07.10.2013 Views

Main trends of research in the social and human ... - unesdoc - Unesco

Main trends of research in the social and human ... - unesdoc - Unesco

Main trends of research in the social and human ... - unesdoc - Unesco

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

Ma<strong>the</strong>matical models <strong>and</strong> methods 565<br />

<strong>in</strong>creased markedly <strong>in</strong> <strong>the</strong> last few years. This <strong>in</strong>crease is on occasion associated<br />

with specific problems. Thus <strong>the</strong> utility measurement <strong>the</strong>ory was born <strong>of</strong> needs<br />

engendered by economic <strong>the</strong>ory. But above all <strong>the</strong> development <strong>of</strong> measurement<br />

<strong>and</strong> classification models got its impulsion from psychology <strong>and</strong> pr<strong>in</strong>cipally<br />

psychometry. Only later did <strong>the</strong> <strong>social</strong> psychologists <strong>and</strong> sociologists<br />

realize <strong>the</strong> value <strong>of</strong> measurement <strong>and</strong> classification models, notably when <strong>the</strong>y<br />

came up aga<strong>in</strong>st <strong>the</strong> problem <strong>of</strong> ‘attitude measurement’.<br />

Spearman, Thurstone, Guttman, Lazarsfeld <strong>and</strong> many o<strong>the</strong>rs have lent <strong>the</strong>ir<br />

names to classification <strong>and</strong> measurement models. In most cases <strong>the</strong> models<br />

consist <strong>in</strong> <strong>the</strong> expression <strong>of</strong> a collection <strong>of</strong> observable variables (success <strong>in</strong> tests,<br />

answers to questions about personal attitudes, etc.) as functions <strong>of</strong> classify<strong>in</strong>g<br />

variables, variously labelled non-observable, latent or genotypic. Thus, <strong>in</strong><br />

Spearman’s classical factorial analysis we write that <strong>the</strong> success zij <strong>of</strong> a subject<br />

i <strong>in</strong> a test j is a l<strong>in</strong>ear function <strong>of</strong> an <strong>in</strong>telligence measure Fi <strong>and</strong> <strong>of</strong> <strong>the</strong> score<br />

eiJ <strong>of</strong> subject i on a factor peculiar to test j. Accept<strong>in</strong>g certa<strong>in</strong> conventions on<br />

how <strong>the</strong> variables t?J <strong>and</strong> Fi are measured, <strong>and</strong> assum<strong>in</strong>g all latent variables to<br />

be <strong>in</strong>dependent, it is possible to test <strong>the</strong> model <strong>and</strong> estimate <strong>the</strong> Fi quantities.<br />

Lazarsfeld’s latent structure analysis can be regarded as an adaptation <strong>of</strong><br />

Spearman’s factorial analysis to <strong>the</strong> cases - frequent <strong>in</strong> sociology <strong>and</strong> <strong>social</strong><br />

psychology - when <strong>the</strong> observed variables are qualitative.<br />

Over <strong>and</strong> above <strong>the</strong> <strong>research</strong>es <strong>in</strong> <strong>the</strong> direct Spearman l<strong>in</strong>eage, <strong>the</strong>re is ano<strong>the</strong>r<br />

<strong>research</strong> tradition which is important <strong>in</strong> <strong>the</strong> field <strong>of</strong> measurement <strong>the</strong>ory,<br />

namely <strong>the</strong> one deriv<strong>in</strong>g from <strong>the</strong> work <strong>of</strong> <strong>the</strong> psychophysicists.<br />

This latter tradition was started by Thurstone <strong>in</strong> <strong>the</strong> 1920s, <strong>and</strong> developed <strong>in</strong><br />

greater ma<strong>the</strong>matical depth by writers such as Mosteller between 1950 <strong>and</strong><br />

1960; Luce’s recent work on <strong>the</strong> <strong>the</strong>ory <strong>of</strong> measurement can also be <strong>in</strong>cluded.<br />

All <strong>the</strong>se <strong>research</strong>es present <strong>the</strong> common characteristic <strong>of</strong> confront<strong>in</strong>g <strong>the</strong><br />

respondent not with isolated questions or stimuli, but with groups <strong>of</strong> stimuli<br />

whose terms he must compare. These models, which presuppose complicated<br />

observation procedures, are usable pr<strong>in</strong>cipally <strong>in</strong> experimental situations,<br />

whereas models <strong>in</strong> <strong>the</strong> Spearman tradition are better adapted to observation<br />

situations (surveys by questionnaire, psychometrical observations, etc.).<br />

One o<strong>the</strong>r tradition, which started from a critical review <strong>of</strong> <strong>the</strong> two just<br />

mentioned, is particularly well represented <strong>in</strong> <strong>the</strong> works <strong>of</strong> Guttman <strong>and</strong><br />

Coombs. Whereas <strong>the</strong> models aris<strong>in</strong>g from <strong>the</strong> <strong>research</strong>es <strong>of</strong> Spearman <strong>and</strong><br />

Thurstone are all statistical models, those <strong>of</strong> Guttman <strong>and</strong> Coombs are ei<strong>the</strong>r<br />

algebraic or comb<strong>in</strong>atorial. They avoid <strong>the</strong> disadvantages <strong>of</strong> <strong>the</strong> assumptions<br />

built <strong>in</strong>to previous models (l<strong>in</strong>ear character <strong>of</strong> <strong>the</strong> relationships between overt<br />

<strong>and</strong> latent variables <strong>in</strong> Spearman’s analysis ; normal distribution hypo<strong>the</strong>ses <strong>in</strong><br />

<strong>the</strong> Thurstone-Mosteller models), but <strong>in</strong>troduce o<strong>the</strong>rs. In effect, <strong>the</strong>y assume<br />

that <strong>the</strong> data conform to <strong>the</strong> model not stochastically but exactly. (To mark <strong>the</strong><br />

contrast between <strong>the</strong> two types <strong>of</strong> model, <strong>the</strong> psychosociological measurement<br />

<strong>the</strong>ory relies on a choice <strong>of</strong> terms which, <strong>in</strong>asmuch as it may be <strong>the</strong> source <strong>of</strong><br />

all sorts <strong>of</strong> confusion, is highly debatable: it calls models <strong>of</strong> <strong>the</strong> latter type<br />

determ<strong>in</strong>istic, contrast<strong>in</strong>g <strong>the</strong>m with probabilistic models <strong>of</strong> <strong>the</strong> Spearman-<br />

Thurstone type.)

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

Saved successfully!

Ooh no, something went wrong!