Table 2. A total of 1357 schoolchildren were individually asked to assess single or double items shown on a screen as being nice or mean (see Fig. 1).

	Dl; F	P
{1} Image category	1; 84.12	0.001 (***)
{2} Sex (within {1})	1; 1.01	0.36 (NS)
{3} Age (within {1})	16; 2.76	0.001 (***)
{4} Adjective order (within {1})	2; 0.14	0.87 (NS)
{5} Image type (within {1})	14; 12.41	0.001 (***)

Images were arbitrarily given a reference depending on the presence or absence of aposematic signals; single item images: 0 or 1 aposematic signal; double item images: 0 versus 1 or 1 versus 2 aposematic signals. We then scored the responses as correct assignment of nice and mean. A General Linear Model was run with age, sex, adjective order and image type as nested categorical factors into item category (single or double). Probability values follow the coding: ***P < 0.001; **0.001 < P < 0.01; *0.01 < P < 0.05; M 0.05 < P < 0.10; NS P > 0.10.