Ch apter I

E v a lu a tin g the validity of

‘s c h i z o p h r e n i a *

Two major claims have been made about the concept of schizophrenia: first,
that it is a scientific concept or, at least, that those who use it work within a
scientific framework (for example, Gottesman and Shields, 1982; W ing, 1988;
APA, 1994; Sartorius, 1994; Ross and Pam, 1995) and, second, that the term
refers to a particular kind of medical pattern known as a syndrome (Gottesman
and Shields, 1982; Kendell, 1991; APA, 1994). These claims imply that the
concept has been developed and is used in a manner similar to that of other
concepts which claim scientific status and that the phenomena to which it
refers are similar to those usually denoted by the medical term ‘syndrome’.
This chapter will describe in detail what is meant by these claims, not least
because their meaning is rarely spelt out and is often misrepresented in the lit­
erature, but also as a necessary prelude to evaluating whether the introduction,
development and use of the concept of schizophrenia actually conforms to the
claims made about it.

‘S c h i z o p h r e n i a ’ as a s c i e n t i f i c c o n c e p t

As Medawar (1984) and Chalmers (1990) have pointed out, attem pts to define
‘science’, to give the term some absolute meaning, have always ended in fail­
ure. With this in mind, all that will be attempted here is a description of some
of the ways in which those who call themselves scientists —particularly natural
and medical scientists, with whom supporters o f ‘schizophrenia have identified
themselves —tend to proceed. This rather pragm atic stance, however, should
not be thought of as second best: the crucial issue is that, given the claims
made about ‘schizophrenia’, then the ways it has been developed and used
ought at least to be recognisable to philosophers and practitioners of natural
and medical science. Kendell (1991), a supporter o f ‘schizophrenia’, concurs,
claiming that ‘[the concept of schizophrenia] can only be judged by the same
criteria as other explanatory concepts’ (60).
The search for patterns and attem pts to describe relationships between phe­
nomena are fundamental to scientific activity. They are central, too, to much
non-scientific activity; of particular interest here is that they are central to lay
2 Evaluating the validity o f ‘sc h iz o p h re n ia ’

or everyday attem pts to understand behaviour and other events. What perhaps
distinguishes the two is the scientist’s persistent demand for the provision of
certain types o f evidence that a pattern has been observed and the imposition
of various publicly demonstrable criteria for evaluating it. This demand for
evidence, and its public evaluation, is crucial in view of our apparent propen­
sity to claim that certain events ‘go together’, in the absence of any direct
evidence that this is the case (see, for example, Shweder, 1977; Chapman and
Chapman, 1982).

Observable s a nd unobservables in scientific theory

Scientific theories contain both observables and unobservables. The term

‘observable’ can give the misleading impression of referring to some outside
reality, but I am using it here simply to mean readily agreed statements about
sense data with a minimum o f interpretation. In scientific research, simple
statements about putative relationships between observables (when certain
metals are brought close together, one moves towards the other) are used to
justify inferring unobservables (magnetic force); as we shall see, these unob­
servables are then used to aid the construction of more elaborate theoretical
statements.
It is readily agreed that ‘schizophrenia’ is an unobservable, an abstract con­
cept inferred from overt behaviour or from verbal reports of behaviour and
experience (Kendell, 1991; APA, 1994). Scientific theories, however, may
contain different types of unobservables and it is important to distinguish
them in assessing the validity o f ‘schizophrenia’. Beck (1953) has distin ­
guished two types of unobservables by applying the concepts of systemic and
real ‘existence’ and the language of logic to differentiate them. The first, he
says, is that ‘mode of existence of an (unobserved) entity all descriptions of
which are analytical within a system of propositions’, while real existence ‘is
the mode o f existence attributed to an entity if there is any true synthetic
proposition that can be made about it’ (369). In other words, statements about
an unobservable said to have real existence will contain words which are not
reducible to the empirical relationships from which it is inferred and will
imply hypotheses about the antecedents of these empirical relationships.
Benjamin (1937) has made the same distinction by contrasting what he calls
abstractive and hypothetical methods in scientific activity. In the abstractive
method, phenomena are grouped by a restrictive set of properties into classes
whose relationship can be discovered empirically; nothing is added to what is
observed. By contrast, the hypothetical method relates observations by ‘invent­
ing a fictitious substance or process or idea in terms of which the experiences
can be expressed. It correlates observations by adding something to them ’
(184). Constructs formed by Benjam in’s abstractive method (e.g. solubility,
resistance, temperature, habit strength; hunger as X hours of food deprivation,
and so on) would be said by Beck to have systemic existence, while those
 Evaluating the validity o f ‘sc h iz o p h re n ia ’ 3

formed by the hypothetical method (electricity; proton; intelligence; diabetes)

