<P> Broadbent showed evidence of early selection using a split - span technique . Participants were presented with a different list of digits in each ear . When asked to report the digits they heard, participants tended to report all digits presented to one ear first and then the digits reported to the second ear, regardless of the order in which the digits were presented . When asked to report the digits they heard in the order they were presented, accuracy dropped . This study suggested that information is filtered according to basic characteristics of the stimuli (e.g., the ear in which it was presented). Participants first reported the information from one ear and then switched filters in order to report the digits presented to the second ear . Therefore, when required to report the digits in the order they were presented, participants had to continuously switch filters, which impacted accuracy . </P> <P> Late selection models argue that information is selected after processing for meaning, as opposed to during the earlier stages of processing . According to these models, all information is attended to, whether intentionally or unintentionally . Information inputs are processed equivalently, until semantic encoding and analysis can be performed . The filter merely acts as an information attenuator; it intensifies the pertinent information and attenuates the intensity of the stimuli deemed to be unimportant . This notion implies that internal decisions of stimuli relevance must be made, before allowing it to gain conscious awareness . </P> <P> Gray and Wedderburn showed evidence of late selection using a split - span technique similar to Broadbent . This time, participants heard a mixture of numbers and words presented to each ear, such as, "Dear--7--Jane" in the left ear and, "9--Aunt--6" in the right ear and were asked to report back what they heard . According to the early selection model, participants should have reported all items presented to one ear first, and the items presented to the other ear second . However, the researchers found that participants reported hearing, "Dear Aunt Jane" and "9--7--6 ." This study suggested that stimuli are not selected based on physical characteristics (e.g., location of sound) determined by the filter but according to meaning . </P> <P> Anne Treisman, a graduate student of Broadbent's, was not fully convinced by the notion of a filter performing decisions as to what stimuli gain conscious awareness . She proposed an alternative mechanism, the attenuation theory . This theory supports an early - selection filter . However, in this case, the filter also attenuates stimuli presented to the unattended channel . If the stimuli pass a threshold, it will leak through the filter and can be attended to . As the unattended channel includes weakly attended to information, to gain conscious awareness this information must surpass a threshold, which Treisman believed was determined by the words' meaning . Important words (such as one's name) would have a low threshold, to easily gain awareness, whereas unimportant words (such as "lamp") would have a higher threshold to prevent them from gaining awareness inappropriately . In this way, the threshold for each word acts as a filtering mechanism, relying on semantic features . </P>

Where does the bottleneck occur in broadbent's filter model