2.12 Modulations of the N400 and morphological priming
3.2.2 Materials
The experimental items consisted of 240 pairs of words. The words in 120 of these pairs were derivationally related to one another (e.g. ‘ruin’ - ‘ruinous’), whilst the words in the second set of 120 pairs were only formally related (e.g. ‘scan’ - ‘scandal’). In order to facilitate discussion of these various word types the following nomenclature will be adopted. Words such as 'ruin' will be referred to as 'root words' and words such as 'ruinous' will be described as 'derived words'. Words such as 'scan' will be referred to as 'formal root words' and words such as 'scandal' will be referred to as 'formal words'. The 120 derivational words were all suffixed forms, and there were 23 different types of suffix used. The derivationally- and formally- related word pairs are given in Appendices I and 2 respectively.
The two sets of 120 word pairs were chosen such that the mean word lengths and mean word frequencies (Francis & Kucera, 1982) of the root and formal root words were approximately equal, and the word lengths and frequencies of the derived and formal words were also
approximately equal (Table 3.1). Because of the limited number of possible experimental pairs, the frequencies of the root items were allowed to vary over a relatively wide range. As a result, the mean frequency of the root words was somewhat higher than that for the derived and formal words. This asymmetry was, however, equally true of both word sets. Within each word type, the critical words were rotated around the unrepeated, partial priming and repetition priming. Thus for each word type, across subjects waveforms in each experimental condition were elicited by the same set of words.
A further constraint on the choice of items was a requirement that the degree of orthographic overlap in the derivationally related pairs was matched by a similar degree of overlap in the formally related pairs. The degree of overlap was estimated by determining the ratio of the number of letters the prime and probe words had in common to the length of the probe word.
Prime Words Probe Words
Length Frequency Length Frequency Overlap Derived 4.5 (0.8) 29.2 (30.6) 6.8 (0.9) 3.2 (3.2) 0.68 (0.04) Formal 4.2 (0.7) 28.1 (34.6) 6.4 (1.1) 4.6 (4.0) 0.64(0.06)
The number of common letters was defined as, proceeding from left to right, the number of letters up to, either the first letter in the root word that differed from the corresponding letter in its paired word, or to the end of the root word. Thus for the pair ‘scan’ - ‘scandal’, there are four letters in common (i.e. 'scan') and the formal word is 7 letters long, producing a ratio of 0.57. For the pair ‘raise’ - ‘raisin’, there are also four letters in common, i.e. ‘rais’. However, since, in this example, the formal word is 6 letters long, the ratio is 0.67.
Each of the two sets of 120 word pairs was divided into three sub-sets of 40 pairs, with each of these sub-sets being matched for word length, word frequency and degree of orthographic overlap between paired words.
In addition to these experimental items a further 82 words were selected from Francis and Kucera (1982). They were selected so as to cover a similar range of word frequency and word length as was found in the experimental items. From these 82 items, 78 were used to produce pronouncable non-words. This was done by changing one or two letters of the word. Two of these non-words together with the 4 words not used to construct non-words were used as filler items. Twenty four of the non-words were used to form a set which were repeated during the experiment. A further set of 24 non-words were used to create a group of so-called 'pseudo derived' non-words. This was done by adding a legal derivational suffix to the non-word. Thus from the non-word 'frab' the pseudo-derived non-word 'frabbage' was formed. These non word/pseudo-derived non-word pairs (i.e. ‘frab’ - ‘frabbage’) were also presented during the experiment. The remaining 28 non-words were presented once during the experiment. The derived items from one sub-set of 40 derivationally related pairs were repeated (e.g.
‘ruinous’ - ‘ruinous’ referred to hereafter as 'derivational repetition priming' ). The formal words from one of the sub-sets were also repeated (e.g. ‘scandal’ - ‘scandal’ ; 'formal repetition priming'). The derived words from a second set of derivationally related pairs were primed by their root words (e.g. ‘ruin’ - ‘ruinous’ ; 'derivational partial priming' ), whilst the formal words from a second of the formally related sets were primed by their formal root word ( e.g. ‘scan’ - ‘scandal’ ; 'formal partial priming' ). Twenty root words from the third sub-set of derivationally related pairs and 20 formal root words from the third sub-set of formally related pairs were also repeated (e.g. ‘ruin’ - ‘ruin’ ; ‘scan’ - ‘scan’). Of the remainder of these final subsets of
derivational and formal word pairs, 10 root words and 10 formal root words were presented unrepeated, as were the remaining 10 derived words and 10 formal words. The repetition of root words and the single presentation of both root, formal root, formal and derived words reduced the degree to which the occurrence of a word of a particular type on one trial was not predictive of the type of word which would occur on the immediately succeeding trial.
Items presented as the first of a pair of words in which the second word is either a repeat of, or is related to, the first, will be referred to as the 'prime words', whilst the repeated or related items presented second will be referred to as the 'probe words'.
The 440 words were combined with the 124 non-words and 6 filler items in a pseudo-random order, producing a sequence of 570 items, sub-divided into 6 blocks of 95 items. This ordering was constrained such that :
(i) prime words and the corresponding probe words and items in the non-word pairs occurred on immediately succeeding trials, e.g. the lag between prime and probe was 0.
(ii) a filler item was presented on the first trial of each block.
(iii) the experimental items were evenly distributed across the blocks, as were the different types of non-word.
Two further lists were produced by rotating the three sub-sets of each word type around conditions. A further three experimental lists of stimuli were produced by again rotating sub sets of items around conditions, but using a different pseudo-random ordering of conditions.