1
- Page 420 and 421: STRUCTURED OBSERVATION 401 Box 18.2
- Page 422 and 423: STRUCTURED OBSERVATION 403 or event
- Page 424 and 425: NATURALISTIC AND PARTICIPANT OBSERV
- Page 426 and 427: NATURALISTIC AND PARTICIPANT OBSERV
- Page 428 and 429: ETHICAL CONSIDERATIONS 409 Box 18.3
- Page 430 and 431: SOME CAUTIONARY COMMENTS 411 ob
- Page 432 and 433: CONCLUSION 413 the data mean. This
- Page 434 and 435: NORM-REFERENCED, CRITERION-REFERENC
- Page 436 and 437: COMMERCIALLY PRODUCED TESTS AND RES
- Page 438 and 439: CONSTRUCTING A TEST 419 achieveme
- Page 440 and 441: CONSTRUCTING A TEST 421 Select the
- Page 442 and 443: CONSTRUCTING A TEST 423 where A = t
- Page 444 and 445: CONSTRUCTING A TEST 425 true/fal
- Page 446 and 447: CONSTRUCTING A TEST 427 short-answe
- Page 448 and 449: CONSTRUCTING A TEST 429 demonstrate
- Page 450 and 451: CONSTRUCTING A TEST 431 (e.g. to as
- Page 452 and 453: COMPUTERIZED ADAPTIVE TESTING 433 H
- Page 454 and 455: 20 Personal constructs Introduction
- Page 458 and 459: PROCEDURES IN GRID ANALYSIS 439 inv
- Page 460 and 461: PROCEDURES IN GRID ANALYSIS 441 Box
- Page 462 and 463: SOME EXAMPLES OF THE USE OF REPERTO
- Page 464 and 465: GRID TECHNIQUE AND AUDIO/VIDEO LESS
- Page 466 and 467: FOCUSED GRIDS, NON-VERBAL GRIDS, EX
- Page 468 and 469: INTRODUCTION 449 Box 21.1 Dimension
- Page 470 and 471: ROLE-PLAYING VERSUS DECEPTION: THE
- Page 472 and 473: THE USES OF ROLE-PLAYING 453 of
- Page 474 and 475: ROLE-PLAYING IN AN EDUCATIONAL SETT
- Page 476: EVALUATING ROLE-PLAYING AND OTHER S
- Page 480 and 481: 22 Approaches to qualitative data a
- Page 482 and 483: TABULATING DATA 463 data set reprod
- Page 484 and 485: TABULATING DATA 465 Box 22.4 Studen
- Page 486 and 487: FIVE WAYS OF ORGANIZING AND PRESENT
- Page 488 and 489: SYSTEMATIC APPROACHES TO DATA ANALY
- Page 490 and 491: SYSTEMATIC APPROACHES TO DATA ANALY
- Page 492 and 493: METHODOLOGICAL TOOLS FOR ANALYSING
- Page 494 and 495: 23 Content analysis and grounded th
- Page 496 and 497: HOW DOES CONTENT ANALYSIS WORK 477
- Page 498 and 499: HOW DOES CONTENT ANALYSIS WORK 479
- Page 500 and 501: HOW DOES CONTENT ANALYSIS WORK 481
- Page 502 and 503: A WORKED EXAMPLE OF CONTENT ANALYSI
- Page 504 and 505: A WORKED EXAMPLE OF CONTENT ANALYSI
ALLOTTING ELEMENTS TO CONSTRUCTS 437<br />
would be strengthened by the additional statement<br />
that ‘persons resemble each other in their<br />
construction of events’.<br />
Can the practice of providing constructs to<br />
subjects be reconciled with the individuality<br />
corollary assumptions A review of a substantial<br />
body of research suggests a qualified ‘yes’:<br />
[While] it seems clear in the light of research<br />
that individuals prefer to use their own elicited<br />
constructs rather than provided dimensions to<br />
describe themselves and others ... the results of<br />
several studies suggest that normal subjects, at<br />
least, exhibit approximately the same degree of<br />
differentiation in using carefully selected supplied<br />
lists of adjectives as when they employ their own<br />
elicited personal constructs.<br />
(Adams-Webber l970)<br />
However, see Fransella and Bannister (1977) on<br />
elicited versus supplied constructs as a ‘gridgenerated’<br />
problem.<br />
Bannister and Mair (1968) support the use<br />
of supplied constructs in experiments where<br />
hypotheses have been formulated and in those<br />
involving group comparisons. The use of elicited<br />
constructs alongside supplied ones can serve<br />
as a useful check on the meaningfulness of<br />
those that are provided, substantially lower<br />
inter-correlations between elicited and supplied<br />
constructs suggesting, perhaps, the lack of<br />
relevance of those provided by the researcher. The<br />
danger with supplied constructs, Bannister and<br />
Mair (1968) argue, is that researchers may assume<br />
that the polar adjectives or phrases they provide<br />
are the verbal equivalents of the psychological<br />
dimensions in which they are interested.<br />
Allotting elements to constructs<br />
When a subject is allowed to classify as many or<br />
as few elements at the similarity or the contrast<br />
pole, the result is often a very lopsided construct<br />
with consequent dangers of distortion in the<br />
estimation of construct relationships. Bannister<br />
and Mair (1968) suggest two methods for dealing<br />
with this problem which we illustrate in Box 20.2.<br />
The first, the ‘split-half form’, requires the subject<br />
to place half the elements at the similarity pole<br />
of each construct, by instructing the subject to<br />
decide which element most markedly shows the<br />
characteristics specified by each of the constructs.<br />
Those elements that are left are allocated to<br />
the contrast pole. As Bannister (1970) observes,<br />
this technique may result in the discarding of<br />
constructs (for example, male–female) which<br />
cannot be summarily allocated. A second method,<br />
the ‘rank order form’, as its name suggests,<br />
requires the subject to rank the elements from<br />
the one which most markedly exhibits the<br />
particular characteristic (shown by the similarity<br />
pole description) to the one which least exhibits<br />
it. As the second example in Box 20.2 shows,<br />
a rank order correlation coefficient can be<br />
used to estimate the extent to which there is<br />
similarity in the allotment of elements on any<br />
two constructs. Following Bannister (1970), a<br />
‘construct relationship’ score can be calculated<br />
by squaring the correlation coefficient and<br />
multiplying by 100. (Because correlations are not<br />
linearly related they cannot be used as scores.)<br />
The construct relationship score gives an estimate<br />
of the percentage variance that the two constructs<br />
share in common in terms of the rankings on the<br />
two grids.<br />
Athirdmethodofallottingelementsisthe<br />
‘rating form’. Here, the subject is required to judge<br />
each element on a 7-point or a 5-point scale,<br />
for example, absolutely beautiful (7) to absolutely<br />
ugly (1). Commenting on the advantages of the<br />
rating form, Bannister and Mair (1968) note that it<br />
offers the subject greater latitude in distinguishing<br />
between elements than that provided for in the<br />
original form proposed by Kelly. At the same time<br />
the degree of differentiation asked of the subject<br />
may not be as great as that demanded in the<br />
ranking method. As with the rank order method,<br />
the rating form approach also allows the use of<br />
most correlation techniques. The rating form is<br />
the third example illustrated in Box 20.2.<br />
Alban-Metcalf (1997: 317) suggests that there<br />
are two principles that govern the selection of<br />
elements in the repertory grid technique. The<br />
first is that the elements must be relevant to<br />
that part of the construct system that is being<br />
Chapter 20