<<Back to Cognition Class Home Page
---------------------------------------------------------------------------------------
LEXICAL DECISION AND SPREADING ACTIVATIONLexical decision is a very widely used task to study language and semantic memory. In this task, the experimenter shows a series of letter strings. On each trial, the subject must look at the letter string and decide whether the letters form a word or not. On any given trial the letter string might either be a true word (MOTOR) or it might be a nonword or pseudoword (MANTY). Subjects are asked to respond rapidly but accurately and the reaction time to each letter string is the main performance measure. Usually the nonwords used in lexical decision are "orthographically correct", that is, they respond to orthographic rule and therefore are possible words, so that people cannot rely on other features to make a response (what would happen if we used consonant string instead?). To correctly respond in a lexical decision task, participants need to access their lexicon or mental dictionary and decide whether the letter string belongs to it. The mental lexicon is our knowledge of words, their name, meanings, and even their pronunciation.
Which mental operations do we use to perform this task? The first step is to encode the stimulus, taking in the visually presented letter string and transferring it to the holding mechanism of short-term or working memory. The central executive system in working memory contacts long-term memory to determine whether that item is stored in the mental lexicon. This is accomplished through a search process that ends when the item is found or after a certain time if the item is not found. The outcome of this memory search is the basis of the individual's decision (yes or not) and consequent motor response.
What do you predict about reaction times? Do you think it will be faster to respond yes or no?
One interesting result in this task is that reaction times are a function of the frequency of usage of the word (WORD FREQUENCY EFFECT). Words that are often used are responded to faster than words that are less used. Which stage of processing is likely to be influenced by word frequency? Search in long term memory seems to be the most likely candidate. This type of evidence has been used to support the idea that word frequency is "stored" in some way in our lexicon. It is possible that more frequent words have a "stronger" memory trace, and therefore can be retrieved faster. Alternatively, frequent words may be encountered earlier, in average, when we search the lexicon (using the "search" metaphor, you can think of it like starting your search from a particular starting point, where more frequent words are stored, and then move to different places where less frequent words are stored.)
The lexical decision task has been used to study priming. In the priming paradigm two strings of letters are presented in sequence in each trial. The first stimulus is called prime and the second stimulus is called target. When both strings are words, they may be related (BREAD-BUTTER) or unrelated (OFFICE-BUTTER). Mayer and Schvaneveldt (1971) showed that lexical decision times were faster when two related words were presented than when two unrelated words were presented. What is the theoretical consequence of this result? Theoretically, in a lexical decision task it is possible to respond without accessing the meaning of the word (we just have to decide whether the stimulus IS a word). However, priming studies showed that we do access the meaning, and that presenting two words which meaning is associated speed up the lexical decision. In some way this result is similar to the Stroop effect: the meaning of a word is accessed automatically, even if it is not necessary to perform the task, and influences task performance.
Another impressive result is that priming effects also occur when we are not aware that a word has been presented, something similar to subliminal perception. Marcel (1980) used a mask to stop processing of the prime word. The use of a scrambled visual pattern is often used to stop visual processing of a previous presented stimulus. If I present a stimulus briefly, we can continue to visually process it even if it is not physically present. Which kind of memory are we using in this case? The stimulus persists in the sensory storage for a certain time, usually sufficient or process the word. However, this visual processing without the stimulus being physically present is blocked by the presentation of a second visual stimulus that "masks" the visual features of the first (usually the mask is a scrambled visual configuration). If the mask is used after a very brief presentation of the first stimulus, we may not have any conscious awareness that something has been presented. Marcel showed that when the prime was masked and not consciously perceived, it could still produce facilitation in the processing of a related word. This result shows that even a brief and nonconscious presentation of a word will automatically activate its meaning in the mental lexicon and facilitate processing of related words.
