Documente Academic
Documente Profesional
Documente Cultură
Abstract
The paper discusses the concept of measurement. Measurement, in the wide sense, is defined as a process of empirical,
objective assignment of symbols to attributes of objects and events of the real world, in such a way as to describe them.
Strongly defined measurement is measurement that conforms to the paradigm of the physical sciences. Weakly defined
measurement is measurement in the wide sense, but which is not strongly defined. Strongly and weakly defined
measurements are analysed and compared. Other forms of symbolic representation are distinguished from measurement.
2003 Elsevier Science Ltd. All rights reserved.
Keywords: Measurement, fundamental concepts; Measurement, widely defined; Measurement, strongly defined; Measurement, weakly
defined
0263-2241 / 03 / $ – see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016 / S0263-2241(03)00018-6
40 L. Finkelstein / Measurement 34 (2003) 39–48
It is not proposed to discuss in any detail the measurement is closely connected. An important
historical development of the theory of measurement development of the work of Helmholtz was the
proper. Guidance to more general topics may be axiomatization of measurement of additive quantities
sought in relevant reference works [1–5]. The works by Hoelder in 1901 [7].
concerned give guidance to more detailed infor- The British physicist, N.R. Campbell provided in
mation. 1920 a lucid and thorough analysis of the fundamen-
The practical pursuit of measurement in com- tal basis of the measurement of physical quantities in
merce, crafts, and surveying arose in the earliest his Physics: the Elements [8]. The book can still be
stages of the material culture of man. It grew in the read with pleasure and profit. Campbell’s theory was
course of the Urban Revolution in Mesopotamia and based on the measurement of physical quantities for
Egypt and in parallel developments in China. It was which an empirical operation of addition could be
based on an intuitive understanding of the process. constructed, what is now known as extensive mea-
The ancient Greeks were the first to investigate the surement. The theory became generally accepted
philosophical foundations of measurement. The under the influence of logical positivism and
school of Pythagoras was concerned with the operationalism [3].
philosophy of the relation between numbers and the The theory of measurement from Helmholtz to
real world and hoped to make arithmetic the fun- Campbell, and those who developed their work, was
damental study in physics. Plato’s Academy de- concerned with physical measurements, which are
veloped an extensive theory of magnitudes, though it based on additive quantities and quantities derived
was concerned more with the nature of numbers than from them. In the social and behavioural sciences,
with an analysis of the nature of measurement. however, there are many properties that cannot be
Aristotle studied the concepts of measurement in his empirically added nor derived from additive quan-
Metaphysics. tities. This created great philosophical difficulty. A
The Middle Ages saw much scholarly study of the report of a committee of the British Association for
theory of measurement, though scholars were not the Advancement of Science, published in 1940,
concerned with the application of measurement to which considered quantitative methods, rejected
observation of the physical world. measurements not based on additivity, and hence the
With the rise of modern science, Galileo put possibility of psychological measurements. It even
mathematics at the heart of physics and established a questioned the status of the thermodynamic scale of
programme of application of measurement to the temperature [9].
observation of the physical world. Newton in his These rigid positions of the classical theory of
development of mechanics provided the first com- measurement were broken down by work in the
prehensive mathematical theory of a domain of social and behavioural sciences. The concern of the
physics. Measurement was increasingly applied to social sciences with the concept of utility lead
scientific observation, culminating in full mathemati- through the work of the early writers, Bentham and
cal theories of domains of physical science such as Pareto, to the classical work of von Neumann and
the celestial mechanics of Lagrange and the electro- Morgenstern in 1944 [10]. Their axiomatic formula-
magnetic theory of Maxwell. These profound de- tion of the theory of utility has been the basis of
velopments of the application of measurement were much work in the theory and practice of measure-
not, however, accompanied by studies of the fun- ment in the social sciences. In psychology S.S.
damental concepts of measurement. Stevens carried out much fundamental work on
The true foundations of the modern theory of developing an appropriate analysis of the nature of
measurement were laid by Helmholtz [6] in a measurement [11–13].
thorough logical analysis of the epistemology of The proceedings of a conference in the United
counting and measuring published in 1887. The work States [14], published in 1959, presented a review of
was part of the beginning of studies in the logical the classical approaches to measurement as extended
foundations of mathematics with which the theory of to needs of the social and behavioural sciences.
