PERCEPTION AND PATTERN RECOGNITION • Making sense of sensation – Local vs. Global scope – Data-driven (sensory, bottom-up) vs. Concept-driven (knowledge, “top-down”) • Organizing the perceptual world – Gestalt “strategies” of grouping • Recognizing familiar patterns – Changes in performance and process as we practice • Impairments of pattern recognition skill GESTALT “STRATEGIES” OF PATTERN ORGANIZATION PROXIMITY SIMILARITY CONTINUITY CLOSURE ASYMMETRIES IN HEMISPHERIC PROCESSING One hypothesis: Left Hemisphere specialized for “local detail,” fine-grained analysis Right Hemisphere specialized for “global form” and wide scope Damaged LH Damaged RH SOME BASIC FACTS ABOUT HUMAN PATTERN RECOGNITION • vast number of distinct patterns can be learned – e.g., over 60,000 spoken or written words • recognition can with practice by very fast and “automatic” – e.g., Lexical Decision Speed BLACK BLARK ? ~600 msec • can succeed in spite of great variability of input (“noise”) PRACTICE AND PATTERN RECOGNITION SKILL • speed and accuracy improve • requires less attention and effort • becomes more “noise resistant” • “distinctive” features are learned • “prototype” patterns may be learned • larger “units of recognition” emerge • skill, and impairment, are “domainspecific” THE POWER LAW OF PRACTICE Speed and accuracy improve, but at an ever-slower rate ( L O Gm ) i n / p a g e Task: reading inverted text (Kolers, 1975) 19 16 13 10 7 4 1 2 4 8 16 32 64 128 # of pages read (LOG) m in / p a g e 100 10 1 2 4 8 16 32 64 128 # of pages read (LOG) Time = 10 a x practice -b LEARNING DISTINCTIVE FEATURES OF PATTERNS Feature Analysis: Define a small set of features whose presence and arrangement defines the patterns e. g.: consonant phonemes FEATURES Voicing Place of Articulation Voiced Unvoiced Bilabial /b/ /p/ Alveolar /d/ /t/ FEATURES and RELATIONS Recognizing Objects by Components (Biederman’s RBC model) Fig. 3-17, p. 69 ABSTRACTING THE ”TYPICAL” PATTERN (PROTOTYPE) Task: learn to categorize faces: (Reed, ‘72) Category 1 Category 2 Then tested on old and new faces: P2 P1 EVIDENCE FOR PROTOTYPE ABSTRACTION (Reed, 1972) • “Studied” prototypes are classified more quickly and accurately than other studied patterns • Even if prototype had not been studied, – it was still the easiest to classify – and was often falsely identified as “studied” in an old/new decision Reed Biederman CARICATURES: Exaggerating distinctive features Celebrity caricatures at About Faces PRINTED WORDS AS UNITS OF PATTERN RECOGNITION Task: letter detection Reicher (1969): where unitizing helps 30 MSEC WARM OR WBPM OR ###M M #$%& N Then… (1 sec) 70%C 58%C 62%C Johnson (1987): where unitizing hurts “is first letter an “R”? BEAN faster than BEAR but BFXN equal to BFXR STIMULUS FEATURES AND SENTENCE CONTEXT (Rueckl & Oden, 1986) Task: read sentence contexts, The { lion tamer / dairy farmer } raised ____ to supplement his income. s e e in g { b e a r } then. . beans ....... bears 100 80 Dairy Farmer 60 Lion Tamer 40 % 20 0 Stimulus feature (in bea_s) IMPAIRMENTS OF PATTERN RECOGNITION SKILL • Skills, and impairments, tend to be “domain-specific” to codes or modality: – ALEXIA WITHOUT AGRAPHIA • can’t read, but can write – PROSOPAGNOSIA • can’t recognize familiar faces –MOTION AND COLOR AGNOSIA •objects appear still, or “grey” –AMUSIA •can’t recognize/match familiar melodies DEVELOPMENTAL DYSLEXIA • defined as a selective slowness in reading acquisition and speed • estimates range from 2% to 10% of school population • similar numbers of boys and girls • not a problem of visual perception • perceiving and representing rapid sequences of speech sounds • predicted by “phonological awareness” tests • reading and complex phonology remain problems into adulthood