Here is a modern rendition of an old audio filter,
the "Selectoject." I think it was originally invented by O. Villard of SRI/Stanford
( 0. G. Villard, Jr., and D. K. Weaver, "The selectoject," QST,
vol. 33, pp. 11-17; November, 1949.)  and published in ARRL Handbooks in the 50's and 60's.
This circuit has been rediscovered on occasion, (Very high-Q insensitive active RC networks
Tarmy, R.; Ghausi, M.;, IEEE Transactions on Circuit Theory, Volume 17, Issue 3,
Aug 1970 Page(s):358 - 366,) for example.
It's much easier to get high Q than with a Sallen-Key or infinite gain multiple feedback filter.
One way this filter is used is to peak an interfering signal, then switch modes to null it out.
Another use is selecting the desired signal in the peak mode.
It seems that National had a Selectoject built into the NC125, as seen in the bama archive schematic.
My only claim is the simplified switching between peak and null from the Handbook versions and
the use of opamps instead of transistors or tubes.

As is my custom, here is a spice-like annotated netlist:

c1 16 3 0.1uf (input/blocking cap)
op1 3 2 1 (+in -in out) (opamp)
r1 1 6
r1 6 7
op2 5 6 7
r2 7 9
r2 9 8
op3 10 9 8
r3 8 13
r3 13 14
op4 12 13 14
r3 2 15
r3 2 15
r4 2 1 (variable, max= 2*r3) (peak/null adjustment)
switch 15 8 13 (common A B) (spdt, peak/null select)
r5 1 5 (variable, 2-gang for tuning)
c2 5 11 (sets tuning range with r5)
r5 7 10
c2 10 11
r6 3 12 1meg (input bias)
r7 12 4 100k (supply splitting)
r7 12 11
c3 12 11 10uf (+,-) (bypass/decoupling cap) 
c4 14 17 1uf (output coupling/blocking cap)
input 16 11
output 17 11
batt 4 11 9v (+,-) (supply voltage to suit opamps)


notes:
repeated reference designators require matched values.
r1, r2, r3 may all be equal; an eight resistor 16-pin dip
pack has adequate matching. Typical useful values here are in the 22k to 100k range.
the parallel combination of r3 across 2,15 is critically proportioned to r3 8,13 and r3 13,14.
the wiring at node 13 needs to be minimized. since it is an opamp summing junction, extra distributed
capacitance affects the amplifier's stability. a small cap across nodes 13,14 (0.1pf to 10pf,say) will
help compensate this issue if necessary; switching the summing junction represents a legitimate
design compromise.
r5 and c2 have less severe tracking/matching requirements than r1, r2 and r3. nominal values will be okay.
typical values might be 100k and 0.47 uf. adding end resistance to r5 of the range 1% to 10% will
improve the tuning functioning at the high frequency end; audio taper pots might be better than linear
pots. the slow taper end should be connected to the wiper for this to work properly; depending on the
direction of the taper, the frequency calibration marks should either be on the panel or else be on the knob
skirt; one or the other will yield the desired frequency increasing clockwise labelling.
r4 is roughly twice the value of r3, (r4=50k, r3=22k, 27k or 33k; r4=100k, r3=47k or 56k; etc.)
lots of quad opamps might work here: lm324 or your choice; put in a socket and try different types;
the space for these notes is way too small to embody a treatise on amplifier selection.
supply voltage might need adjusting based on the particular type.
low impedance headphones definitely need a buffer of some sort, perhaps an emitter follower,
lm386, lm380, etc.