The research area was Gazipur district which is an average agricultural productivity area. The selection of the study sites and sample respondents were done purposively. There were some salient features in the selection procedure. First one, the selected district includes some important infrastructures, such as BARI, BIM, and BSMRAU, etc. Second one, total numbers of selected villages were ten by taking two villages from five upazilas. Of the two villages in each upazila, one village is selected comparatively near to the upazila headquarters and the other one is selected comparatively away from the upazila headquarters. The selected comparatively nearer villages were namely, Samantapur (Sadar), Bagnahati (Sreepur), Dushya Narayanpur (Kapasia), Katalia (Kaliakoir) and Poinlanpur (Kaliganj). The selected villages which were comparatively away from the upazila headquarters, namely, Bara Bhabanipur (Sadar), Saitalia (Sreepur), Noyanagar (Kapasia), Poshim Chandpur (Kaliakoir), and Bhatgati (Kaliganj). Third one, the total households were more than one hundred in the selected villages (BBS, 1993). It was then decided to collect one hundred samples from each village. The total numbers of investigated farmers were one thousand (2 villages x 5 upazilas x 100 farmers) and multistage random sampling technique was followed. Primary data was collected using survey method and personal interviews were conducted through pre-tested questionnaires with a view to collecting data. The survey was administered with the help of staffs of BARI in 2002. Fourth one, each upazila has some characteristics: Sadar upazila is completely urban type; Sreepur, Kapasia, and Kaliganj upazilas are rural type and headquarters of these upazilas are the only urban areas while Kaliakoir upazila headquarter is the only urban area and Safipur is the other urban area of this upazila (BBS, 1993). This study can be comprehensive compared to many research works due to the above salient features which will explore the actual situation of agriculture extension either in Bangladesh or elsewhere from the grassroots levels. Many of the previous researches used the productivity index representing the amount of production per unit of farm land, that is, the value added of production, which is found by deducting production costs from gross income. By using that index, it is possible to convert the specific quantities of products into given amounts of money to be added up; therefore, it represents a considerable analytical benefit. The method of settling the type of variables from which the index is determined, expected to be discussed. As is commonly used in analyzing production function, chemical fertilizer, farm buildings, irrigation facilities, family and hired labours should be considered as important investment functions (Everson et a1., 2001; Owens et a1. 2003). Haq et al.(2003) considered crop income per unit of land as dependent variable and chemical fertilizer cost per unit of land, irrigation cost per unit of land, experience of farmers, farm area, number of times extension contact as independent variables. Therefore, it summarized the model, Ln crop income= f (In chemical fertilizer, In irrigation, In experience, In farm area, In labour, extension contact dummy, extension contact dummy 2). Haq et al. (2004) interpreted total income as dependent variable, while age of farmers, years of schooling of farmers, family size, number of educated family members, number of earners of a farm family, rural institutions dummy, number of times extension contact, proportionate effect of flood to crop land, distance between crop land to market, homestead area, size of farm, irrigation cost, village dummy were taken as independent variables. The income function was solved by applying ordinary least squares. The above concepts provide to run an empirical model which is found in the ensuing section. The model applied here is the input-output model. The heart of the input-output model is the concept of the production function [Y=f (Capital, Labour)] which helps us in understanding the role of important variables like capital and labour in determining the crop productivity. But only two factors have no reflection on the productivity of a major crop like rice. Therefore, based on related past studies and logical analysis, some important explanatory variables which are considered in this study namely age of the farm household head (Ag), number of family earners in the household (Fea), number of times extension contact received by the farmer for the sample crop season (Et), proportionate effect (%) of flood to crop land (Fec), distance from farm land to market in miles (Mr), actual size of cultivated land in acre (Fs) , per acre total cost of chemical fertilizer (Chem), per acre total labour cost (Lab), per acre total money spent for irrigation (Irr), village dummy (dummy) = 1 if near village; otherwise = 0 and upazila dummy (Udummy) i.e. Sadar upazila = 1, otherwise = 0. The yield of rice (maund/acre) is the dependent variable in the present paper as it is the major food crop in the country. It includes boro rice because it is hardly affected by the natural disaster compared to other rice crops and it was cultivated by all sample farms. In the objective of this research, the most important variable is that of the activities of the agricultural extension services. In Bangladesh T&V system, farmlands are divided into blocks and the T&V workers target the representative farmers of the different blocks, who are referred to as "contact farmers" (Haq et a1, 2003). Although the T&V workers can directly get in touch with ordinary farmers, they mainly train the contact farmers, who afterwards transmit the training results to the other farmers, in a progressive system (ibid). Considering this situation in Bangladesh, the current paper used the frequency of contacts on the basis of actual number of times contacted between ordinary farmers and T&V workers or contact farmers. Note that the combination of T&V workers and contact farmers is hereinafter referred to as "extension agents" (ibid). Most of the farmers of Bangladesh are either illiterate or unskilled. Thus with the knowledge derived from extension services through extension contact, farm operators may increase their production (Haq et al., 2003; Owens et al., 2003). Relevant importance of other selected variables can be found in related literatures (Haq et al., 2004; Everson et al., 2001; Begum et al., 1998). Except for the variables of contact frequency, proportional effect of flood to crop land, village dummy and upazila "dummy, all the variables have been evaluated with a logarithmic converter to avoid disparities of the figures (Haq et a1., 2003; Owens et a1., 2003). Data have been analyzed by correlation and regression analyses. The productivity expressed in terms of physical quantity is as follows for village level analysis: Ln Rice yield = f (LnAg, Ln Fea, LnFs, Et, Fec, LnMr, Ln Lab, LnIrr, LnChem, Vdummy). For the upazila level analysis, the above function is, Ln Rice yield= f( LnAg, Ln Fea, LnFs, Et, Fec, LnMr, Ln Lab, Lnlrr, LnChem, Vdummy, Udummy).