The view is proposed that where conditions for psychological
testing are favourable the nature of the conventional pencil-and paper
group test sets a limit on the quality of assessment achievable.
This is so because it insists on standardisation through uniformity.
This procrustean template is a handicap to the assessment of the
Selection and allocation in the Army is a favourable testing
situation - centralised, stable and with high volume. Tailored
testing is an alternative to conventional testing that allows
variation to suit the ability of the person being assessed. The
present thesis proposes and tries out a tailored testing procedure
aimed at selection and allocation in the Army and other like circumstances.
The research review shows tailored testing to be a post-war
interest with statistical antecedents in many non-psychological
areas. In the last five years research has grown rapidly, stimulated
by the increasing possibilities of online computer-assisted testing.
A tailored testing procedure is proposed that makes few assumptions, makes full use of prior information, conducts a test item
by item, and reports its outcome in decision risk terms. The aim
is for a coping procedure without critical item requirements.
Real-data simulations are carried out using a large sample of
recruits' answers to vocabulary items. An independent conventional
verbal test provides a basis for item calibration.
The procedure uses two novel Indices of item performance concerned with the tails of the empirical item characteristic curves
and their interaction with the normative recruit distribution of
verbal attainment. These indices are held to be more relevant to
effective convergence of the tailoring process.
Good accuracy of convergence is demonstrated by the procedure,
and savings in test length for the average recruit as well as greater
savings for the non-average.
Empirical studies are needed to investigate the temporal dimension of tailored testing.