By tracing the flow of computations in the iterative decoders for low-densi
ty parity-check codes, we formulate a signal-space view for a finite number
of iterations in a finite-length code. On a Gaussian channel, maximum a po
steriori (MAP) codeword decoding (or "maximum-likelihood decoding") decodes
to the codeword signal that is closest to the channel output in Euclidean
distance. In contrast, we show that iterative decoding decodes to the "pseu
dosignal" that has highest correlation with the channel output. The set of
pseudosignals corresponds to "pseudocodewords," only a vanishingly small nu
mber of which correspond to codewords, We show that some pseudocodewords ca
use decoding errors, but that there are also pseudocodewords that frequentl
y correct the deleterious effects of other pseudocodewords.