"... In PAGE 10: ...g., cen- tered around the vowel; Daugherty amp; Seidenberg, 1992) but it is not eliminated completely (see Table1 ). Adequate general- ization still requires learning the regularities separately across several slots.... In PAGE 11: ... Unfortunately, however, the knowledge of spelling-sound correspondences is still dispersed across a large number of different contexts, and adequate generalization still requires that the training effec- tively covers them all. Returning to Table1 , although the words LOG, GLAD, and SPLIT share the correspondence L a1 /l/, they have no triples of letters in common. A similar property holds in phonology among triples of phonemes or phonemic features.... ..."

"... In PAGE 10: ...g., cen- tered around the vowel; Daugherty amp; Seidenberg, 1992) but it is not eliminated completely (see Table1 ). Adequate general- ization still requires learning the regularities separately across several slots.... In PAGE 11: ... Unfortunately, however, the knowledge of spelling-sound correspondences is still dispersed across a large number of different contexts, and adequate generalization still requires that the training effec- tively covers them all. Returning to Table1 , although the words LOG, GLAD, and SPLIT share the correspondence L )/l/, they have no triples of letters in common. A similar property holds in phonology among triples of phonemes or phonemic features.... ..."

"... In PAGE 14: ... We also show the results of the third assessment (assignment) for information purposes. Table3 shows various measures of dispersion for these tests, over the three years. The first line shows the mean and, in brackets, the standard deviation.... ..."

"... In PAGE 3: ... Our dispersion definitions give a spectrum of modeling options, although many other dispersion choices are also possible. Table1 shows the chosen dispersion relationships. The Poisson and Chisquare families have fixed dispersion val- ues of a0a2a1a4a3 and a0a2a1a6a5 respectively.... In PAGE 8: ... The actual effort, a55 a3 a58 , was already known. We investigated the distributions in Table1 for fitting the probability model a55 a3 a58 a16 a18 a3 a20a23a22 a3 a58 a24 . The data were hard to fit because the CBR method tended to underpredict ef- fort, often substantially, so a distribution with extreme pos- itive skew was required.... ..."

"... In PAGE 11: ... Our scale provides for comparisons between developing and developed nations. Geographic Dispersion Geographic dispersion ( Table2 ) describes the physical dispersion of the Internet within a country. In addition to just having the network accessible through- out the country, there are benefits to having multiple points-of-presence within an area, redundant transmission paths, and multiple international access points.... ..."

"... In PAGE 31: ... The measurement of the extent and persistence of cyclical divergence between regions and nations is not a straightforward affair. Table5 shows the dispersion of regional unemployment rates in the UK, and Table 6 the dispersion of national unemployment rates, output gaps and national inflation rates for 11 EU member states and the UK. TABLE 5 HERE TABLE 6 HERE It is clear that the dispersion of regional unemployment rates in the UK (which ranged, in April 1999, from 10.... ..."

"... In PAGE 8: ... Combining the frms in which either two ftuds, one domestic, one foreign investor, or the state could effectively have control, I find 992 fuins or 67% of all firms, with effective control by one party or two investment funds (relctrt). The large fraction of firms in which there is effective control by one (or two) investors indicates that nominal shareholding--on average, these investors combined own only 51% (nonwide in Table2 )--is not necessarily a good indicator of effective control. quot; Dispersion A useful measure of the degree of ownership dispersion is the Herfindahl index (the sum of squared ownership shares).... In PAGE 8: ... The share held by any individual is set to zero for the calculation of H as any single individual owns only a marginally small amount of a given firm. Calculating this indicator for the 1491 firms, I find that the mean H-indicator of dispersion is 14%, with a low close to 0% and a high of 87%, and a standard deviation of 13% ( Table2 ). A similar index for dispersion among the funds only (scaled, such that when one fund owns all fund shares, the index is one) has a mean of 33%, with a low of 7% and a high of 100%, and a standard deviation of 23 %.... In PAGE 9: ... I therefore also use the share of equity held by all (four) strategic investors combined (nonwide). The average share held by all strategic investors combined is 51 %, with a standard deviation of 23% ( Table2 ). The low is close to zero and the high of 93 %.... In PAGE 9: ... But, the correlation between nonwide and H is relatively low (0.41, Table2 ), suggesting that they are separate indicators of ownership structure. Prices I use three kind of prices: the prices of the last (5th) bidding round for both Czech and Slovak firms,9 and secondary market prices for firms traded on the Prague Stock Exchange (PSE) and the Czech RM-system.... ..."