EDPSY 520: Psychology of Reading Virtual Overview and Introduction Welcome This introduction is intended to provide a brief history of the field and familiarize you with major issues in the psychology of reading You will also have the opportunity to participate in some classic experiments in the study of reading processes Click your way through this presentation via the mouse or spacebar; during some of the experiments the computer may take control of the screen pacing Brief History Psychological research on reading was done in the early 1900s by E. B. Huey and others According to Huey (1908): “ . . .to completely analyze what we do when we read would almost be the acme of a psychologist’s achievements, for it would be to describe very many of the most intricate workings of the human mind . . . .” With the rise of Behaviorism in the early part of the last century, especially within American psychology, the focus shifted to studying behaviors that were more observable than reading (more simple learning and more simple learners) As cognitive psychology became increasingly dominant during the latter half of the century, reading (and human thinking in general) again became a phenomenon to study How might we think about reading? Perhaps because of lasting effects of Behaviorism, many early models of reading still had a very “stimulus-driven,” or “text-driven” feel In these early models, the assumption was that information in the text is taken in by the reader Information was seen to flow from the “bottom up” An Early Bottom-Up Model of Reading, LaBerge & Samuels (1972) Letters are perceived as visual patterns, and with learning, come to be recognized as letters Letters are mentally P+A+T PAT combined and recognized as words Words are successively PAT THE DOG ON recognized as phrases, THE HEAD. sentences, etc. A Not the whole story To test such a text-driven model, you can now be a subject in an experiment that Edmund Burke Huey performed 100 years ago (with slightly different technology) When you click the mouse or spacebar, you will see an asterisk ( * ) appear on the screen to orient you. A vertical list of items (letters or words) will then appear under the asterisk See how many of the items you can read in the time you are given (if possible, read out loud, in a rapid but comfortable pace) * * y w u s q o m k i g e c Did you finish the whole list? Did you have time left over? Click when you are ready to start the next list. * * pool rugs mark send list more pick stab neck your dice font Did you finish the whole list? Did you have time left over? Click when you are ready to start the next list. * * analysis habitual occupied inherent probable summoned devotion remarked overcome resolute elements conclude Did you finish the whole list? If you are like most adult readers, you had time to read all the items on all the lists Click to advance each screen again What does this demonstrate? We do not read letter by letter, building up words from individual letters The lists contained successively more letters List 1: 1 letter per line List 2: 4 letters per line List 3: 8 letters per line However, the lists were not displayed for proportionately longer times All lists were displayed for 8 seconds You just replicated classic findings You read the short words (4 letter words in List 2) just as quickly as you read single letters (List 1); both were displayed for only 8 seconds Even though List 3 contained eight times more letters than List 1, you did not need eight times longer to read it In studies with more precise timing, people even read short words more quickly than single letters, a phenomenon that has come to be called the “word superiority effect” Conclusions Your reading times just demonstrated that not all information flows from the text to the reader; not all information flows from the bottom up. Alternative models of reading Theoretical models emphasizing the expectations and prediction abilities of readers have come to be known as “top-down models” Top-down models Perhaps the purest of top-down models was offered by Ken Goodman, based on his analysis of readers’ errors as they read aloud, which he called “miscues” Goodman noted that when they misread a word, good readers are more likely than poor readers to substitute a word that makes sense in the sentence context Examples of reading errors, or “miscues” When reading the sentence, “I walked up the sidewalk, across the porch, and knocked on the door of the house” Good readers may misread the word house as home (but note how the sentence still makes sense) Poor readers are more likely to misread the word house as horse, or how, or even as a nonsense word like “hoose” (rhyming with choose) Other evidence of top-down processing Semantic priming effects We read the word doctor more quickly following the word nurse, chair more quickly following the word table, compared to when these words follow words with which they share no meaningful relationship Other evidence of top-down processing Prior knowledge effects When interpreting sentences containing ambiguous words, such as “ When Jerry, Mike, and Pat arrived, Karen was sitting in her living room writing some notes” College students majoring in music interpreted the word notes as musical notes Other students interpreted the words notes as brief letters or reminders to oneself Not the whole story either Of course, top-down processing can’t be the whole story either because we do read, not simply daydream in front of books The context that drives expectations comes from somewhere Interactive Models of Reading Most reading researchers now adopt a theoretical model that includes interaction from both top-down and bottom-up processes Implications of an Interactive Model Interactive models are not a simple compromise, not a little of both Interactive reading processes require complex coordination of multiple sources of information The text on the page Context from previous text Prior knowledge Reading goals Implications of an Interactive Model Where do we store all these sources of information? How do we coordinate multiple information sources, accessing them seemingly simultaneously? Unlimited Long-term Memory (LTM) Every kind of knowledge we use as part of reading must be stored in our long-term memory (LTM) Letters Words Word meanings Memory for general story structures Memory for a specific text Prior knowledge of the topic Limited Short-term Memory (STM) and Attention The amount of information we can hold in immediate STM at any given time is limited The storage capacity of STM has been estimated between 5 and 9 (7+2), but there are strategies to increase STM Still, STM limits generally exist; to convince yourself, try this multiplication problem in your head: 1683945 x 6939 Implications of an Interactive Model During reading, attention can’t be everywhere at once, even in an interactive system If attention is required to identify individual letters, then less attention is available to devote to recognizing words If attention is required to recognize words, then less attention is available to devote to building up story context How do Interactions Work? The key to getting around limits