Word Perception Stages

• Communication is not always language-based.

Stages in Word Perception

• Features: The Visual Search task demonstrates the importance of features in word perception.
• A limited subset of the alphabet is used.
• Preattentive processing uses distinct features such as O, S, U vs. N, M, X
• Letter: Familiarity and extensive perceptual experience lead to automaticity in letter recognition.
• Consistent mapping also supports automaticity of letter recognition
• Word: Words can be perceived as units based on their shape.
• Word shape uses ascending letters (h, b), descenders (p, g), and half-line letters (e, r).
• Word recognition is holistic (anl vs. and).

Top-Down Processing: Context & Redundancy

• Context is important; "Guess" is easier to perceive given the context of what has appeared before.
• An example is the phrase 'Move the lever to the rxxxx.'.
• Letter recognition relies on knowledge of surrounding letters and examples such as CAT and THE.
• Redundancy reduces the top-down contribution to pattern recognition.
• Tachistoscopic word recognition shows that when contextual constraints are reduced, bottom-up processing becomes more important (Tulving et al., 1964).
• The Word Superiority Effect (Reicher, 1969) posits that letters within a word mutually facilitate one another's analysis versus unrelated letters supporting automaticity.

Tachistoscopic Word Recognition Demo

• Prior word context (redundancy) results in higher percent correct.
• A longer flash duration (stimulus quality) results in a higher percent correct.
• With 8 words the slope is shallow, but without context, the slope is steep because of a lack of context, and the stimulus quality is more important.

Reading: From Words to Sentences

• Eye anatomy constrains the perceptual span to approximately three words from the fixation point
• The perceptual span is 3-4 letter spaces to the left and 14 letter spaces to the right of an advanced reader's fixation point (Rayner 1986).
• Four propositions are carried from a previous sentence.

Applications of Unitization

• Unitization is used in warning signs, maintenance and instruction manuals, and system displays.
• Automatic processing (based on training) is better than increasing physical intensity.
• Automatic processing is less distracting and personalized to alert only those whom the alert is relevant.
• It is compatible with the visual representation of the unit in memory.
• Font use differs between sentences and isolated words, with sentences using lowercase letters and words using uppercase letters
• Truncated abbreviations are preferred over contracted abbreviations.

Context-Data Trade-offs

• Bottom-up quality and top-down quality are balanced.

Code Design: Economy vs. Security

• Greater p(x) (less information) is transmitted with less sensory evidence.
• Frequent signals lower beta and can be detected even at lower sensitivity (lower d').
• High-probability, low information messages should be short, while low-probability ones should be longer.
• Shannon-Fano principle states that message length is proportional to information content (Sheridan & Ferrell, 1974).
• In natural languages (Zipf's law), frequent words are shorter, while rare words are longer.
• Violating those principles may produces errors in copying, memory, and reading

Code Design Sacrifices

• Economy can be sacrificed by including redundancy in high-frequency messages.
• Examples of redundancy are: alpha, bravo, charlie, etc. for a, b, and c in voice communications.
• Redundacy can ensure absolute security by preventing information loss.

Recognition of Objects

• Geons (Biederman, 1987) are a small number of basic features consisting of simple geometric solids (ions).
• Context is important.
• Objects are detected well at peripheral vision if context-appropriate.
• If the object isn't appropriate, then performance rapidly declined with an increased visual angle from fixation
• Simple pictures of objects are understood as rapidly as words (Potter & Faulconer, 1975).
• Cultural features (man-made objects) act as landmarks, with advantages in scene recognition due to greater familiarity.
• Recognition is viewpoint independent (Biederman, 1999).

Pictures & Icons

• Icons or pictorial symbols can be recognized as rapidly as words.
• Legibility is not always viewed under ideal conditions.
• Discriminability is based on global shape, not fine detailed features.
• Interpretation of symbols involves two separate questions as to what is depicted and what it represents
• Examples of symbols are earcons or musicons.
• Redundant verbal labels increase costs with space issues and clutter.

