The fact that the span of apprehension averages only four or five... probably results from the high rate of encoding. In a tachistoscopic experiment … - Ulric Neisser

To cope with this difficulty [of limited capacity], even a mechanical recognition system must have some way to select portions of the incoming information for detailed analysis.

If we allow several figures to appear at once, the number of possible input configurations is so very large that a wholly parallel mechanism, giving a different output for each of them, is inconceivable.

The attentive synthesis of any particular letter or figure takes an appreciable time, of the order of 100ms...If a whole row of letters is to be identified, they must be synthesized one at a time... To "identify" generally means to name, and hence to synthesize not only a visual object but a linguistic-auditory one... Hence the span of apprehension is limited to what can be synthesized, and then verbally stored.