Equivalence relations may be attested through recombination of elements previously and arbitrarily related (stimuli and responses) via reinforcement contingencies. Thus, once relations have been established between specific sets of stimuli and responses, these relational will be considered equivalence relations if the recombination of these elements is attested without additional training. Previous research has confirmed that equivalence classes are difficult to obtain with non-human subjects. This difficulty is not necessarily due to subject’s deficiency linked to absence of language, but may be due to the incoherence between the control relations planed by the experimenter and the control relations effectively developed by the subjects. The development of non-programmed control by stimuli position, for example, is recurrently pointed as determining of stimulus control topography different from that one planed by the experimenter, and may generate negative results in equivalence tests. With the purpose to explore the control by position, previous research had found the workability of using positions has stimuli in conditional discriminations training with primates as subjects. The purpose of the present paper was continue this kind of research, by checking the possibility of obtaining emergent conditional relations, as symmetry, after a history of direct training of exemplars of this kind of relations. Two capuchin monkeys (Cebus apella) served as subjects, and an experimental chamber equipped with a computer and touch-screen was used. The stimuli were positions of sixteen white identical squares designed on the computer screen and presented as a 4x4 matrix. The procedure included the establishment of conditional relations with position-stimuli via an adapted matching –to-sample procedure. The relations AB and BC were trained and the properties of symmetry (BA and CB), transitivity (AC) and symmetry of transitivity (CA) were tested. The same procedure was used with more two set of stimuli (sets DEF and GHI). No evidences of equivalence classes formation were found. Positive results in potencially emergent relations were found only in some symmetry tests (when topographically similar- parallel-relations were trained before) and transitivity (conducted among mixed and reinforced baseline and with some evidences of mediators responses, that is, response to nodal stimuli). The recombination of the same positions-stimuli in a new set of conditional relations the training of parallel conditional relations was not adequate to testing the hypothesis of experience with exemplars of symmetric and transitive relations would benefit the emergence of the relations that document these properties, after the training of new conditional relations. The data obtained in the present work shows that conditional relations with position-stimuli, trained in a specific baseline, interfere the training of a posterior baseline. It is possible that the use of another kind of visual stimuli be more appropriated to this kind of research. The data presented here, nevertheless, identifics clearly the variables that control the subject’s performance in the tests, contributing to discuss the previous works on the possibility of equivalence classes formation with position-stimuli.Relações de equivalência podem ser atestadas através da recombinação de elementos (estímulos e respostas) previamente e arbitrariamente relacionados via contingências de reforçamento e raramente são obtidas com sujeitos não-humanos. O desenvolvimento de controle não-programado pela posição dos estímulos pode ser um determinante dessa dificuldade. O objetivo do presente trabalho foi verificar se se obtém recombinações de relações condicionais entre posições-estímulo, após uma história de exposição dos sujeitos a alguns exemplos de treino direto de recombinação. Dois macacos-prego (Cebus apel/a) e uma câmara de condicionamento equipada com computador e tela sensível ao toque foram utilizados. O procedimento consistiu no estabelecimento de relações condicionais entre posições-estímulo. através de uma adaptação do procedimento de "matching-to-sample". Foram treinadas relações AB e BC e testadas as propriedades de simetria (BA e CB), transitividade (AC) e simetria da transitividade (CA). O mesmo procedimento foi adotado com mais duas combinações de estímulos (conjuntes DEF e GHI). Não foram encontradas evidêocias de formação de classes de equivalência. A recombinação das mesmas posições de uma matriz em novas relações condicionais e o treino de conjuntos de relações condicionais entre posições paralelas não se mostraram adequados para testar a hipótese de que a experiência com exemplos de relações simétricas e transitivas pode favorecer a obtenção deste tipo de relações após tremo de novas relações condicionais.