Some have taken to respected pages to discuss the Administration’s purported next set of nuclear weapons decisions.  As the fourth anniversary of the President’s Prague speech nears, advocates are petitioning and pushing the President to make deep cuts in American nuclear weapons, perhaps down to 1,000 to 1,100 deployed warheads.  Given difficulties in verifying nuclear warhead eliminations, as well as Russian resistance to new, legally-binding limits for reduction (of offensive, not defensive weapons, and shorter-range, or theater nuclear weapons) what appears likely is a unilateral announcement that by a date certain (but before January 2017, or for some period of time?) the United States will deploy not more or less than a number or range of numbers the interagency decides it can handle and Russia can answer, relying on existing verification to do the work of providing confidence about lower limits.

The fact that public advice now argues for an executive agreement is revealing.  The presumption is you cannot obtain Senate consent to go lower.  That was revealed in 2010–the price for going lower then, and not until 2018, was deemed too high by advocates of going lower, even while they supported 1,550 deployed warheads by February 2018.  Now, one is too high, the other, too far off (for some).

It is not without approximate precedent to do as described above.  There is a rough analogue:  The Vladivostok Agreement.  Carter faced an expensive modernization choice he didn’t like and ultimately insufficient support in his own party for going lower.  There is little difference between Obama and Carter when one considers the Prague speech.  There is some degree of commonality in their economic woes.  The terms of Vladivostok sound very similar to what is guessed at, above.  Yet, if the motivating factor in seeking an executive agreement, instead of an amendment to New START, is to avoid the need for two-thirds of the Senate, then we ought to ask ourselves why and what that would mean–it is not clearly the case that Congress would have no role and it is more certainly the case that there is not a majority in the House willing to bless lower numbers.

Whether Russia might accede is also in doubt, so to go lower without obtaining something in return strikes me as an incomplete idea.  It might be wise to step back and ponder what’s happened since Vladivostok in terms of extant practice in order to better understand why it may be the case that New START could end up as the strategic warhead limitation treaty or SWHaLT, and why that might not work.  It might also be wise to focus on verification needed to go lower before going lower, particularly given the amount of time we have before 2018 and the uncertainty we have about our own modernization, replacement or maintenance of the weapons we have now.  We might also wonder about what it is we are trying to (a) reduce, (b) limit or (c) eliminate, how each is verified, and (d) what is to be retained and/or limited inside and outside limits.

Arms control aims to reduce the risk of use, but it does not proscribe it.  Nor does it mandate eliminating anything even if it permits taking a look to make sure an item has been eliminated from accountability.  The best task of arms control has been to detect the signals of intention in order to reduce the risk of nuclear war, not to end all wars nor to get rid of all nuclear weapons.

I take exception with the view that “arms control is back” because an Administration might want to go lower.  There is a push for unilaterally-imposed reduction without enough regard for verifiable limitation, at least from the outside.  One could argue this is not so bad:  Unilateral decisions can be reversed, no?  Not really, at least not for nuclear weapons.  Nor for the verification of decisions involving them, to say nothing of postponing the verification questions for later, after disarmament.

The choice to restore a weapon or a number of them covertly or overtly sends a signal that may be detected or announced.  To the extent such a choice violates an agreement, it could be reversed, but that is harder to do than one might suppose.  The military significance of a violation is ultimately a policy decision.  But any suitable arms control regime ought to try to detect violation before militarily significant advantage can result based on the limits imposed.  At very low numbers, the policy judgement might sway toward significance more often than not, unless one never wants to admit the possibility, or the means have not been provided to make sound judgments regarding significance.

Four brief questions arise if Vladivostok II/SWhaLT I is the option:

(1) Does it require amending the New START Treaty?

(2) Under U.S. statute, does it require a treaty be submitted to the Senate–i.e., is such a reduction militarily significant?

(3) Does it require a new assessment as to its degree of verifiability under U.S. statute (same link as previously placed, in the same law)?

