The Mental Lexicon Flashcards
What is phonological information?
This is information about how a word is pronounced, which can get extremely complicated. Think about accents, which can sometimes get in the way of our understanding even though it is the same language. Let’s take an example from Slovene ‘apple’, which can be said as ‘ jabolko/jabolk/japko/japka/jabuk’, all distinct sounds that carry the same meaning
What is morphological information?
We have morphemes that we add to the root/stem. According to this, we have simple words (walk, sleep, …) and complex words (sleeping, sleepwalking, walked, …). Morphology can be derivational or inflectional. In the former, the suffix creates a new word, in the latter, we add grammatical and functional info, we modify but do not change the meaning of a word.
What is syntactic information?
When we know a word, we know its grammatical class.
John bought a new notebook. *John bought a new write. → we can observe an EEG peak in an experiment with these two sentences, this is ungrammatical (because it’s the wrong class and our brains recognize this).
Remember that verbs like company and in our mental lexicon, a sentence needs the agent, recipient and location, otherwise, it’s ungrammatical in English: Philip put the ball. Moreover, we have further restrictions on what kinds of roles these participants will have. Not just any word can be an agent (The thunder fears the boy → the subject needs to be an experiencer but thunder isn’t because it can’t be afraid).
But why is syntactic information understood as part of meaning? Syntactic information is a huge help in interpretation. In some languages (like English), we need to know what a certain thing stored is because there is no phonology to help us distinguish a word class (like in Slovene).
What is semantic information?
This is information about the range of senses of a word. Most words have more than one meaning. The lemma ‘’attaches meaning to a word’’. All of this information is stored in our mental lexicon. There are two major questions we want to answer about semantic information in the mental lexicon:
How are words stored in the mental lexicon? (do we store all words 1:1; just stems? Are verbs and nouns in different places?)
How much do we activate when encountering a certain word? (do we activate all words with similar meaning? With a similar sound? What do we do?)
TOT (tip-of-tongue) → shows that in lexical access, access for meaning happens first and then phonology (we can recall what we want to say but not the phonological form) — but this holds true for mispronunciation, and cannot be necessarily generalized to the whole mental lexicon.
Which of these in incorrect?
There are several possible factors that can influence the organisation of the mental lexicon:
A) Age
B) Frequency of use
C) Part of speech
D) Ortography
A) Age
Maybe it is a factor but we didn’t explicitly state it is soooooo…..
Explain the hierachichal network model of the mental lexicon (Collins, Quillian)
This model talks about the organization of concepts in general (not words!) and postulates that concepts are organized into a hierarchy. In each node of the model, we have taxonomic and attributive information. For example:
A canary -> is an animal (eats, reproduces) → bird (it belongs to the concept animal and we don’t have to list info
about eating and reproducing again, we add info that it has wings and feathers) → canary (can sing, is yellow).
This model follows the principle of cognitive economy.
However!!! This model cannot accommodate for phonological, orthographical info. In other words, it’s not a linguistic model.
Explain the Spreading Activation Model (Collins& Loftus)
This model is related to associations and postulates that words are connected with each other. Within this theory, the mental lexicon is a network, but the organization is not strictly hierarchical. It’s a web of interconnected nodes, and the distance btw the nodes is determined by various structural characteristics, e.g. taxonomic relations and typicality (red has a short distance to other colours, the fire engine, and a more distant connection with apples, roses, car, sunset, …). Seeing a word triggers the activation of related words. It would be impossible to determine all the connections for one node. Additionally, this is not a linguistic model, but a model of concepts (like the hierarhichal network model). Words refer to concepts but have other properties as well.
Explain the Extended Network Model
The model has three levels: the conceptual level (word meaning), lemma level (syntactic properties) and the lexeme/sound level (phonological and morphological properties). This model was based on French. The argument for this model is that you cannot think of a word without these linguistic properties.
The model has been important in psycholinguistics, especially in explaining pathological cases (people with deficits at the lemma level and cannot recall syntactic information, but they know meaning and pronunciation). The support for the lemma and lexeme levels comes from the tip-of-the-tongue (TOT) phenomenon.
