diff --git a/src/TableReader.jl b/src/TableReader.jl
index b45f8d9..fa5bc95 100644
--- a/src/TableReader.jl
+++ b/src/TableReader.jl
@@ -531,7 +531,7 @@ function readdlm_internal(stream::TranscodingStream, params::LexerParameters)
                     # not a date column
                 end
             elseif is_datetime_like(col)
-                hasT = occursin('T', col[1])  # check delimited by T or space
+                hasT = is_T_delimited(col)  # check delimited by T or space
                 try
                     columns[i] = parse_datetime(col, hasT)
                 catch
diff --git a/src/parser.jl b/src/parser.jl
index 76dd75d..75d38b9 100644
--- a/src/parser.jl
+++ b/src/parser.jl
@@ -267,6 +267,15 @@ function parse_datetime(s::String, hasT::Bool)
     end
 end
 
+function is_T_delimited(col::Vector{<:Union{String,Missing}})
+    for x in col
+        if !Base.ismissing(x)
+            return occursin('T', x)
+        end
+    end
+    return false
+end
+
 function is_datetime_like(col::Vector{<:Union{String,Missing}})
     # Check if the first three strings (if any) are datetime-like.
     i = 1
diff --git a/test/runtests.jl b/test/runtests.jl
index 7ba580d..121984d 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -854,6 +854,18 @@ end
         """
         @test_throws TableReader.ReadError("invalid file format at line 2, column 1 (found 0xff)") readtsv(IOBuffer(data), quot = nothing)
     end
+
+    @testset "datetimes" begin
+        buf = IOBuffer("""
+        col1,col2
+        1,
+        2,2018-04-10T08:19:30.000
+        """)
+        df = readcsv(buf)
+        @test df[:col1] == [1, 2]
+        @test ismissing(df[1,:col2])
+        @test df[2,:col2] == DateTime(2018, 4, 10, 8, 19, 30)
+    end
 end
 
 @testset "readdlm" begin