L. Finkelstein / Measurement 34 (2003) 39–48 41
Ellis has provided, in 1966, a most useful and 3. Wide sense definition of measurement
interesting philosophical analysis of measurement. It
is not formal in approach and is principally It is proposed here to adopt a wide definition of
concerned with physical measurement, though it measurement, as a basis of discussion.
takes into account non-classical theory [15]. Measurement, in the wide sense, is thus defined as
Modern formal measurement theory may be said a process of empirical, objective assignment of
to be based the work Tarski on relational systems symbols to attributes of objects and events of the real
and model theory. This contribution is well described world, in such a way as to represent them, or to
in Ref. [16]. The theory, which may be termed describe them [27]. Measurement theory provides the
representational theory of measurement, considers logical basis for the analysis of the conditions of
measurement, loosely speaking as the establishment representation.
of a correspondence between a set of manifestations The essential elements of this wide-sense defini-
of a property and the relations between them, and a tion will be analysed informally.
set of numbers and the relations between them. Firstly, measurement is defined as the assignment
Suppes and Zinnes provided in 1963 an early clear of symbols to properties of objects and events, and is
exposition of the theory, in the development of thus the description of properties of objects or events
which Suppes has been one of the key workers [16]. and not of the objects or events. Measurement is
The representational theory of measurement was then based on a clear concept of a property, as an abstract
developed and well documented in a group of aspect of a whole class of objects of which in-
important books published between 1971 and 1990. dividual instances, or manifestations, are the subject
Pfanzagl published in 1968 the first monographic of measurement.
work devoted to the theory of measurement with a The definition states that the assignment of sym-
representational approach [17]. Krantz et al. pub- bols in measurement is such that the symbols
lished between 1971 and 1990 a very detailed and describe the property of the object or event. The
thorough account of the representational foundations meaning of this can be explained as follows. Consi-
of measurement, in three volumes [18]. Other der that a symbol, or measure, is assigned by
thorough accounts are those of Roberts [19] and of measurement to the property of an object, and other
Narens [20]. symbols are assigned by the same process to other
The representational theory has now been accepted manifestations of the property. Then the logical
widely in the social and behavioural sciences. How- relations between the symbols or measures, in the
ever, while it embraces physical measurement, the symbolisation system adopted, imply and are implied
theory has not found a recognised place in the theory by empirical relations between the property mani-
and practice physical sciences. festations.
Modern measurement theory has been introduced The next aspect of the definition of measurement,
to the measurement and instrumentation technology which requires discussion, is the fact that measure-
community by Finkelstein, who has regularly re- ment is an objective process. By this is meant that
viewed its development and progress [21–27]. The the numbers assigned to a property by measurement
topic has been taken up and debated in the communi- must, within the limits of error, be independent of
ty, particularly by some of the authors represented in the observer.
this issue. The informal definition of measurement presented
Representational measurement theory has not pro- above stresses the fact that measurement is an
gressed significantly in recent years, as it seems to empirical process. This means first that it must be the
have proved adequate for the purposes for which it result of observation and not, for example, of a
has been developed. However the rapid development thought experiment. Further, the concept of the
of symbolic computation and soft computing has property measured must be based on empirically
offered prospects for its further development and determinable relations and not, say, on convention.
extension [26,27]. It should be noted that all measurement involves a
42 L. Finkelstein / Measurement 34 (2003) 39–48
comparison of the measurand with a standard. either (ii) mapping on the real number line on which an
material, or conceptual. Measurement theory pro- operation of addition is defined,
vides the logical basis for that comparison. (iii) well-formed theories for broad domains of
knowledge.
Z is termed the symbol set. Let there be on Z a 7. Comparison of strongly and weakly defined
family P of relations measurement
P 5 hP1, . . . ,Pn j. (13)
Having defined strongly and weakly defined mea-
We may term Z 5 kZ, Pl the symbol relational surement it is now proposed to compare their
system. characteristics.
Let there be a mapping Strongly defined measurement is based on the
mapping of an empirical relational system into a well
M:Q → Z (14) defined theory in a mathematical language. The
so that z i 5M(q i ). Further let there be a mapping terms language and theory are used in the senses
explained above.
F:Q → Z (15) In strongly measurement:
so that Pn 5 F(R n ).
We may define (i) the definition of the quantities is based on the
theory,
C 5 kQ, Z, M, Fl (16)
(ii) the empirical relational system is also defined
where C is the representation code, and its inverse, by the theory,
the interpretation code. z n is termed symbol of or for (iii) the symbolic relational system is rich,
qn . (iv) the theory for the domain is complete.
