As a result, the assembly errors of a modular robot are usually larger than those of a robot having fixed configuration.
The two last methods showed the smallest errors when no intermediate states were taken into account.
Further analysis examined three types of possible errors at each level of the task.
A total lack of tone-splitting errors would support (19a), while a preponderance of tone-splitting errors would provide evidence for model (19b).
The analysis of tone featural errors will be discussed in detail below.
Looking at them another way, we see that there were 34 errors (41 %) in which the environment for feature spreading was present.
Thirdly, errors which involve the addition or omission of a whole segment are quite common, and these have no obvious featural explanation.
Recall that no lexical blend or telescoping errors produced hybrid tones.