Acoustical Society of America
ASA/EAA/DAGA '99 Meeting
Lay Language Papers

No Role For Syllables in English Speech Production

Niels O. Schiller - nschille@wjh.harvard.edu
Cognitive Neuropsychology Laboratory, Department of Psychology, Harvard University, William James Hall 918, 33 Kirkland Street
Cambridge, MA 02138

Popular version of paper 5aSCb6
Presented Friday morning, March 19, 1999
ASA/EAA/DAGA '99 Meeting, Berlin, Germany

One of the most intuitive units in speaking is the syllable. If you were to ask an illiterate person how many sounds or letters the word 'syllable' has, the probability that you would get the correct answer is pretty low. However, asked for the number of syllables, even illiterates are likely to know that the answer is three (Morais, Content, Cary, Mehler, Segui, 1989). Furthermore, people are able to manipulate syllables in a metalinguistic fashion. Imagine you heard the word 'pilot' and you were asked to articulate the two syllables of the word in the reversed order. Do you think this is possible? You would have to syllabify the word first - and so far we have no idea about how this works at a cognitive level. Your next step would be to reverse the order of the two syllables. This would probably be accomplished by putting the two syllables into some kind of articulatory buffer and then spelling them out again. Sounds complicated, doesn't it? Speakers like you and me, however, do not consider this a difficult task. Speakers can reverse the two syllables of a word in less than half a second and make very few errors (Schiller, Meyer, & Levelt, 1997; Treiman & Danis, 1988). Theories of speech production assume that syllables are the smallest articulatory motor unit that can be produced in continuous speech. However, theories differ in several important respects concerning syllables. Some theories claim that words are already syllabified in the mental dictionaries in our brains (e.g., Dell, 1986, 1988). Accordingly, when we produce a word, the syllables are already there and we just have to put them together in order to say them. Other theories claim that words are not syllabified in our mental dictionaries. Rather, they propose that syllables are computed on-line - as speech unfolds itself over time (Levelt, 1992; Levelt & Wheeldon, 1994). The most important argument for this latter view comes from the fact that sometimes - more often than we actually think - sounds of adjacent words influence the syllabification of a word. Take, for example, the three words "sounds of adjacent" from the previous sentence and say them aloud to yourself in a casual manner. Do you notice what happened? Listen carefully! Now, say it again. Did you hear it? What you did - to some extent at least - is compute a series of syllables that deviates from the syllables of the individual words spoken carefully. In fact, you said something like "sound so fadjacent". This is called "resyllabification" in linguistic terms. We human beings are lazy creatures and therefore we try to minimize our physical efforts whenever we can - even in speaking. Resyllabification helps us keep our articulation as economic as possible by computing optimal syllable structures. Since the syllable structure of words may vary depending on the words that surround them, what's the point in having the syllabic structure of words stored in the mental dictionary?

Although this may sound like a good argument against the view that the syllables of words are stored in the brain, recently some experimental results have been published that favor this view - although they are not incompatible with the other view. The result is the following: When people are asked to say the name of a picture presented on a computer screen in front of them as fast as possible and the presentation of the picture is preceded by a visual letter stimulus that is presented for a time too short to be visually identified, participants are faster to say the name of the picture when the letter stimulus matched the first syllable of the picture's name than when it was a letter string of equal length not matching the first syllable (Ferrand, Segui, & Grainger, 1996).

To give an example, French participants were faster to say ca-rotte in response to a picture of a carrot when the picture was preceded by the brief representation of the string CA than when the letters CAR preceded the picture. Similarly, they were faster to say car-table to a picture of a schoolbag when this picture was preceded by the letter string CAR than when CA preceded the picture. Note that both words overlap in the first three letters C-A-R, so it cannot be the length of the letter string that mattered. This result was obtained in French, i.e., a language with a relatively clear syllable structure. However, if the effect is real, it should also work for Germanic languages such as German, English, or Dutch. In Dutch, however, there is no such effect (Schiller, 1998). The study presented here at this conference reports data from English. But English behaves just like Dutch - no syllabic effect. Since a study in Spanish, another language with a relatively clear syllable structure, has also failed to replicate the syllabic effect (Carreiras, personal communication), the question is: Why is it so hard to obtain psycholinguistic evidence for a unit everybody believes exists?