The Midstream Order Deficit
When many impressions follow in excessively rapid succession in time,
although we may be distinctly aware that they occupy some duration, and are not simultaneous, we may be quite at a loss to tell which comes first and which last- William James (1890), The Principles of Psychology, ch. 15.
The phenomenon James informally reported may be the same as what we have experimentally documented and term
the Midstream Order Deficit (
Holcombe, Kanwisher, & Treisman, 2001;
is a difficulty in apprehending the
order of successive items in the middle of a rapidly presented stream.
Subjects know that in the cycling condition the same sequence is
presented over and over and that they can begin their
report with any letter. Thus in the above cycling condition
DMPG, MPGD, PGDM, and GDMP were all counted as correct, yet still
subjects did much worse in the cycling condition.
What it means
- Endogenous attention is slow
In the cycling condition, if you could selectively attend to any
four successive letters and ignore the rest, there would be no
problem. But people can't, even at rates as slow as 4
items/second (4 per second is slow by the
standard of "RSVP" experiments. For example, when you don't put
a mask between the items, as I didn't, it's too slow to get an
attentional blink). At rates significantly slower than 4 per
second, the task becomes extremely easy. This may be because
such rates are slow enough for endogenous attention to segment
- An anchor or subjective
beginning is needed to encode order into memory
that in the cycling condition shown above, the letters begin
extremely quickly and then gradually slow down. This is
crucial to the effect. When the cycling condition does
not have this "fade-in" at the beginning, it is
not harder than the single condition. In fact,
in my experiment, when there was no fade-in in the cycling
condition, observers did even better than in the single
condition! The presence of the fade-in makes the cycling
condition hard by eliminating the clear subjectively-first item
or anchor that usually occurs at the beginning of lists.
When there is no identifiable first item, the order is extremely
difficult to encode. Our experiment suggests that the ending
item is not nearly as important, thus we believe that an anchor
is needed at the beginning, but not the end, of a list to encode
The Midstream Order Deficit does NOT occur due to an
The data from our experiments suggests that in the conditions
of our experiments, no attentional blink occurred. See the paper for more info.
A similar effect occurs for repeated sequences of
auditory stimuli (Warren, Obusek, Farmer, & Warren, 1969),
and for repeated sequences with one visual, one auditory, and one tactile stimulus (Hirsh, 1976).
This suggests that a central attentional limit prevents ordering of