Illusions, Perception, and Lab Analysis: Top-Down Influences and Paired-Samples Testing

Perceptual Illusions: Overview

  • Illusions are visually perceived images that are deceptive or do not match real-world visuals.
  • They arise because perception is constructed by the brain using top-down influences that read and interpret sensory input.
  • Common examples discussed:
    • Musician vs. face illusion: some people see a musician first, others a face; many can see both when looking closely.
    • “Two buttons” illusion: a pattern with two button-like features (described as part of the session).
    • Black dots illusion: some people see black dots appearing in the middle of patterns that do not actually exist; perception varies.
    • Green dot illusion: a (pink) dot pattern makes a green dot appear to move around the circle; no actual green dot exists; fixation point affects perception.
    • Poppy nose (face-related) illusion: some elements appear to pop out while middle elements recede; covering parts of the image can change what’s perceived.
  • Key point: perception is influenced by top-down processes, not just raw sensory input.

Top-Down Influences on Perception

  • Top-down influences discussed: memory, expectations, and context.
  • These influences are interrelated; memory, expectations, and context can shape how raw sensory information is interpreted.

Memory

  • Defined as prior knowledge and experiences that inform current perception.
  • Types of knowledge:
    • Factual knowledge or explicit knowledge.
    • Prior exposure or familiarity with stimuli.
  • How memory can distort perception:
    • When there are gaps or blanks in the sensory input, expectations can fill in missing details, potentially leading to misinterpretation.
  • Examples mentioned:
    • Faces: we have a memory of faces, so prior experience with faces influences what we expect to see when presented with ambiguous stimuli.
    • Relatedness study: when participants were told that a baby and adults were related, they rated relatedness higher than when told they were unrelated, illustrating memory/expectation effects on interpretation.

Expectations

  • Expectations are predictions about what we will perceive based on prior knowledge, context, or cues.
  • Influence on perception is demonstrated in several experiments:
    • Taste study: participants told a food will taste bad rated it worse after tasting than those not told anything.
    • Social judgment: referees who were told one team was very aggressive gave more penalties to that team even though there was no actual difference in play; context altered perception/behavior.
  • Expectations and memory are closely tied; they shape how we interpret sensory data and fill in missing information.

Context

  • Context includes environmental cues, surrounding information, and situational factors.
  • Examples from the session:
    • Color contrast in a figure provides contextual information that can alter perception of the same stimulus.
    • Focusing on a fixation point can change what we see, illustrating how attention and context influence perception.
    • Context compounds with memory and expectations to modulate perceptual outcomes.
  • Context is broad and overlaps with memory and expectations: it often acts as a scaffolding for interpretation.

Why Illusions Occur

  • Process overview:
    • Senses gather information from the environment.
    • The brain uses attention to process this information and constructs a perceptual experience.
    • This constructed perception can differ from the actual physical stimulus, especially when top-down influences override or fill in missing details.
  • Summary takeaway: we tend to see what we expect, even when pieces of the stimulus are missing or misleading.

Illusions as Research Tools

  • Illusions are used by scientists to study perception and the mechanisms of top-down processing.
  • They are not just classroom curiosities; they provide insights into how memory, context, and expectations shape perception.

Labs and Experiments (Overview)

  • Three experiments on illusions were conducted; identical overall design across experiments.
  • Focus: how context affects perception in a perceptual task (e.g., misalignment/“mean error” across angles).
  • Data collection approach:
    • Results captured from screens; participants take pictures of results for the first experiment.
    • For the other experiments, participants record their data (context vs no context, and narrow vs wide lines) on handouts.
    • Data and materials are organized in the desktop documents folder (e.g., a folder named something like "authorware experiments").
  • Data entry workflow:
    • Enter all angles and corresponding mean errors into a spreadsheet (Angle in column A, Mean error in column B).
    • For the last two experiments, fill in context/no-context and narrow/wide entries on the handout.
    • After data collection, upload/enter data into the designated data-entry system (Jamovi or similar) for analysis.
  • Class workflow during the session:
    • Short break for data entry.
    • Then proceed to statistical analysis workflow (paired-samples tests, descriptives, and plotting).
  • Homework integration:
    • Each of the three experiments will be used for homework datasets.
    • Students will run analyses and produce graphs as described by the handouts.
    • The third illusion (Bob and Dirac illusion) involves running paired-samples t-tests between two of the three possible conditions.
  • Important logistics:
    • Use Canvas for accessing sample data and handouts; a specific sample data file is provided (e.g., named sample data).
    • All numbers must be rounded to two decimal places (see rounding rule below).
    • Always ensure proper figure labeling and formatting per the instructor’s guidelines.

Data Analysis Workflow (Paired and Descriptives)

  • Descriptives for context vs no context:
    • In Jamovi: Analysis > Descriptives; place Context and No Context into the variable box; outputs appear with each condition as separate columns (within-subject design).
    • For angles: run Descriptives with each angle as a variable to obtain a comprehensive table of statistics per angle.
  • Paired-samples t-test:
    • In Jamovi: Tests > Paired samples t-test; set the pair as Context vs No Context.
    • Output interpretation format (APA-like):
    • t(df) = tvalue, p = pvalue
    • Example: t(df) = -3.50, p < 0.001 (significant if p < typical alpha, e.g., 0.05).
    • Interpretation guidance:
    • If the mean in the context condition is higher than No Context, report that context increased the mean error.
    • Write up: “The context condition had a significantly higher mean error than the no-context condition, t(df) = tvalue, p = pvalue.”
  • Graphing the data:
    • Use a line graph to depict mean errors across angles for contextual vs. no-context conditions.
    • Construction steps (example workflow):
    • In Excel: put angles in column A and mean errors in column B (per angle, across participants).
    • Select both columns and insert a line graph with markers (scatter plot with lines).
    • Title the graph in Word (do not embed the title in the graph image itself).
    • Axis labeling: X-axis = Arrow angle (or similar), Y-axis = Mean alignment error (or mean error).
    • Ensure the graph title is separate from the graph image (caption/title outside the figure).
  • Graph formatting notes:
    • Some graphs may start at zero; others may not, depending on the data.
    • When preparing the final report, include axis labels for the X and Y axes and a figure title on the Word document.
    • The figure caption (Figure 1, etc.) should be italicized in text (formatting note from instructor).

