20.08.2013 Views

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

Suppes and Zinnes - basic measurement theory.pdf - Ted Sider

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

BASIC MEASUREMENT THEORY<br />

One might be tempted to conjecture that the first four axioms of Def. I 6<br />

would characterize all finite difference systems for which a numerical<br />

representation could be found (the representations of a given system<br />

would not necessarily be related by a linear transformation). The resulting<br />

<strong>theory</strong> would then represent one forma1ization of Coombs' ordered metric<br />

scale. However, Scott <strong>and</strong> <strong>Suppes</strong> (1958) have proved that the <strong>theory</strong> of<br />

all representable finite difference systems is not characterized by these<br />

four axioms <strong>and</strong>, worse still, cannot be characterized by any simple finite<br />

list of axioms.<br />

The f. d. systems are not as artificial or as impractical as they may seem.<br />

One <strong>theory</strong> for approximating these systems is to be found in Davidson,<br />

<strong>Suppes</strong>, & Siegel (1957, Chapter 2). However, these systems can have<br />

more general usefulness if they are used to establish a "st<strong>and</strong>ard set" of<br />

stimuli. In the case of tones, for example, a set of tones may be selected in a<br />

successive manner so that the set satisfies Axiom 5. If this st<strong>and</strong>ard set of<br />

tones also satisfies the remaining four axioms, then we know from Theorem<br />

IS that the tones may be assigned numbers that are on an interval scale.<br />

Arbitrary tones that are not in the st<strong>and</strong>ard set but that satisfy the first<br />

four axioms may then be located within intervals bounded by adjacent<br />

tones in the st<strong>and</strong>ard set. This means that by decreasing the spacing<br />

between the st<strong>and</strong>ard tones any arbitrary tone may be measured within<br />

any desired degree of accuracy. This is in fact what a chemist does in<br />

using a st<strong>and</strong>ard set of weights <strong>and</strong> an equal arm balance to determine<br />

the weight of an unknown object. His accuracy of <strong>measurement</strong> is limited<br />

by the size of the smallest interval between the st<strong>and</strong>ard weights or, if he<br />

also uses a rider, by the gradations on the rider.<br />

Other relational systems closely related to f.d. systems may appropriately<br />

be mentioned at this point. Among the simplest <strong>and</strong> most appealing are<br />

the bisection systems "21 = (A, B), where B is a ternary relation on the set<br />

A with the interpretation that B(a, b, c) if <strong>and</strong> only if b is the midpoint of<br />

the interval between a <strong>and</strong> c. The method of bisection, which consists in<br />

finding the midpoint b, has a long history in psychophysics. The formal<br />

criticism of many experiments in which it has been used is that the variety<br />

of checks necessary to guarantee isomorphism with an appropriate<br />

numerical system is not usually performed. For example, if B(a, b, c)<br />

implies that aPb <strong>and</strong> bPc, where P is the usual ordering relation, then<br />

from the fact that B(a, b, c), B(b, c, d), <strong>and</strong> B(c, d, e) we should be able to<br />

infer B(a, c, e). But the explicit test of this inference is too seldom made.<br />

Without it there is no real guarantee that a subjective scale for a stimulus<br />

dimension has been constructed by the method of bisection.<br />

Because of the large number of axiomatic analyses already given in this<br />

section, we shall not give axioms for bisection systems. The axioms in<br />

1.<br />

.(.<br />

.•<br />

'EXAMPLES OF FUNDAMENTAL MEASUREMENT<br />

:any case are rather similar to those of Def. 16, <strong>and</strong> the formal connection<br />

between the difference relation D <strong>and</strong> the ternary bisection relation B<br />

should be obvious:<br />

B(a, b, c) if <strong>and</strong> only if abDbc <strong>and</strong> be Dab.<br />

As an alternative to giving general axioms for bisection systems, it may<br />

be of some interest to lqok at the problem of characterizing these systems<br />

in a somewhat different manner, namely, by simply listing for a given<br />

number n of stimuli the relations that must hold. In perusing this list it<br />

should be kept in mind that we assume that bisection systems have the<br />

same property of equal spacing possessed by f. d. systems. As examples,<br />

let us consider the cases of n = 5 <strong>and</strong> n = 6.<br />

For n = 5; let A= {a, b, c, d, e) with the ordering aPbPcPdPe. We<br />

then have exactly four instances of the bisection relation, namely, B(a, b, c),<br />

B(b, c, d), B(c, d, e), <strong>and</strong> B(a, c, e).<br />

For n = 6, we may add the element f to A with the ordering<br />

aPbPcPdPePf To the four instances of the bisection relation for n = 5,<br />

we now add two more, namely, B(d, e,f) <strong>and</strong> B(b, d,f). We may proceed<br />

in this manner for any n to characterize completely the bisection system<br />

with n stimuli, none of which is equivalent with respect to the property<br />

being studied. Establishing the representation <strong>and</strong> uniqueness theorems<br />

is then a trivial task. The disadvantages of this approach to characterizing<br />

those relational systems for which numerical representation theorems<br />

exist are twofold. In the first place, in contrast to the statement of general<br />

:axioms, the listing of instances does not give us general insight into the<br />

structure of the systems. Second, for systems of <strong>measurement</strong> that have a<br />

more complicated or less sharply defined structure than bisection systems,<br />

the listing of instances can become tedious <strong>and</strong> awkward-semiorders<br />

:provide a good example.<br />

3.5 Extensive Systems<br />

We consider next a relational system leading to a ratio scale. Since this<br />

relational system contains an operation a that corresponds to an addition<br />

operation, we may justifiably call this system an extensh'e system (see<br />

Sec. 1.4). The axioms that we shall use to define an extensive system<br />

(<strong>Suppes</strong>, 1951) are similar to those first developed by Holder (1901).<br />

HOlder's axioms, however,. are more restrictive than necessary in that they<br />

require the homomorphic numerical relational systems to be nondenumerable<br />

(<strong>and</strong> nonfinite). The present set of axioms applies both to<br />

denumerable <strong>and</strong> nondenumerable but infinite relational systems.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!