Comprehension

• Word strings or sentences are comprehended.
• Poorly written instructions can decrease the rate of comprehension
• Commonsense principles which increase the rate of comprehension include:
• Stating directly without excess words
• Using familiar vocabulary
• Avoiding pronouns such as “it” and “this”
• Numbering or separating different points
• Highlighting key points

Context for Comprehension

• Top-down processing affects comprehension in two ways:
• Influencing probability on response bias
• Influencing the context on information
• Effects of the context on comprehensibility and recall is important and can be affected by showing a picture or including a thematic title
• Organizing material before the verbal presentation improves comprehension and retrieval,
• Preparation is key for organizing or storing the material.

Command vs. Status

• "Your speed is too high" is a status compared to "Lower your speed," which is a command.
• Results are inconsistent from research.
• Command displays are preferable if under high stress and time pressure as there is no extra processing step
• Status displays are preferable under relaxed conditions as they are not fully trusted
• Redundancy can act as a solution
• Commands vs. status should be clear (e.g., verb: You are left vs. Turn left).
• Commands are given verbally, while status is visually presented.

Linguistic Factors

• Logical reversals, negatives, and falsifications act as key linguistic factors
• Positive statements are preferrable to negative - on is better than not off
• Negatives always take longer to process
• True propositions are understood faster than false propositions, if there are no negatives
• Conversely, if there are negatives, false statements are processed faster than true ones due to the Principle of Congruence.

Working Memory

• Instructions should avoid unnecessary information. – Kintch & Van Dijk (1978) showed that the capacity of working memory with text comprehension is around 4 bits
• Avoid ambiguous pronouns.

The Optimal Multimedia Medium

• Pictures or graphics work best for analog spatial relations and complex patterns
• Verbal formatting is better for abstract information
• The Function also determines, factoring in cognitive and thinking style (Riding, R. J.)
• Wholist-Analytic (W-A) and Verbal-Imagery (V-I) are examples of cognitive approaches to be aware of

Redundancy Gain

• Redundant media use in combination (e.g., graphics and text) is benificial
• The Picture provides overall context or frame
• The Text provides details of the procedures or instructions
• Procedures can be shown with pictorial emphasis/redundant print
• Airline safety cards integrate words with diagrams
• The best way to learn assembling a model is in pictorial instructions with redundant text.
• There are advantages for auditory-pictorial combination over text-pictorial combination
• Each channel is supported by separate attentional or cognitive resources (Ch 11)
• Length and difficulty matter.

Realism of Pictorial Material

• More realism is not necessarily better (Spencer, 1988).
• Photographs led to significantly worse performance than schematic line drawings.
• Naive Realism is similar to simulator fidelity.

Successful Product Warnings Need to:

• Be Noticed: Auditory warnings are more noticeable than visual ones, especially if visually encountered, close to power switch, or embedded within operating instructions.
• Be Read: Readability and legibility are key (prioritize between warnings).
• Be Understood: Calibration of the signal word is important.
• Minimize effort to comply cognitively and physically (avoiding uncomfortable or expensive goggles)

Speech Perception

• Speech is frequently transmitted from a less-than-perfect audio transmission (e.g., KLM 747).
• Speech is similar to print text involving top-down and bottom-up processing with a hierarchical structure.
• Speech consists of phonemes, combined into syllables and words.
• Also relies upon Analog-to-Digital conversion.

Speech Perception: Top-Down Processing

• Bottom-up processing is especially difficult because of speech's serial and transient nature.
• The Invariance Problem states that the physical form of a phoneme is highly dependent on the context in which it appears.
• The Segmentation Problem notes a lack of correspondence with the physical speech sound.
• Anna Mary candy lights since Imp Pull lay Things An American delights in simple things as examples where semantic understanding of sentances affect comprehension.
• Context is necessary and the semantic and syntactic level provide that context.

Voice Recognition Research Applications:

• Voice recognition research acts to compensate for bottom-up processing by top-down processing
• This can be achieved by using standardized vocabulary.
• Voice recognition relies on providing redundant “carrier” sentences (“fuel low” < “Your fuel is low”).

Improving Communications

• Face to face improves auditory only communications (multimodality)
• Nonverbal Communications such as:
• Visualizing the mouth (or lipreading)
• Nonverbal cues which decrease problem solving time and increase convo with: pointing, gesturing, facial cues (e.g., puzzled look, nod of acknowledgment, etc.)
• Disambiguity: Face-to-face makes conversation more flexible and less formal
• Shared knowledge of action: seeing the actions can change a mental set of the coworker in the cockpit (e.g., distributed cognition