diff --git a/src/composition/learning_networks/signatures.jl b/src/composition/learning_networks/signatures.jl
index b224cd47..d49aace9 100644
--- a/src/composition/learning_networks/signatures.jl
+++ b/src/composition/learning_networks/signatures.jl
@@ -356,7 +356,7 @@ function fitted_params(signature::Signature; supplement=true)
 end
 
 """
-    output_and_report(signature, operation, Xnew)
+    output_and_report(signature, operation, Xnew...)
 
 **Private method.**
 
@@ -375,3 +375,6 @@ function output_and_report(signature, operation, Xnew)
     report = MLJBase.report(signature_clone; supplement=false)
     return output, report
 end
+# special case for static transformers with multiple inputs:
+output_and_report(signature, operation, Xnew...) =
+    output_and_report(signature, operation, Xnew)
diff --git a/src/operations.jl b/src/operations.jl
index efa275ac..7bd4da4d 100644
--- a/src/operations.jl
+++ b/src/operations.jl
@@ -222,9 +222,9 @@ for operation in [:predict,
                   :transform,
                   :inverse_transform]
     quote
-        function $operation(model::NetworkComposite, fitresult, Xnew)
+        function $operation(model::NetworkComposite, fitresult, Xnew...)
             if $(QuoteNode(operation)) in MLJBase.operations(fitresult)
-                return output_and_report(fitresult, $(QuoteNode(operation)), Xnew)
+                return output_and_report(fitresult, $(QuoteNode(operation)), Xnew...)
             end
             throw(err_unsupported_operation($operation))
         end
diff --git a/test/composition/models/network_composite.jl b/test/composition/models/network_composite.jl
index 87e064df..fd493a0d 100644
--- a/test/composition/models/network_composite.jl
+++ b/test/composition/models/network_composite.jl
@@ -645,6 +645,39 @@ end
 
 end
 
+# # STATIC MODEL WITH MULTIPLE INPUTS
+
+mutable struct Balancer <: Static end
+MLJBase.transform(::Balancer, _, X, y) = (selectrows(X, 1:2), selectrows(y, 1:2))
+
+struct ThinWrapper <: StaticNetworkComposite
+    balancer
+end
+
+function MLJBase.prefit(wrapper::ThinWrapper, verbosity)
+
+    data = source() # empty source because there is no training data
+    Xs = first(data)
+    ys = last(data)
+
+    mach=machine(:balancer)
+
+    output = transform(mach, Xs, ys)
+
+    (; transform = output)
+
+end
+
+balancer = Balancer()
+wrapper = ThinWrapper(balancer)
+
+X, y = make_blobs()
+mach = machine(wrapper)
+Xunder, yunder = transform(mach, X, y)
+@test Xunder == selectrows(X, 1:2)
+@test yunder == selectrows(y, 1:2)
+
+
 
 # # MACHINE INTEGRATION TESTS