would be said to have real existence. MacCorquodale and Meehl (1948) have
suggested the terms intervening variables’ and ‘hypothetical constructs’ to dis­
tinguish these two types of unobservables, and these terms (with ‘construct’
and ‘concept’ used interchangeably) will be adopted here. Other names have,
however, been suggested: Beck uses constructions and inferred entities, while
Carnap (1937) uses the term ‘dispositional concepts’ rather than ‘intervening
variables’. Although all hypothetical concepts refer to unobservables, it is
worth noting a distinction between two different types. The first is that which
becomes, in Beck’s phrase, an ‘object o f search’ and which is postulated by sci­
entists to show certain characteristics. The constructs of atom or proton are
examples of this type; although unobserved at the moment, they are claimed
to exist as a result of certain observations and mathematical calculations and
may be observed at some future point, given appropriate technology. The
second type o f unobservable remains for ever an abstraction, which cannot be
postulated to exist except in the most abstract sense. ‘Intelligence’, ‘memory’,
‘diabetes’ and ‘m ultiple sclerosis’ are examples of this type of hypothetical con­
struct.
W hich kind o f concept is ‘schizophrenia? Kraepelin clearly did not see his
concept of dementia praecox, from which the modern concept of schizophre­
nia is derived, as being reducible to a statement about correlations between
behaviours. Instead, he postulated a ‘m etabolic disorder’ to account for the
putative correlations. Similarly, statements made today about the concept con­
tain terms not reducible to statem ents about behavioural correlations.
‘Schizophrenia’ is said to be an illness; a biochemical disorder; a genetic dis­
order. W hat is implied by statements like these is not always specified but it
is clear that something is being added to a statement about behavioural cor­
relations. 'Schizophrenia' therefore functions as a hypothetical construct rather
than an intervening variable; and, as an abstraction from observables, rather
than an 'entity' postulated to exist as a result of observations, it also functions
as one of the second type of hypothetical construct - no one ever expects to ‘see’
schizophrenia, any more than they expect to ‘see’ diabetes, regardless o f how
advanced technology becomes. It must be emphasised that this status as an
eternal abstraction is perfectly respectable and has nothing to do with the pre­
sent lack of a clear relationship between biochemical events and behaviour said
to be symptomatic of schizophrenia. Even if such events were to be observed
in the future, the concept o f schizophrenia would remain an abstraction
because there is no identity between it and any biological, genetic or behav­
ioural event, any more than there is identity between the concept o f diabetes
and a specific biological event.
The most important distinction between hypothetical constructs (of both
types) and intervening variables lies in the predictive function of hypothetical
constructs (Benjamin, 1937; MacCorquodale and Meehl, 1948; Beck, 1953;
Carnap, 1974). They have, or should have, the power to generate predictions
4 Evaluating the validity o f ‘sc h iz o p h re n ia ’

about events which have not yet been observed. This power derives partly from
the fact that hypothetical constructs, unlike intervening variables, are not
reducible to statem ents about what has already been observed, but imply
hypotheses about the antecedents of what has been observed. This distinction
has important implications for evaluating the validity of the two types of con­
struct. The validity o f an intervening variable can be questioned only by
denying the observations from which it was originally inferred, i.e. by claim ­
ing that such and such did not actually happen. The concept of solubility, for
example, requires only that substances be observed to dissolve at different
rates and to different extents in water or other liquids; it implies nothing about
the variables which control this. These observations are therefore said to be
both necessary and sufficient conditions for asserting the validity of the concept.
For a hypothetical construct, the actual occurrence of the empirical relation­
ships from which it was inferred is, of course, a necessary condition for
asserting its validity, but it is not sufficient. The demonstration o f predictive
power is central to assertions of the validity of hypothetical constructs and is
the sufficient condition for inferring them.

The corre s p o n d e n ce rules o f hypothetical constructs

To say that a construct has predictive power is to say that it is capable of pre­
dicting events which, though observable in principle, have not yet been
observed. Thus, the concept o f intelligence may be capable o f predicting
performance differences in a laboratory task. It is possible to use unobserv­
ables to make and investigate statem ents about observables because o f what
are variously called correspondence rules (Carnap, 1974), operational rules
(Bridgm an, 1927) and The Dictionary (Cam pbell, 1920; cited in Carnap,
1974). These rules, which may be very simple or highly complex, specify
what must be observed before a concept can be inferred and may specify
quantitatively the relationship between variation in what is observed and
variation in the inferred construct. The correspondence rules for the concept
o f intelligence, for exam ple, specify the relationship between observable
responses to items on a standardised test and 'am ounts’ of the unobservable
concept, intelligence. Any investigation of the predictive power of a hypo­
thetical construct therefore involves examining the relationship between two
sets o f observables.

The process o f ch a n g e o f c o rre s p o n d e n ce rules

Inferred constructs usually start as relatively vague concepts associated with

certain observations. It is then discovered, by empirical investigation, that
these observations vary systematically with another observation which can be
made more reliably and is more strongly correlated with relevant experimen­
tal changes. The concept of temperature, for example, was originally associated
 E va lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’ 5

