diff --git a/src/pomdps.jl b/src/pomdps.jl
index d6712ed..ac0197d 100644
--- a/src/pomdps.jl
+++ b/src/pomdps.jl
@@ -16,9 +16,7 @@ function predict!(pm, m::POMDP, b, a, rng)
     all_terminal = true
     for i in 1:n_particles(b)
         s = particle(b, i)
-        if isterminal(m, s)
-            pm[i] = undef
-        else
+        if !isterminal(m, s)
             all_terminal = false
             sp = generate_s(m, s, a, rng)
             pm[i] = sp
diff --git a/test/runtests.jl b/test/runtests.jl
index d4c0a4c..4f8cb6e 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -4,6 +4,7 @@ using POMDPModels
 using Test
 using POMDPPolicies
 using POMDPSimulators
+using POMDPModelTools
 using Random
 using Distributions
 using NBInclude
@@ -77,6 +78,17 @@ struct ContinuousPOMDP <: POMDP{Float64, Float64, Float64} end
     end
 end
 
+struct TerminalPOMDP <: POMDP{Int, Int, Float64} end
+POMDPs.isterminal(::TerminalPOMDP, s) = s == 1
+POMDPs.observation(::TerminalPOMDP, a, sp) = Normal(sp)
+POMDPs.transition(::TerminalPOMDP, s, a) = Deterministic(s+a)
+@testset "pomdp terminal" begin
+    pomdp =  TerminalPOMDP()
+    pf = SIRParticleFilter(pomdp, 100)
+    bp = update(pf, initialize_belief(pf, Categorical([0.5, 0.5])), -1, 1.0)
+    @test all(particles(bp) .== 1)
+end
+
 @testset "alpha" begin
     # test specific method for alpha vector policies and particle beliefs
     pomdp = BabyPOMDP()