Table 2.
Automatic classification of global error patterns: operational definitions.
(A) Place-sparing patternsa | |
Error pattern 1 (place identity) | There are permutations with or without omissions. |
Error pattern 2 (place approximation) | There are permutations and substitutions, with or without omissions or insertions; matching each substitution in response to a substituted counterpart in stimulus (the nearest substituted locationb), the distance between the two never exceeds the max. threshold distancec; if there are insertions, each of them is within the max. threshold distance from a correct location. |
Error pattern 3 (place intrusion) | There are permutations, plus substitutions and/or insertions not matching criteria for pattern 2. |
(B) Path-sparing patternsa | |
Error pattern 4 (path approximation) | No permutations; there are substitutions and/or insertions, with or without omissions; each incorrect location (substitution or insertion) lies within the max. threshold distancec from the correct pathd; |
Error pattern 5 (path shortening) | There are only omissions. |
Error pattern 6 (path deviation) | No permutations; there are substitutions and/or insertions not matching criteria for pattern 4. |
(C) Minimal or random response | |
Error pattern 7 | Omissions involve half or more of the locations and/or there are not two correct locations in correct order. |
aFirst always check criteria for pattern 7: all other patterns require that omissions spare more than half of the locations and at least two correct locations in correct order are retrieved.
bMatching substitutions with substituted counterparts is accomplished by trying all possible permutations of the incorrect locations. The permutation that minimizes the mean distance between each substitution and its substituted counterpart provides the best matching. If there are insertions (i.e., response is longer than stimulus), these are distinguished from substitutions by trying all possible combinations of incorrect locations: the combination yielding the minimal mean distance from correct counterparts identifies substitutions, whereas the other incorrect locations are identified as insertions. This method also allows to detect permutations combined with substitutions.
cThe maximum threshold distance was set to 90.14 cm on the MC and 8.50 cm on the Table Corsi; this is the maximum distance between any two adjacent locations (notably, it is distance between number 7 and 8). All couples of adjacent locations are shown in Figure 2.
dThe path is formed by all straight segments connecting two consecutive locations, starting from number 0.