diff --git a/src/parsing.jl b/src/parsing.jl
index 09d237e6..68a6f68f 100644
--- a/src/parsing.jl
+++ b/src/parsing.jl
@@ -421,6 +421,31 @@ fun_to_vec(ex::QuoteNode;
            outer_flags::Union{NamedTuple, Nothing}=nothing,
            allow_multicol::Bool = false) = ex
 
+# Catch-all method for literal values (Bool, Int, String, etc.).
+# Behaviour mirrors the underlying `DataFrames.subset` exactly, keyed on the
+# `@byrow` flag:
+#   * byrow  -> `[] => ByRow(Returns(ex))`, a per-row constant vector, same
+#               as `subset(df, [] => ByRow(Returns(true)))` (which works).
+#   * !byrow -> a bare literal is a scalar predicate, which `subset` rejects
+#               (`subset(df, [] => Returns(true))` errors). Raise a clear
+#               ArgumentError in place of the #259 MethodError.
+# ByRow is applied only when byrow is set; applying it unconditionally would
+# make `@subset(df, true)` accept what the function rejects.
+function fun_to_vec(ex;
+                    no_dest::Bool=false,
+                    gensym_names::Bool=false,
+                    outer_flags::Union{NamedTuple, Nothing}=nothing,
+                    allow_multicol::Bool = false)
+    byrow = outer_flags !== nothing && outer_flags[BYROW_SYM][]
+    if byrow
+        return :([] => $ByRow($(Base.Returns)($ex)))
+    end
+    throw(ArgumentError(
+        "literal value `$ex` is not a valid column expression; pass a " *
+        "column reference or a vectorised predicate (e.g. `:x .> 0`), " *
+        "or use `@rsubset` for a per-row literal"))
+end
+
 
 """
     rename_kw_to_pair(ex::Expr)
diff --git a/test/subset.jl b/test/subset.jl
index 505f1f9b..05f2067d 100644
--- a/test/subset.jl
+++ b/test/subset.jl
@@ -173,4 +173,24 @@ end
     @test @subset!(groupby(copy(df), :g), :c .== :g) ≅ df[[], :]
 end
 
+@testset "@subset / @rsubset with literal values" begin
+    df = DataFrame(A = [1, 2, 3], B = [4, 5, 6])
+
+    # @subset (not byrow): a bare literal is a scalar predicate, which
+    # subset rejects (subset(df, [] => Returns(true)) errors). It must error,
+    # not be silently broadcast. The error is raised during macro expansion
+    # (as the #259 MethodError was), so it surfaces via @eval as LoadError.
+    @test_throws LoadError @eval @subset($df, true)
+    @test_throws LoadError @eval @subset($df, false)
+    @test_throws LoadError @eval @subset($df, true, :A .> 1)
+    @test_throws LoadError @eval @subset!(copy($df), true)
+    @test_throws LoadError @eval @subset(groupby($df, :B), true)
+
+    # @rsubset (byrow): a per-row literal is well-defined and matches
+    # subset(df, [] => ByRow(Returns(true))), which works.
+    @test @rsubset(df, true) ≅ df
+    @test @rsubset(df, false) ≅ df[Int[], :]
+    @test @rsubset(groupby(df, :B), true) ≅ df
+end
+
 end # module