These illustrations, or behavioral anchors, are derived from a job analysis, namely,
the critical incident technique or CIT (Flanagan 1954 ). The appraiser keeps
a diary of what each employee is observed doing on the job. At the time of a
formal appraisal, the person reads the diary and then rates the employee, typically
on a 7 – 9 point scale, using the behavioral anchors as a guide for determining
whether the performance is highly acceptable, highly unacceptable, or somewhere
in between.
BOS are algebraic, summated,Wve-point Likert scales where the items compris-
ing a scale, originally derived from the CIT, are item or factor analyzed. As noted
earlier, an appraiser records the frequency with which an employee was observed
engaging in the behaviors listed on the scale (e.g. ‘praises a person for performing
well’ Almost Never 01234 Almost Always).
Despite the care that goes into developing BARS, subsequent studies have found
no evidence that they are superior to trait scales with regard to minimizing rating
errors (Schwab et al. 1975 ; DeCotiis 1977 ). In addition, no signiWcant diVerences
have been found between BARS and BOS with regard to minimizing halo and
leniency error, or increasing inter-observer reliability.
TheseWndingsWnally led to the hypothesis that perhaps more important than
the rating scale that is used is the training that appraisers receive in using it. Fay and
Latham ( 1982 ) found that rating errors were reduced signiWcantly regardless of
whether BARS, BOS, or trait scales were used when people received rater error
training (Latham et al. 1975 ). However, behavioral scales, following rater training,
were more resistant to rating errors than trait scales. BOS and BARS were equally
resistant to rating errors. With regard to practicality, BOS were evaluated as sign-
iWcantly better than BARS and trait scales by the users.
- 2 Practicality
Practicality is important because to the extent that a manager feels the appraisal
scale is useful, it is likely to be used as the basis for setting goals which in turn
directly aVect an employee’s motivation (Locke and Latham 2002 ). In too many
instances, psychometrically sound instruments either are not used or are soon
abandoned because of failure on the part of researchers to take user reactions into
account.
Consequently, Wiersma and Latham ( 1986 ) examined the practicality of BOS,
BARS, and trait scales. American managers and their white-collar subordinates
preferred BOS to BARS in all cases, and in all but two cases to the use of trait scales.
Lawyers who specialize in human resource litigation also preferred BOS to the two
alternatives in terms of defensibility in the courtroom. These results have been
replicated in the Netherlands (Wiersma et al. 1995 ). BOS were preferred by Dutch
managers over the other two methods for providing feedback, diVerentiating
368 gary latham et al.