PSY 369: Psycholinguistics Language Comprehension

Some of the big questions Production “the horse raced past the barn” How do we turn our thoughts into a spoken or written output?

Some of the big questions Production Comprehension “the horse raced past the barn” How do we turn our thoughts into a spoken or written output? How do we understand language that we hear/see?

Some of the big questions Comprehension Production Representation How do we store linguistic information? How do we retrieve that information? Lexicon Semantic Analysis Syntactic Word Recognition Letter/phoneme Formulator Grammatical Encoding Phonological Encoding Articulator Conceptualizer Thought

Conceptualizer Formulator Lexicon Articulator Thought Semantic Analysis Syntactic Word Recognition Letter/phoneme Formulator Grammatical Encoding Phonological Encoding Articulator Conceptualizer Thought

Overview of comprehension Language perception c a t /k/ /ae/ /t/ cat dog cap wolf tree yarn claw fur hat Word recognition Syntactic analysis cat S VP rat the NP chased V Semantic & pragmatic analysis The cat chased the rat. Input

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning

The Comprehender’s Problem Different signals Reading and listening are very different If reading were like listening whereareyougoing Where are you going? Different speakers speak differently Lots of differences in written/printed language

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning The cat chased the rat. The cat chased the rat. The cat chased the rat. The cat chased the rat.

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning Oronyms I scream for ice scream The stuffy nose can lead to problems. The stuff he knows can lead to problems. Why don’t you take a nice cold shower? Why don’t you take an ice cold shower? See here for more oronyms

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning Groucho Marx shot an elephant in his pajamas Good shot How he got into my pajamas I’ll never know

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning “Oh no, Lois has been hypnotized and is jumping off the bank!” Money “bank” River “bank”

The Comprehender’s Problem Ambiguity Must take a potentially ambiguous serial acoustic (or visual) input, and recover the intended meaning Uncle Bob “Uncle Bob kicked the bucket last night” “Sure as soon as I’m done using it.” “Can you pass the salt” “Nope, somebody glued it to the table.”

Overview of comprehension Lexical Access Language perception c a t /k/ /ae/ /t/ cat dog cap wolf tree yarn claw fur hat Word recognition Syntactic analysis cat S VP rat the NP chased V Semantic & pragmatic analysis The cat chased the rat. Input

Lexical access How do we retrieve the linguistic information from Long-term memory? How is the information organized/stored? What factors are involved in retrieving information from the lexicon? Models of lexical access

Lexical access How do we retrieve the linguistic information from Long-term memory? How is the information organized/stored? What factors are involved in retrieving information from the lexicon? Models of lexical access

Storing linguistic information Tale of the tape: High capacity: 40,000 – 60,000 words Fast: Recognition in as little as 200ms (often before word ends) How do we search that many, that fast!? – suggests that there is a high amount of organization Or something much more complex “The world’s largest data bank of examples in context is dwarfed by the collection we all carry around subconsciously in our heads.” E. Lenneberg (1967) Excellent reading: Words in the Mind, Aitchison (1987, 2003)

Storing linguistic information Interesting questions: How are words stored? What are they made up of? How are words related to each other? How do we use them? Some vocabulary Mental lexicon The representation of words in long term memory Lexical Access: How do we access words and their the meanings (and other properties)?

Theoretical Metaphors: Access vs. Recognition Often used interchangeably, but sometimes a distinction is made Recognition - finding the representation Here it is dog cap wolf tree yarn cat claw fur hat Search for a match cat Select word cat The magic moment Balota (1990) cat cat

Theoretical Metaphors: Access vs. Recognition Often used interchangeably, but sometimes a distinction is made Recognition - finding the representation Access - getting information from the representation dog cap wolf tree yarn cat claw fur hat Search for a match Select word Access lexical information Cat noun Animal, pet, Meows, furry, Purrs, etc. Open it up and see what’s inside cat cat