(4) Would Congress have to approve it by joint House-Senate resolution?–If so, could Congress insist on changes to the agreement, or to New START?

Backup Material     

New START did away with limits and attendant obligations determined to be applicable only to the Cold War, whether because of perception or reality.  I don’t mean aggregate force levels, I mean detailed sub-limitations on throw-weight, re-entry vehicle attribution to declared missile types, full telemetric data exchange, on-site continuous monitoring of a solid rocket motor integration facility and more.  They were not perfect, but they were the best that could be done to count, declare and confirm under the INF Treaty and START I.

As change unfolds, a calculation based on the numbers and types of weapons (in which category launchers and delivery vehicle used to matter more) and numbers and types of inspections (which is all that New START has, next to declarations) for the probability of detecting violation becomes complicated by going lower and not changing the treaty.

In this scenario, I am assuming the Parties make independent, unilateral statements that they will deploy no more than 1,000 to 1,1000 warheads, relying on the New START treaty to verify that, with no change in the obligations under that treaty and/or pending a new one.  Ability to “upload” in response might not necessarily be eliminated under a reasonable approach to reduction, but that is not clearly what the advocates want, at least in the United States.  At 1,000 the variances in new weapons might make noncompliance more severe because the limits are lower and because daily deterrence, extended deterrence and reassurance are tighter.  Noncompliance involving tens of missiles could amount to hundreds of warheads that count for more advantage outside lower limits.  This is, of course, true at 1,550, too, but the significance of unilateral American limitation and/or violation of the other party grows large if we mean to eliminate capability and warheads, as is proposed, on our own.

Moreover, the preoccupation with warheads imposes difficult verification questions.  If you want confidence regarding limited warheads, then you probably need a better idea of how many fit on a deployed delivery vehicle than you might have in the future.  That could mean moving away from flexible rules for bomber-accountable weapons.  It could mean you fiddle, again, with the idea of what a deployed weapon really is, but that will mean choices involving tradeoffs for sovereignty that nobody has yet made.  It does mean, however, in my opinion, that you’d need more than we presently have available under existing obligations for a confident judgement about 1,000 total deployed warheads.  New START was sold on the basis that it treated non-deployed weapons lightly and didn’t require any choices for a long time that could result in asymmetric reduction before its Article II obligations had to be met.  That was made crystal clear in 2010.

I have appended a portion of the Senate Committee on Foreign Relations START I report to this post to give the reader a sense for how one might begin to better understand what is meant.  Note well, due to the combinations of START I inspections in New START, you cannot make a direct extrapolation based on the inspection number in the 1992 report excerpted because New START combines a data-update inspection with re-entry vehicle onsite inspection or RVOSI.  You cannot directly compare since there is no warhead attribution in New START.  So, why bother?

Because it might give you a more accurate sense of what START I inspections could reasonably do and might lead you to wonder what the successor to New START might try to do.  My own view is that you should try to get more if you are contemplating lower deployment with newer systems in Russia not previously declared under START I but that have been deployed since 2009.  You might even want to do more than START I allowed, depending on how many new types of items and locations are involved.  You might want to focus on the purpose of inspections, too, and the fact that there only ten per treaty year for deployed weapons under New START.

Even though Russia may only have 35 sites as compared to almost 70 in the past, if one adds new types of missiles with no declared throw-weights and less than full telemetric data exchanged regarding flight testing (and if one is a glutton for probability calculations, non-strategic nuclear warheads, too, governed also in new limits) the analysis would presumably support a need for more verification that involved more than warhead counting but also delivery vehicle capacity.  More inspections would not hurt, but inspections do nothing more in present practice than confirm what is known, based on treaty declarations and national technical independent verification.  If one is contemplating going lower than 1,550, with no change in extant obligations with respect to new, lower limits, then one is not serious about the lower limit.

