Language 1 Flashcards

Question

Models of Speech Production (1) DELL (1986) SPREADING ACTIVATION THEORY Thing to remember about Dell's theory...

Answer 1

According to Dell we are spreading activation across the mental lexicon both semantically (to items related by meaning) and phonologically (items related by sound).

Answer 2

*** Dell (1986) Theory: - Spreading activation results in activation of related items at all levels of processing - Processing items that overlap in semantics and/or phonology will result in higher error rates than processing items that do not overlap Predictions: - As all of this spreading activation is occurring all at once, processing items that overlap in terms of meaning and sound, should mean we make a lot of errors - easy to switch in the wrong thing. - BUT… we don’t make that many errors in real life… slightly problematic for this model of speech production.

Answer 3

**Procedure and Predictions: Participants asked to read the sentences, one at a time, but not to say the end word, then asked to name the picture (the pope). Then the next sentence, name the picture and so on. The second sentence activates the concept of a “nun” which is semantically related to the picture of the pope, which should make is harder to say “pope” when asked to identify the picture. The first sentence activates the word “none”, this is not semantically or phonologically related to related to the picture of the pope. But is phonologically related to the word “nun” which is related to the picture of the pope semantically, so should make it harder to inhibit the word “nun” when naming the picture. The third sentence isn’t related to the picture at all. It should be easier to inhibit the word “match” when identifying the picture of the pope (compared to the other 2 sentences) because it's not phonologically or semantically related. **Findings: Successfully inhibit ‘match’ to say ‘priest’ but fail to inhibit ‘nun/none’ and say ‘priest’ Semantically and phonologically related items are active at the same time and can be selected for output This data suggests it is harder to inhibit saying the wrong word when it is semantically and/or phonologically related to the picture that's shown - supports Dell's model

Answer 4

In terms of the speech errors we looked at last year, Dell’s model accounts for them really well. Errors where you switch words in, exchange and blend things as so many related concepts are active in the lexicon all at once are accounted for. > Errors may occur due to more than one concept receiving the same amount of activation e.g. ‘spoon’ being selected INSTEAD OF ‘fork’. > Errors may also occur due to more than one concept receiving the same amount of activation and ‘spoon’ and ‘fork’ BOTH being selected e.g. saying “spork”. Both concepts become active resulting in a cascade of processing to the level of phonology. Resulting in ‘the spoon’ instead of ‘the fork’ or ‘spork’! **Anticipation error becomes an exchange. You want to say “the prongs of a fork”, but everything is active at once and you could end up saying “the forks of a prong”. Described as an anticipation error becoming an exchange because what Dell proposed is that what happens is… You’ve got fork and prong active, you’ve said fork instead of prong, once you’ve slotted fork into the beginning of a sentence you can’t put it at the end of the sentence too because you can’t reactivate it, so you have prong left, so you say it at the end instead. This model can account really easily for these kind of speech errors.

Answer 5

WEAVER: "Word-form Encoding by Activation and VERification" Similar model to Dell but "discrete", so where Dell's model proposed that activation was happening all at once, in paralel and in more than one direction... Levelt proposes that activation can only go in one direction and a process must be complete before another can begin Levels: - Conceptual level - LEMMA - Morphemes - Phonemes

Answer 6

Discrete – activation is “feed forward” can only move from meaning to sound > LEMMA – function that codes some of the syntax and the word form (morpheme) > Morphemes/morphological level – put different morphemes together, so the morpheme in the LEMMA would come down here and be added to morphemes that mark plural, or tense etc > Phoneme level – activate the phonemes that we use to express the word

Answer 7

The idea that on a conceptual level the selection process of words happens via spreading activation is the same as Dell's model However, Levelt says selection of items at each level is based on competition between items. So, at the beginning there is semantic spreading of activation between words with similar meaning, but after we’ve selected what it is we are going to say then inhibitory signals go out to all the other concepts that were activated in the semantic space, in order to stop us making speech errors.

Answer 8

Levelt et al., (1999) Processing of items which are semantically similar will be inhibitive - When hearing the word "doctor" activation is spread to "nurse" but is inhibited, when you hear "nurse" you have to release that inhibition in order to activate it. If "nurse" was followed by "sheep" the activation would be quicker, because they aren't semantically related Dell (1986) Processing of items which are semantically similar will be facilitative - Priming experiments from last year showed when people presented with a words like “nurse” and “doctor”, they were quicker to process doctor, than if it was “nurse, sheep”, this means semantically similar words can act as facilitators for processing.

Answer 9

Wheeldon & Monsell (1994) If you are producing a word then naming something semantically related, you should be slower to say the semantically similar word - because you've activated one word you've inhibited semantically similar words Levelt would say you should be slower to say these related words. Dell would say you should be quicker to say these related words. ***Results and Conclusions: The data from this experiment supports Levelt - similar words are harder to say quickly because they activate similar networks

Answer 10

So far in the lecture we have seen... > Evidence that phonological competitors interfere with word production does not fit with Levelt et al.’s (1999) discrete model. - Supports Dell's idea that semantically and phonologically related info is active when trying to process words > Evidence that selection of concepts is competitive does not fit with Dell’s (1986) spreading activation model. - Supports Levelt's model that you inhibit semantically related info once you've selected a word for production

Answer 11

Levelt says... - Errors may occur if the wrong concept is selected and processed to the phonological level (if the concepts are really similar) - Two concepts that are both viable options for the utterance could be selected and processed to the phonological level (if both are things that could both be feasibly put into the phrase) Conclusions: > Speech errors can be accounted for by Dell’s model but do not appear to fit with Levelt et al.’s model > Where Dell’s model easily accounts for Speech errors Levelt et al.’s model has to suggest that these are ‘special cases’ to explain these effects

Answer 12

Lemma's (dictionary form of a word) contain syntactic information Tip of the tongue state > Italian speakers in tip-of-the tongue states where they can't quite grasp the word they want to say, they know the grammatical gender of the word (Vigliocco et al. 1997) This provides strong support for the lemma stage of the model

Answer 13

According to Levelt phonological similarity should not affect processing of competing items but a number of researchers have found evidence that it does … If processing of spoken words is discrete and feedforwards (Levelt) phonological processing should only occur for items that have been selected and should not be affected by ‘non target’ information. If processing of spoken words is interactive and parallel (Dell, 1986) phonological processing is likely to occur for all items that are related by semantics and/or phonology to the target item.

Language 1 Flashcards

(37 cards)