Young-of-year (YoY) Atlantic silversides were collected from Mumford Cove (41.32N, 72.02W), a small, salt-marsh dominated embayment in eastern Long Island Sound (Connecticut USA, 0.5 km², 1-3 m depth). From a biweekly beach seine survey (30×2m seine with 3mm mesh), three samples on 23 October, 20 November, and 18 December 2015 were used. On each date, samples were pooled from two independent seine casts. All silversides were euthanized, enumerated, measured for total length (TL) to the lower 5 mm and preserved frozen (-20°C). From each collection, ~100 individuals were selected for otolith microstructure analysis using a TL-stratified random sampling design. For each individual, we again measured TL (nearest 0.1 mm) and determined its sex via visual inspection of gonads, before extracting both sagittal otoliths. Sagittae were mounted on microscope slides using CrystalBond 509 thermoplastic cement. If equally suitable, the left or right sagitta was randomly chosen for analysis. Otoliths were hand-polished using 9 µm then 3 µm lapping films (3M) until daily increments along the entire reading axis were clearly visible under 400x magnification (Nikon Eclipse E400 compound microscope). Otoliths were measured and read across the sagittal plane from the nucleus to the dorsal or ventral otolith edge, because the more conventional core to post-rostrum axis proved to be too curved to be reliably interpretable. Increments were enumerated and measured using Image Pro Premier (V9.1) connected to a Luminera Infinity2-2 digital camera. For each otolith section, multiple focal planes were captured and merged into multi-layer images to aid the reader interpreting the otolith microstructure. The radius of the hatch-check (µm) was measured. The last growth increment was presumed to be incomplete and thus excluded from growth analyses. Increment number was assumed to correspond to an individuals’ age in days post hatch (dph). Hatch date was calculated by subtracting age from the date of collection, while the formation date of each increment was calculated by adding the increment number to the hatch date.

To estimate the thermal history of October YoY we used continuous temperature data recorded by a Manta Sub2 probe (Eureka© Water Probes) at our study site. Three small gaps in the record (<8 days) were linearly interpolated. Daily temperatures during a larger data gap (34 days, 18 June - 22 July) were estimated via linear regression between the Mumford Cove dataset and temperatures in nearby Niantic Bay (2015-2016) that were strongly correlated (TMumford= 1.07*TNiantic – 0.54 r²= 0.92, F = 6024.9, p < 0.001). Average daily temperatures were then linked to the corresponding day of increment formation for each individual. To quantify the thermal dependency of larval growth in October YoY, we calculated the mean experienced temperature during the first 30 dph for each individual and correlated it to its mean daily growth rate over the first 30dph (GR30) and its back-calculated TL at day 30 post hatch (TL30, a proxy for the end of the larval stage).