The paper by Cramer and colleagues illustrates how a network approach can model personality systems without positing causal latent factors such as the Big Five. We applaud this effort but argue that nodes should be distinguished on more than quantitative grounds (e.g. displayed centrality or connectivity). To realistically model the affects, cognitions and behaviours that constitute real personalities, organizing constructs such as needs and comparators seems necessary. Incorporating them requires greater consideration of functionalist personality theories that link together environmental features and adaptive behaviour in meaningful and stable ways. Copyright © 2012 John Wiley & Sons, Ltd.