Poll tracker: How the parties compare
Check the latest opinion polls trends with the poll tracker, which measures general election voting intention.
Note on margin of error: Polling companies generally claim that 95% of the time, a poll of 1,000 people will be accurate within a margin of error of +/-3 percentage points. This means that a figure in the poll could be up to three percentage points higher or lower than that shown.
Election 2017 updates from senior elections and political analyst Peter Barnes
13 June: How accurate were the polls?
Once again the polls, taken as a whole, were not a good guide to the election result.
Over the course of the campaign the gap between the main two parties narrowed but, with one exception, the final polls all suggested a clearer Conservative lead than the actual outcome.
7 June: Final polls
Nearly all the final polls are now out. Ipsos Mori will publish on Thursday and ICM are expected to update their preliminary figures from earlier today. Apart from that, we're there. What should we take from them?
First, it's worth making the obvious point that every single poll throughout the campaign has put the Conservative Party ahead. If Labour receives more votes tomorrow it will be a bigger polling failure than in 2015.
- Manifesto guide: Where the parties stand
- How the BBC reports polling day
- How to follow results night and what to watch for
Second, the range of vote shares for the parties in polls published on Tuesday and Wednesday are CON 41-46%, LAB 34-40%, LD 7-10%, and UKIP 2-5%. Not all the polls separate out the SNP, Plaid Cymru or Green Party.
We've looked a lot at the gap between the two top parties - arguably too much. But the polls suggest it could be anywhere between a one and 12-point Conservative lead, with Survation suggesting the closest result.
That suggests anything from a small swing to Labour to a 2-3% swing to the Conservatives. And it's doubly difficult to estimate how that could play out in terms of the number of seats each party wins. So much will depend on variations in different parts of the country and in different types of constituency. There could be no national swing between Conservatives and Labour but still a significant shift in seats.
The BBC/ITV/Sky exit poll will be published at 10pm on Thursday, when polls close. It would be an understatement to say it's eagerly awaited.
Scotland and Wales
The final polls in Scotland suggest a fairly settled picture - although there have only been seven of them during the whole campaign so there's not much to go on. The SNP are clearly out in front on 40% or a little higher. The Conservatives and Labour are in a battle for second place at around 25%, perhaps with the Conservatives' ahead by a nose.
That would represent a significant shift from the actual 2015 result of SNP 50%, LAB 24%, CON 15%.
In Wales, YouGov are the only company to have produced polls throughout the campaign so we have even less evidence. Their last, released on Wednesday, showed little change from the previous two. Labour remain on 46% and the Conservatives fall back one point to 34%. Plaid Cymru are on 9% with both the Lib Dems and UKIP on 5%.
That implies a modest swing to Labour with UKIP losing well over half their 2015 votes.
7 June: Final polls - what to expect
Three polls have been published since the weekend - by Opinium, Survation and ICM. None of them showed dramatic changes from what they'd had before.
The race is still very close according to Survation - their phone poll showed similar figures to their weekend internet poll - but the Conservatives have a handsome lead according to ICM.
Opinium's poll (CON 43, LAB 36, LD 8, UKIP 5, SNP 5, GRN 2) was their last of the campaign.
Lots of other polls are expected today.
We should have final figures from ComRes, ICM, Kantar Public and YouGov. Others are possible as well. We'll update the poll tracker as they are released.
Ipsos Mori are expected to publish their final figures on Thursday.
That won't get added to the BBC poll tracker because of editorial guidelines, which state that no opinion poll on any issue relating to the election can be published on polling day.
3 June: Last weekend of polls
The normal array of Saturday night polls have been published - six in total. All of them still have the Conservatives in the lead but by a widely-varying degree. The gap over Labour ranges from just one point in Survation's poll to 12 points according to ComRes.
Clearly, the discrepancies between the polling companies are very large. And that obviously makes it hard to draw firm conclusions.
