This article presents data from visual search experiments that challenge the standard feature integration model of visual search. The model, proposed by Treisman, suggests that searches for conjunctions of basic features (e.g., color and form) require serial, self-terminating search. However, the results from these experiments show that for many unpracticed subjects, the slopes of RT × Set Size functions are shallow, inconsistent with the model's predictions. In particular, searches for triple conjunctions (e.g., color × size × form) are easier and can be independent of set size, suggesting that parallel processes can guide attention more effectively than two. The guided search model, similar to Hoffman's two-stage model, is proposed as an alternative. This model suggests that parallel processes use information about simple features to guide attention in the search for conjunctions. Triple conjunctions are found more efficiently because three parallel processes can guide attention more effectively than two. The results also show that searches for targets among distractors can be easier for some stimuli than others. For example, targets defined by a unique color or orientation are found easily. However, searches for a target among a field of distracting items (e.g., a T among Ls) are not so effortless, and the time required to find the target increases with the number of distractors. The article also discusses the differences between the results of these experiments and previously published results. These differences are attributed to differences in the stimuli, not to learning or a general ability to perform all searches in parallel. The results suggest that the standard feature integration model needs to be modified to account for the role of parallel processes in guiding serial search. The guided search model is proposed as a modification of the feature integration model. It suggests that parallel processes guide the "spotlight of attention" toward likely targets. This model can explain the results of the experiments, including the efficiency of searches for triple conjunctions and the shallow slopes observed in some cases. The article also discusses the implications of these findings for the understanding of visual search. The results suggest that the standard feature integration model may not fully account for the efficiency of visual search, and that the guided search model provides a better explanation of the data. The guided search model predicts that searches for triple conjunctions should be more efficient than searches for simple conjunctions, and this prediction is supported by the experimental results.This article presents data from visual search experiments that challenge the standard feature integration model of visual search. The model, proposed by Treisman, suggests that searches for conjunctions of basic features (e.g., color and form) require serial, self-terminating search. However, the results from these experiments show that for many unpracticed subjects, the slopes of RT × Set Size functions are shallow, inconsistent with the model's predictions. In particular, searches for triple conjunctions (e.g., color × size × form) are easier and can be independent of set size, suggesting that parallel processes can guide attention more effectively than two. The guided search model, similar to Hoffman's two-stage model, is proposed as an alternative. This model suggests that parallel processes use information about simple features to guide attention in the search for conjunctions. Triple conjunctions are found more efficiently because three parallel processes can guide attention more effectively than two. The results also show that searches for targets among distractors can be easier for some stimuli than others. For example, targets defined by a unique color or orientation are found easily. However, searches for a target among a field of distracting items (e.g., a T among Ls) are not so effortless, and the time required to find the target increases with the number of distractors. The article also discusses the differences between the results of these experiments and previously published results. These differences are attributed to differences in the stimuli, not to learning or a general ability to perform all searches in parallel. The results suggest that the standard feature integration model needs to be modified to account for the role of parallel processes in guiding serial search. The guided search model is proposed as a modification of the feature integration model. It suggests that parallel processes guide the "spotlight of attention" toward likely targets. This model can explain the results of the experiments, including the efficiency of searches for triple conjunctions and the shallow slopes observed in some cases. The article also discusses the implications of these findings for the understanding of visual search. The results suggest that the standard feature integration model may not fully account for the efficiency of visual search, and that the guided search model provides a better explanation of the data. The guided search model predicts that searches for triple conjunctions should be more efficient than searches for simple conjunctions, and this prediction is supported by the experimental results.