6.4. Theory
8. Concept definition
A theory T in a language L is a set of sentences in
L. The establishment of a scale of measurement of a
A theory may be a representation, or model, of a quality must be based on a satisfactory concept of
domain in the extra-linguistic world under a set the measurand.
appropriate codes C1 , C2 , . . . ,Cn . The formation of a quality concept is normally a
L. Finkelstein / Measurement 34 (2003) 39–48 45
recursive process. The establishment of a basic course, but its definition is vague. Ref. [31] provides
concept leads to a scale of measurement defined on an excellent introduction to the problem of the
it. Observations using the scale of measurement lead formation of the concept of intelligence and its
to a refinement of the concept and hence to a assessment, and cites more detailed literature. Many
development of the scale. Measurement with a tests are available for measuring intelligence, but
refined scale may ultimately lead to a rich theory of there is no certainty as to what intelligence is, or
the domain, which establishes a definition of the what intelligence tests are measuring. It may be that
concept and strongly defined measurement scales. they are merely measuring the ability to score on the
An example of such recursive concept formation same form of intelligence test [32]. The measure-
in the physical sciences is the development of the ment of intelligence by intelligence tests, is an
concept of temperature. Starting with a vague con- objective and empirical process, but is, in the sense
cept of ‘the degree of hotness of a body’, scales of defined above, a weakly defined measurement.
temperature based on practical instruments were
developed. Such scales led to the refinement of the
concept of temperature and improved thermometers. 9. Uncertainty
Observations with improved thermometers led to
theories of heat and ultimately to an integrated All measurement is accompanied by uncertainty.
theory of thermodynamics. It is on such an integrated However, there are differences in kind between
theory of thermodynamics that the temperature scale strongly and weakly defined measurement with
is currently defined in physical science [28,29] respecy to the nature of the uncertainty involved.
Problems of concept formation arise commonly in Referring to the notation above we may distin-
social science and psychology. guish between uncertainty in the mapping M and
Consider for example the concept of poverty, uncertainty in the empirical relational system R.
which is much discussed in political discourse in In strongly defined measurement R is defined by a
natural language. It is extensively discussed in well-founded and well-formulated theory. M is again
sociological literature, and is the subject of much theoretically founded. Uncertainties are considered to
sociological measurement. The development of the be imperfections of the measurement process, and
concept is clearly considered in [3] and [4], which are estimated by corresponding analysis.
cite more detailed literature. Sociologists distinguish In weakly defined measurements, both R and M
between absolute and relative poverty. Absolute are based on empirical observations and empirical
poverty occurs when people fail to receive sufficient laws, in the determination of which there is uncer-
resources to maintain health and efficiency, often tainty,
expressed in terms of calories, or other nutritional Good examples of uncertainty in weakly defined
measures. Relative poverty is defined by the general measurement are provided by psychophysical mea-
standard of living in a society, and by what is surements of perceptions of sensory stimuli, such as
culturally deemed as poor. The relative poverty of a brightness and loudness. The problem is reviewed in
household may be described by reference to some Ref. [32], which gives references too more detailed
percentile of income distribution, which is deemed to literature. It makes clear that human judgements of
represent the poverty line. Relative poverty thus stimuli are poor, and that substantial uncertainty in
varies between societies and for the same society both the concept and scale formation arises.
over time. It is not possible here to consider the
measurement of poverty in any detail, but it can be
seen from the above that the definitions of poverty 10. Strength of measurement scale
are theory-laden and hence measurement of poverty
is weak in the sense defined above. 10.1. In strong measurement
Another example of the problem of concept
formation is the notion of intelligence [30]. The term In the classical approach to strictly, or strongly,
intelligence is much used in natural language dis- defined measurement such measurement is confined
46 L. Finkelstein / Measurement 34 (2003) 39–48
to fundamental, or extensive measurement [24]. In The scale is invariant under any one-to-one trans-
such measurement, the empirical relational system in formation. Statements implying order, or distance,
the quality Q, is are thus not meaningful in the scale.
An example of a nominal measurement is the
classification of members of a population according
R 5 ( ¯ , p, o),
to their first language, if it is objective and based on
empirical observation. Such nominal measures may
consisting of an empirical partial order, and an be used for empirical observations, relating first
empirical concatenation operation. The measurement language to, say, income, religion and educational
operation M maps this Q into the real number line level. The class of meaningful statements in terms of
Re, with 5, 1 defined on it. nominal measures is limited. In statistics it is mean-
For qualities for which this is not possible, we ingful to state a mode of a distribution, but not a
may define conjoint measurement structures, on the median, or a mean.
basis of empirical relational systems established for It should be noted that wide measurement admits a
qualities conjointly [18]. very wide class of empirical relations R, and corre-
Alternatively, it is possible to treat qualities that spondingly a variety of symbolic systems Z. There
are not fundamentally measurable as derived vari- are, as yet, no measures developed for the strength of
ables [24]. The scales of measurement are then diverse wide symbolisation systems.
derived from an empirically established relation Staying with linguistic examples, it is possible to
between two fundamentally measurable quantities, classify languages by empirical and objective criteria
which maps manifestations of the variable into [33]. in such a way that it is possible to draw a
numbers, which have the same order as the mani- graphical tree describing a genetic classification of
festations have in the concept of the quality. languages to determine say, a distance of one
The approach to measurement adopted in the language from another. For example Polish is close
paradigm of the physical sciences is to establish a to Slovak, more distant from Russian, yet more
theory for a domain. The theory is mathematical. The distant from German, and even more distant from
concept of a variable is then defined by the theory. Arabic.