with global, subjective judgem ents o f hot and cold. It was later shown that
these judgem ents and a range of other phenomena were reliably correlated with
m easurable changes in the height o f a mercury column. The correspondence
rules for inferred constructs (i.e. what m ust be observed before the construct is
inferred) therefore change over tim e as a result o f increasing specification and
elaboration o f events which can be shown to be associated with the original set
o f observations from which the construct was inferred. That correspondence
rules do change in this way may o f course dem onstrate a construct’s predictive
power.
This process o f change in correspondence rules is well illustrated by the
development o f the medical concept o f diabetes mellitus. The construct o f dia­
betes was origin ally inferred by the G reeks from observations o f the
co-occurrence o f inordinate thirst and urine production, lethargy and em acia­
tion. These observations therefore constituted the first set o f correspondence
rules for the hypothetical construct o f diabetes. In the seventeenth century it
was noted that this cluster was frequently and reliably associated with sweet-
tasting urine. Thus, a new correspondence rule could be set up between an
observable event and the unobservable ‘diabetes’. Later, it was discovered that
sw eet-tasting urine contained glucose which could be detected independently
o f the original observation o f inordinate thirst, and so on, or judgem ents
about the sweetness o f urine. Yet another set o f correspondence rules could
therefore be set up. These, in turn, were superseded by rules specifying that
diabetes m ellitus was to be inferred when a certain relationship between intake
o f glucose and its level in the blood at certain time intervals was observed. (The
‘division’ o f the concept o f diabetes into m ellitus and insipidus was necessi­
tated by the observations that inordinate thirst and urine production were not
always associated with sw eet-tasting urine and, later, that they could be asso­
ciated with low levels o f anti-diuretic hormone.)
Every attem pt to examine the fate o f predictions from a hypothetical con­
struct involves the use o f its correspondence rules. Each attem pt to show that
the sufficient conditions for inferring the construct are fulfilled therefore
involves invoking the necessary conditions for inferring its existence. In the very
early stages o f a construct’s developm ent, the necessary conditions —the p uta­
tive regularities from which the construct was originally inferred - will also be
the correspondence rules. Later, the necessary conditions will be invoked indi­
rectly as the correspondence rules change. B u t it would be quite wrong to
depict this process of change in correspondence rules as one of finding out what
a hypothetical construct ‘is’. Young (1 9 5 1 ) and Carnap (1 974) have noted
how often the question, ‘W hat is X ? ’, where X is a hypothetical construct, is
put to scientists by lay people. They point out that the question is unanswer­
able and is based on a m isunderstandin g o f the function o f hypothetical
constructs. These concepts can only be described in terms o f the observed
events from which they are inferred and the predictions which have been
made from them, in terms, that is, of the theoretical network in which they are
6 Eva lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’

embedded. The question ‘W hat is diabetes?’ would be answered (and attem pts,
however m isguided, are often made to answer such questions) very differently
in classical Greece and in the modern Western world and w ill, no doubt, be
answered very differently again in the twenty-second century as the theoreti­
cal network is further elaborated. Each answer is ‘true’ but m isleading; to
pose the question at all is to reify the construct and to im ply that there is a
final, concrete answer.

D e r iv in g p re d ict io n s from h y p o t h e t ic a l c on stru c ts

The correspondence rules which tie unobservable to observable events provide

the general means for exam ining predictions from a construct. The content of
these predictions can be specified only by exam ining the whole theoretical net­
work which surrounds the concept, the assum ptions on which it is based,
indeed, everything that is asserted about it (Cronbach and Meehl, 1955). The
statem ents which make up this network may relate observables to each other,
unobservables to observables, and/or unobservables to one another. But it is
axiom atic that any construct which claim s scientific status be em bedded in a
network in which at least some o f its statem ents contain observables. The ease
with which researchers can derive predictions from this network, and then
investigate them, depends not on its sim plicity but on its specificity. One
obvious result o f a lack o f specificity is disagreem ent over the content o f pre­
dictions; another is disagreem ent over the results o f attem pts to test an agreed
prediction. There is no question o f a concept’s validity depending on a partic­
ular number o f the predictions derived from it being upheld; there is certainly
no quantitative answer to the question, ‘How valid is construct X ? ’ Rather,
validity is usually assessed from a utilitarian stance by asking in what ways and
for what purposes the construct has proved useful. The term ‘u tility ’ is there­
fore often used instead o f ‘v alid ity ’ to reflect the fact that, as a rule, a
hypothetical construct has no claim to validity unless it can be used to predict
events which, without the construct, would probably have gone undetected. If
the same events are predicted by different constructs, then that which carries
fewest assum ptions is usually preferred.
It may happen, however, that research fails to detect events predicted from
a construct. The question then arises o f whether the 'fault' lies with the con­
struct, and its theoretical network, or with the method o f inquiry. As will be
seen in Chapter 2, the nineteenth-century m edical profession sought to retain
the concept o f mental disease even though they were unable to detect the pre­
dicted differences between the brains o f those said to be mentally diseased and
the brains o f those who were not. It was argued that existing methods o f mea­
suring brain function were not sufficiently advanced to detect the postulated
differences. T h is may well have been true, but such a post hoc argum ent could
be used indefinitely to justify the continued use o f a concept, one o f whose
m ajor predictions had no em pirical support. It can be argued instead that if a
 Evaluating the validity o f ‘sc h iz o p h re n ia ’ 7

construct has already led to the detection of previously unobserved events and
if it predicts specific events observable in principle but undetectable by exist­
ing methods, then the failure to detect such an event m ight well indicate that
a more sophisticated methodology is required and not that the construct is
invalid. But if a concept is in the early stages of development, if it lacks a his­
tory of ‘successful’ predictions or is embedded in a loose theoretical network
which cannot be used to predict specific events, then the ‘failure’ o f a major
prediction should direct attention to the concept itself. In particular, such fail­
ures should direct our attention to the fundamental question of whether the
necessary condition for inferring the concept has been fulfilled; that is, was a
set of regularities which would justify bringing the concept into existence ever
observed? The question is crucial, as this necessary condition is invoked,
directly or indirectly, every time a prediction is ‘tested’. This failure to make
reliable predictions to new observations, being embedded in a loose and ever-
changing theoretical network and failing to specify what would be observed
were technology to improve, is precisely the problem which besets the concept
of schizophrenia (Bentall, 1990b; Boyle, 1994; British Psychological Society,
2000). The question, then, of whether the necessary conditions for inferring
schizophrenia have been fulfilled — of whether anyone has ever observed a
pattern of regularities which would justify the original and continued existence
of the concept —will be central to this book.