Another researcher that explored the priming phenomenon was Neely. Neely demonstrated the very important finding that presenting a word not only facilitates processing of associated words but may also make more difficult processing unrelated words. In one experiment, Neely used related pairs (the target word was a strong associate of the prime), unrelated pairs (the target word was semantically not related to the prime) and neutral pairs (the prime stimulus was a string of Xs). He found, as Mayer and Schvaneveldt, that related primes facilitate lexical decision on targets as compared to the neutral condition. He also found that unrelated primes produced a small cost in the reaction time to the target as compared to neutral primes.
Does this experimental paradigm remind you of another paradigm that we studied? Well, this task has a very similar structure to the spatial cueing paradigm that we studied in spatial attention. We have a valid condition (related prime) and invalid condition (unrelated prime) and a neutral condition. The results are also very similar: Neely found benefit in the related prime condition and costs in the unrelated prime condition.
Neely also studied the effect of expectancy, a conscious strategic factor in a lexical decision task. In this experiment, Neely told subjects what kind of targets had to be expected after each prime. For example, if the prime was BIRD, the target would be probably a type of bird. However, if the prime was BUILDING, the participant should have expected a body part and if the prime was BODY, the participant should have expected a building part. These instructions were often correct but in a small percentage of the trials the opposite occurred: After BIRD a non-bird was presented and after BUILDING a building, instead of a body part, was presented.
Also this experiment is very similar to the cueing paradigm, except that instead of cueing spatial positions Neely cued semantic categories. The interesting condition is the "category shift" condition, in which subjects expected to see not a member of prime category but a member of a different category (BODY-window). In this condition there are two different tendencies in contrast: one is the automatic activation of the prime category (BODY -> arm, leg, heart) and one is the strategic activation of a different category due to the expectancy (BODY-building parts). The results showed that when no category shift was expected (BIRD-robin) and the expected prime was presented, facilitation was observed. When the category shift was not expected by occurred (BIRD-arm) there was inhibition. The interesting aspect of these results is the time course of these effects. When we vary the time interval between prime and target (SOAs = stimulus onset asynchrony), facilitation and inhibition effects change. In the "no shift condition", facilitation stays pretty much the same whereas inhibition increases as a function of SOA. What happens in the category shift condition? When a category shift is expected and it does occur (BODY-window) there is no facilitation at short SOAs but an increasing facilitation for longer SOA. When the category shift is expected and it does not occur (BODY-heart) there is a small facilitation at short SOAs and an increasing inhibition at longer SOAs.
How can we interpret these results? Neely proposed that there are two processes that operates in this task. One is a fast and automatic process of facilitation that occur for semantically related category and it is not influenced by strategic factors. This process is responsible for the early facilitation observed in the category shift condition when semantically associated but not expected targets are presented, and in the no category shift condition when semantically associated and expected targets are presented. The second is a slower process that is controlled by strategic factors and is responsible for the slower building of inhibition for the non expected outcomes (whether semantically related or not) and the facilitation of the expected but not semantically related outcome in the category shift condition.
The first process corresponds to what is usually called "spreading activation": the activation of a representation immediately creates activation in related representations so that this activation "spread" through the semantic network. The second process corresponds to the effect of controlled/strategic mechanisms that allow us to have a flexible behavior controlled by our intentions and expectations.
-------------------------------------------------------------------
READING
DIFFERENCES BETWEEN READING AND LISTENING TO SPEECH
There are several differences between reading and listening to spoken language.
1. Writing is spread out across space, whereas speech is spread out across time.
2. Readers can control the rate of input, whereas listeners usually cannot.
3. Readers can re-scan the written input, whereas listeners must rely much more heavily on their working memory.
4. Writing shows discrete boundaries between words, whereas speech does not.
For example, try to read this:
INWRITINGWEUSEDISCRETEBOUNDARIESBETWEENWORDS
5. Writing is confined to the words on a page, whereas speech is supplemented by additional auditory cues (such as stressed words and variation in pace) that enrich the linguistic message.
HOW DO WE STUDY READING
It makes sense to study reading in an obvious way: asking people to read a passage of text and then asking them to recall what they read. However, does this task reveal something about the process of reading? Does it tell us something about what happens DURING reading?