Survation's poll has the narrowest gap of any poll so far during the campaign. It's the only company that has not changed its methodology since 2015. But up to now they've not been especially favourable to Labour.
Their poll was conducted today, after last night's Question Time special, where Theresa May and Jeremy Corbyn faced questions from a live audience. And they found that more people who watched the programme said it made them more likely to vote Labour than Conservative.
It's very unclear whether this could really have caused a significant shift though. A further complication is that most of Survation's polls in this campaign have been conducted by phone. This was an internet poll so it makes comparisons with their previous figures harder.
Ipsos MORI also published a poll on Friday with a Conservative lead of 5%. In addition to the voting intention numbers, they reported a dramatic change in attitudes towards the two main party leaders.
Since she became Prime Minister, Theresa May has always enjoyed positive approval figures - more people have said that they are satisfied with her performance than dissatisfied. But in their latest poll she has slumped to a negative overall figure of -7% (43% satisfied, 50% dissatisfied).
Meanwhile, Jeremy Corbyn's rating has shot up. He's still in negative territory and, on -11%, still behind Theresa May. But that contrasts with -41% in March, before the election was called.
Other pollsters have seen a similar trend. ComRes have also reported Theresa May's first negative negative rating (-3%). Opinium put Jeremy Corbyn on -12%, up from -35% at the start of the campaign. Theresa May has fallen to +6%.
More polls to come
For ORB tonight's poll will be their last of the campaign. Their final figures are CON 45%, LAB 36%, LD 8% and UKIP 4%.
Other companies will be publishing further polls in the final week of the campaign though. YouGov have also been updating their seat projection - that should also continue in the coming days.
31 May update: How close is the gap now?
After their striking seat projection published last night, YouGov has a new poll this evening which has the narrowest gap between the main two parties so far during the campaign. It has the Conservatives on 42% and Labour on 39% - their highest rating in any poll since 2014.
It's not all good news for Labour though. Kantar Public also published a poll on Wednesday with a slightly increased Conservative lead compared to their previous poll. They now have the Conservatives on 43% and Labour on 33%.
Its not the only company which still shows a large gap. In fact, it's more than 10% according to ICM and ComRes.
The main reason for this disparity is the different ways that the pollsters estimate turnout.
Polling companies are sometimes accused of "herding" - manipulating their figures so they all say the same thing. Nobody could accuse them of that at this election. If anything, the differences have become more stark as we approach election day.
There was also a rare Scottish poll published on Wednesday afternoon. The SNP, on 43%, maintained the commanding lead that they've had throughout the campaign and for a long time before that. More interesting was that Labour and the Conservatives were on level pegging at 25%. That's the first time this year that the Conservatives haven't been in a clear second place.
As with the other nations, though, it's hard to draw conclusions when there have been so few polls.
31 May: YouGov seat projection
The polls covered in the BBC tracker are standard voting intention polls. That is, they try to gauge levels of support for each of the parties across the country. They do not attempt to say how many seats each party would win.
There's a good reason for that.
The first-past-the-post electoral system means that there is no straightforward relationship between votes and seats won.
The 2017 election could be almost a repeat of the 2015 result in terms of vote share but produce a very different outcome in terms of the number of MPs returned for each party.
Clearly, though, people are interested in knowing how votes might translate into seats.
So YouGov have attempted to model how that could work. Their projection of a hung parliament has, understandably, caused a stir.
The model used to make this projection is complicated.
It's not based on a single poll, but on the responses of 50,000 people who have taken part in other YouGov polls. Nor does it work by just counting up how the respondents in each constituency said they would vote.
Even with 50,000 people there wouldn't be nearly enough to have a reliable sample in each of the 650 constituencies in the UK.
Instead, it uses demographic information about the people who take part in the polls, as well as information about where they live and how they've voted before, to try to work out how different types of people are likely to vote.
Then it looks at the demographic make-up of each constituency and comes up with an estimate of the likely outcome.
YouGov themselves have indicated that their projection is subject to quite a lot of uncertainty.