Lexical access How do we retrieve the linguistic information from Long-term memory? How is the information organized/stored? What factors are involved in retrieving information from the lexicon? Models of lexical access

Studying Lexical Access Generally people ask: what makes word identification easy or difficult? The assumption: Measures of identification time are usually indirect Time spent identifying a word can be a measure of difficulty

Common methodologies Measure how long people take to say a string of letters is (or is not) a word (lexical decision) Measure how long people take to categorize a word (“apple” is a fruit) Measure how long people take to start saying a word (naming or pronunciation time) Measure how long people actually spend looking at a word when reading Word by word reading Line by line reading Using eye movement monitoring techniques

Factors affecting lexical access Morphological structure Role of prior context Phonological structure Lexical ambiguity Concretness/abstractness Grammatical class Imageability Frequency Semantic priming Some of these may reflect the structure of the lexicon Some may reflect the processes of access from the lexicon

Words or morphemes? Word primitives Morpheme primitives horse horses barn barns Need a lot of representations Fast retrieval Morpheme primitives horse -s barn Economical - fewer representations Slow retrieval - some assembly required Decomposition during comprehension Composition during production

Words or morphemes? Lexical Decision task (e.g., Taft, 1981) See a string of letters As fast as you can determine if it is a real English word or not “yes” if it is “no” if it isn’t Typically speed and accuracy are the dependent measures

table

vanue

daughter

tasp

cofef

hunter

Words or morphemes? table Yes vanue No daughter Yes tasp No cofef No Lexical Decision task table Yes vanue No daughter Yes tasp No cofef No hunter Yes

Words or morphemes? Lexical Decision task daughter hunter

Words or morphemes? daughter Pseudo-suffixed daught -er hunter Lexical Decision task This evidence supports the morphemes as primitives view daughter Pseudo-suffixed daught -er Furthermore, a word like indecision (which has three morphemes) takes longer to process than a word like deciding (two morphemes) hunter Multimorphemic Takes longer hunt -er

Words or morphemes? May depend on other factors What kind of morpheme Inflectional (e.g., singular/plural, past/present tense) Derivational (e.g., drink --> drinkable, infect --> disinfect) Frequency of usage High frequency multimorphemic (in particular if derivational morphology) may get represented as a single unit e.g., impossible vs. imperceptible Compound words Semantically transparent Buttonhole Semantically opaque butterfly Try to strike a balance between cognitive economy of number of representations and how much assembling needs to be done

Phonology What word means to formally renounce the throne? abdicate Words that sound alike may be stored “close together” Brown and McNeill (1966) Tip of the tongue phenomenon (TOT) What word means to formally renounce the throne? abdicate Look at what words they think of but aren’t right e.g, “abstract,” “abide,” “truncate”

Phonology Words that sound alike may be stored “close together” Brown and McNeill (1966) Tip of the tongue phenomenon (TOT) Letters at Word beginning Word end 10 2 3 1 20 30 40 50 % of matches Similar-meaning words Similar-sounding words More likely to approximate target words with similar sounding words than similar meanings The “Bathtub Effect” - Sounds at the beginnings and ends of words are remembered best (Aitchison, 2003)

Imageability Imageability, concreteness, abstractness Try to imagine each word Umbrella Lantern Freedom Apple Knowledge Evil

Imageability Imageability, concreteness, abstractness Try to imagine each word Umbrella Lantern Freedom Apple Knowledge Evil How do you imagine these?

Imageability Imageability, concreteness, abstractness Umbrella Lantern Freedom Apple Knowledge Evil More easily remembered More easily accessed Actually interacts with frequency in interesting ways Bleasdale (1987) found that concrete primed concrete and abstract primed abstract, but not cross category priming

Frequency Lexical Decision Task: Gambastya Revery Voitle Chard Wefe Cratily Decoy Puldow Raflot Oriole Vuluble Chalt Awry Signet Trave Crock Cryptic Ewe Mulvow Governor Bless Tuglety Gare Relief Ruftily History Pindle Develop Gardot Busy Effort Garvola Match Sard Pleasant Coin

