Purpose: In India, a national survey conducted in 2003 showed that only 40% of farmers accessed extension. But little is known of the characteristics of farmers who did not access extension. However, this understanding is needed in order to target approaches to farmers, who differ in their access and use of information, that is their information search behaviors. The main objective of this paper is to segment farmers from this survey based on their information search behaviors and identify the factors that determine farmers' information search behaviors. Design/methodology/approach: Cluster analysis is applied to the number of sources accessed and frequency of source used, to define farmers' information search behaviors. Findings: The four groups that emerged are: "no search", "low search", "moderate search" and "high search". Sixty percent of farmers had no search behavior, which means they had not accessed any extension that year. By state, the largest group of these farmers was in Rajasthan. By comparison, the largest group of high searchers was in Kerala. Using Rajasthan and Kerala as case studies, these search behaviors differ by landholding size and education. "No search" farmers had the smallest landholdings, lowest education, used fewer inputs and relied on groundwater for irrigation. By comparison, "high search" farmers had the largest landholdings, most education, used more inputs and irrigated using canals. Practical implications: The difference in search behaviors between the case study states, and within the states, shows that targeted extension approaches are needed to reach different farmers, particularly the no, low and moderate search groups, with programs customized to address their context-specific information needs. Originality/value: Using information search strategy as the basis for analysis, this paper provides additional evidence of the need to consider the context-specific situations of farmers when designing extension services. (Contains 11 tables.)