Sequence motif search remains an important problem in molecular biology. Many current tools for motif search utilize the expectation –maximization algorithm. Commonly, the algorithm assumes that a sub-sequence is produced by either the background distribution or the motif distribution. Likelihoods of having been sampled from the two distributions are computed for a sub-sequence. The ratio of the likelihoods is the subsequence score. Is background really needed? Definition of motif does not require background. Thus, the existence of background is hypothetical information that can lead the search astray. This work shows that the expectation-maximization algorithm does not have to utilize the background distribution to be practical. The relation between the number of logL minima and the number of sequences is exponential. Nevertheless, the time to find the constant fraction of motifs depends on the number of sequences linearly. Thus, relatively large sequence sets can be analyzed by the EM without the introduction of background.