And Stefan Shakespeare, the company's Chief Executive, has said that it would only take a small shift in favour of the Conservatives to see them win a healthy Conservative majority.
So it should definitely be treated with caution.
There's been no shortage of Britain-wide polls but in the individual nations it's a different story.
With so few polls focused on the election in Scotland, Wales or Northern Ireland it's impossible to identify any well-supported trends.
That said, there has been one Welsh poll since I last mentioned them which showed a very different picture from the two that had gone before.
YouGov gave Labour a 10 point lead in Wales whereas previously they'd shown a clear Conservative advantage.
The direction of shift is consistent with what we've seen in the national polls but it's very difficult to know what to make of the particular numbers. It's a huge turnaround and we really would need more evidence before drawing any firm conclusions.
27 May: Conservative lead continues to shrink
Five polls have been published this evening. Overall, they reinforce Thursday night's polls which suggested that the Conservative lead over Labour has shrunk again since last weekend but there are quite large differences between the pollsters' figures.
ICM and ComRes have a larger gap between the two main parties than the other companies. They put it at 14% and 12% respectively but that's still down from what they were showing two weeks ago. The others have it at 6%-10%.
There will no doubt be some talk about the fact that Labour has dropped two points in YouGov's poll compared to the one they published on Thursday. But that still means that the gap is narrower than they found last weekend and it would be unwise to read too much into a small change between two individual polls.
The general trend is clear. The Conservatives are still ahead but Labour has closed the gap.
Is the shift believable?
The godfather of psephology or the study of elections, Sir David Butler, said this week that the movement in the polls over this campaign is bigger than in any election he's covered since 1945. But many experts think that election campaigns don't make much difference - voters' views are set years or months before polling day.
So does that mean the polls must be wrong? Not necessarily. Voters have become much more volatile than they used to be - much less loyal to a single party.
Chris Prosser and Jane Green from the British Election Study have shown that over the last 50 years the proportion of voters who switch parties between elections has risen from around 15% in the 1960s and 1970s to 43% in 2015.
We can't be sure that the same will happen this time and many of the people who do switch might have made up their minds before the campaign started. It's also possible that people who said they were going to switch at the start of the campaign have gone back to the party they supported before.
But if voters are more likely to change allegiance than in the past then perhaps we shouldn't be too surprised if the polls show large movements.
How reliable is Labour's vote?
One reason for caution about Labour's advance is that their improved polling numbers are based on very high levels of support among young voters and people who didn't vote in 2015.
A big part of the problem with the polls in 2015 was that they failed to estimate accurately the difference in turnout rates among different groups of voters. In particular, they underestimated the turnout gap between young and old voters.
As I've mentioned before, many of the pollsters have adjusted their methods to try to overcome this problem.
But concerns remain about whether the people who say they'll vote Labour will actually do so on 8 June. If they don't then Labour's vote share could be lower than the polls suggest.
26 May: Gap narrows further
Two new polls published on Thursday night suggest a further continuation of the trend we've seen throughout the campaign so far. The gap between the Conservatives and Labour has narrowed again. It's down to eight points in Kantar Public's poll and just five points in YouGov's - smaller than the actual gap between the parties at the 2015 election.
The Manchester attack has, rightly, meant that the country's attention has been focused away from the election. But as the campaign resumes, it's inevitable that people will ask whether it's had an effect on how people might vote.
So it's important to note that Kantar Public's poll was conducted before the attack, but YouGov's was done afterwards - on Wednesday and Thursday.
That doesn't mean we can say that the latest shift is related to the attack.
Theresa May has consistently had a strong lead over Jeremy Corbyn on the issues of defence and security. That hasn't changed. YouGov found that 55% of respondents trusted Theresa May to make the right decisions to keep Britain safe from terrorism compared to just 33% who trusted Jeremy Corbyn.
