diff --git a/src/kernal/iec/iec_return.s b/src/kernal/iec/iec_return.s
index c80323b7..8be97a46 100644
--- a/src/kernal/iec/iec_return.s
+++ b/src/kernal/iec/iec_return.s
@@ -36,8 +36,8 @@ iec_return_success:
 
 iec_return_DEVICE_NOT_FOUND:
 
-	; Set IEC line to normal idle state
-	jsr iec_set_idle
+	; Nobody answers: release all lines, so the bus is idle - see https://www.pagetable.com/?p=1135
+	jsr iec_release_atn_clk_data
 
 	; BSOUR is certainly not valid anymore
 	lda #$00
diff --git a/src/kernal/iec/unlsn.s b/src/kernal/iec/unlsn.s
index 227ff3fb..2ef74cdf 100644
--- a/src/kernal/iec/unlsn.s
+++ b/src/kernal/iec/unlsn.s
@@ -42,7 +42,15 @@ UNLSN:
 	; Buffer empty, send the command
 	lda #$3F
 
-	jmp common_open_close_unlsn_second
+common_unlsn_untlk: ; common part of UNLSN and UNTLK
+
+	jsr common_open_close_unlsn_second
+	bcs @1
+
+	; The transmission is over: release CLK and DATA too, so the bus is idle - see https://www.pagetable.com/?p=1135
+	jmp iec_release_clk_data
+@1:
+	rts
 
 } else {
 
diff --git a/src/kernal/iec/untlk.s b/src/kernal/iec/untlk.s
index f26fdd22..c7998dfa 100644
--- a/src/kernal/iec/untlk.s
+++ b/src/kernal/iec/untlk.s
@@ -39,10 +39,11 @@ UNTLK:
 	sta IECPROTO
 }
 
-	; Buffer empty, send the command
+	; Buffer empty, send the command; the talker releases the bus - see https://www.pagetable.com/?p=1031
 	lda #$5F
+	jmp common_unlsn_untlk
 
-common_untlk_tksa: ; common part of UNTLK and TKSA
+common_untlk_tksa: ; the rest of TKSA
 
 	; XXX - for TKSA, is it really the right place to do a turnaround? I have some doubts
 	; (it forces us to send the TKSA command 'manually' sometimes), but