Savants are extremely rare individuals who, although often severly brain impaired,
frequently by autism, can display islands of astonishing excellence in the same peculiarly
restricted areas, across all cultures. Their skills are literal, non-symbolic, and appar-
ently not derived from practice. They often emerge `spontaneously' and do not improve
qualitatively with time, even though the skill might be better articulated. Savants typically
have no idea how they do it (Rimland 1963; Treffert 2000, 2005).
In the words of one pioneering researcher, their ``gift springs so to speak from the
ground, unbidden, apparently untrained and at the age of somewhere between five and
eight years of age. There is often no family history of the talent'' and it ``is apparently
not improved with practice'' (O'Connor 1989, page 4).
It has been hypothesised (Snyder and Mitchell 1999) that savants have privileged
access to raw sensory details, before these details are assembled into concepts, mean-
ingful labels, and holistic pictures. All brains possesses this same raw information but,
without some sort of brain dysfunction, or altered states of mind (Humphrey 2002;
Sacks 2003), it is normally beyond conscious access. We tend to see the whole and
not the parts (Howe 1989, page 83). Savants tend to see the parts and not the whole.
They are literal with an inclination to focus on local, rather than global aspects
of the scene, and to recall detail without meaning (Rimland 1963; Frith and Hill 2003;
Snyder 2004).
4.2 Why does being literal enhance numerosity?
How does being literal enhance numerosity? We argue that it removes our unconscious
tendency to group discrete elements into meaningful patterns, like grouping stars into
constellations, which would normally interfere with accurate estimation. By being literal,
a savant sees elements as discrete and disconnected, thus removing this interference.
This explanation is consistent with the fact that the accuracy of estimating numbers
of elements depends on their arrangement (Ginsburg 1976, 1991; Boles 1986; Ginsburg
and Goldstein 1987; Dehaene 1997) and even on the sensory properties of the stimulus
(Barth et al 2003). As Krueger (1984) concluded, ``...perceived numerosity depends more
on higher level cognitive factors...than on lower level perceptual or sensory factors''.
Indeed, the healthy normal brain makes hypotheses in order to extract meaning
from the sensory input, hypotheses derived from prior experience (Gregory 1970, 2004;
Snyder and Barlow 1988; Snyder et al 2004). If perceived numerosity depends on
higher-level cognitive factors, then the estimation of number is likely to be performed
on this hypothesised content, not on the actual raw sensory input, thus exaggerating
errors in estimation that would otherwise be absent.
In sum, we argue that the estimation of number by normal people is performed
on information after it has been processed into meaningful patterns. The unconscious
meaning we assign to these patterns interferes with our accuracy of estimation, whereas
savants, by virtue of being literal, have less interference. This, together with the fact
842 A Snyder, H Bahramali, T Hawker, D J Mitchell
that it takes only a handful of precise measurements to calibrate our number estimation
system (Dehaene 1997, page 71), could explain the reported numerosity feats of savants.