on attention and STM is to have some processes (such as letter and word recognition) go on automatically Automaticity is the underlying cause of the word superiority effect you demonstrated for yourself some while back How do Automatic Processes Work? With the next few slides you can take part in another classic reading experiment Click to begin Name the color of each letter cluster Read the following clusters from left to right, as though you were reading connected sentences. Remember to name the color of the ink. Click to begin Name the color of each letter cluster Name the color of each letter cluster XXXXX XXXX XXX XXXXX XXXXX XXX XXXX XXXX XXX XXXX XXXX XXXXX XXX XXXX XXX XXXXX XXX XXXXX XXXX XXX Did you finish the whole set? Did you have time left over? Click when you are ready to start the next set. Name the color of each letter cluster Read the following clusters from left to right, as though you were reading connected sentences. Remember to name the color of the ink. Click to begin Name the color of each letter cluster Name the color of each letter cluster GREEN BLUE RED GREEN GREEN RED BLUE BLUE RED BLUE BLUE GREEN RED BLUE RED GREEN RED GREEN BLUE RED Did you finish the whole set? If you are like most adult readers, you named the color of ink far more easily in the first set than the second In the second list, you may have experienced strong interference because you wanted to say aloud the word that you saw, rather than name the color of the ink in which the words were printed What does this demonstrate? Your word reading processes are automatic The good news: Automatic processes run without conscious attention or effort The bad news: Automatic processes are also hard to “turn off” You just replicated another classic study in reading research, illustrating the “Stroop effect” (named for the experimenter who first reported it) Implications for an Interactive Model If some of the multiple processes we are trying to coordinate are automatic, they can operate without tying up much attention or STM capacity Thus, automatic lower level (bottom up) process such as letter and word identification can free up limited STM capacity for higher level (top down) processes, such as comprehension, interpretation, etc. Reading Problems What are the most common sources of reading problems? Are less skilled readers poorer at bottom up processes such as word identification? Or, are less skilled readers necessarily poorer at using context (as Goodman suggested)? Another classic experiment in reading Perfetti & Roth examined the use of context by skilled and less skilled readers in Grade 4 Children listened to short stories and were timed as they read aloud just the final word Thus, the beginning of each story provided context for the final word, which was the only word that children had to read However, the predictive quality of the context provided by the story varied Perfetti and Roth’s study Some contexts were highly predictable Children heard: The garbage men had loaded as much as they could onto the truck. They would have to drop off a load at the garbage _________. Children read: dump Perfetti and Roth’s study Some contexts were less predictable Children heard: Albert didn’t have the money he needed to buy the part to fix his car. Luckily he found the part he wanted at the _________. Children read: dump Perfetti and Roth’s study Some contexts were anomalous (misleading) Children heard: Phil couldn’t decide whether to go to the movies or to the party. Both sounded like lots of fun, but he finally decided to go to the _________. Children read: dump How do skilled readers use context? Perfetti & Roth measured in milliseconds how long children in the two groups needed to read aloud the final word of each story If skilled readers use context to help identify words, what do you predict will happen to their reading times as they go from highly predictable to less predictable contexts? What about as they go to anomalous contexts? Click to see their actual results Time to Read Final Word Aloud 1000 950 900 850 800 750 700 Skilled 4th grade readers 650 600 550 500 High Less predictable Anomalous Predictability of Context What do these results tell us? Skilled readers use context Skilled readers read words fastest when the context was highly predictive When contexts were less predictive or anomalous, skilled readers took almost 100 milliseconds longer to read the same word How do less-skilled readers use context? If less skilled readers do not use context to help identify words, what do you predict will happen to their reading times as they go from highly predictable to less predictable contexts? What about as they go to anomalous contexts? Click to see their actual results Time to Read Word Aloud 1000 950 900 850 800 750 700 Skilled 4th grade readers 650 600 550 500 High Less predictable Anomalous Predictability of Context Time to Read Word Aloud 1000 950 Less skilled 4th grade readers 900 850 800 750 700 Skilled 4th grade readers 650 600 550 500 High Less predictable Anomalous Predictability of Context What do these results tell us? Less skilled readers use context too! Less skilled readers read words fastest when the context was highly predictive When contexts were less predictive they took almost 150 milliseconds longer to read the same word When contexts were anomalous, they took another 150 milliseconds longer to read the same word Who uses context more? Less skilled readers were even more sensitive to context than skilled readers The slope of the less skilled readers’ reading times was steeper as context became less predictive Less skilled readers were especially hesitant to read a word when it violated the context Who uses context more? This does not mean that skilled readers don’t use context In all but the most predictable contexts, skilled readers identify words so fast that they don’t need to use context Who uses context more? How do these results fit with Goodman’s observations about poor readers not using context as they read? In Perfetti and Roth’s study, the less skilled readers did not have to read the context; they got it “for free” by listening The problem for less skilled readers is keeping context in mind when they have to read it for themselves Implications of an Interactive Model During reading, attention can’t be everywhere at once, even in an interactive system If less skilled readers’ attention is required to identify individual words, then less attention is available to devote to building up story context Less skilled readers are in a double bind; they rely on context more than skilled readers, but can’t maintain it while reading The story is more complicated Reading is a complex interaction between bottom up and top down processes The experiments you have reviewed here do not tell the complete story Not all readers have difficulties for exactly the same reasons Readers differ in terms of the processing skill and the knowledge they bring to reading More to come Much of our work for the rest of the term will examine: On what knowledge does reading build How do knowledge and processes interact How do skilled and less skilled readers differ See you at our first meeting