One afternoon we had a meeting which was intended to help Huang Yiyi come up with a new avenue of approach after the recent failure. She didn’t say a word from start to finish, and so I was left to mention the two main points: one was to do with the rate at which analysis was proceeding, which reflected the standard of our performance. To begin with we were having problems even reaching the rate of two in a thousand, but now we were achieving five in a thousand, and both the speed and extent of this increase were very pleasing. However, looking at it from a cryptographic point of view, even though the rate of analysis was improving rapidly, the gold content wasn’t very high. What did I mean by that? Even though we were picking out some words and phrases, and some numbers, we were getting only a tiny proportion of the key words and phrases. The majority of what we were finding were the code-names for army units, their designations, people’s names and dates. I would guess that we were getting roughly 87 per cent of the people’s names. That means that we were picking out the ordinary stuff but not what actually mattered, and that was not a good state of affairs. In a good situation maybe the rate of analysis isn’t so high, but you get what matters, even if the rest is a tissue of holes. Right now we had far too many holes, and what is worse the holes were all the important ones. The second thing was a request, or perhaps it would be better to say a demand. I wanted the comrades in the analysis department to take back all the intercepts that they had worked through and analyse them again. There was a good reason for doing that, which was as follows: we were getting our newspapers about ten days to two weeks after they had been published, which means that we were losing any connection with events on the ground. If we did the analysis in the light of what was being reported on the day, we might make new discoveries.

As the facts subsequently demonstrated, my idea was right, and there was a great improvement in the quality of analysis of the radio intercepts. Comrade Chen was the immediate beneficiary of this: a couple of days later he burst into my office to tell me the good news. He had decrypted one entire message. The contents of this secret communication were: ‘Operation Wolf’ was under way, and the agents should wait in the usual place for ‘Banana’ …