Speech production

• Speech is a complex process involving thinking what one wants to say, selecting the appropriate words, combining these words grammatically, and turning the combination into a series of articulatory gestures.
• Meaning is conveyed not just by the words which we utter, but also the way in which we utter them.
• Lea (1973) found that syntactic boundaries were often marked by so-called prosodic cues like pauses and stress.
• Pauses might not just help in comprehension, they might also allow the speaker time to plan the next part of their utterance.
• Lee, Brown-Schmidt & Watson (2013) asked whether the grammatical complexity of a sentence influenced the time it took to start speaking.
• People planning an utterance might do so at a clause level or at a smaller, phrase level.
• Word exchanges are more often within clauses and pauses in spontaneous speech are at clause boundaries.
• Speakers describing pictures took longer to initiate speech with complex initial phrases than with simple phrases.
• Sound exchange errors seem to have a shorter range than word exchange errors, suggesting that sounds are planned at a shorter scale than syntax.
• Ferreira and Swets (2002) found evidence for planning that responded flexibly to task demands.
• Freud maintained that parapraxes reflected a speaker’s true desires, that had been suppressed.
• Some data from Motley (1980) suggest that speech errors, or “slips of the tongue”, can reveal a speaker’s true desires, that have been suppressed.
• Phoneme deletions, perseverations, and anticipations are types of speech error.
• Collections (or corpora) of spontaneous speech errors have been assembled, often comprising the speech errors which the researchers themselves have heard in casual circumstances.
• methodological problem as some types of speech error may be easier to spot in a laboratory setting.
• One can also generate speech errors under laboratory conditions, by asking people to speak fast or to repeat “tongue-twisters”.
• Laboratory studies support the conclusions drawn from actual error collections.
• Various kinds of speech error include semantic substitution of a word, such as “where's my tennis bat”, where the inserted word is nearly always of the correct class (i.e., noun for noun, adverb for adverb, etc.), though verb for verb substitution is rare.
• Another type of speech error is phrase blending, for example, “the sky is shining”, as a blend of “the sky is blue” and “the sun is shining”.
• Word exchange, such as “To let the house out of the cat”, is also a type of speech error.
• Morpheme-exchange errors, like “He has trunked the packs” in place of “he has packed the trunks”, are also types of speech error.
• Morphemes are units of meaning/grammatical function, such as the morpheme “pack” and the suffix morpheme “ed” to indicate past tense.
• In morpheme-exchange errors, the morpheme tends to adapt to its new circumstances.
• In 1987, it was shown that the plural morpheme “s” in “the forks of a prong” is pronounced in a manner appropriate to “forks” (an ess sound) rather than a manner appropriate to “prongs” (a zed sound).
• Spoonerisms, such as “boon spending” in place of “spoon bending”, or “you hissed my mystery lecture” in place of “you missed my history lecture”: a part of one word swaps with a part of another word
• Harley (2014) also lists word blends
• Such swaps often involve the onsets of words (or sometimes syllables) changing places rather than their codas (the part after the vowel) or their rimes (the nucleus/vowel plus the coda).
• Swaps are more likely when they result in real words like boon spending/ spoons bending (e.g
• In these cases the onsets (the parts before the vowel) have swapped.
• The phonemes involved in such Spoonerisms are often featurally similar (i.e
• Spoonerisms tend not to bridge across phrase boundaries, suggesting that the phrase is a unit of speech planning.
• This suggests that inflections (e.g., endings, sometimes called suffices) are added at a point in the speech production process once the words to which they attach have already been fixed in place.
• People monitor their upcoming output.
• Dell (1986) and Dell & O'Seaghdha (1991) developed a spreading activation theory of speech production based on principles of neural network/connectionist modelling.
• The main assumptions of Dell & O'Seaghdha's model are a semantic level at which the meaning of the intended utterance is represented, a syntactic (grammar) level specifying the grammatical structure of the utterance, a morphological level specifying the morphemes in the planned utterance, and a phonological level specifying the phonemes or basic sound units in the utterance.
• In Dell & O'Seaghdha's model, a representation is formed at each level, with representations developing simultaneously and interactively at the four levels.
• Higher levels (e.g., semantic) are likely to have progressed further at any given point in time.
• At each level in Dell & O'Seaghdha's model, there are categorical specifications of the structure of the sentence at that level.
• The syntactic (grammar) level in Dell & O'Seaghdha's model will specify the permissible grammatical categories of items in the planned utterance.
• The semantic, morphological and phonological levels in Dell & O'Seaghdha's model contain a lexicon (dictionary) comprising units for the concepts, words, morphemes and phonemes relevant to each entry.