THE REAL ERROR OF CYRIL BURTjections. He identified as group factors the clusters of positive and
negative projections themselves.
Burt's identification of group factors may seem, superficially, to
challenge Spearman's theory, but in fact it provided an extension
and improvement that Spearman eventually welcomed. The
essence of Spearman's claim is the primacy of g, and the subordi-
nation of all other determinants of intelligence to it. Burt's identi-
fication of group factors preserved this notion of hierarchy, and
extended it by adding another level between g and 5. In fact, Burt's
treatment of group factors as a level in a hierarchy subordinate to
g saved Spearman's theory from the data that seemed to threaten
it. Spearman originally denied group factors, but evidence for
them continued to accumulate. Many factorists began to view this
evidence as a denigration of g and as a wedge for toppling Spear-
man's entire edifice. Burt strengthened the building, preserved the
preeminent role of g, and extended Spearman's theory by enu-
merating further levels subordinate to g. The factors, Burt wrote
(1949, p. 199), are "organized on what may be called a hierarchical
basis.... There is first a comprehensive general factor, covering
all cognitive activities; next a comparatively small number of broad
group factors, covering different abilities classified according to
their form or content.... The whole series appears to be arranged
on successive levels, the factors on the lowest level being the most
specific and the most numerous of all."
Spearman had advocated a two-factor theory; Burt proclaimed
a four-factor theory: the general factor or Speai man's g, the partic-
ular or group factors that he had identified, the specific factors or
Spearman's s (attributes of a single trait measured on all occasions),
and what Burt called accidental factors, or attributes of a single trait
measured only on a single occasion. Burt had synthesized all per-
spectives. In Spearman's terms, his theory was monarchic in rec-
ognizing the domination of g, oligarchic in its identification of
group factors, and anarchic in recognizing s-factors for each test.
But Burt's scheme was no compromise; it was Spearman's hierar-
chical theory with yet another level subordinate to g.
This accidental variance, representing peculiarities of particular testing situations,
torms part of what statisticians call "measurement error." It is important to quan-
tify, for it may form a basic level of comparison for the identification of causes in a
'amily of techniques called the "analysis of variance." But it represents the peculi-
arity of an occasion, not a quality either of a test or a testee.