It's possible that the polls are still reacting to the fallout of the manifesto launches last week - and the revisions to the Conservatives' social care plans announced on Monday (they denied it was a U-turn). There could also just be an element of natural polling variation.
Whatever the cause, the important thing to remember is to look at the general trend rather than concentrating too much on individual polls. The Conservatives retain a clear lead over Labour but the gap has closed significantly.
20 May: Labour closing the gap
After the week of the manifesto launches for the main GB-wide parties, the trend of Labour improvement that we've seen throughout the campaign is continuing.
In fact, four polls released on Saturday, for the Sunday papers, suggest that the Labour advance has strengthened. And, while the Lib Dems and UKIP are still below where they were when the election was announced, for the first time it looks as though the latest Labour rise is coming at the expense of the Conservatives.
The Conservative lead was generally around 20 points or a little above in the middle of April. It's down to between 9 and 13 points in today's polls.
As ever, the figures should be treated with caution. It's particularly worth noting that none of the most recent polls come from the companies that tend to put Labour on a lower figure.
ICM and Kantar Public both make adjustments based on turnout rates for different groups of people at previous elections. They've had Labour at a lower level than some of the other companies throughout the campaign. So it will be interesting to see whether their next polls follow a similar pattern.
What's causing the shift?
Looking beneath the voting intention figures there are some clues about what might be causing the shift towards Labour that we've seen.
One is voters' responses to the manifestos. For example, Survation found that nearly half of respondents opposed Conservative plans for changes to the system of funding for social care. By contrast, Labour's plan to increase income tax for people earning over £80,000 was backed by more than six out of ten people and opposed by under a quarter.
Jeremy Corbyn has also seen an improvement in his approval ratings although he still lags a long way behind Theresa May. For example, Opinium have him on a net figure of -18 this week, up from -28 last week.
Still a big gap
Despite the significant moves we should still remember that the Conservatives retain a large lead by historical standards.
If today's polls were repeating on 8 June it would be the biggest gap between the two parties since Tony Blair's landslide in 1997 in terms of share of the vote.
There's also some evidence in the polls that the Labour vote is "softer" than the Conservative vote. In other words, more Labour voters than Conservative voters say they might change their minds.
15 May: Labour improvement
Four polls over the weekend reinforced the picture of a Labour improvement during the course of the campaign so far. ORB, Opinium, ComRes and YouGov all had them at 30% or above - clearly above the levels seen at around the time the election was announced.
However, this increase has not come at the expense of the Conservatives who remain in the mid-to-high 40s with a commanding lead.
The main losers have been UKIP, who are down in the 3-6% range.
If the current polls were reflected in the final result it would mean the two main parties between them capturing a significantly larger share of the vote than at recent elections.
In 2015, they received a total of 69%. The polls suggest a joint share of almost 80%.
You have to go back to 1992 to find an election where the total Conservative and Labour share was close to that - the figure was 78%. The last time it was above 80% was 1979.
Scotland, Wales and Northern Ireland
All the polls shown in the tracker report vote intention share across Great Britain, but that's not very helpful for working out what's going on in parts of the UK where other parties stand.
However as we've only seen two campaign polls in Scotland, two in Wales and one in Northern Ireland so far, there's not very much evidence to go on.
Having said that, both of the Scottish polls, conducted by Survation and YouGov, have given the SNP over 40% with a clear lead over the Conservatives on 28% and Labour down in third place on 18%.
In Wales, two YouGov polls have put the Conservatives ahead of Labour, but with a smaller gap than across Britain as a whole.
Like in the national polls, Labour's share has seen an improvement from the beginning of the campaign. Plaid Cymru are back in third place
In Northern Ireland, the single Lucid Talk poll gave the DUP a narrow lead over Sinn Fein with the UUP, SDLP and the Alliance all some way back.
Most important issue
As well as asking people which party they intend to vote for, pollsters also ask which issues are the most important. In 2015 the three biggest issues were the NHS, immigration and the economy.
