One lasting controversy in attention concerns whether or not unattended stimuli are identified. The late selection view assumes that unattended stimuli receive full analysis leading to identification (van der Heijden, 1992). In contrast, the early selection view proposes that unattended stimuli receive a coarse analysis that does not include identification (Broadbent, 1982). The "flankers task" of Eriksen and Eriksen (1974) has frequently been used to investigate this issue. The subject sees a linear array of letters and decides whether, say, H or I has appeared in the center of the array (target). Flankers on either side of the target must be ignored. The flankers are either compatible (HHH) or incompatible (IHI) with the target. Typically, subjects are slower under the incompatible than under the compatible condition. On the basis of this finding, it has often been stated that unattended flankers are processed to the point of identification.Alternatively, Yantis and Johnston (1990) argued that flanker effects could be the product of attentional leakage to the flanker location. They used uncrowded (3.2º centerto-center separation between items) circular arrays, and they predirected attention to the target with a 100% valid exogenous precue. Although residual flanker effects were found for flankers adjacent to the target, no flanker effect was apparent for flankers located farther away from the target. Similarly, Paquet and Craig (1997) reported flanker effects for near (.3º) but not for far flankers (5º away from the target). These converging results suggest that flanker identification may not occur when the possibility of attentional leakage to the flanker location is reduced.In contrast with the preceding interpretation, Driver and Tipper (1989) pointed out that null flanker effects do not necessarily reflect a lack of identification. In their experiment, two circular alphanumeric arrays were presented in rapid succession, and subjects had to report, for each array, the number of red items (target) and to ignore the black items (flankers). The first array was referred to as the prime and the second was termed the probe. For the prime displays, flanker identification was inferred by examining whether or not performance was affected by the identity relationship between the digit flankers and the number of red items. In addition, a second index of prime flanker identification (labeled negative priming) was provided by varying the identity of the prime flankers in relation to the probe target. More specifically, the prime flankers and the probe target were identical in an ignored repetition condition, whereas in a control condition, the probe target had not appeared in the prime display. The key result was that although prime display performance was unaffected by prime flankers, negative priming was obtained for the probe trials (i.e., latency was longer in the ignored repetition than in the control condition). The investigators concluded that unattended flankers are identified but that flanker effects are avoide...