Lay a nd scientific con cep ts

These methods o f developing and investigating constructs can be compared

with the development and use of what are often called ‘lay’ or ‘folk’ concepts.
Our language is rich in terms, particularly those referring to people, which are
superficially sim ilar to hypothetical constructs. Some of them (deep; hard) are
obviously metaphorical and others (anxious; normal) less obviously so. There
are some major differences between these concepts and those whose claims to
scientific status are rarely disputed; two of these are of particular interest here.
The first is that the correspondence rules of lay concepts are often many and
varied, and vary from user to user. They also tend to change in idiosyncratic
ways over time: the referents of the concept ‘nice’, for example, used to be very
similar, when applied to women, to those of later concepts like ‘fast’. The
second difference is that, although these terms may appear to be derived from
patterns of behaviour, no systematic attem pts are usually made to check this;
or, when they have been made (for example, Shweder, 1977; Kahneman et a l.,
1982) the apparent patterns have been shown to be more closely related to cul­
tural beliefs about ‘what goes with what’ than to reliable observations. It must
be emphasised that attem pts to demonstrate that lay concepts are not neces­
sarily ‘messy’ (e.g. Cantor and Mischel, 1977; Cantor et a l., 1980) do not
affect their status as lay constructs but at most can be said to demonstrate the
orderliness of cultural beliefs and stereotypes.
8 Evaluating the validity o f ‘sc h iz o p h re n ia ’

None o f this implies criticism of lay concepts or demonstrates their inferi­

ority. They are not intended to serve the functions of scientific concepts but to
be used in day-to-day discourse. It would place an intolerable burden on our
interaction if we had to assess the reliability and validity of every concept
before we could use it. Problems arise only when lay concepts are used as i f
they had been derived in a way quite different from they actually were (see, for
example, Sarbin’s 1968 discussion o f ‘anxiety’ and W arburton’s 1985 discus­
sion o f ‘addiction’). I am making these points in some detail here because they
may be important in understanding some of the problems encountered by
‘schizophrenia’ —not least because our sheer familiarity with lay concepts, in
all their ‘looseness’, may blunt our critical faculties when it comes to assessing
‘schizophrenia’.

‘S c h i z o p h r e n i a ’ as a s y n d r o m e

The claim that ‘schizophrenia’ is a scientific concept implies that it was derived
from the observation o f a pattern o f regularities. The claim that the term
refers to a syndrome implies that it was derived from, and now refers to, a par­
ticular type of pattern. The nature of this pattern, and the ways in which it
differs from that originally envisaged by Kraepelin, can perhaps best be clar­
ified by describing some of the historical background to modern medical ideas
about pattern description.
The aim of medicine, as of all branches of science, is to describe patterns and
relationships. In medicine, however, ideas about pattern identification, about
the grouping of phenomena, are inseparable from ideas about the concept of
disease. Engle (1963) has described two important and contrasting views on
diseases which can be traced to classical Greece. The Platonic tradition taught
that reality was universal and unchanging, unlike perceptions received through
the senses, which were relative and imperfect. Applied to medicine, these
ideas led to a search for unvarying universals —individual diseases ‘out there’
and separable from the person. The Aristotelian tradition, by contrast, did not
scorn sense data and encouraged the detailed study, for its own sake, of what
were assumed to be the manifestations of disease in individuals, rather than the
search for abstract (or metaphysical?) universals.
In practice, o f course, it was not always easy to distinguish those who
claimed adherence to one school or the other. The theories were expressed in a
way which made their implications for research and practice less than clear, and
in any case adherents to the Platonic school, their distrust o f the senses
notwithstanding, were forced to work with phenomena as they presented to
the senses. They did this to such an extent that it seemed that every phenom­
enon they observed was classified as a disease - skin rashes, swellings, fevers,
etc. Indeed, de Sauvages was able, in 1763, to list 2,400 diseases, most of
which would now be considered to be individual sym ptom s (Zilboorg, 1941).
It would therefore be naive to expect the Platonic and Aristotelian views to be
 E va lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’ 9

