Wainwright, Leatherdale, and Dublin (2007) evaluated the advantages and disadvantages of using an ANOVA to analyze data with animal models. They explained that animal pups are often raised or clustered together. In terms of the assumptions for an ANOVA, they state, 

An analysis that is based on individual pups without taking this clustering into account is likely to over-estimate the statistical significance of any observed effect. Similar concerns apply to [within-subjects] analyses, where the errors in an individual over time are likely to be [related], violating statistical assumptions of independence. (Wainwright et al., 2007, p. 665) 

Explain why housing rat pups in clusters might violate the assumption of independence for a within-subjects design.