Psycholinguistics

Question 1

What is word recognition?

Answer 1

The perception of speech sounds.

How linguistic knowledge, context & expectations help us to perceive words.

Question 2

CATEGORICAL PERCEPTION
Liberman et al (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology, 53, 358-368.

Answer 2

The sound varies in steps as the onset frequency of the second formant changes from -6 to +9. But we hear a sudden shift from /ba/ to /da/ to /ga/. Further experiments by Liberman et al showed that people cannot reliably tell the difference between sounds within categories (e.g. between a -6 and -5). Yet the same frequency differences are perceptible when non-speech simple tones are used.

Question 3

How does the idea of categorical perception contribute to our model of word recognition.

Answer 3

Acoustic features are recognised as phonemes, but phonemes exert a top down influence on acoustic features.

Question 4

How does the McGurk effect contribute to our model?

Answer 4

Visual speech gestures AND acoustic features have a bottom up path to phonemes. Interestingly the visual signal overrides the acoustic, it is possible the visual signal activates the phoneme which has a top down effect on the acoustic signal.

Could be taken as evidence for the Motor Theory (see Li1 notes).

Question 5

Samuel, 1997: When we hear what we expect

Answer 5

The stimulus is made by replacing the /d/ with noise, yet we still hear the /d/ (and maybe the noise just after it?). This is known as a “perceptual restoration” effect. Our lexical knowledge tells us that there has to be a /d/ there, so that is what we hear. But what we hear is a mental construction that is only partly based on the signal.

Question 6

How does Samuel, 1997 contribute to our model?

Answer 6

the TRACE model of speech perception (McClelland & Elman, 1986), where there is a bottom up and top down relationship between the input phonological word form

Question 7

The word superiority effect (Reicher, 1969)

this is visual recognition

Answer 7

The word superiority effect is like the phoneme restoration effect. It is easier to perceive letters in words than in isolation or in consonant strings. But, logically speaking how can this happen? If we are looking up a word in a dictionary we first need to recognise the letters. So how can knowledge of the word affect recognition of the letters?

You might think that perhaps we are accessing the word on the basis of just some of the letters, e.g. L A _ E, and that we infer the ‘N’ from lexical knowledge. But the alternative choice, ‘K’, would make a word too, so we would be at chance in this case.

Question 8

How does the word superiority effect contribute to our model?

Answer 8

Interactive activation model of visual word recognition.
McClelland & Rumelhart (1981)

Visual features lead to letters (which also have top down), letters lead to input orthographic form (which also have a top down)

Question 9

What is shadowing?

Answer 9

Repeating as somebody is speaking. Marslen-Wilson & Welsh, 1978, found that fast shadowers were repeating words before they had finished. They would also restore erroneous phonemes in recognised words with no delay (unconsciously).

Question 10

What is shadowing evidence for?

Answer 10

The Cohort Model.

Early recognition in shadowing is more likely when the word is predictable context. There are more fluent restorations in later syllables of words and where there is high contextual constraint.

Question 11

How does this context info influence our model?

Answer 11

Meaning has a top down and bottom up relationship with input ortho and phono form.

Question 12

Interactive activation model of speech production

Dell, 1986

Answer 12

Meaning > output phonological word form (with back up) > phonemes (with back up) > articulatory gestures

Question 13

Dual-route model of reading.

Coltheart et al (2001)

Answer 13

There are two ways of converting print to sound in reading aloud. The “sublexical” route goes from letters to phonemes. This mapping must utilise rules of spelling (as in phonics training). This is good for reading regularly spelled words (e.g. cat) and nonwords (e.g. blemp). The ‘lexical’ route simply reflects associations between written word forms and pronunciations. This is good for reading irregularly spelled words (e.g. have, cough). The Dual-Route model reflects the division between knowledge of rules and knowledge of exceptions.

Question 14

What is global ambiguity?

Answer 14

2 different structures, no one word is ambiguous

Question 15

What is local ambiguity?

Answer 15

1 word is ambiguous > garden path sentences

Question 16

Evidence for incremental parsing:

Answer 16

Eye tracking (fixations and saccades) 
Frazier & Clifton (1982) found reading time was longer on the ambiguous element if the sentence was a GPS.

Question 17

What is the principle of minimal attachment?

Frazier, 1988

Answer 17

: Attach incoming material into the phrase marker being constructed using the fewest nodes consistent with the well-formedness rules of the language.

Question 18

What’s the principle of Late Closure?

Frazier, 1988

Answer 18

If consistent with the rules of the grammar attach each incoming word into the structure currently being built.

Question 19

What’s the garden path model?

Answer 19

incremental parsing according to late closure and minimal attachment.
Frazier, 1988

Question 20

The garden path model and modularity?

Answer 20

input > lexical processor > syntactic categories > syntactic parser > syntactic structure > thematic interpreter > general knowledge

if general knowledge finds it to be unlikely you go back to syntactic parser and reanalyse

Question 21

Fodor: modularity of mind?