clearly distinguish able in the actual activities o f medical men (as distinct
from what they said they were doing); nevertheless, two distinct themes,
roughly corresponding to the two ancient traditions, can be discerned in the
w ritings o f philosophers and physicians on ‘disease’. These have been variously
characterised as ontological (disease-entity) versus biographical; qualitative
versus qu an titative; discon tin uous versus continuous (K en d ell, 1975b).
Thom as Sydenham was one o f the most articulate proponents o f the ontolog­
ical view o f disease. W ritin g at the end o f the seventeenth century, he
reiterated the belief in the existence o f natural and unvarying disease entities,
separable from the person, and whose presentation was uniform across suffer­
ers. These entities, he maintained, m ust be separated because each required a
different treatm ent. Sydenham ’s ideas, which were heavily influenced by the
classification system s o f botan ists and by the ‘disease’ o f m alaria, were
extremely popular because, as Kraupl-Taylor (1979) suggests, they encouraged
medical men to participate fully, via clinical observations, in the search for
clinical data on which to base a classificatory system. Indeed, this was done
with such enthusiasm that for alm ost every physician there was a classificatory
system. These efforts were seriously hampered not only by tenuous acquain­
tance with the principles o f classification (Zilboorg, 1941) but also by the fact
that observations were subjective and lim ited to what could be observed ‘at the
bedside’.
Sydenham ’s ontological theories were strongly challenged in the late eigh ­
teenth century and the early nineteenth. In particular, critics attacked his
view o f disease entities as having a separate existence and suggested that d is­
ease was a quantitative and not a qualitative deviation from the norm, and that
the course m igh t vary from individual to individual. In 1 8 4 7 , Virchow
declared that ‘Diseases have no independent or isolated existence; they are not
autonom ous organism s, nor beings invading a body, nor parasites grow ing on
it; they are only the m anifestations o f life process under altered conditions’
(Kraupl-Taylor, 1979: 11). But later in the century Virchow was to alter his
views to the extent that he was able, in 1895, to call him self a ‘thoroughgoing
on tologist’. H is concept o f disease entity was, however, very different from
Sydenham ’s. The introduction o f the microscope and the practice of histology
in the late nineteenth century allowed the detailed investigation o f the various
bodily derangem ents which accom panied overt sym ptom s. For Virchow, a
disease entity was a particular pathological abnormality. A disease entity there­
fore became an altered body part, a significant change which avoided the
m etaphysical overtones o f Sydenh am ’s theory. C lassification consisted o f
descriptions o f the various types o f change which could be observed in differ­
ent body organs; diagnosis was the m atching o f these to the changes observed
in a particular patient.
Virchow’s views in turn were challenged by bacteriologists at the end o f the
nineteenth century. The nature o f their dispute is clearly illustrated by an
exam ple given by K raupl-Taylor (1979). Virchow began from the prem ise
10 E va lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’

that a pathological abnorm ality constituted a disease entity and that the d iag­
nosis had to name the abnormality, no m atter how it was caused. D iphtheria,
for example, would be diagnosed whenever surface necrosis and membrane for­
m ation were observed in an organ. That these changes could have causes other
than the presence o f the diphtheria bacilli was not im portant to Virchow. To
the bacteriologists, diagnosis consisted in identifying the kind o f organism
which had invaded the body, regardless o f the nature o f the pathological
changes it m ight produce in any individual. The bacteriologists’ view's won the
day because they proved more useful than those o f Virchow. But the debate
between them and Virchow was between one kind o f ontological theory and
another; there is no doubt that the b acterio lo gists’ fin din gs served to
strengthen the already popular ontological theories to an extent never achieved
by Sydenham and his many followers. Once again, a disease entity had become
a discrete and separate unit, with its own distinctive cause, sym ptom s, course
and outcome.
It is interesting to note the strength o f ontological theories in popular d is­
course about physical ailm ents. As Kraupl-Taylor (1 9 7 9 ) has pointed out,
much o f the language surrounding disease im plies an ontological theory —we
talk o f ‘catching’ or ‘gettin g rid o f a disease’; we are ‘attacked by diseases’; we
‘carry diseases’ which we ‘pass on’ to our children or to others. It is also notable
that ontological theories seem to offer a reassuringly sim ple solution to the
apparent chaos o f physical suffering: a number o f seem ingly disparate phe­
nom ena can be accounted for and perhaps abolished by reference to one
underlying cause.

From disease entity to s y n d ro m e

K raepelin’s ideas about the phenomena with which he was confronted in asy­
lum s were derived from a Platonic view o f the w'orld. H is belief in natural
disease entities with an independent existence was, o f course, strengthened by
progress in understanding the so-called infectious diseases: the finding that
single m icro-organism s, with an independent existence, were responsible for
certain clusters o f phenomena with their own, apparently natural, course and
outcome. Kraepelin accepted unquestioningly that the behaviour o f asylum
inm ates w'as a manifestation o f biological events, just as were the fevers, rashes
and at tim es odd behaviours o f those infected with various m icro-organism s.
H e therefore assum ed that the inm ates’ behaviour would be found to fall into
natural clusters, representing qualitative deviations from whatever he thought
o f as norm ality and that each cluster would have its own distinct antecedent,
both necessary and sufficient to produce the cluster.
But as I noted earlier, there were conflicting views as to the nature o f these
entities. Some said the entity was the observed cluster; others that it was the
antecedent o f the cluster; still others that it was the anatom ical pathology
which accompanied the overt cluster. Indeed, as Kraupl-Taylor (1 982) has
 Evaluating the validity o f ‘sc h iz o p h re n ia ’ II

