word or syllable errors are said to be easier to hear than phoneme errors, and anticipations are perceived more readily than perseverations. I think it is interesting that in every naturalistic corpus I have encountered, the 'difficult-tohear' phonological errors always outnumber other types of errors; thus it is possible that these biases are somewhat overstated (see Poulisse 1999:9 for a list of corpora in which this is true). g. Thai, Mandarin, Japanese, Spanish, Hindi), anticipations and perseverations tend to be collected in equal numbers (see, for example, Wan 1999, Wells-Jensen 1999, Muansuwan 2000, Kawachi 2002), so that in languages where they are produced in equal numbers, they seem to be perceived as such.
In these cases subjects would often catch themselves before they spoke the error aloud, but would blush and look embarrassed because of what they had been about to say. These studies attest to the existence of an internal monitor which scans planned output prior to speaking. However, luckily for us, this internal monitor is far from perfect, and erroneously planned utterances can fool the monitor and be spoken aloud, producing SOTs. At this point a second monitor, the 'external' monitor, comes into play.
Al: 'Look! ' 'Alice, come here! red]; age 3;5) c. ves]; age 3;3) d. ' (age 5;6) In (3a), Alice substituted the word 'tomorrow' for 'yesterday'; however, this was not a SOT, since she did not have the meanings of these words clearly distinguished, and was fluctuating between the two randomly at the time. e. rather than 'went' she added the regular past tense suffix to 'go', producing 'goed', then double marked it by adding the suffix again, now requiring the syllabic form, producing 'goeded'. This was an overgeneralization she was making regularly at this time, and not a one-time SOT.