Researchers have chosen other methods that allow us to analyze reading on line. One of these methods is the study of eye movements.
EYE MOVEMENTS
A very important experimental methodology in studying reading processes is the study of eye movements.
The way our eyes move when we read is not smooth and continuous. Our eyes move in a series of little jumps called saccades. As we know, there is a small region of our visual field where our acuity is very high, the fovea. A saccade allows one to bring words in the fovea during reading. Saccadic movements are very fast and are followed by a fixation: during fixation the eyes do not move and the visual system acquires the information that is used to read.
In an eye movement study, subjects are asked to read a passage silently on a computer monitor and to prepare for some sort of comprehension test after they have finished reading. The computer monitor permits the investigator to control exactly what visual stimulus is presented, its duration, and other variables. During reading, a camera records eye movement and it is usually interfaced with the computer, so that what it is displayed on the monitor depends on the subject's eye position. The most important dependent variable in this task is the duration of eye fixations.
To illustrate some of the uses of these techniques, we can examine studies that have analyzed how many letters we can process in a single fixation. One technique to study this particular aspect of reading is the moving window technique. In this experimental paradigm, people read text on a computer screen and their eye movements are controlled. The text on the screen changes as the person is reading. Usually, the text around the fixation point is normal, but at same distance on the left and on the right the text is replaced by random letters and spaces. Experimenters manipulate the distance between fixation and the random letters. The main result of these experiments is that reading is influenced when the distance between fixation and random letters is less than 4 position to the left and 15 position to the right. This region of the visual space is called "perceptual span" and indicates how many letters we can perceive in one fixation (this is sort of similar to the concept of "span of apperception" that we studied in perception). The "asymmetry" of the perceptual span depends on the fact that we read left to right, and therefore information on the right is very important, especially in guiding a new saccade.
Eye movements' studies are based on two assumptions:
(1) the immediacy assumption states that we process the text while we are reading it. We already encountered this assumption talking about Fillmore's case grammar and we also showed that this assumption is supported by the study of garden-path sentences such as, for example:
Although he spoke softly, yesterday's speaker could hear the little boy's question.
(2) The eye-mind assumption states that the eyes remain fixated on a word as long as that word is being actively processed during reading. The alternative hypothesis is that we gather information during reading and we store this information in a buffer and we process the entire sentence at the end. In this case, which pattern of fixation times should we observe?
Just and Carpenter (1980) studied eye movements during reading. In class I presented a figure that shows number and duration of fixations during reading. What can you observe? First, saccadic movements during reading are very precise. A saccade tends always to land in the middle of a word. Second, unusual words are fixated a lot longer than frequent words. Third, short words like function words (a, of, the, into) are often skipped or receive very short fixation, whereas content words always receive a (longer) fixation. As you can notice, this experimental paradigm gives us a lot of information about the reading process on line.
Interestingly, saccadic pattern can discriminate good readers from poor readers. (Figure p. 292). Reading patterns of good and poor readers are very different. Good readers are faster and have a better comprehension of the text than poor reader. If we look at their reading patterns we can observe that good readers make longer saccades and fewer regressions (backward saccades that go back to material already fixated). Good readers are also faster: each fixation lasts less.
The duration of fixations is very variable and according to Just and Carpenter can allow us to understand the reading process in depth. In particular, Just and Carpenter proposed that during reading we assign case roles (Fillmore's semantic roles) and that eye movements during reading can illustrate this process. For example, consider these two sentences:
(1) The tenant complained to his landlord about the leaky roof. The next day, he went to the attic to get his luggage.
(2) The tenant complained to his landlord about the leaky roof. The next day, he went to the attic to repair the damage.
Just and Carpenter observed that in sentence 1 the readers' eye moved along until the word "luggage" was encountered and then "bounced" back to the word "tenant". In sentence 2 the same thing happened when "repair the damage" was encountered, and the eyes bounced back to the word "landlord". These eye movements, according to the eye-mind assumption, illustrated the underlying mental processes of finding antecedents and determining case roles.
---------------------------------------------------------------------------------------