"... In PAGE 7: ... Here we report only the sensitivity effects for the two populations (separately for grammatical, semantic and discourse items); these were calculated by subtracting the mean normalized response time (for all subjects from each population) for good continuations from the mean response times for bad continuations. Examples of the stimuli from each of the three constraint types (grammatical, semantic, and discourse) are shown Table2 . A detailed presentation of the methodology and the tasks can be found in Kempler et al.... In PAGE 8: ... Grammatical Constraints Forty grammatical sentences were constructed and then altered to create ungrammatical counterparts by the substitution, addition or deletion of one word. Two types of grammatical violations were included: (1) subject noun-verb agreement errors and (2) errors of transitivity (see Table2 ). Preliminary analysis of the subject-verb and transitive sentence structures did not demonstrate any effect due to construction type (F lt; 1), so the data from both constructions were combined for analysis.... In PAGE 8: ... In half of the sentences, the final word was grammatically and semantically appropriate. In the other half, the final word was anomalous due to a semantic (or pragmatic) violation ( Table2 ). The procedure and subject populations were identical to the grammatical experiment described above.... In PAGE 11: ... Andersen, 1998). Short subject-verb agreement stimuli were taken from the subject-verb agreement items described above ( Table2 ) and long stimuli were constructed by adding an intervening 10-15 word clause between the subject and the verb. For instance, the subject-verb agreement item The young girl was/*were was transformed into a longer item The young girl, who improved greatly every day because of the excellent teaching and good books was/*were.... In PAGE 13: ... This is also true for the on-line discourse items: in order to determine whether the pronoun him or them was coherent in the discourse, patients would have to understand the entire discourse. However, full processing of the sentences was not necessary for the on-line grammar and semantic items: in the case of our on-line grammatical and semantic items ( Table2 ), patients could perform well on the task (i.... In PAGE 15: ...asks. We also do not think that on-line tasks are inherently preferable. For example, we suspect that our on-line cross-modal naming task may be very unnatural, insofar as it requires substantially less processing than normal language comprehension. That is, in the case of our short grammatical and semantic items in the on-line experiments ( Table2 ), patients could perform well on the task by attending to only minimal information in the sentence concerning the relationship between verbs and nouns. As stated in the body of this chapter, simple grammatical relations (e.... ..."

"... In PAGE 6: ... Procedures and Conditions (Video Stimulus): The video stimuli include two types of combinations of expressive strength: (i) pose changes from 0 to 1 when the singing voice changes from 0 to 1; and (ii) the reverse order of (i) with respect to the singing voice. Considering the order effect, we also prepared video stimuli in which: (iii) pose changes from 1 to 0 when the singing voice changes from 1 to 0; and (iv) the reverse order of (iii) with respect to the singing voice ( Table7 ). These stimuli are understood to reflect changes in expressive strength along the diagonal slopes of Figs.... ..."

"... In PAGE 7: ... Two different types of processing are used for each of the methods discussed earlier: off-line supervised training where the transformation matrices are generated before hand using groundtruth data, and on- line dynamic processing where the transformation matrices are generated on the fly using the input testing video directly. Table3 provides the performance comparison of different methods. Overall, CFA achieves 91.... ..."

"... In PAGE 11: ... This acoustic-only model may be thought of as a rough approximation to a minimum description length approach to finding highly repeated speech patterns which are likely to be words of the language. Results of the evaluation shown in Table1 indicate that the cross-modal algorithm was able to extract a large proportion of English words from this very difficult corpus (M2), many associated with semantically correct visual models (M3). Typical speech segments in the lexicons included names of all six objects in the study, as well as onomatopoetic sounds... ..."