A number of points remain unclear from the studies presented so far. Study One suggested that priming may occur between both associated and non-associated
collocations. However, the small number of ‘non-associated’ collocations used in that study left this conclusion in need of further support. Moreover, Study One used only very high frequency collocations. It is still not clear, therefore, if priming can also be found between more moderate-strength collocations or if it is particular to these extreme cases. Study Two attempted to address these issues, but failed to find any evidence of priming. I have suggested that the failure to find any priming in this study was due to the fact that semantically-plausible control items were used, but this interpretation stands in need of direct experimental confirmation. Furthermore, if it is true that using semantically-plausible controls attenuates priming effect, this suggests that semantically-plausible non-collocations may themselves exhibit priming, in comparison to less plausible word pairings. The existence of priming between such items would be of considerable interest, suggesting that priming is at least in part a result of participants’ being able to construct a semantically-plausible context for a word pair, rather than simply a matter of co-occurrence frequencies. This possibility also requires further investigation.
The present study aims to clarify these issues. The experiment reported here is similar to that in Study Two, but differs in that it uses non-attested word pairs as controls and adds an additional level of attested, but low frequency, pairs as test items. The use of non-attested controls aims to test whether the failure to find priming in Study Two was due to our use of attested controls in that study. Since the procedure and
collocating items used in the current study remain the same as those in Study Two, if we find any evidence of priming here, we can conclude that our previous failure to find priming was indeed the result of using attested controls. It will also enable us to ask again whether priming exists for moderate-strength collocations and for
collocations which are not normative associates, as well as for very high frequency, associated collocations. The addition of a level of attested, but low frequency, word pairs aims to test whether semantically-plausible pairs which are not collocations themselves demonstrate priming.
Materials
The present study will compare facilitation between four types of word combinations, relative to that between non-attested pairs:
• Level 0 pairs are low frequency pairs, being attested in the BNC between two and four times and having MI scores of less than 2 and t-scores of less than 1.5; • Level 1 pairs are moderately strong collocations, having MI scores of 4-5
(Mdn = 4.47) and t-scores of 4-8 (Mdn = 5.52), and are not strong normative associates;
• Level 2 pairs are strong collocations, having MI scores of more than 6 (Mdn = 7.65) and t-scores of more than 7.5 (Mdn = 10.95), and are not strong
normative associates;
• Level 3 pairs are strong collocations, having MI scores of more than 5.5 (Mdn = 7.01) and t-scores of more than 6 (Mdn = 10.63), and are strong normative associates.
As in Study Two, there are 16 test pairs for each level. The test items for levels 1-3 are identical to those used in Study Two. Control items were created by re-pairing
adjective-noun combinations from the test items such that the new pairs are not attested as occurring adjacent to each other in the BNC. The final lists used in this study are shown in Appendix Aiii.
Participants
32 students at the University of Nottingham. All were native speakers of British English.
Procedure
The procedure for this experiment was similar to that in Study Two. The only difference was that, since the extra experimental level meant that a larger number of items were used, items were presented in two blocks, with a self-paced break between blocks. Each block contained an equal number of items from each level, and the order of presentation of the blocks was counterbalanced between participants.
Results and discussion
Reaction times of less that 250ms and greater and 1,250ms (4.3% of the total) were excluded from analysis. Mean accuracy by participant, collapsed across conditions was 96%. There was no effect of condition upon accuracy. Average reaction times for collocations and non-collocations at each of the three levels are shown in Table 12. Reaction times were not normally distributed within conditions.
Table 12: average reactions times in each condition
collocations non-collocations Level 0 Median RT (ms) 522.23 528.82
Level 1 Median RT (ms) 511.97 525.07
Level 2 Median RT (ms) 520.50 521.32
Level 3 Median RT (ms) 506.75 529.07
Though median reaction times were somewhat lower for collocating than for non- collocating pairs at all levels, this difference was statistically significant only at level 3: i.e. between normatively associated word pairs, where a strong and highly significant facilitation effect was found (analysis by participants: T = 72.0, p < .001, r = -.45; analysis by items: T = 0, p < .001, r = -.62).
This result clarifies a number of points which had remained ambiguous after Studies One and Two. Firstly, it appears that high frequency collocations which are not strong normative associates (i.e. items at Levels 1 and 2) do not prime each other in the same way as do pairs which are strongly associated. The priming seen in Study One may therefore have been an effect restricted to the few items used in that study. Secondly, it appears that our failure in Study Two to find priming between strongly associated pairs was due to the nature of the control items used. It seems that priming can be found between associated pairs only when relatively implausible pairs are used as controls. Finally, low frequency but semantically plausible pairs do not appear themselves to exhibit priming in these circumstances. This result is unsurprising in light of the fact that priming was also not detected between much higher frequency pairs, unless they were psychological associates.