Do conductors make a difference? Answer: yes, of course, but on a sliding scale of relative responsibility that depends on any number of factors; not least, the kind of music- making it turns out to be. With average performances you can assume that the conductor isn't doing much beyond police work, keeping everything in order. With good ones (or bad), something beyond a mere reading of the notes has happened and for better or worse it's usually down to the man with the stick - as it was in most of the concerts I heard this week.

Three involved large-scale choral and orchestral works with equivocally ritual qualities that challenge the conductor to take a view on how they should be presented. The first was Bach's B Minor Mass: a Frankenstein piece assembled from spare parts in the composer's catalogue of old scores, some of them written 25 years before the new material that holds them together. How, in performance, do you make stylistic integrity out of that? And how do you fix the tone of a Catholic Mass set by a Protestant composer with - arguably - no intention that it should ever be performed in a liturgical context? Do you follow Karajan's example on disc and factor in the swing of the censer, an implied pontifical grandeur? Or do you stand by the Reformation and keep the glamour quotient low?

Last Sunday's performance at the Barbican came from the period-specialist Ton Koopman conducting his Amsterdam Baroque Orchestra, and the tone was decidedly Protestant, with modest forces - four singers sharing the solo parts, 24 voices in the chorus - and few attempts at sensual assault. Bach, of course, would have expected no greater numbers, and could only have been delighted by the clarity and focus Koopman gets, conducting from the console of a portable organ like a pharmacist dispensing gentle purgatives across the counter. His approach is interventionist - exaggerating rather than diminishing the stylistic variety in the numbers - but likeable and largely satisfying.

I just wish his purgatives weren't quite so gentle. Conductors like Eliot Gardiner give more bite to the attack, more drive to the rhythms, and more transcendence to the spirit of the piece. However modest Bach's original resources, the ambition of, say, the Sanctus looks beyond them. I also wish Koopman wouldn't cut the final chords so short. It's probably his response to a non-ecclesiastical acoustic - if the sound isn't going to linger in the air, why try to sustain it? - but there are ways to suggest resonance when it doesn't actually exist. Clipped cut-offs strike me as defeatist.

Simon Rattle's magnificent performance of Britten's War Requiem with the CBSO at the Festival Hall on Wednesday was, by contrast, a triumph over acoustic limitations. No one on the near side of insanity would think the dry, dead Festival Hall a good place to hear this monumental score: it needs the Albert Hall, or a cathedral. But Rattle engineered the performance so expertly that he capitalised on the particularity a dry sound offers, with incredible finesse of detail in the chamber sections, and elsewhere fabricated an illusion of spaciousness from careful measurement of the accumulating climaxes. The only thing that didn't work was the offstage children's choir: too distant and too un-celestial. Otherwise this Requiem was electrifying. Rattle led it like a journey down to the abyss and back, with an unflagging nervous energy and three strong soloists in Robert Tear, Simon Keenlyside and Andrea Gruber (a voice of spectacular liquidity, well-suited to Galina Vishnevskaya impersonations in the Sanctus).

The ancient legend of Persephone is another return-journey to the abyss; and when Stravinsky set it for chorus and orchestra as part of his 1920s/30s "Greek Trilogy", the intention was to create a piece that would be staged like a static ritual half- way between oratorio and theatre. But where the half-way line would fall was never quite agreed with his librettist Andre Gide, who had different ideas about the piece. Minds didn't meet. And Gide's mind closed against the whole project when he realised how free Stravinsky's music would make with his French prosody. From then on, Persephone acquired the reputation of a misfire in the Stravinskyan canon: pale, pedestrian and lifeless.

But Andrew Davis's revelatory performance with the BBCSO and Chorus at the RFH last weekend should have changed some minds. The music still seemed pale, its tonal values registering like shades of grey in half-light, but with unexpected subtleties and an alluring, muted gorgeousness that Davis cultivated lovingly. I'd never realised the piece could offer quite so much; or that the speech-role of Persephone could register with such unaffected tenderness as the Swiss actress Irene Jacob managed.

Mstislav Rostropovich is 70 this year and celebrating the event (as Pierre Boulez did before him) in a birthday series with the LSO. Running at the Barbican and elsewhere, it features the great man in his dual capacities as cellist and conductor; and the printed programme flags his unique, extra-musical visibility in world affairs with a portfolio of birthday greetings from the great and good, including two Royals. Baroness Thatcher writes with surprising passion, Mr Major with perfunctory blandness. No doubt he has other things on his mind.

Tuesday's instalment was the least glamorous in the series but potentially interesting in that it comprised three works written for Rostropovich to conduct, which all present problems. Walton's Prologo e Fantasia is his last score, written when he was ill and desperate for ideas; Lutoslawski's Novelette is a misleadingly titled score of violent but empty contrasts; Schnittke's Sixth Symphony is thin-textured, and fragmentedly short-breathed. How you make a vital performance from any of these I don't know, but I thought Rostropovich might. Hence the interest.

It turned out that he didn't know either; nor did the LSO who got lost, several times, and seemed insecure, even when they and the conductor were together. But at least there was a sense of trial and effort to it all, enriched by Rostropovich's endearing personality; and to that extent it was a more rewarding experience than the premiere of Michael Nyman's Concerto for Saxophone & Cello at the Festival Hall. Inflated by fatuous razzmatazz at the instigation of the Japanese car company that commissioned the score, this dismal evening had more to do with marketing than music; and the best I can say is that behind it was a good idea worth exploring: namely, the close sound-relationship between the top register of the cello and bottom register of the sax. What exploration actually took place is anybody's guess, because the whole piece was a half-hour blur of brazen noise - grotesquely amplified with no concern for balance. I barely heard the cello; and the orchestra had clearly been told to play with all the relentless, triple-forte pump-action it could muster. For the record, the orchestra was the Philharmonia, the soloists John Harle and Julian Lloyd Webber. Good musicians, wasted.