Представлены результаты исследования степени устойчивости одного из видов коллокаций русского языка - биномиалов, определено соотношение степени устойчивости, частотности и предсказуемости элементов. По каждому из параметров выделены ядерная, околоядерная и периферийные группы. Выявлены пересечение ядерных групп по составу биномиалов и характер корреляционной связи групп по трем параметрам. Исследование проведено с применением корпусных, психолингвистических, статистических методов.

The article presents the results of a research of the degree of fixedness of one of the collocation types of the Russian language, binomials, which are currently poorly researched by Russian linguists. At the previous stages of the study, a set of 128 Russian binomials was composed by the authors of the publication according to the Russian National Corpus. All units were ranked according to their objective and subjective frequency. The research tasks described in the article are the following: to rank binomials according to the degree of fixedness and to find out the parameters correlated with this important property of the class of collocations of the Russian language under study: the frequency of the whole unit and the predictability of the second element. The degree of fixedness is determined on the basis of the (ir)reversibility score, which is calculated by the formula: the number of occurrences in the corpus (data from the Russian National Corpus) of a binomial with the direct word order divided by the sum of occurrences of direct word order binomials and reverse word order units, and multiplied by 100. Statistical objective data analysis was accompanied by the psycholinguistic method of obtaining subjective estimates using the scaling method. A comparison of the results of the objective and subjective estimations of the degree of fixedness of binomials allowed the authors to distinguish the nuclear part (38 units), the near-nuclear part (52), and the peripheral part (38) of the sampling. The frequency of a binomial was determined using the data of the Russian National Corpus in accordance with the formula adopted in this paradigm: the number of uses of the binomial is divided by the total number of words, and multiplied by a million. According to this criterion, 16 units (frequency from 3.447 ipm to 1.234 ipm) were allocated into the nuclear part of the sampling, 47 into the near-nuclear part (from 0.991 ipm to 0.102 ipm), and the periphery consisted of 65 (0.095 ipm to 0.003 ipm). During the comparison of the units of the nucleus of frequency and the nucleus of fixedness, only three coincidences were found: muzh [husband] and zhena [wife], dobro [good] and zlo [evil], zhizn' [life] and smert' [death]. However, the analysis using Pearson's chi-square test revealed the statistical significance of the correlation for the objective frequency of a binomial and the (ir)reversibility score of the nucleus of this sample. At p-value = 9.861e-05, it was 0.5894538. To determine the degree of predictability of binomials, eight questionnaires were developed, each containing 32 words. Each word was followed by a conjunction i [and], revealing the link with the potential next component, for example, sudit' i . . . [judge and . . .]. The analysis of the coincident units of the nuclear parts of the two samples based on the fixedness and predictability coefficients revealed a coincidence of 19 units. However, statistical analysis revealed the absence of a correlation dependence of the stability factor and the predictability coefficient of the nuclear part of the sample with respect to the stability parameter, p-value < 0.05. Thus, the analysis revealed that the features examined are relatively independent of each other, the perception of the fixedness of the binomial complex can be facilitated by highlighting one of the features in speech functioning. However, in a binomial as an invariant, an extra-textual unit, none of the characteristics can be recognised as dominant.