Академический Документы
Профессиональный Документы
Культура Документы
perception Inventory
61
Stephen G. Fisher, W.D.K. Macrosson and Gillian Sharp
University of Strathclyde, Glasgow, UK
Introduction
In an extensive survey of middle management[1] it was reported that, in the UK,
middle managers anticipated more of their work being undertaken in teams
than is presently the practice. Indeed, in that survey, sponsored by the British
Institute of Management, it was suggested that the possession of good team-
working skills will soon become a precondition for entry to the ranks of
management rather than an added bonus that some managers bring to their job.
Given the importance that employers currently appear to accord to team
performance, the validity and the reliability of the instruments employed to
gauge team potential, in particular the Belbin Team Role Self-perception
Inventory (BTRSPI), is now a matter of significance.
An assessment of the BTRSPI has appeared[2] which has questioned its
psychometric properties and cast doubt on the ability of the inventory to
provide a reliable measure of team role preferences. In a series of experiments
using not only subjects drawn from university courses, but also subjects drawn
from the ranks of full time managers, Furnham et al.[2] calculated from the
ipsative as well as the non-ipsative form of the questionnaire values for
Cronbach’s alpha (the statistic most widely accepted[3] as the best measure of
internal consistency). They found the alpha coefficients “not impressive” and
judged that the factor analysis results did not provide clear evidence of the
proposed team role structure. In his reply to the doubts voiced by Furnham et
al.[2], Belbin[4] has criticized their departure from the recommended method of
scoring the BTRSPI and, furthermore, has criticized the creation by Furnham
and his co-workers of their own “de-ipsativized” version of the BTRSPI. Indeed,
Belbin[4], in his reply to Furnham et al.[2], protests that the BTRSPI was never
intended to be a formal psychometric test and was meant only to be “a quick
and useful way of intimating to readers what their own team roles might be”.
The readers of Belbin’s first book[5], however, appear to have used the BTRSPI
as an established psychometric test. This is not particularly surprising in light
of Dulewicz’s[6] report that the BTRSPI was designed in order to attempt to
measure an individual’s team role scores without recourse to psychometric
instruments. Our experience aligns closely with that of Furnham et al.[7], who
have certain knowledge that a number of commercial organizations and
management consultancies use it not only in training courses, but in actual Personnel Review, Vol. 25 No. 2,
1996, pp. 61-67. © MCB
team building and development. University Press, 0048-3486
Personnel In view of the unresolved differences between Furnham et al.[7] and Belbin [4,8]
Review it appeared that further evidence on reliability and validity would be helpful.
25,2 Also, Furnham et al.[4] had noted that the relatively small samples in their
experiments meant that part of their results may well have been unstable. As the
test-retest reliability of the BTRSPI appears not to have been described in the
literature, we undertook to examine it for a sample of university students; this was
62 the first part of our study.
In the second part we examined the team role scores derived from a different
instrument. Belbin[5], in his original experiments, had employed several
dimensions of the 16PF as predictors of each team role and had evolved a method
for combining the 16PF scores to obtain team role scores; the equations for
calculating these scores are available in the open literature[6,9,10]. Thus, by
collecting the 16PF profiles, we were in a position to examine the closeness of the
match of team roles derived from the BTRSPI with those derived from 16PF data.
The results of that comparison are also given in this paper.
Method
A total of 149 undergraduate and 43 postgraduate students drawn from the
faculties of business, engineering and science at the University of Strathclyde
completed Cattell’s 16PF Personality Questionnaire, Form A[11] and Belbin’s[5]
Team Role Self-perception Inventory (BTRSPI). The students were all members
of classes taught by one of the research team and were told that they were
helping an enquiry into team roles. They were given the chance not to
participate, but none withdrew. They were not told at the first administration
that the Belbin Self-perception Inventory would be re-administered at a later
date. No feedback was given until all the field work had been completed. Seven
months later, 103 of these students, undergraduates only, completed the Self-
perception Inventory for a second time, the other 89 subjects having graduated
and left the university. The 16PF was administered only at the first round;
subjects were not asked to complete the 16PF at the retest administration
because data are available[12] to indicate that 16PF profiles are stable over
similar and longer periods. Test-retest reliability coefficients were calculated for
each of the eight team roles measured by the BTRSPI. Similarly, a test-retest
reliability coefficient for each individual item in the inventory was obtained.
Values for Cronbach’s alpha were calculated for each of the team roles for both
the test and the retest sets of data. The values for alpha were compared with
those published in the literature[2] by means of Friedman’s two-way analysis of
variance on the ranked data. Finally, using the 16 PF data, team role scores were
computed[10] and then used to calculate the correlation coefficients with the
team role scores derived from the BTRSPI data.