Frequency Typically the more common a word, the faster (and more accurately) it is named and recognized Typical interpretation: easier to access (or recognize) Lexical Decision Task: Low frequency High(er) frequency Gambastya Revery Voitle Chard Wefe Cratily Decoy Puldow Raflot Oriole Vuluble Chalt Awry Signet Trave Crock Cryptic Ewe Mulvow Governor Bless Tuglety Gare Relief Ruftily History Pindle Develop Gardot Busy Effort Garvola Match Sard Pleasant Coin

Frequency Typically the more common a word, the faster (and more accurately) it is named and recognized Typical interpretation: easier to access (or recognize) However, Balota and Chumbley (1984) Frequency effects depend on task Lexical decision - big effect Naming - small effect Category verification - no effect A canary is a bird. T/F

Semantics Free associations Semantic Priming task Most associates are semantically related (rather than phonologically for example) Semantic Priming task Meyer & Schvaneveldt (1971) For the following letter strings, decide whether it is or is not an English word Tasp Nurse Doctor Fract Slithest Shoes no yes nurse shoes Responded to faster Related Unrelated “Priming effect” Evidence that associative relations influence lexical access doctor 940 msecs 855 msecs

Role of prior context Cross Modal Priming Task: Listen to short paragraph. At some point during the paragraph a string of letters will appear on the screen. Decide if it is an English word or not. Say ‘yes’ or ‘no’ as quickly as you can. Hear: “Rumor had it that, for years, the government bulding has been plagued with problems. The man was not surprised when he found several spiders, roaches and other bugs in the corner of his room.”

ant Role of prior context “Rumor had it that, for years, the government building has been plagued with problems. The man was not surprised when he found several spiders, roaches and other bugs in the corner of his room.” ant Hear: “Rumor had it that, for years, the government building has been plagued with problems. The man was not surprised when he found several spiders, roaches and other bugs in the corner of his room.”

Role of prior context Swinney (1979) Lexical Decision task Context related: ant Context inappropriate: spy Context unrelated: sew Results and conclusions Within 400 msecs of hearing "bugs", both ant and spy are primed After 700 msecs, only ant is primed “Rumor had it that, for years, the government building has been plagued with problems. The man was not surprised when he found several spiders, roaches and other bugs in the corner of his room.” Tabbossi (1988) found that this effect depends on which meaning is the dominant meaning of ambiguous word. Using a similar task, if the context strongly biases the dominant meaning of the ambiguous word, then only the dominant word meaning is found activated.

Lexical ambiguity Hogaboam and Pefetti (1975) Words can have multiple interpretations The role of frequency of meaning Task, is the last word ambiguous? The jealous husband read the letter (dominant meaning) The antique typewriter was missing a letter (subordinate meaning) Results: Participants are faster on the second sentence. The results may seem counterintuitive The task is the key, “is the final word ambiguous” In the first sentence, the meaning is dominant and the context strongly biases that meaning. So the second meaning may not be around, which in turn makes the it harder to make the ambiguity judgment in the first sentence Things are a little complex. Here the results may seem counterintuitive (or at least at odds with the previous results). The task is the key, “is the final word ambiguous” In the first sentence, the meaning is dominant and the context strongly biases that meaning. So the second meaning may not be around, which in turn makes this the harder sentence to make the ambiguity judgment.

Lexical access How do we retrieve the linguistic information from Long-term memory? How is the information organized/stored? What factors are involved in retrieving information from the lexicon? Models of lexical access

Models of lexical access Serial comparison models Search model (Forster, 1976, 1979, 1987, 1989) Parallel comparison models Logogen model (Morton, 1969) Cohort model (Marslen-Wilson, 1987, 1990) Connectionist models Interactive Activation Model (McClelland and Rumelhart, 1981)

Logogen model (Morton 1969) Auditory stimuli Visual stimuli Auditory analysis Visual analysis Context system Semantic Attributes Logogen system Available Responses Output buffer Responses

Logogen model The lexical entry for each word comes with a logogen The lexical entry only becomes available once the logogen ‘fires’ When does a logogen fire? When you read/hear the word

When the bell rings, the logogen has ‘fired’ Think of a logogen as being like a ‘strength-o-meter’ at a fairground When the bell rings, the logogen has ‘fired’

‘cat’ [kæt] What makes the logogen fire? seeing/hearing the word What happens once the logogen has fired? access to lexical entry!