pointed out, Sydenham him self appears to have used the term 'disease entity’
in two quite different ways — to refer both to an independent, God-given
species and to the body’s reaction to invasion by external agents. Kraepelin was
well aware of the fact that any attem pt to postulate antecedents for putative
clusters of behaviour was pure speculation; similarly, it made no sense to su g­
gest that the term 'disease entity’ be applied to morbid anatomy when none
could be found in many asylum inmates. However, as Jaspers (1963) has
remarked, Kraepelin ‘embarked on a new approach which hoped to arrive at
disease entities in spite of everything’ (5 66). This approach consisted in observ­
ing inm ates’ behaviour in an attem pt to discover sim ilarities amongst the
m ost frequently appearing behaviours and sim ilarities in the ways they
changed over time. Charting behaviour changes over time (called ‘the whole
course of the illness’) presented considerable problems as asylum doctors had
direct access to inm ates’ behaviour only from the time of their incarceration.
A ttem pts were made, however, to reconstruct the past from discussion with
inmates or their relatives. Strictly speaking, K raepelin’s approach was not
new; followers of both the Platonic and the Aristotelian traditions had stressed
the importance of careful observation o f morbid manifestations, although
attem pts to describe similarities amongst these phenomena in different indi­
viduals, and the view that a disease entity accounted for them, were largely
confined to the Platonic school. The novelty of Kraepelin’s approach lay in its
emphasis on investigating the ‘whole course of the illness’ in an attempt to dis­
cover natural groupings, entities or patterns. Kraepelin believed that the
behavioural clusters he hoped to discover would be found to have distinctive
biological antecedents and cerebral pathology as well as course and outcome.
The totality of this pattern would be called a natural disease entity.
It is well documented that Kraepelin’s hopes of finding such patterns were
never realised, either by him or by any of his successors (Jaspers, 1963; W ing,
1978a; Gottesman and Shields, 1982; Kendell, 1991). Instead, as I noted ear­
lier, the term ‘schizophrenia’ is now said to refer to a syndrome, which is
usually described as a clustering o f sym ptom s and signs (Morris, 1978). More
generally, the term ‘syndrome’ refers to the very basic fact that certain phe­
nomena appear to cluster at greater than chance level. It is not assumed that
the clusters of events denoted by the syndrome name are distinctive natural
groupings, and people who display the cluster m ight vary widely in the way
it changes over time. The antecedents of the cluster are generally unknown.
The connection between the medical idea of syndrome and more general sci­
entific research —a connection which is crucial in evaluating ‘schizophrenia’ -
can be seen in the fact that most syndrome names (e.g. rheumatoid arthritis)
refer to hypothetical constructs: they are abstractions inferred from the obser­
vation o f patterns of regularities. They should therefore fulfil the functions of
summ arising a pattern of observations and of allowing predictions to as yet
unobserved events. Syndrome names are thus theoretical abstractions which are
thought to be useful in research but which may be abandoned if and when they
12 E va lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’

cease to fulfil this function. The construct o f D ow n’s syndrome, for exam ple,
has proved extremely useful in that it allowed predictions to previously unob­
served features o f chromosom es. The idea o f a syndrome, unlike that o f a
disease entity, can encom pass both qualitative and quantitative deviations
from a norm: possessing certain characteristics which represent a quantitative
deviation from some standard can still have im plications which are different in
im portant ways from those associated with the possession o f a standard set of
characteristics.

The id e ntifica tion o f s y n d r o m e p a tterns

One function o f the construct denoted by a syndrome name is to sum m arise a

meaningful cluster o f phenomena; that is, a cluster unlikely to have occurred
by chance and therefore likely to signify other, as yet unknown, events. But
whereas it is easy to im agine that certain occurrences form a pattern (Shweder,
1977; Chapm an and Chapm an, 1982), it is much more difficult to determ ine
whether a pattern actually has been observed. The m ethod favoured by
Kraepelin —that of charting the progress and outcome o f a postulated cluster -
has a number o f problem s associated with it, whether or not it involves ideas
about disease entities. The use o f the term ‘outcom e’ in this context im plies,
erroneously, that a specific end-point can be identified. In practice, however,
the term usually refers to a com plex series o f events or processes with no obvi­
ous end-point. Given the problem s of deciding which variables to measure and
when to end the process, and the finding that most o f the clusters which are
the source o f syndrom e names show considerable variability over tim e, it
would be unproductive to rely on outcome as a criterion o f meaningfulness for
any postulated cluster. The problem is com pounded by the fact that identifi­
cation o f possible new instances o f the cluster (diagnosis) is im possible until
their progress is charted, to check that it matches that o f the earlier instances.
A reliance on ‘outcom e’ is apt to lead to post hoc diagnosis (and, as will be
shown, K raepelin h im self fell into this trap) where a diagnostic label may be
changed several tim es, depending on the progress o f the phenomena from
which it was inferred. For these reasons, reliance on progress or outcom e to
identify m eaningful patterns has generally been rejected by m edical
researchers. A much less problem atic and more productive method o f ju dgin g
whether a postulated cluster is m eaningful —and the one usually adopted in
medicine —is to search for a phenomenon which is reliably associated with the
cluster and which has two other im portant characteristics: first, it can be mea­
sured reliably and independently o f the cluster it is associated with and,
second, it can be shown or reasonably assum ed to be an antecedent o f the clus­
ter. In turn, this phenomenon may later be found to be associated with another
reliably and independently measurable event, and so on along an assum ed
‘causal’ chain. It is in this way, o f course, that correspondence rules for an unob­
servable may becom e more reliable and m ore specific and its theoretical
 Evaluating the validity o f ‘sc h iz o p h re n ia ’ 13