How would you test the effect of frequency on lexical access?
I would use a lexical decision task in which I would measure reaction times of participants comparing low-frequency and high-frequency words (controlling for other factors like length, age of acquisition, part of speech, …). Research shows that the more frequent a word is, the faster we make a lexical decision. Other research showed that the effect of facilitation by frequency is smaller in naming task and category verification tasks, so frequency likely occurs after lexical acces (decision process, pronunciation).
There are many ways the mental lexicon could be organised. One idea is that it is organised by meaning. How would you test this?
One way to test if the mental lexicon is organised by meaning is through priming. Priming is an experimental method in which participants make lexical decisions in prime - target pairs. The idea is that a prime facilitates reaction times in lexical decision because the mental lexicon is organised by meaning. As such, when items with similar meanings like ‘apple’ or ‘pear’ act as primes for things like ‘lemon’ ‘peach’, while primes like ‘pen’ or ‘truck’ would inhibit reaction times for the same targets. What I described is called semantic priming, because you prime the target with a semantically related word (it could also be syntactic, phonetic, …)
Why are polysemous words challening in the priming experiment paradigm?
Because one word can activate multiple meanings and it is hard to know which meaning acted as the prime or how the mental lexicon is really organised.
This isn’t a great answer but smth along these lines
Explain the cross-modal priming technique
The cross modal priming technique is similar to the prime-target paradigm but the prime or target can be either seen or heard. For example, you hear ‘nurse’ and then read ‘doctor’ or vice-versa.
Explain the concept of active search in the context of word recognition
when we are confronted with words, our system actively looks for matches (as soon as we hear nur-). This is why we recognize words before we reach their end, and cross-modal priming experiments show that. It might only be necessary to hear nur- to activate doctor, the recognition point is earlier.
But all great things come at a cost. The cost of active search is that we activate many inappropriate words and these must be rejected. The stages of lexical acces are activation vs. competition, recognition vs. selection.
Explain Marslen-Wilson and Tyler’s Cohort theory (model of lexical access). Which four stages of speech perception does it state?
They state that we activate many words when we hear something (words that start with the same phoneme). This process is automatic and autonomous and is based on cohort only, but the selection is non-autonomous (we can use contextual information as well).
When there is just one more plausible word, we call this a word uniqueness point (spinafor spinach). “Our cohort is completed and there is one soldier less.”
To show this process: we see the letter ‘s’ which activates many words (song story sparrow saunter slow secret sentry etc.) → SP (spice spoke spare spin splendid spelling spread etc. ) → SPI (spit spigot spill spiffy spinaker spirit spin etc.) → SPIN (spin spinach spinster spindle) → SPINA (spinach) = word uniqueness point
Stages of speech perception:
1) activation of the cohort candidates based on phonological, morphological, syntactic or semantic factors
2) competition between candidates
3) recognition
4) selection of the optimal candidate
→ Lexical activation happens upon hearing the beginning of the word, the parser doesn’t wait till the end of the word
Provide some evidence for Marslen-Wilson’s Cohort Model
In the experiment, they did cross-modal priming, people heard the prime and saw the target. They showed the target in 2 different positions: one was after kapi-; once they say boot and once they say geld. At this position, there was facilitation for both geld and boot (because they haven’t reached the word uniqueness point; the cohort is open and both are active). In position 2, the target was shown one syllable later after kapitein. There was facilitation for boot, but not geld. Kapitaal got deactivated; the processor forgot about it. The experiment provides robust evidence for the Cohort Model.
Later they did a similar experiment but before the word of interest, there was some more context.
‘’After so many years at the sea, the kapi-”. What they expected was that they would not find facilitation for both geld and boot, but they found facilitation for both! Kapitaal was also activated. The model says that activation is independent and autonomous (the cohort is open despite the context, we activate both kapitein and kapitaal) but the selection isn’t constrained. This holds true even in polysemous words. In the activation step, context plays a more significant role.