Reporting and Formatting Guidelines for Homework

  • Rounding:
    • All numerical values should be rounded to two decimal places, e.g., 0.55, 1.23, 2.00.
    • If p < 0.001, report as p = 0.001 (or follow the instructor’s specific convention discussed in class).
  • Statistical notation:
    • In-text/statistical notation, italicize statistics (e.g.,
      $t$, $r$, and $p$ should be italicized).
    • Report results in the standard format: t(df) = value, p = value (or p < value).
  • Figure and table formatting:
    • Figures should be referenced as Figure 1, Figure 2, etc., with a separate, properly formatted caption.
    • Do not place the figure title inside the figure image; include it in the document text or caption.
    • Use italicized t, r, and p when referring to test statistics in the write-up.
  • Feedback and iteration:
    • The instructor will provide retrospective feedback by email on common issues spotted in submitted homework, so students can adjust before the next assignment.
  • Timeline and structure:
    • Only a couple of classes remain with this style of activity; one library session is entirely in-person.
    • The group assignment details will be discussed next week; topics should be psychology-related and viable.

Group Work and Next Steps

  • Group activity:
    • Students will be assigned to groups to exchange contact information and discuss potential topics.
    • The goal is to identify a psychology-related, viable topic that all group members find interesting.
  • Upcoming topics and prep:
    • Next week, the instructor will discuss the group project details, including constraints and allowable topics.
  • Availability of resources:
    • A document for class is uploaded to Canvas for reference, and the homework data sharing is coordinated through that platform.
  • Final reminders:
    • If there are questions about the analysis or homework, ask during class or reach out for clarification.
    • Ensure you are working on the correct class and dataset during the labs.

Quick Reference: Key Formulas and Notation

  • Independent samples t-test (brief reminder):
    • t = rac{ar{X}1 - ar{X}2}{
      oot{2} ext{(}sp^2 igl( rac{1}{n1}+ rac{1}{n_2}igr)igr)^{1/2}}
    • where s<em>p2=(n</em>1−1)s<em>12+(n</em>2−1)s<em>22n</em>1+n2−2s<em>p^2 = \frac{(n</em>1-1)s<em>1^2 + (n</em>2-1)s<em>2^2}{n</em>1+n_2-2}
    • df = n1 + n2 - 2
  • Paired samples t-test:
    • Define differences $Di = X{1i} - X_{2i}$ for each participant.
    • t = rac{ar{D}}{sD / abla}{ abla} = rac{ar{D}}{sD /
      sqrt{n}}
    • df = n - 1
  • Descriptive statistics: mean, standard deviation, standard error, etc., per condition or per angle as appropriate.
  • Reporting format (APA-like):

    • t(df) = t{value}, p = p{value}
    • Example: t(28) = -3.50, p < 0.001

Connections to Broader Concepts

  • The illusion studies illustrate foundational principles of perception:
    • Constructivist view: perception is constructed by the brain from sensory data, not a direct readout of the world.
    • The interplay of bottom-up input and top-down interpretation leads to misperceptions.
  • Real-world relevance:
    • Bias in judgments based on contextual cues (e.g., judging performance, taste, or fairness) can influence decisions even without changes in actual stimuli.
    • Understanding these processes can help design better interfaces, educational tools, and experiments that account for bias.

Ethical, Philosophical, and Practical Implications

  • Ethical: awareness of context effects in expert judgments (e.g., refereeing, ratings) highlights potential biases that can affect fairness; need for blind or controlled contexts when possible.
  • Philosophical: perception is a constructive process, blending memory, expectation, and context with sensory input; what we perceive is not a guaranteed reflection of reality.
  • Practical: when conducting experiments on perception, control context and provide clear instructions to minimize unintended top-down influences unless they are part of the experimental manipulation.

References to the Transcripted Lab Details

  • Three experiments focused on a perceptual task with context vs no-context conditions and angle manipulations.
  • The teacher-guided workflow included:
    • Descriptive statistics for context vs no-context and for angles.
    • Paired-samples t-test to compare means between the two conditions.
    • Creation of line graphs showing mean error across angles.
    • Emphasis on formatting, rounding, and proper figure labeling in reports.
  • Example interpretive guidance from the session:
    • If context increases mean error, report that the context condition produced significantly higher errors than no-context (t(df) = tvalue, p = pvalue).
    • Graphs should illustrate the relationship of mean error with angle and the difference between conditions.

Quick Summary Takeaways

  • Illusions reveal how top-down processing (memory, expectations, context) can distort perception.
  • Memory and context can fill in missing information, shaping our interpretations in predictable ways.
  • Experiments in perception use descriptive stats and paired-samples t-tests to quantify the influence of context.
  • Data visualization (line graphs of mean error across angles) helps interpret how perception changes with angle and condition.
  • Homework and group work are designed to reinforce statistical literacy, data handling, and communication of results with appropriate formatting and reporting conventions.