network more elaborate, as was described earlier for the concept of diabetes
mellitus.
The identification o f patterns known as syndromes is therefore a two-stage
process: a researcher or clinician first identifies a cluster of features which
they believe hang together’. They, or someone else, then needs to show that
this cluster is reliably associated with another feature which can be measured
independently. The importance of this two-stage process can be seen by exam­
ining the type o f bodily features usually involved at each stage. Those features
which make up the cluster at the first stage are usually those most easily and
directly perceptible to an observer or to the person themselves (for example,
pallor, sweating, rashes, vom iting, pain, and so on). Such phenomena are usu­
ally called symptoms and share a number of characteristics. First, they are often
not directly observable to an onlooker (e.g. nausea) but are made available by
verbal report. Second, the reliability with which they can be observed may be
low. Third, they are overdetermined in that each may have many antecedents.
The clustering of any of these events in an individual m ight therefore be a
chance occurrence; worse, if the reliability of reports of their occurrence is low,
then reports of their co-occurrence may be false. It is therefore very unwise to
assume that any reported co-occurrence of events like these is meaningful or to
infer from it the existence o f an unobserved process. Instead, it is necessary to
go to the second stage and to demonstrate that any supposed cluster of sym p­
toms is reliably associated with another independently measurable event. In
medicine, such events are called ‘sign s’ and they differ from sym ptom s in a
number of important ways. First, they can be observed with a much higher
degree of reliability. Second, they are directly available to an observer, rather
than being available only through introspection and verbal report or by infer­
ence from, for example, shaking or moaning. Third, although signs are also
overdetermined, the number of antecedents is thought to be fewer than for
symptoms. An unselected population will therefore show signs far less often
than symptoms. Fourth, there should be plausible, even if speculative, theo­
retical links between those signs and symptoms whose co-occurrence is said to
be meaningful, to justify the assumption that the signs are antecedents of the
symptoms.
The distinction between signs and sym ptom s can be clarified using the
example of glucose in the urine (Kraupl-Taylor, 1979). This is called a sign
because it can be reliably measured by an external observer, its frequency of
occurrence is less than is the individual frequency of events with which it may
be associated (excess urine production, thirst, tiredness, and so on) and
because there are plausible grounds for assum ing that it is not a consequence
of these. The term ‘sig n ’ suggests that an event signifies, or is indicative of,
another event which can be independently observed. The presence o f glucose
in the urine may signify high blood-glucose (itself an overdetermined event)
or dim inished glucose reabsorption in the renal tubes. Because hypergly-
caemia fulfils the criteria listed above, it too is a sign, in this case perhaps of
14 E va lu atin g the v alid ity o f ‘s c h iz o p h r e n ia ’

pancreatic abnorm alities. Sign s, however, are often m istakenly said to be

indicative o f a syndrom e name as, for exam ple, when high blood-glucose
levels are said to be a sign of diabetes. Statem ents like these are obviously tau­
tological because syndrome names are concepts inferred from clusters o f signs
and sym ptom s; they are not observable events which can be measured inde­
pendently o f the sign.
People who receive the same syndrome name as a diagnostic label will usu­
ally be hom ogeneous for at least one sign in the cluster from which the
syndrome name is inferred. They w ill, how'ever, be heterogeneous for sym p­
tom s because the sam e sign may be associated with a variety o f sym ptom s.
This heterogeneity su ggests that unknown m ediating variables are operating
in the chain from sign to sym ptom s. Similarly, because signs are overdeter­
m ined, people who show the same sign may be given different diagnostic
labels to reflect the fact that the antecedents o f the sign they share —probably
unknown in the case o f those given the syndrome name as a diagnostic label -
appear to be different. These people are also likely to share a number o f sym p­
tom s. B u t even those who do not share sign s, and who receive different
diagnostic labels, may share sym ptom s because different sign s can be associ­
ated with the same sym ptom s (for exam ple, nausea, abdom inal pain, fever, and
so on).
In spite o f the heterogeneity am ongst those given the same diagnostic label
and the overlap am ongst those given different labels, it seems useful to m ain­
tain the separation between some gro u p s who share sign s and sym ptom s
because they do not seem to share antecedents. Only some o f those who show
glucose in the urine, for exam ple, will show hyperglycaemia, and those who do
will, in turn, show a variety of pancreatic abnormalities. It is therefore assumed
that different processes led to the appearance o f the signs in the various groups.
This assum ption may be supported by the observations that between-group
variability in progress over tim e is considerably greater than w ithin-group
variance; that the groups, without intervention, reach obviously different end­
points; for exam ple, early death versus average lifespan, or that response to the
same intervention is quite different between groups. Taken together, these dif­
ferences provide good grounds for separating the groups for research purposes.
Even if such differences are not observed, the separation o f the groups for the
present can be justified if the proxim al antecedents o f the sign they share
seem to be different.

E v a l u a t i n g t h e v a l i d i t y o f ‘s c h i z o p h r e n i a ’

If we return to the claim s made about ‘schizophrenia’ with which the chapter
began — that it is a scientific concept and a syndrome - then the claim s can
now be restated in this way: ‘schizophrenia’ is a hypothetical construct. It is
inferred from a pattern, a set o f regularities which conform to a syndrome (i.e.
a m eaningful cluster o f sign s and sym ptom s). These claim s can be specified
 Evaluating the validity o f ‘sc h iz o p h re n ia ’ 15

further in terms of the necessary and sufficient conditions for inferring hypo­
thetical constructs.