The United States may take its unilateral chances as it may, but it ought to get something for them.  If present policy is to rest on narrow, amalgamated limitation and verification, we might be more honest about what we have, now, before going lower.

Verification is not a form of disorder–it is not a mania.   At very low numbers it becomes more, not less, important.  It is notable that the Four Statesmen have now put some focus back on verification.

Trust is also not a predicate for verification, even if it may result from it.  But if you purposefully seek reduction without it, or put priority on it after you go lower, you may find yourself in the unenviable position of losing trust as well as finding yourself with no means to restore it.  The focus on Russia is due to the fact that we have no other agreements nor any substantial record of verification, implementation and compliance on which to rely.  What are the things we seek to limit there, now?  If we are only focused on reduction, are we sure they are, as well?  Do we understand, now, where they will be in 2018?  If we do not have something like Nunn-Lugar, how much will we know about what is eliminated?

There are two agendas described here–one that might be the Administration’s choice and another that it has embraced.  The third, arms control, is not a focus of either.

Appended Material

U.S. Congress. Senate Committee on Foreign Relations. The START Treaty, Report with Additional Views to Accompany Treaty Doc. 102-20 and Treaty Doc. 102-32. 102nd Congress, 2d Session, 1992.  Senate Exec. Rept. 102-53, pp. 46-48.  (Material has been edited to correct typos in the original, published report):

How many inspections are enough?

Data update inspections.  The Treaty allows 15 data update inspections each treaty year.  If the number of C.I.S. declared sites available for these inspections remained at about 68, there would be on average a visit to one-fourth of the site[s] each year.  If Russia remains the only nuclear-weapons state republic, and it has some 45 sites, then up to one-third of the sites would be inspected each year.  Since information from NTM would guide these OSIs to the more suspicious, or the more important declared sites, the relative effectiveness of OSIs would be enhanced beyond the one in three ratio.  If the Russians were to attempt a break-out of the Treaty, they would undoubtedly need to cheat at several sites in order to gain a large numerical advantage, thus increasing their risk from the additional sites.  From the point of view of the Russians, they would have to face this considerable risk since they would not know if they would be discovered by NTM, OSIs [,] human intelligence or defectors.

Reentry vehicle inspections.  If the Russians were to consider violating the RV-counting rule provisions, they would probably do it in large scale on just a few types of systems (SS-18, or downloaded SLBMs) in order to gain many extra warheads.  In order to gain a statistical perspective on this type of cheating, the Committee asked the following question for the record: Assume that the Russians deploy 14 RVs, instead of 10, on 25 of its 154 SS-18s. How many OSIs would it take to have a 50% (75%, 95%) confidence of detection of one violation?  What are the limits of this type of calculation?

In January 1992, the Administration responded: The following table shows the number of OSIs required to detect one violation out of 25 at different confidence levels:

Confidence Level Number of Inspections
50 percent 4
75 percent 8
95 percent 16

The key factor for this type of calculation is the number of violations assumed.  The number of violations regarded as significant is a policy judgment, not a technical one.  However, two important assumptions in assessing OSI are: 1) that the selection of missiles for RV OSI is completely random (the Russians cannot influence our selection in any way), and 2) that inspections of “illegal” silos cannot be foiled, or prevented on some pretext, by the inspected Party.  START provisions concerning pre-inspection restrictions, time limits for transporting inspectors to the designated site, and use of satellite navigation aids, add to the difficult[y] a Party would confront in attempting to foil RV OSI.

The probability of detection of this violation of 100 warheads should be compared to the national security implication of the violation.  The violation discussed above would increase the number of actual Russian warheads under START from a projected 7,000 warheads under START to 7,100 warheads with this level of cheating.  The effectiveness of this violation would be calculated by the additional effect of the 100 warheads on U.S. targets.  These results by the Executive Branch are consistent with the attached graphs from the Office of Technology Assessment.

