THE RESEARCH REPORTED HERE CONCERNED THE RELATIONSHIP BETWEEN LANGUAGE AND COGNITION IN THE CONTEXT OF THE WHORF HYPOTHESIS ABOUT LANGUAGE AND CULTURE. EXPERIMENTS WERE CARRIED OUT INVESTIGATING THE EFFECT OF SOME LINGUISTIC AND REFERENT DIMENSION VARIABLES ON THE LEARNING OF MINIATURE LINGUISTIC SYSTEMS. A MINIATURE LINGUISTIC SYSTEM (MLS) IS A LIMITED, ARTIFICIAL LANGUAGE MADE UP OF NONSENSE SYLLABLES AND NONSENSE FIGURE REFERENTS ARRANGED IN RELATIONS ANALOGOUS TO THOSE IN NATURAL LANGUAGES. A TOTAL OF 210 ADULTS WERE ASKED TO LEARN THE NAMES OF THE REFERENTS WITHOUT BEING TOLD THERE WAS A LINGUISTIC SYSTEM UNDERLYING THE NAMES. THE CORE EXPERIMENTAL GROUPS EXPLORED THE LEARNING OF MLSS WHERE GRAMMATICAL STRUCTURE WAS HELD CONSTANT, AND THE REFERENT DIMENSIONS WERE PAIRED WITH STRUCTURAL UNITS IN ALL OF THE POSSIBLE COMBINATIONS ALLOWED BY THE SET OF DIMENSIONS AND UNITS USED. THE LEARNING OF THESE SYSTEMS WAS SIGNIFICANTLY AFFECTED BY BOTH GRAMMARS "AS A WHOLE" AND MORPHEME TYPES, AND THESE TWO VARIABLES INTERACTED SIGNIFICANTLY. THE RESULTS WERE DISCUSSED IN TERMS OF A MILLER AND CHOMSKY PERFORMANCE MODEL. THE MAJOR DETERMINANT APPEARED TO BE THE NATURE OF STRUCTURAL UNITS, WITH LESSER EFFECTS DUE TO REFERENT DIMENSIONS. SEVERAL ALTERNATIVE INTERPRETATIONS OF THE LEARNING PROCESS WERE GIVEN. THE AUTHOR CONCLUDES THAT ANY STRONG FORM OF THE WHORF HYPOTHESIS OVERSIMPLIFIES THE LANGUAGE-COGNITION RELATIONSHIP. (AUTHOR/JD)