Selective Attention: Endogenous vs Exogenous Orienting; Covert vs Overt; Theoretical Models (Lecture Notes)

Lab context and data quality

• The instructor notes the assessed lab is running this week with three tutorial groups so far, all going smoothly.
• Bonus data is added because groups are small (some groups have only 4–5 people).
  • The experiment is real and run in the instructor’s lab; data cannot be used as class data due to ethics and to preserve research integrity.
  • Running in Windows introduces variability in reaction times; the lab uses older DOS-based machines for more stable reaction times.
  • Even though the bonus data exist, it is not used to publish results; differences observed are small (roughly $ext{16–25 ms}$) and the overall pattern matches the lab machines' data.
• Power and precision
  • More data from the class increases statistical power and reduces variability, making the combined results more interpretable.
• Practicalities of the lab session
  • The session is longer than expected (30–35 minutes planned, possibly 40–45 minutes) to run enough trials so data are stable.
  • Initial lab groups’ data look interesting and interpretable, which should make the assignment easier.
• Ethical and methodological safeguards in real experiments
  • It is unethical to publish class data as if the class were a single subject pool.
  • In the actual lab, no personal details are collected (no biological sex, age, handedness, etc.).
  • The data aim to answer a real research question while protecting participant privacy.
• Broader goal of the lab
  • Students act as researchers on a genuine question of interest, which is more engaging than a classic replicated study.

Key concepts: what attention is and how it is studied

• Attention can be directed (or oriented) to a location in space; we don’t attend to the whole world at once.
  • Visual attention is often illustrated with green arrows/markers showing where attention is focused; eyes may move or stay fixed.
• The premotor theory of attention
  • Evidence suggests attention shifts before eye movements (a saccade). Attention may lead eye movement; this is debated.
  • Implication: attention can move without moving sensory organs (covert attention).
• Covert vs overt orienting
  • Covert orienting: attention shifts without moving eyes/head, observable by behavior but not by gaze.
  • Overt orienting: attention shifts with movement of eyes/head toward a stimulus.
• Endogenous vs exogenous orienting
  • Endogenous (internal) orienting: voluntary control of attention, driven by goals or expectations (e.g., a cue that you interpret as indicating where the target will appear).
  • Exogenous (external) orienting: attention captured by an externally salient event (e.g., a flash or sudden sound).
  • Mental shorthand: endogenous can be thought of as internal, exogenous as external.
• Interplay between orienting type and mode of attention
  • The same overall label "attention" may cover different underlying processes (endogenous/exogenous) and different modes (covert/overt).
  • A metaphor: two engines (two control mechanisms) may lead to the same destination, or two separate cars may arrive with different drivers, implying potentially different properties of the attentional processes

Demonstration paradigms: Posner cueing tasks

• Posner cueing tasks come in two main forms, developed by Michael Posner:
  • Endogenous cueing (arrow cue): participants are told that the arrow indicates where the target will appear with higher probability (80% valid).
  • Setup: four boxes in a square with a central fixation point. An arrow cue points to one location while eyes stay fixed (covert, endogenous).
  • Target: appears with higher probability at the cued location; reaction time faster for valid vs invalid cues.
  • Exogenous cueing (flashes): a cue that is meaningless or neutral but captures attention via a flash or sudden onset.
  • Setup: fixation point with a brief flash at one location before the target appears; eyes may stay fixed (covert, exogenous).
  • Target: appears at the cued location more quickly if the cue was valid, but the cue’s meaning is not informative (the cue is not predictive).
• Parallels across modalities
  • The same cueing logic can be used in auditory tasks (e.g., high vs low tones directing attention to a location in space) with similar effects.
  • Attention effects generalize across modalities: visual and auditory tasks show similar endogenous and exogenous cueing patterns.
• Likely outcome of cueing tasks
  • With cued locations (valid trials), responses are typically faster than on uncued trials (invalid trials).
• The 2x2 framework for Orienting and Response
  • Control mechanism: endogenous vs exogenous
  • Type of orienting: covert vs overt
  • The experiment can be arranged to test all combinations (e.g., endogenous-covert, exogenous-covert, endogenous-overt, exogenous-overt).
• Meaningful vs meaningless cues
  • Meaningful cues activate both endogenous and exogenous processes (you choose to attend to the cue location, and the cue also captures attention).
  • If cues are meaningless, exogenous capture can be isolated (attention shifts without a helpful cue).

