The purpose of the present study was to assess the farmers’ behavioral domains regarding optimal farm water management practices and to determine the most important discriminating factors on their behavior with an approach to compare two groups of farmers in irrigation networks with water users’ cooperative (WUC) and without this cooperative. The methodological approach was a descriptive-correlational and causal-comparative study of the survey type. The target population in the study was the entire farmers of irrigation and drainage networks in Khorasan-e-Razavi Province, Northeast Iran (N=2551). A stratified random sampling technique was used to select the sample population. Finally, 330 questionnaires were collected and analyzed. The content and face validity of the instrument was specified after several times of review and correction by the Faculty members at University and several experts of administrative offices. The reliability analysis was conducted through 30 questionnaires and Cronbach’s Alpha values for the different sections of questionnaire were estimated to be between 0.73 and 0.91 using SPSS software. Considering Venn diagram in relation to the triple domains of studied farmers’ behavior showed that only 39.2% (62 farmers) of all farmers in irrigation network with WUC and 19.2% (33 farmers) of all farmers in irrigation networks without WUC were identified as “Qualified Farmers”. The results of t-test indicated that there were significant differences between the means of farmers’ knowledge, attitudes and skills regarding sustainable practices of agricultural water management when two groups of respondents in irrigation networks with WUC and without it are compared together. The results of the stepwise discriminant analysis by using of Wilks’ Lambda procedure revealed that the most important discriminating variables of farmers’ behavioral categories included farmers’ attitude toward the WUC, status of regional irrigation system, social confidence, education level and annual farm income. Generally, these variables could correctly classify some 66.9% of all subjects based on the discriminant functions. Furthermore, by means of the obtained equations of this analysis can predict that the intended farmer to which the category of farmers belongs.