A significant change at this election is the emergence of Brexit. Polls conducted since the election was called have put it at the top of the list of important issues ahead of the NHS in second place with the economy and immigration battling it out for third.
This may help to explain, at least in part, the Conservatives' lead in the polls.
So far as we can tell, they've managed to attract the support of a large number of new voters who backed leave at last year's referendum whilst holding on to most of their own supporters who backed remain - many of whom now think that the Government has a duty to implement the outcome of the referendum.
11 May 2017: Should we ignore the polls?
As everybody knows, the polls got the 2015 general election wrong.
They suggested that the likely outcome was a hung parliament but, as we know, the Conservatives won an overall majority. So is it worth paying attention to them this time?
Well, we certainly shouldn't assume that the result will be exactly what the polls say. But that doesn't mean they're completely useless.
For one thing, critics have perhaps exaggerated other polling "disasters".
The belief that the polls were just as bad at the EU referendum and in the US Presidential election is widely held. However, whilst some polls gave a misleading picture at the referendum, others were pretty close.
We reported at the time that the polls overall indicated a very narrow race in the weeks running up to referendum day.
Similarly, at the US election, the national polls weren't that far off in terms of the share of the vote won by Donald Trump and Hilary Clinton - Clinton actually won the popular vote by over 2%.
It was state polls that were unreliable and led to a misleading impression as to who would emerge from the Electoral College system as president.
In both cases, if all we'd taken from the polls was that the votes would be close, we would have been on the right lines.
The polling companies have also made adjustments to try to prevent the problems of 2015 from happening again and to resolve issues that arose at the referendum.
These methodological changes vary from pollster to pollster but there are some general trends.
Several of them now ask the people who take part about their educational background. The aim, as with questions about class, age, gender and region is to get a sample of people who are representative of the population as a whole.
Others have developed more sophisticated ways to estimate how likely it is that somebody who takes part in a poll will actually vote. Just asking people whether they will vote is not a good guide.
Of course, we can't be sure whether these adjustments will make the polls more accurate. So some people will no doubt decide to ignore them all together.
But there's still clearly an appetite for them.
No fewer than 30 have been conducted since the Prime Minister made her surprise announcement on 18 April.
That's more than one a day.
What's happened since the election was announced?
It's now three weeks since the election was announced and the official campaign is well under way.
After Theresa May's surprise statement, the Conservatives saw their poll rating jump with several polls suggesting a comfortable 20 point lead.
Since then, nothing very dramatic has happened. There has been a modest uptick for Labour, who are generally up to the high 20s or around 30 - up from the mid 20s just after the announcement.
But that still leaves a very large gap between the main two parties.
UKIP seem to have slipped a little further down and perhaps the Lib Dems have also fallen back a bit, although these trends are not clear.
How are polls actually carried out?
Most opinion polls, and all of the ones covered in the BBC poll tracker, are either conducted by telephone or online.
For phone polls the polling company rings up landline and mobile numbers.
In principle, anyone with a phone could be asked to participate.
For internet polls, the company maintains a panel of people who are prepared to take part. For each poll they will contact the required number of panel members.
In both cases the company will aim to survey a sample of people who are representative of the country as a whole - in terms of age, gender, social class, etc.
They will generally then apply weighting adjustments if one or other group is over-represented or under-represented in their sample.
It's also common to seek a representative sample or apply a weighting based on past-voting behaviour.
Polls included: All polls conducted by companies which are members of the British Polling Council. This includes: BMG, ComRes, GfK, ICM, Ipsos-Mori, Opinium, ORB, Populus, Panelbase, Survation, Kantar Public (TNS-BMRB) and YouGov.
Sample area: Polls record voting intention for Great Britain (England, Scotland and Wales)
Dates: Polls are ordered by latest date of fieldwork.
Margin of error: Polling companies generally claim that 95% of the time, a poll of 1,000 people will be accurate within a margin of error of +/-3%. This means that a figure in the poll could be up to three percentage points higher or lower than that shown.