The necessary conditio ns

I noted earlier that the necessary condition for inferring a hypothetical con­
struct is the observation o f a pattern o f regularities. K raepelin’s original
inference of dementia praecox and then Bleuler’s of schizophrenia therefore
imply that such a pattern was observed in the behaviour of asylum inmates.
And, because ‘schizophrenia’ is said to be a syndrome, the claim is implicitly
made that this original cluster or, at least, the cluster which is the source of the
modern concept of schizophrenia, conforms to a meaningful pattern o f events
which can reasonably be called signs and events which can reasonably be called
symptoms.
One of the problems of discussing the status of the necessary conditions for
inferring schizophrenia is that various sets of regularities have been put for­
ward as the source of the construct. But because it is generally accepted that
K raepelin’s construct o f dem entia praecox marked the beginning o f the
modern construct of schizophrenia (for example, Gottesman and Shields, 1982;
M. Bleuler, 1991; Kendell, 1991; APA, 1994) —indeed the terms ‘dementia prae­
cox’ and ‘schizophrenia’ were often used interchangeably at least until the
1930s —the regularities in the behaviour of asylum inmates allegedly observed
by Kraepelin will be taken here as the first set of necessary conditions for infer­
ring schizophrenia. These, together with the clusters suggested by Bleuler and
Schneider, will be examined in Chapter 3 to assess how far they justify the
claim that ‘schizophrenia’ came into existence through the observation of a
meaningful pattern which could reasonably be described as a syndrome.

The sufficient conditio ns

The sufficient condition for inferring a hypothetical construct is that it can be

used to predict new observations. But it can hardly be overemphasised that if
the necessary conditions for inferring a hypothetical construct have not been
fulfilled, then the sufficient conditions cannot be either; you cannot, after all,
expect to make new observations about the antecedents of a non-existent pat­
tern. It must be said, however, that if the cluster from which schizophrenia was
and is inferred did conform to the pattern denoted by ‘syndrome’, it would
amount to the fulfilment of the necessary and just sufficient conditions for
inferring it, because an originally postulated cluster o f sym ptom s would have
been shown to be reliably associated with an independently and reliably mea­
surable event (a sign) and thus to have some predictive power.
But considerably more than this is claimed for ‘schizophrenia’. A number of
predictions have been derived from the construct but it is adm itted that data
relating to many of them, for example, that particular patterns of cognitive or
16 Evaluating the validity o f ‘sc h iz o p h re n ia ’

psychophysiological or biochemical functioning will be observed, are in some

disarray (for example, W ing, 1988; Chua and McKenna, 1995; McGrath and
Emmerson, 1999)- What users o f ‘schizophrenia’ claim not to be in disarray are
data relating to predictions about genetic inheritance. It has been claimed, for
example, that this hypothesis has been ‘proven’ (APA, 1980), and that the evi­
dence is ‘incontrovertible’ and ‘beyond challenge’ (Kendell, 1991). These
claims will be examined in Chapter 6. They will be examined in some detail:
first, because of their strength and near unanimity; second, because ‘genetic’
data may be used to justify both the continuing search for supporting bio­
chemical data and the retention of a particular theoretical model, and third,
because the presentation o f this literature could offer examples of the ways in
which the presentation of data in secondary sources m ight function to main­
tain the concept o f schizophrenia.
Another way o f implying that the sufficient conditions for inferring a con­
struct have been met is by changing the correspondence rules. I noted earlier
that the regularities from which a hypothetical construct is originally inferred
form the first set of correspondence rules which tie the construct to observable
events. I also noted that these rules may change as the initial set of events is
shown to be reliably associated with other, reliably and independently mea­
surable phenomena (i.e. the construct is shown to predict new observations).
The correspondence rules for inferring schizophrenia have been changed a
number of times since the concept was introduced (these are usually called
diagnostic criteria, but the term correspondence rules’ will be used here to
help keep in mind the fact that the validity o f ‘schizophrenia’ is to be evaluated
by applying criteria used by the scientific com m unity in general). These
changes in correspondence rules will be examined in Chapters 4 and 5 in
order to assess how far they reflect the processes described earlier. It is of
course the case that the use of any set of correspondence rules implies that these
refer to a pattern of phenomena. In other words, the publication of diagnostic
criteria amounts to a public claim that these represent a pattern of regularities
which justify inferring ‘schizophrenia’. The correspondence rules/diagnostic
criteria set out in DSM -III, D SM -IIIR and DSM -IV will therefore be exam­
ined, in Chapter 5, to assess the extent to which this is the case. This chapter
will therefore be examining both the necessary and sufficient conditions for
inferring ‘schizophrenia’.
But before these criteria for assessing validity are applied to ‘schizophrenia’,
it is important to set the introduction and use of the concept in their histori­
cal and social context. It is particularly important that this should have been
done if the concept is found wanting, so as better to understand not only the
problems it presents but why these should have happened in the first place.
The next chapter will therefore consider the historical background to the
introduction of the concepts of dementia praecox and schizophrenia.