Tasks and terminology in attention research

• Focus vs divided attention tasks
  • Focused attention: respond only to targets at the cued location; ignore targets at other locations.
  • Divided attention: respond to targets in attended and unattended locations; processing may be distributed.
• What counts as a focus task
  • In a focus task, non-cued locations should be ignored, and targets appearing there should not trigger responses.
• The concept of buffers and bottlenecks
  • Buffers: temporary memory stores that hold information for later processing (a buffer for unattended information that may be processed later).
  • Bottleneck: a stage where information processing slows or stalls; only a limited amount of information can be processed at a time.
• Sensory channels and information flow analogy
  • Information from multiple sensory channels (e.g., left vs right ear, or different visual streams) enters processing streams with a bottleneck; attention acts like a selective filter.
• Dichotic listening and monaural tasks (auditory domain)
  • Dichotic listening: different streams of information presented to each ear simultaneously.
  • Monaural listening: information presented to only one ear; the other ear is quiet.
  • Hemispheric processing nuances: right ear input tends to engage left hemisphere language areas more quickly; left ear input more strongly engages right hemisphere.
  • Shadowing task: participants repeat back what they hear in one ear to ensure they are focusing attention there.
• Broadbent’s early selection model
  • Idea: unattended information is blocked from semantic processing; attention operates early in the processing stream, near detection rather than semantics.
  • Metaphor: pipes and a filter; only information that passes through the "hole" (attention) is semantically processed.
  • Buffers are needed to hold unattended information until it can be processed; if information arrives too fast, we may buffer more on one side and shift attention later.
  • Predictions: unattended information is not semantically identified unless it breaks through the filter via attention.
• Breakthrough events and semantic influence on unattended channels
  • Breakthrough events: instances where unattended information is semantically processed and can influence responses (e.g., your own name can break through the unattended channel).
  • Problem for early selection: such semantic breakthroughs occur more often when unattended material is semantically related or personally relevant; this challenges strict early selection.
• Attenuation theory (Treismann)
  • Proposed as a refinement: attention reduces (attenuates) but does not completely block unattended information; processing can occur at a reduced level, possibly reaching semantic processing for highly salient items (e.g., own name, high-frequency words).
  • Thresholds: some words have lower thresholds to activate meaning even when attenuated; high-frequency words are easier to activate.
• Late selection theory (Deutsch & Deutsch)
  • Proposes that all input is fully processed semantically; attention later selects what remains to be held or reported.
  • This model accommodates semantic processing of unattended information and explains breakthrough events without requiring early filtering.
• Experimental investigations and contrasts
  • Shadowing with biased context: Johnson and Wilson experiments on divided vs focused attention with ambiguous words (e.g., bank can mean river bank or financial bank).
  • In divided attention, semantic biases from the unattended context can influence responses; in focused attention, such biases are reduced or eliminated.
  • This supports a view that some semantic processing occurs even when attention is focused, aligning more with attenuation or late selection than strict early selection.
• Synthesis: where attention acts in the processing stream
  • Evidence supports a spectrum: attention can bias processing early (Broadbent-like) or later (Treisman-like), depending on stimuli, task demands, and relevance.
  • There is no single universal bottleneck; different tasks reveal different bottlenecks and degrees of processing for unattended information.

How researchers interpret and use these models

• Research logic and theory testing
  • Scientists test how attention explains current data, predict what should not happen if a theory is true, collect data to see whether the impossible happens, and then revise theories accordingly.
• Examples of predictions and experiments
  • If unattended information is truly blocked from semantics (Broadbent), you should not see semantic influences from unattended streams under focused attention unless there is a breakthrough.
  • Attenuation or late selection predicts that unattended information can influence processing if it is semantically related or personally relevant.
• Experimental design considerations
  • Endogenous vs exogenous cues help isolate voluntary vs automatic attentional control.
  • Covert vs overt tasks help differentiate motor involvement from attentional shifts.
  • Shadowing and dichotic listening are used to probe how attention filters affect perception and reporting.
• Real-world relevance and applications
  • The tasks model how people monitor environments and ignore distractions in daily life (e.g., driving, classroom attention).
  • Understanding how attention filters information can inform training, safety, and UX design.

Real-world implications and debates raised in the lecture

• Attention and safety: driving and cell phone use
  • Real-world example discussed: talking with a passenger can be safer than holding a phone because a passenger can respond to traffic, unlike a hands-free phone that eases cognitive load but still diverts attention.
• ADHD and breakthrough events (speculative discussion)
  • The lecturer raises a question about whether individuals with ADHD might experience more breakthrough events due to differences in attention control.
  • The complexity of ADHD is noted: strong focus on topics of interest versus diffuse attention when not engaged.
• Ethics of data collection and publication
  • Emphasizes protecting participant privacy and avoiding publishing class data as representative of a broader population.
• Philosophy of science in psychology
  • Emphasizes iterative theory development: build the theory to explain current data, predict impossible outcomes, test, and revise.

Summary of key takeaways

• Attention is a selective process that filters and prioritizes information because processing capacity is limited.
• There are different forms of orienting attention:
  • Endogenous (voluntary) vs Exogenous (stimulus-driven)
  • Covert (no overt movement) vs Overt (eye/head movements toward the stimulus)
• Posner cueing tasks reveal robust effects where cues bias attention and speed up responses at cued locations.
• Auditory attention studies (dichotic vs monaural listening) illustrate early filtering, buffering, and semantic processing of attended vs unattended channels.
• Competing theories explain how attention filters information:
  • Broadbent’s early selection (filtering at or before semantic processing)
  • Treisman’s attenuation theory (unattended information is attenuated, not fully blocked; semantic access possible for salient items)
  • Deutsch & Deutsch’s late selection (semantic processing occurs for all inputs; attention selects what to report)
• Experimental results often support a hybrid view: attention can operate at multiple stages depending on stimuli, task demands, and salience.
• Ethical and methodological considerations are essential in laboratory work: ensure data quality, protect participants, and be mindful of how results are used and reported.

Key numerical and methodological references (selected)

• Class data variability and reaction times
  • Observed differences between lab machines vs Windows: roughly ext{16 ms}
    ightarrow ext{25 ms}; patterns remain similar across data sets.
• Power considerations
  • Adding class data increases statistical power; small bonus data by itself may be non-significant due to limited sample size.
• Timing of sessions and trial stability
  • Longer sessions (up to $40 ext{–}45$ minutes) yield more stable data than shorter sessions.
• Cue validity and timing in Posner tasks
  • Endogenous cue validity typically around 80 ext{%}; cued locations yield faster responses on valid trials than invalid trials.
• Buffering and processing capacity
  • Conceptual buffering spans can be on the order of seconds in the auditory system (e.g., two-second window observed in some studies).

Closing reminder from the lecture

• The instructor emphasizes that the study of attention is about testing theories against data, predicting what should not happen, and refining theories when empirical evidence contradicts them.
• Thursday’s session will further explore the empirical evidence and distinctions among endogenous/exogenous and covert/overt attention, as well as delve into focus vs divided attention and memory buffering in more detail.