Effects of negative instances in concept acquisition using a verbal learning task.

Abstract

Concept acquisition was promoted by manipulating positive and negative instances. Also, stimulus simdlarity variables produced the concept classification errors -of overgeneralization, undergeneralization, and misconception. The value of negative instances in concept instruction-was, investigated in a second adVerb learning task 5y removing the negative instances from the six treatment conditions. Subjects (260 `total) were seventh grade students from three school -districts. Subjectt 'in experiment one responded according tothe hypothesized outcomes (p < .01). In experiment two, the subjects responded -randomly on the posttest. The results indicate that negative instances are an integral .part of concept acquisition. The relationship between the, positive and negative instances was based upon similarity of irrelevant attributes and sentence di fftcuf ty. Effect of of Instances in Concept Acquisition Using a Verbal Learning Task Robert D. Tennyson Florida State University The earliest research dealing with the relationship of negative instances (nonexeirplars) to positive instances (exemplars) was Smoke's (1933),. Smoke used an artificial task in which the exemplars and nonexemplars were randomly orderedwith the order changing after each succession through the list: No logical relationship was est4biishE,:: between exemplars andnonexemplars, resulting in his conclusion that negative instances were of no value in concept learning. A-study which looked at the relationship of exemplars based on critical attributeswas Marri sett and Hovland's (1959) replication of Adams' (1954) study of single task vs. multiple task. They found that a variety of positive instances was necessary to effect a transfer of concept attainment, However,. there waS..no operational definition of the relationship between exemplars according to their irrelevant attributes or any criteria. In investigations of combined instances, the equivalent attributes of positive and negative instances, were found to be poorly utilized by huMan subjects (Bruner, Goodnow, & Austin-, 1956; Donaldson, 1959; Hovland & We4S, 1953), These studies show the lack of an operational control'betWeen exemplars and nonexemplars, as does the Smoke study. The oOnOoPtSP-used-in, these.prior studies were finite withthe subject attempting to gyeSs--thivrule (critical attribute) fn3m a-ries of instances. 2 Callentine and Warren (1955) studied positive instances and concluded that repetition of one or two instances increased attainment.' Luborsky (1945) indicated that eight exposures was more effective than three. These last two studies show that a series of instances is needed, but no mention is made of the difficulty of the instances or that discrimination of negative instances could be affected by a series which includes a combination of exemplars and nonekeMplars. Irrelevant attributes as measures of difficulty has been shown in the studiesdealing with ease of attainment of concept classes(Archer, Bourne, & Brown; 1956; Brown & Archer, 1966; and BoUrne, 1967). 'Each: of these :Studies' found that as the number of irrelevant attributes indreatedthelearriing: latency and number of errors also increased. They concluded that the: number of irrelevanfattribUtes *has a linear relatiorishi0 with 'difficulty of instances. Negative instances facilitate learning 'of :conceptt by requiring the subject to concentrate on the critical attribute when Presented' a,matched relationship of exemplars -and ncinexeinplars. When the eXenpThr and nonexenplar are as i.imtrar,;;at possible in' their irrelevant attribUtesi the noticeable differences are among the critical Concept: -.._ attainment research' does not provide, 'the subject with the op-Ortunity to focus on the critical attributes by using negative' instances :. ' ...negative instances in concept adqUiSitiOn the subject thight:conteive as a critical attribute -an irreTeyaki'atirlbute.` In a study by-Tennyson; Woolley -, and 1 1 11,974;,-4ndePendent variables were investigated Oa t:cjire0i4ed* tb:riCelit54-64140:iaciki afidltpeci fied classification errors. The resUltSOf' their study se iz,n% reek