Conjunction search is the search for a target defined by a unique combination of features rather than a single unique feature. For example, finding a red vertical bar among red horizontal bars and green vertical bars requires detecting the conjunction of "red" and "vertical" — neither feature alone distinguishes the target from all distractors. Feature integration theory predicted that such searches should require serial attention to each item, and while this prediction has been partially supported, the reality is more nuanced.
Classic Findings
Treisman and Gelade (1980) found that conjunction search produced steep search functions (approximately 20-30 ms per item), in stark contrast to the flat functions for feature search. The target-absent slopes were typically about twice the target-present slopes, consistent with a serial self-terminating search process: on average, the target is found after searching about half the items (present trials) or all items (absent trials).
Conjunction search: ~20-30 ms/item target-present
Conjunction search: ~40-60 ms/item target-absent
Absent:present slope ratio ≈ 2:1 → serial self-terminating search
Challenges to the Serial Account
Later research revealed that conjunction search is not always strictly serial. Nakayama and Silverman (1986) found that conjunctions involving stereoscopic depth could be searched efficiently. Wolfe, Cave, and Franzel (1989) showed that many conjunction searches are more efficient than strict serial search predicts, with slopes well below 30 ms/item. These findings motivated the development of guided search models, which propose that pre-attentive feature information can guide attention toward likely targets, making conjunction search more efficient than item-by-item inspection.
The difficulty of conjunction search reflects the binding problem: the need to correctly associate features that belong to the same object. Pre-attentive feature maps register that "red" and "vertical" are present somewhere in the display, but without focused attention, the system cannot determine which features co-occur at the same location. This is why conjunction targets do not pop out and why illusory conjunctions (misattributions of features between objects) occur under conditions of divided attention.
Factors Modulating Conjunction Search
Several factors affect the efficiency of conjunction search. Target-distractor similarity in each feature dimension matters: the more discriminable each feature, the more efficiently top-down guidance can narrow the search. The number of feature dimensions defining the target also matters: triple conjunctions can be searched more efficiently than double conjunctions when each feature provides useful guidance. Practice can also improve conjunction search efficiency, though it typically does not achieve the flat slopes characteristic of feature pop-out.
Real-World Applications
Many real-world search tasks are conjunction searches. Airport security screeners must find targets defined by combinations of shape, size, and density in X-ray images. Radiologists search for tumors defined by conjunctions of shape, contrast, and location. Understanding the factors that make conjunction searches efficient or inefficient has practical implications for training, display design, and the evaluation of human search performance in safety-critical domains.