Results
The test-retest reliabilities and the associated probability levels for each of the
items in the Self-perception Inventory are shown in Table I. The coefficients of
reliability of individual items ranged from 0.17 to 0.58, for the 48 correlations
Belbin Team Role
I II III IV V VI VII
Self-perception
SH 32* 64* 35* 44* 48* 37* 53* Inventory
TW 44* 44* 36* 38* 42* 31* 34*
CF 28* 37* 22* 37* 18* 57* 41*
ME 28* 58* 40* 26* 31* 41* 04* 63
PL 33* 17* 44* 54* 31* 48* 24*
CW 17* 22* 25* 35* 43* 09* 16*
RI 08 35* 17* 38* 10* 08* 45*
CH 16 21* 35* 33* 15* 43* 20*
Notes:
* p ≤ 0.05 (one-tailed); n = 103
Table I.
Roman numerals at column heads correspond to the items in the Belbin Self-perception Test-retest reliability
Inventory. Decimal points omitted from the reliability coefficients in the table coefficients for each
Team roles: SH = shaper; TW = team worker; CF = completer-finisher; ME = monitor-evaluator; item in the Belbin
PL = plant; CW = company worker or implementor; RI = resource investigator; CH = chairman Self-perception Inventory
or co-ordinator for team roles
Alphas
Team type Test Test re-administration Furnham Test-retest reliability
Discussion
Test-retest reliabilities
The test-retest reliabilities have here been calculated from the original ipsative
form of the BTRSPI. The claims made by Saville and Willson[13] that
individuals can be as validly compared on a scale-by-scale basis on ipsative as
normative, that ipsative data can be factor analysed legitimately and that
neither reliabilities nor validities appear to be overestimated have re-energized
the ipsative versus normative data debate and have drawn adverse criticism[14].
Doubts expressed regarding the Saville and Willson[13] claims may lead
readers to conclude that the use of scores derived from ipsative questionnaires
for the calculation of reliabilities is universally unsound. However, using such
scores for the purposes of intra-individual comparisons has been found
legitimate[15] and is confirmed by Cronbach in personal correspondence[13].
The reliability coefficients obtained from the test-retest data are shown in
Table I and clearly signal a lack of stability over the seven months between the
first and second administrations. This result is perhaps not surprising in view
of the criticisms of Furnham et al.[2], but the size of the reliability coefficients Belbin Team Role
does prompt the question as to whether Belbin’s[4,8] objective for the BTRSPI is Self-perception
achieved. Inventory
In addition to high test-retest reliability, internal consistency is essential for
the validity of tests[15]. In this regard, Furnham et al.[2] have reported values
for the BTRSPI in both ipsative and “de-ipsativized” form, values which they
considered “unsatisfactory” for Cronbach’s alpha. Notwithstanding the fact 65
that the derivation of values for Cronbach’s alpha from data obtained from an
ipsative questionnaire may well be inappropriate, we have calculated these
values following the lead of Furnham et al.[2] in an effort to provide comparable
data. Our results yielded values for Cronbach’s alpha which were also well
below the level which is generally recognized as acceptable, namely 0.7, thus
confirming the results and conclusions of Furnham and his co-workers. Only
the value for the team role of “Shaper” approached that level, but, overall, the
internal consistency of the BTRSPI does not reach an acceptable level, median
values of C. 0.4 being obtained.
The majority of subjects used by Furnham et al.[2] were all full-time
employees of large, private or public organizations and were used to working in
teams, whereas the subjects used for our work were all students; our results
must always be viewed in light of this fact. Nonetheless, the Friedman analysis
of variance by ranks of the values of Cronbach’s alpha obtained by ourselves
and by Furnham et al.[2] suggested that both groups reacted similarly to the
items in the BTRSPI. This result is, perhaps, less surprising than initially might
be thought, for two reasons. Project work in teams is commonplace in the
undergraduate curriculum of the subjects used in this study and, second, the
majority of the subjects are of management calibre as is evidenced by the fact
that many of these students subsequently proceed into managerial careers.
Thus, although the sample used in our study was not drawn from practising
managers, the use of undergraduates is by no means indefensible.
Conclusion
Two conclusions may be drawn from this study. First, our results appear to
confirm and extend the findings of Furnham and his co-workers and suggest
that parties interested in establishing their team role preference would do well
to reconsider the value of the BTRSPI as published[5] and examine the
alternatives, perhaps one of the formulae utilizing an established personality
questionnaire or the new “Interplace” system[8]. Second, this study suggests a
cause for the confusion that Dulewicz[6] has encountered among managers
(namely, variability in team roles scores from the BTRSPI which neither Belbin Team Role
correlate with those derived from personality questionnaires nor remain steady Self-perception
over time). However, since Belbin’s team role model has both intuitive appeal Inventory
and empirical support it would be a great pity if this confusion pitched his
constructs into disrepute, causing them to be set aside. More rigorous studies,
encompassing both managers and non-managers, designed to further validate
the constructs in commercial and industrial settings, would help reduce this 67
confusion as well as being welcome to personnel practitioners.