Week 8: Introduction to Cognition & Attention

• Cognition

  The study of the workings of the human mind by studying human behaviour, often using experimental methods (and also computational modelling, neuroscience, and neuropsychological/patient data)
• Focused (auditory) attention

  • Studies of auditory selective attention
  • Dichotic listening task
  • Findings by Cherry, Moray ("cocktail party effect")
• Bottleneck models

  • Broadbent's filter model
  • Treisman's attenuation model
  • Deutsch & Deutsch's late selection model
• Early vs. late selection

  • Findings by Treisman & Riley
  • Flexible bottleneck
• Studies of auditory selective attention: the cocktail party problem
• How do we follow one conversation at a noisy party? Use selective attention to focus on one stimulus input and ignore others
• Dichotic listening task

  Participant asked to shadow (repeat back) message played to one ear and ignore (not be distracted by) the other
• Cherry (1953) findings


  • Subjects could report physical characteristics but not semantic content of unattended message
  • Only the physical characteristics, and not the semantic content (meaning), are processed in the unattended message
• Moray (1959) findings

  • Subjects did not notice repetition of the same word 35 times in unattended message, but 33% detected their own name
  • Meaning is processed in the unattended message
• Moray (1959) referred to it as the "identification paradox" and it is generally referred to as the "cocktail party phenomenon/effect"
• Only about 1/3 of Moray's (1959) participants showed the cocktail party effect (detected their own name in the unattended message)
• Conway, Cowan & Bunting (2001) investigated which individuals are more likely to demonstrate the cocktail party effect
• Naveh-Benjamin et al. (2014) followed-up this work to investigate whether older participants are more/less likely to show the cocktail party effect than younger participants, and why
• Bottleneck models of attention

  • They all assume the multistore model of memory architecture
  • They differ in where they regard the bottleneck is, and the nature of the bottleneck
• Broadbent's filter model

  • Stimuli gain access in parallel to a sensory register
  • Selective filter (all-or-none) blocks processing of unattended information to prevent overloading of limited-capacity STM store
  • Input remaining in STM not blocked by the filter (i.e., attended) undergoes semantic processing
• Broadbent's filter model is consistent with Cherry's findings but inconsistent with Moray's findings (the cocktail party effect)
• Treisman's attenuation model

  • Instead of an all-or-one filter, an attenuator turns down the amount of processing of unattended information
  • In addition, the thresholds of context-appropriate stimuli are lower
• Treisman's attenuation model can explain Moray's findings (the cocktail party effect)
• Deutsch & Deutsch's late selection model


  • Information is analysed fully (physical, semantic), even for unattended message
  • Bottleneck is late: at selection for action (e.g., cannot say two things at the same time)
• Deutsch & Deutsch's late selection model can explain what the attenuation model can explain, so the critical difference is how early/late the selection occurs
• Treisman & Riley (1969) experiment

  • Participants shadowed one message and made a tapping response to a target word in either message
  • Supported the attenuation model: early selection
• Flexible bottleneck view

  • The location of bottleneck is flexible (it may be early or late)
  • Unattended message is not always processed fully to the level of meaning
• Johnston & Wilson's (1980) experiment

  • Participants detected a target word (member of semantic category) in either ear, with non-target words biasing the meaning of ambiguous target words
  • In the focused attention condition, no effect of type of non-target on target detection, indicating non-targets were NOT processed to the level of meaning
• Target


  Member of semantic category (e.g., musical instrument) detected in either ear
• Non-target


  Word presented coincidentally with target
• Critical targets


  Ambiguous meaning (e.g., ORGAN)
• Meaning interpretation of critical target

  Biased by non-target (word in the other channel)
• Meaning interpretation of critical target

  • Appropriate: church - ORGAN (musical instrument)
  • Neutral: paper - ORGAN
  • Inappropriate: kidney - ORGAN
• If unattended message is processed to the level of semantics, meaning of non-target should influence detection of ambiguous target word
• Focused attention condition


  Participants told which ear targets would arrive
• Divided attention condition


  Participants did not know which ear targets would arrive
• In the focused attention condition, no effect of type of non-target on target detection. Meaning: non-targets were NOT processed to the level of meaning
• In the divided attention condition, (target detection: appropriate, neutral, inappropriate) non-targets were processed to the level of meaning
• Meaning (of non-target)

  Was processed (late selection) when attention was divided over two ears but not when attention was focused on the other ear (early gating)
• Flexible Bottleneck View

  • The more stages of processing (physical --> semantic), the greater the demands on attentional capacity
  • Selection occurs as early in processing stages as possible to minimise attentional demands
• Dual task performance


  Determined by task (dis)similarity and practice
• Automaticity


  With practice, the task becomes automatic
• Characteristics of automaticity

  • Fast
  • Require little attentional capacity
  • Inflexible (once learned, difficult to modify) = habits
  • Unavoidable (occur without intention)
  • Unavailable to consciousness
• Shiffrin & Schneider's Memory Search Experiment

  1. Participants memorise 1, 2, 3, or 4 targets (memory set)

  2. Participants are shown a display containing 1, 2, 3, or 4 items (display set)

  3. Respond as quickly as possible whether the display contained a target from memory set
• Consistent mapping (CM) condition


  Target set and distractor set do not overlap from trial to trial