Measurement is an objective empirical operation, In the case of biological classification it is possible
which maps manifestations of a variable into sym- by empirical observation a living organism as Homo
bols of the theory. sapiens, or Rana rana, and to determine the distance
between them, by graph theoretical measures [34].
In weakly defined measurement, scales of mea- 11. Theory for strongly and weakly defined
surement are classified in the literature according to measurement
the uniqueness conditions and their invariance under
a class of transformations [17–20,24]. Usually scales In strongly defined measurement the measure of a
of measurement are classified as nominal, ordinal, quantity is related to other measures of other quan-
interval and ratio. The wider the class of admissible tities on the same, or other, objects by a well
transformations of a scale, the narrower is the class formulated and extensive theory. For example mea-
of meaningful statements in terms of the symbols of sures of force on an object are related to measures of
the scale. mass and acceleration.
For example, in nominal measurement the quality In weakly defined measurement such a theory is
Q has only an empirical indifference relation ¯ on commonly lacking. For example, intelligence test
it. The symbol system Z is one in which the symbols scores may not be related with confidence with other
Z may be very general and the symbolic relational measurements. To give an example from physical
system P consists only of the identity relation 5. measurements, measures of hardness on the Mohs’
L. Finkelstein / Measurement 34 (2003) 39–48 47
scale [24] are not related by an adequate theory to 13. Measurement and natural language
other measures of properties of the object.
There is a strong relation between description by
measurement, in the weakly defined sense, and
description by natural language, which is in some of
12. Symbolisations which are not measurement its functions a general form of symbolic representa-
tion.
It is useful for the sake of completeness to It is not possible within the scope of this paper to
consider here assignments of numbers, or other consider natural language, even in outline. The barest
symbols, to properties in such a way as to describe skeleton of the comparison between measurement
them, but which are not measurements. and description by natural language must suffice.
One example is classification of industries by a In measurement the meaning of a symbol is its
standard industrial classification. Assignments, on reference. In natural language there are other views
such a classification, of numbers to enterprises, are of meaning. Meaning may be related to an idea in
objective and based on an empirical observation. the mind of the originator or receiver of an utterance,
They represent the relation of the enterprise to others or it may be considered as determined by conven-
symbolised using the same classification. However, tional use [37].
they are not measurement, to the extent that the The function of measurement is informational.
system of classification is not based on an objective Natural language has other functions such as aes-
empirical system of relations of industries [35]. thetic or phatic [37].
An important class of the descriptive assignment A linguistic symbol, even in its informational
of numbers, the measurement status of which is function may, in addition to its denotation, convey
problematic, arises in educational testing [36]. Marks other meanings, such as emotional colour.
in examinations may be objective, and are based on As discussed above the essence of measurement is
an empirical process, but it is problematic what they that it is an objective, empirical process. Description
measure, other than the performance in a particular by natural language may be derived from empirical
test. It is doubtful whether, when marks are treated observation, though it is not necessarily so. It may be
as measures on a ratio scale, they are not in fact subjective, though it may have a high level of
measures on an ordinal scale. This affects the objectivity.
meaningfulness of statistics on marks, such as t Description by natural language has often a degree
calculations of averages and the like. The conflation of ambiguity, and vagueness. Description by mea-
of marks, such as the calculation of weighted sums surement has generally a high degree of precision.
of marks, contains an element of subjectivity in the Finally description by measurement is generally
conflation scheme, which probably disqualifies such more concise than description by natural language.
conflated marks from being considered measure-
ments.
As a final example of descriptive numerical as- 14. Conclusions
signments that have doubtful measurement status, it
is important to consider utility evaluation in formal The paper has proposed the concepts of widely,
decision making. Problems of decision sciences were strongly and weakly defined measurement. It has
an important stimulus to the development of wider analysed their respective characteristics.
concepts of measurement [19]. Formal concepts of
utility, with considerable power, have been de-
veloped. However, it appears, that utility, however Acknowledgements
strongly defined, is only a description of the subjec-
tive judgement of the decision-maker. It is not a The author wishes to thank the Directors of the
measure of any objectively defined characteristic of Measurement and Instrumentation Centre of City
the object evaluated. University for their encouragement in his work.
48 L. Finkelstein / Measurement 34 (2003) 39–48