That was just added by someone who's primary language probably isn't English. What they meant to say is "QPSK31, a mode that is related to PSK31, uses..."
Other than that, the claim "it uses more constellation points, that's a means error-correction coding", is plain wrong.
The error coding used is this, just one sentence later:
A sliding window of five bits is used to select one of the four possible phase shifts, providing a means of error correction by spreading each bit of data across adjacent bits.
I'd word this completely differently. Maybe like this.
Error correction is used: a rate-1/2 channel code is implemented with a constraint length of 5 bits (i.e. the past 5 input bits are used to determine the next output two bits). Two bits of the resulting bitstream are then grouped into one QPSK symbol. Using QPSK instead of BPSK allows the doubled data rate to be sent through exactly the same bandwidth without reducing robustness.
The whole paragraph is confusingly worded, to put it mildly. I've reworded it to be clearer (and not make incorrect claims, like that you need 20 2-bit symbols to decode a constraint length 5 convolutional code – that's plain wrong).