An unusual study reports the effects of emoticons on human brain activity: Neural correlates of text-based emoticons South Korean neuroscientists Ko Woon Kim et al. used fMRI to record brain activation in 18 volunteers who were shown various expressive text symbols, in both the Asian 'vertical' and Western 'horizontal' styles:
However, it turned out that the brain doesn't really respond to emoticons at all: there was no significant difference in the brain response to the real emoticons compared to the 'scrambled' stimuli, which were meaningless combinations of punctuation marks. :-( Despite this, there was some evidence that sad emoticons produced lower activity than the other stimuli in certain brain regions:
The authors conclude that
In this study, we found that text‐based emoticons were processed in the face‐specific areas: FFA (fusiform face area) and OFA (occipital face area). This result implies that the text‐based emoticons are processed like facial expressions in the human brain
Although this seems a bit of a stretch, since scrambled stimuli also activated these areas. This isn't the first emoticon-based fMRI study. I found three others (2008, 2011, 2014) - two of them also from South Korea, which is clearly the world leader in neuroemoticonology. For what emoticons reveal about personality, see here.
Kim KW, Lee SW, Choi J, Kim TM, & Jeong B (2016). Neural correlates of text-based emoticons: a preliminary fMRI study. Brain and Behavior, 6 (8) PMID: 27547502