Mate preference learning, including imprinting-like learning, is pervasive across animal taxa, and can affect the selection and maintenance of certain phenotypes. However, not much is known about the temporal dynamics behind imprinting-like learning, or the genetic underpinnings underlying it. To uncover the temporal dynamics of imprinting-like learning, from both a behavioural and transcriptional perspective, we conducted learning and RNA-Seq time series using the butterfly Bicyclus anynana, a species where both sexes learn mate preferences. We exposed females to an unfamiliar, unpreferred male phenotype (4-spotted male) for five different exposure (training) periods, ranging from 30 minutes to three hours, and recorded their choice between the preferred, familiar male phenotype (2-spotted) and 4-spotted males in mate choice trials conducted two days after training. We also assessed differential gene expression of naive females and females exposed to trainer males for these same five exposure periods, to identify temporal patterns in gene expression associated with learning. While we found that the longest exposure had the strongest effect on preference learning, we did not observe a linear effect of exposure time on learning. We also show that the highest peak of differentially expressed genes (DEGs) was after one hour of exposure. While a number of genes were uniquely DE at each time point, one gene, associated with transcription initiation, was differentially expressed during learning across all five time points. We observed a similar decreasing trend in both gene expression and learned response after 1.5 hours of exposure, offering new insights to possible attention and forgetting mechanisms. Therefore, our results indicate that both gene expression and learning are temporally dynamic, and not linear through time. They also highlight the role of transcription in mate preference learning and memory formation in Lepidoptera, and illustrate that imprinting-like learning may exhibit similar molecular temporal dynamics as associative learning.