1. In order to reappraise work by Stokes (1962a, b) and subsequent int
erpretations of his results, a data set involving 906 dyadic Blue Tit
interactions was collected and analysed using video playback. 2. A num
ber of shortcomings in Stokes' original approach were highlighted, inc
luding his broad definition of attack and a lack of information on the
true frequency of behaviours and of combinations of elements at an in
stance in time. 3. A total of 16 behaviours and actions with a clear a
gonistic context were identified. Similarity indices (Jaccard Index) w
ere calculated based on the co-occurrence of behaviours at an instance
in time. No behaviours were associated at a high enough level to warr
ant their consideration solely as compound displays. 4. True attack wa
s rare, unpredictable and effective. Escalated fights occurred in < 1%
of interactions, with peck attacks which were not escalated into figh
ting in a further 2% of interactions. The rarity of true attack makes
it more useful to look at the relationship between behaviour and inter
action outcome (win/lose), than between behaviour and subsequent actio
ns (attack/stay/escape).