Statistics of inspections.  In the above discussion we have quoted the chance of being caught on a random, statistical basis.  Some of the limitations of this approach are as follows:  These calculations do not depend on the size of the anomaly on each missile.  What is intended here is that these are clear-cut anomalies that are observable.  There are a number of factors that can make this calculations both more optimistic, or more pessimistic, requiring either more or fewer OSIs to obtain a condition of reality for the necessary confidence.  For example: (1) For random statistics, we have assumed that each violation will be detected if the RV OSI picks the particular missile for inspection.  In order to determine whether the Russians could hide the violation, one must examine:  The time lines of the inspection to determine the time available to the Russians to cheat, the availability of NTM assets to observe the cheating, and the availability of equipment to remove the extra RVs.  If the Russians were able to cheat under certain circumstances, more OSIs would be needed. (2) The site of the OSIs will most likely not be chosen randomly, but their choice will be aided by NTM, human intelligence, other data or by detectors.  This kind of information would allow the targeting of OSIs to locations where the likelihood of violations would be higher. This effect would improve probabilities beyond that of random statistics and, thus, reduce the number of OSIs that would be needed to obtain the required degree of certainty.

Are there enough OSIs?  In order to determine whether the number of on-site inspections is sufficient to effectively verify the START Treaty, it is necessary to compare the monitorability as determined by intelligence estimates or by statistical arguments given above, and then assess whether the militar[y] significance of the additional warheads changes (1) deterrence as measured in the survivability of U.S. strategic forces, or (2) the threat to U.S. population[.]


Questions for the Record Submitted to

Assistant Secretary Gottemoeller by

Senator Roland W. Burris (#4)

Senate Armed Services Committee

July 29, 2010



Secretary Gottemoeller, in the New Strategic Arms Reduction Treaty (START) there will be an annual quota of 18 inspections, instead of 28 as under the START I, and only 35 sites are eligible for inspection instead of the 70 sites under the START I.  Can you describe why these reductions are in our best interest?


The New START Treaty verification provisions are tailored to verify the requirements of the New START Treaty, which are different from the START Treaty requirements.

The New START Treaty provides for an annual quota of up to 18 short notice, on-site inspections to aid in verifying Russian compliance with its Treaty obligations.  These inspections will provide U.S. inspectors with 18 opportunities per year to select from among declared Russian strategic forces facilities to verify the accuracy of Russian data declarations and to deter cheating.  Although the new Treaty provides for fewer inspections than the annual quota of 28 permitted under the original START Treaty, the number of facilities for which Russia provided site diagrams and which will therefore be inspectable under the New START Treaty (35) is also significantly lower than the number of inspectable facilities in the former Soviet Union when the START Treaty entered into force (70).  As explained in the response to QFR #5 below, this is due to the fact that Belarus, Kazakhstan and Ukraine are not Parties to New START, as well as that Russia now has fewer facilities where strategic offensive arms are located than it had when START entered into force.

The New START Treaty annual inspection quota includes 10 Type One inspections of deployed and non-deployed strategic offensive arms, which will be conducted at operating bases for ICBMs, ballistic missile submarines (SSBNs), and nuclear-capable heavy bombers.  Type One inspections combine many of the aspects associated with two different types of inspections that were conducted separately under the START Treaty; specifically, reentry vehicle on-site inspections and data update inspections.  Thus, fewer inspections annually at the operating bases will achieve many of the results of the previous START Treaty inspection regime.  The quota also includes eight Type Two inspections focused on non-deployed, converted, or eliminated strategic systems, which will be conducted at facilities such as storage sites, test ranges, and conversion or elimination facilities, as well as formerly declared facilities.

The Administration assessed the number of Type One and Type Two inspections needed annually to meet U.S. inspection objectives as the nature of these inspection types emerged during the New START negotiations.  These assessments ultimately concluded that an annual quota of 18 such inspections would be adequate to meet U.S. inspection needs.