Twenty preschool children were asked to describe pairs of pictures, the second of which contrasted with the first in terms of agent, action, or object. The children showed a clear tendency to stress the part of the description corresponding to the contrasting element. This demonstrates mastery of contrastive stress patterns (in absence of any formal teaching) by young children. The technique provides a controlled means of eliciting this aspect of speech in children.

Shimcr CollegeP SYCHOLOGISTS often make comparisons between naturally occurring groups, for example, ghetto versus middle-class children, monolingual? versus bilinguals, males versus females, 60 year olds versus 70 years olds, catatonics versus compulsives, and Norwegians versus Sudanese. Sometimes such comparisons are made between species, under the label of "comparative psychology"; but we will concentrate in the following pages on human populations. The questions to be addressed concern the nature of controls appropriate to these studies which are sometimes called "correlational" (Cronbach, 19S7) and, in general, the relation between correlational studies and another class of studies in which the experimenter has control over differences between groups he is comparing, which are sometimes called "experimental" studies. CONTROL PKOCEDUKKS IN Guoui'-Coivti'AiusoN STUDIESThe obvious research design problem which strikes one immediately in considering groupcomparison studies is the following: The groups to be compared are probably always unequal in a variety of respects which are not being studied, as well as in the differences signaled by their labels with which the student is concerned. For example, bilinguals are likely to differ from monolinguals in any community in terms of national origin, food preferences, height, attitudes, family size, etc., as well as in terms of "linguality." No sample, then, of the two populations, regardless of si/e, will ever be free of these "extraneous variables" nor of potential effects from them on obtained differences.

Uncertainty scores (H and relative H) were computed for the verbs, noun phrases, and sentences produced by 180 children between the ages of 5 and 13 years on a story-telling task. Linear age trends occurred for Sentence H and relative H and for Noun-phrase H (though not relative H), but not for verb H and relative H. There were no overall sex differences, but measure-specific age-by-sex interactions and higher-order age trends. Uncertainty scores of the type proposed, if suitably refined, may measure the developmental progression from stereotyped reliance on individual syntactic forms to flexible utilization of productive rules.

Language produced by 180 children (aged five through 13 years) on a story-telling task was analyzed in terms of 57 variables (part-of-speech frequencies and proportions, syntactic elaboration indexes, and constructional variety measures). A factor analysis revealed five dimensions of syntactic usage: general fluency, embeddedness, finite verb structure, noun phrase structure, and qualified speech. The embeddedness dimension was the only one with a sizable relation to age and is interpretable in terms of a developmental progression in the inclusion of transformationally processed content in the sentence. Among the other stylistic dimensions, fluency, and verbal vs nominal emphases in language appear to be in evidence at various age levels and on different types of variables.