Answer 21

Fodor (1983) proposed the modularity of mind: the idea that the mind is composed of multiple distinct mechanisms (modules) used for making inferences but with limited access to background information due to specific cognitive architecture.

Question 22

Syntactic processing:

Answer 22

See Essay

Question 23

Meaning > output phonological form?

Answer 23

Non-verbal concept > lemmas (an abstract representation which controls the mapping between concepts and word forms) > output form

Question 24

What causes blends and substitutions?

Answer 24

lemma competition: if non-verbal concepts correspond to multiple lemmas. (Process of lexical selection involving parallel access and competition).

Question 25

What information goes into non-verbal concepts?

Answer 25

tactile and visual

shown by Shallice 1988 - optic and tactile aphasia (only one aspect of the concept is available)

Question 26

Wilson (1987)

Answer 26

visual recognition wasn’t possible but they could relate verbal definitions to their words and vice versa. shows that verbal semantics is also something with a top and bottom relation to phonological output.

Question 27

What are category specific impairments?

Answer 27

Can define or identify words from some ‘categories’ e.g living things but not from others.

Theoretical accounts:
different distribution of concepts over different brain regions.
Caramazza & Mahon (2003). Neural specificity of representations of different categories (animals, fruit/vegetables, and artefacts). Evolutionary argument.
Tyler & Moss (2001): shared versus distinctive features. Living things – more shared than distinctive; artefacts – more distinctive than shared. Living things more vulnerable. Computational argument.

Question 28

Embodied conceptual representation

Hauk et al, 2004

Answer 28

With action words, activity in motor cortex area related to specific body part.

Zwaan & Taylor (2006)

Participants turn a dial clockwise or counterclockwise to advance through a text. If the meaning of a phrase (e.g., “he turned the volume down”) conflicted with the required hand movement, reading of that phrase was slowed.
(action compatibility effect)

also the pen experiment.

Question 29

(Weinreich) Coordinate bilingual

Answer 29

2 semantic systems, 2 language systems

Question 30

(Weinreich) compound bilingual

Answer 30

1 semantic system, 2 language systems

Question 31

(Weinrich) subordinate bilingual

Answer 31

1 semantic system, 1 language system, from which derives a second language system

Question 32

Dijkstra et al, 1999: Recognition times for L1 English L2 French speakers.

(visual)

Answer 32

True cognates are faster identified than a non-cognate French word. False friends (non-cognate interlingual homographs) are slow compared to control words. Shows that English system is still in play.
(Image A, convergence of activation from 2 langs on same meaning facilitates responses, Image B, Forms in both languages are activated but since they have different meanings there is a conflict between them (horizontal dotted line with round ends) and this slows down the response. )

Question 33

Elston-Guettler et al 2005 psycholinguistic priming

visual

Answer 33

Priming was obtained from the german meaning of the false friend prime. Both languages in play, L1 influence strongest but can be ‘turned off’.

Question 34

Blumenfeld and Marian (2007)

Spoken word recognition

Answer 34

‘desk’

brief look at dustbin lid (deckel in German)

Question 35

Diikstra et al (1999) BIA (Bilingual Interactive Activation) model of visual word recognition

Answer 35

See image.
Note that the BIA model contains a level of language nodes, and a separation between words of different languages at the word level (even though words in different languages have unique representations, they are contained in the same system, which explains why they can interact). These assumptions are necessary to account for the fact that for a false friend such as STEP, when reading in English one meaning is activated, and when reading in Dutch another meaning is activated, despite the fact that they have similar orthographic and phonological forms. These would seem to require the existence of two different representations at the lexical level which can project to distinct semantic representations, and also the ability to control which one of these is the more active (see Dijkstra et al, 1999, for further arguments).

Question 36

Access to meaning.

Answer 36

Direct relation between words and concepts, but also L1 mediation?
Seems to be direct relation in proficient speakers, but translation in progress for beiginners. See Chen & Leung (1989) - on slide.

Question 37

Thierry & Wu (2007)

Covert L2 to L1 translation even at high level of proficiency?

Answer 37

The experiment described on p.64 shows that even with semantically unrelated stimuli in an L2, the orthographic form in the L1 is still activated during processing (as shown with brain responses, ERPs). It has been found more often that L2 activates L1, but that it activates the orthographic word form in L1 even when the L2 words are heard is quite remarkable. The experiment was done fully in English (L2) with high proficiency speakers.

Question 38

More evidence for interaction between langs (distinct but within one system)? This time for output

Answer 38

Stroop effects, see slides.

Question 39

So they are part of the same system, how do people stick to 1 language?

Answer 39

‘Supervisory Attentional System’.

Evidenced by Meuter & Allport (1999) who found that switching from L2 to L1 is harder than the other way. The idea being that to speak L2 you have to massively inhibit L1 which takes more undoing. But see Finkbeiner et al (2006) for a failure to get this effect.

Question 40

The Simon effect

Answer 40

slower on incongruent trials. displayed less in bilinguals (more inhibitory control)