The magnitude and timing of Holocene maximum warmth in the Arctic and sub-Arctic has been the subject of considerable recent interest, particularly in the context of future climate change. Although lying at a crucial location in the North Atlantic close to significant atmospheric and oceanic boundaries, terrestrial Holocene climatic data from Iceland are few and predominantly derive from glacial and palaeoecological evidence. Here we present new datasets from Tröllaskagi, based on chironomid-inferred temperatures (CI-T), using sub-fossil chironomids from the same lake sediments supplemented by pollen data. July air temperatures have been derived using an Icelandic training set, and the data suggest optimal temperatures at sea level up to 1.5 °C above current levels around 8 k cal. yr BP, a time when birch woodland was well developed in Tröllaskagi, but when woodland had still not fully developed in the more isolated NW peninsula. Our data thus suggest that optimal summer warmth did not occur in Iceland until 8 kcal. yr BP at the earliest, possibly lasting until 6.7 kcal. yr BP. The amount of warming for July was therefore at least 1.5 °C, but possibly up to 2–3 °C higher than the 1961–1990 average on the basis of the tree-line data. Comparison with data from elsewhere in adjacent Arctic regions, Greenland and Eastern Arctic Canada show peak warmth to be later in Iceland, and less pronounced. It also appears that there were enhanced temperature gradients during the first half of the Holocene between the two study areas Tröllaskagi and the NW Peninsula and that they influenced patterns of vegetation colonisation, with current spatial temperature patterns only developing as Holocene climate deteriorated after around 6 kcal. yr BP.