This paper presents an experimental study on strategic sophistication in normal-form games, analyzing how subjects' decisions and information searches reflect their ability to predict others' actions. The study uses a computer interface to monitor subjects' searches for hidden payoff information, allowing researchers to better understand how decisions are determined and how deviations from equilibrium decisions relate to deviations from search patterns suggested by equilibrium analysis. The paper reports experiments on 18 normal-form games with various patterns of iterated dominance and unique pure-strategy equilibria, aiming to assess the sophistication of subjects' decisions and information searches. The study introduces a mixture model where each subject's behavior is determined by one of nine decision rules or types, with each subject's type drawn from a common prior distribution. The types include nonstrategic types like Altruistic, Pessimistic, and Naïve, and strategic types like L2, D1, D2, Equilibrium, and Sophisticated. Strategic types exhibit varying degrees of sophistication, with Sophisticated representing the ideal of a game theorist who can predict others' behavior. The study finds that subjects' decisions are too heterogeneous to be described by a single decision rule, even allowing for errors. Compliance with equilibrium decisions is high in games solvable by one or two rounds of iterated dominance but much lower in games solvable by three rounds or those with unique equilibria without dominance. Subjects' information searches are even more heterogeneous than their decisions. The study uses an econometric analysis to estimate the frequencies of different types and error rates, finding that the most frequent types are Naïve and L2, with strategic types like L2 or D1 better describing most subjects' sophistication than Equilibrium or Sophisticated. The analysis also shows that subjects' deviations from equilibrium decisions are closely related to their deviations from search implications of equilibrium analysis. The study concludes that the results suggest a simple view of subjects' behavior, with Naïve and L2 comprising a large portion of the population, and D1 making up a smaller portion. The study provides a framework for analyzing subjects' decisions and information searches, highlighting the importance of cognitive processes in strategic behavior.This paper presents an experimental study on strategic sophistication in normal-form games, analyzing how subjects' decisions and information searches reflect their ability to predict others' actions. The study uses a computer interface to monitor subjects' searches for hidden payoff information, allowing researchers to better understand how decisions are determined and how deviations from equilibrium decisions relate to deviations from search patterns suggested by equilibrium analysis. The paper reports experiments on 18 normal-form games with various patterns of iterated dominance and unique pure-strategy equilibria, aiming to assess the sophistication of subjects' decisions and information searches. The study introduces a mixture model where each subject's behavior is determined by one of nine decision rules or types, with each subject's type drawn from a common prior distribution. The types include nonstrategic types like Altruistic, Pessimistic, and Naïve, and strategic types like L2, D1, D2, Equilibrium, and Sophisticated. Strategic types exhibit varying degrees of sophistication, with Sophisticated representing the ideal of a game theorist who can predict others' behavior. The study finds that subjects' decisions are too heterogeneous to be described by a single decision rule, even allowing for errors. Compliance with equilibrium decisions is high in games solvable by one or two rounds of iterated dominance but much lower in games solvable by three rounds or those with unique equilibria without dominance. Subjects' information searches are even more heterogeneous than their decisions. The study uses an econometric analysis to estimate the frequencies of different types and error rates, finding that the most frequent types are Naïve and L2, with strategic types like L2 or D1 better describing most subjects' sophistication than Equilibrium or Sophisticated. The analysis also shows that subjects' deviations from equilibrium decisions are closely related to their deviations from search implications of equilibrium analysis. The study concludes that the results suggest a simple view of subjects' behavior, with Naïve and L2 comprising a large portion of the population, and D1 making up a smaller portion. The study provides a framework for analyzing subjects' decisions and information searches, highlighting the importance of cognitive processes in strategic behavior.