So how does this help us to explain the frequency effect? ‘cat’ [kæt] ‘cot’ [kot] Low freq takes longer So how does this help us to explain the frequency effect? High frequency words have a lower threshold for firing e.g., cat vs. cot

nurse doctor ‘doctor’ [doktə] ‘nurse’ [nə:s] Spreading activation from doctor lowers the threshold for nurse to fire Spreading activation network So nurse take less time to fire doctor nurse nurse doctor

Search model Visual input Auditory input Mental lexicon /kat/ Pointers Access codes Entries in order of Decreasing frequency cat Serial search of the bins Most frequent words searched first When a match is found, then a pointer goes to the mental lexicon to access the information corresponding to that word Newer versions of the model also have syntactic/semantic bins, used for language production Go over how frequency effect come into play How semantic priming effects happen Has trouble explaining context effects mat cat mouse Mental lexicon

Cohort model Specifically for auditory word recognition (covered in chapter 9 of textbook) Speakers can recognize a word very rapidly Usually within 200-250 msec Recognition point (uniqueness point) - point at which a word is unambiguously different from other words and can be recognized Three stages of word recognition 1) activate a set of possible candidates 2) narrow the search to one candidate 3) integrate single candidate into semantic and syntactic context

Cohort model /s/ /sp/ /spi/ /spin/ time Prior context: “I took the car for a …” /s/ /sp/ /spi/ /spin/ … soap spinach psychologist spin spit sun spank spinach spin spit spank … spinach spin spit … spin time

Interactive Activation Model (IAM) Previous models posed a bottom-up flow of information (from features to letters to words). IAM also poses a top-down flows of information Nodes: (visual) feature (positional) letter word detectors Inhibitory and excitatory connections between them. McClelland and Rumelhart, (1981)

Interactive Activation Model (IAM) Inhibitory connections within levels If the first letter of a word is “a”, it isn’t “b” or “c” or … Inhibitory and excitatory connections between levels (bottom-up and top-down) If the first letter is “a” the word could be “apple” or “ant” or …., but not “book” or “church” or…… If there is growing evidence that the word is “apple” that evidence confirms that the first letter is “a”, and not “b”…..

The Word-Superiority Effect (Reicher, 1969) + Until the participant hits some start key

The Word-Superiority Effect (Reicher, 1969) COURSE Presented briefly … say 25 ms

The Word-Superiority Effect (Reicher, 1969) &&&&& A Mask presented with alternatives above and below the target letter … participants must pick one as the letter they believe was presented in that position.

The Word-Superiority Effect (Reicher, 1969) + + + E KLANE PLANE E & T E &&&&& T E &&&&& T Letter only Say 60% Letter in Nonword Say 65% Letter in Word Say 80% Why is identification better when a letter is presented in a word?

IAM & the word superiority effect We are processing at the word and letter levels simultaneously Letters in words benefit from bottom-up and top-down activation But letters alone receive only bottom-up activation. And of course, the same explanation applies to the advantage for reading letters that are part of “word-like” strings…higher level units (perhaps bigram units) provide top-down activation for letter units.

Comparing the models Each model can account for major findings (e.g., frequency, semantic priming, context), but they do so in different ways. Search model is serial and bottom-up Logogen is parallel and interactive (information flows up and down) Cohort is bottom-up but parallel initially, but then interactive at a later stage AIM is both bottom-up and top-down, uses facilitation and inhibition Note: There are other models out there (TRACE, FLMP, various connectionist models, and more)