# Introducing random slopes in models with nested random effects

I'm trying to see how latency to emerge (response variable) is varies with time (trials). Individuals (ID) are nested within colonies. The nesting is such that individuals 1-20 belong to colony 1, 21-40 to colony 2, and so on. I would like to see how latency of individuals vary with time by fitting a random slope model, and at the same time accounting for the non-independence of individuals from colonies. I'm confused which is the correct formula because the outputs are different:

m1<-lmer(latency~Trial+(1|colony:ID)+(Trial|ID),data=mydata)

m2<-lmer(latency~Trial+(1|colony/ID)+(Trial|ID),data=mydata).

Output of m1 gives interactions of all individuals with trials. However, output of m2 says "singular fit".

Thanks much!